diff options
35 files changed, 15964 insertions, 27 deletions
diff --git a/libgfortran/ChangeLog b/libgfortran/ChangeLog index 4ada8b8..3e6c0ad 100644 --- a/libgfortran/ChangeLog +++ b/libgfortran/ChangeLog @@ -1,3 +1,43 @@ +2017-05-25 Thomas Koenig <tkoenig@gcc.gnu.org> + + PR libfortran/78379 + * Makefile.am: Add generated/matmulavx128_*.c files. + Handle them for compiling and setting the right flags. + * acinclude.m4: Add tests for FMA3, FMA4 and AVX128. + * configure.ac: Call them. + * Makefile.in: Regenerated. + * config.h.in: Regenerated. + * configure: Regenerated. + * m4/matmul.m4: Handle AMD chips by calling 128-bit AVX + versions which use FMA3 or FMA4. + * m4/matmulavx128.m4: New file. + * generated/matmul_c10.c: Regenerated. + * generated/matmul_c16.c: Regenerated. + * generated/matmul_c4.c: Regenerated. + * generated/matmul_c8.c: Regenerated. + * generated/matmul_i1.c: Regenerated. + * generated/matmul_i16.c: Regenerated. + * generated/matmul_i2.c: Regenerated. + * generated/matmul_i4.c: Regenerated. + * generated/matmul_i8.c: Regenerated. + * generated/matmul_r10.c: Regenerated. + * generated/matmul_r16.c: Regenerated. + * generated/matmul_r4.c: Regenerated. + * generated/matmul_r8.c: Regenerated. + * generated/matmulavx128_c10.c: New file. + * generated/matmulavx128_c16.c: New file. + * generated/matmulavx128_c4.c: New file. + * generated/matmulavx128_c8.c: New file. + * generated/matmulavx128_i1.c: New file. + * generated/matmulavx128_i16.c: New file. + * generated/matmulavx128_i2.c: New file. + * generated/matmulavx128_i4.c: New file. + * generated/matmulavx128_i8.c: New file. + * generated/matmulavx128_r10.c: New file. + * generated/matmulavx128_r16.c: New file. + * generated/matmulavx128_r4.c: New file. + * generated/matmulavx128_r8.c: New file. + 2017-05-19 Paul Thomas <pault@gcc.gnu.org> Jerry DeLisle <jvdelisle@gcc.gnu.org> @@ -14,7 +54,7 @@ (st_endfile): Likewise. (st_rewind): Likewise. (st_flush): Likewise. - + 2017-05-15 Jerry DeLisle <jvdelisle@gcc.gnu.org> PR libgfortran/80727 diff --git a/libgfortran/Makefile.am b/libgfortran/Makefile.am index 2857af5..9088812 100644 --- a/libgfortran/Makefile.am +++ b/libgfortran/Makefile.am @@ -460,6 +460,21 @@ $(srcdir)/generated/matmul_c8.c \ $(srcdir)/generated/matmul_c10.c \ $(srcdir)/generated/matmul_c16.c +i_matmulavx128_c= \ +$(srcdir)/generated/matmulavx128_i1.c \ +$(srcdir)/generated/matmulavx128_i2.c \ +$(srcdir)/generated/matmulavx128_i4.c \ +$(srcdir)/generated/matmulavx128_i8.c \ +$(srcdir)/generated/matmulavx128_i16.c \ +$(srcdir)/generated/matmulavx128_r4.c \ +$(srcdir)/generated/matmulavx128_r8.c \ +$(srcdir)/generated/matmulavx128_r10.c \ +$(srcdir)/generated/matmulavx128_r16.c \ +$(srcdir)/generated/matmulavx128_c4.c \ +$(srcdir)/generated/matmulavx128_c8.c \ +$(srcdir)/generated/matmulavx128_c10.c \ +$(srcdir)/generated/matmulavx128_c16.c + i_matmull_c= \ $(srcdir)/generated/matmul_l4.c \ $(srcdir)/generated/matmul_l8.c \ @@ -641,7 +656,7 @@ gfor_built_src= $(i_all_c) $(i_any_c) $(i_count_c) $(i_maxloc0_c) \ $(i_iparity_c) $(i_norm2_c) $(i_parity_c) \ $(i_matmul_c) $(i_matmull_c) $(i_shape_c) $(i_eoshift1_c) \ $(i_eoshift3_c) $(i_cshift1_c) $(i_reshape_c) $(in_pack_c) $(in_unpack_c) \ - $(i_pow_c) $(i_pack_c) $(i_unpack_c) \ + $(i_pow_c) $(i_pack_c) $(i_unpack_c) $(i_matmulavx128_c) \ $(i_spread_c) selected_int_kind.inc selected_real_kind.inc kinds.h \ $(i_cshift0_c) kinds.inc c99_protos.inc fpu-target.h fpu-target.inc @@ -796,7 +811,12 @@ intrinsics/dprod_r8.f90 \ intrinsics/f2c_specifics.F90 # Turn on vectorization and loop unrolling for matmul. -$(patsubst %.c,%.lo,$(notdir $(i_matmul_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 +$(patsubst %.c,%.lo,$(notdir $(i_matmul_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 + +if HAVE_AVX128 +# Turn on AVX128 for AMD-specific matmul, but only if the compiler understands -mprefer-avx128 +$(patsubst %.c,%.lo,$(notdir $(i_matmulavx128_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 -mprefer-avx128 +endif # Logical matmul doesn't vectorize. $(patsubst %.c,%.lo,$(notdir $(i_matmull_c))): AM_CFLAGS += -funroll-loops @@ -936,6 +956,9 @@ $(i_sum_c): m4/sum.m4 $(I_M4_DEPS1) $(i_matmul_c): m4/matmul.m4 m4/matmul_internal.m4 $(I_M4_DEPS) $(M4) -Dfile=$@ -I$(srcdir)/m4 matmul.m4 > $@ +$(i_matmulavx128_c): m4/matmulavx128.m4 m4/matmul_internal.m4 $(I_M4_DEPS) + $(M4) -Dfile=$@ -I$(srcdir)/m4 matmulavx128.m4 > $@ + $(i_matmull_c): m4/matmull.m4 $(I_M4_DEPS) $(M4) -Dfile=$@ -I$(srcdir)/m4 matmull.m4 > $@ diff --git a/libgfortran/Makefile.in b/libgfortran/Makefile.in index 4914a6f..e47d6eb 100644 --- a/libgfortran/Makefile.in +++ b/libgfortran/Makefile.in @@ -289,15 +289,20 @@ am__objects_32 = unpack_i1.lo unpack_i2.lo unpack_i4.lo unpack_i8.lo \ unpack_i16.lo unpack_r4.lo unpack_r8.lo unpack_r10.lo \ unpack_r16.lo unpack_c4.lo unpack_c8.lo unpack_c10.lo \ unpack_c16.lo -am__objects_33 = spread_i1.lo spread_i2.lo spread_i4.lo spread_i8.lo \ +am__objects_33 = matmulavx128_i1.lo matmulavx128_i2.lo \ + matmulavx128_i4.lo matmulavx128_i8.lo matmulavx128_i16.lo \ + matmulavx128_r4.lo matmulavx128_r8.lo matmulavx128_r10.lo \ + matmulavx128_r16.lo matmulavx128_c4.lo matmulavx128_c8.lo \ + matmulavx128_c10.lo matmulavx128_c16.lo +am__objects_34 = spread_i1.lo spread_i2.lo spread_i4.lo spread_i8.lo \ spread_i16.lo spread_r4.lo spread_r8.lo spread_r10.lo \ spread_r16.lo spread_c4.lo spread_c8.lo spread_c10.lo \ spread_c16.lo -am__objects_34 = cshift0_i1.lo cshift0_i2.lo cshift0_i4.lo \ +am__objects_35 = cshift0_i1.lo cshift0_i2.lo cshift0_i4.lo \ cshift0_i8.lo cshift0_i16.lo cshift0_r4.lo cshift0_r8.lo \ cshift0_r10.lo cshift0_r16.lo cshift0_c4.lo cshift0_c8.lo \ cshift0_c10.lo cshift0_c16.lo -am__objects_35 = $(am__objects_4) $(am__objects_5) $(am__objects_6) \ +am__objects_36 = $(am__objects_4) $(am__objects_5) $(am__objects_6) \ $(am__objects_7) $(am__objects_8) $(am__objects_9) \ $(am__objects_10) $(am__objects_11) $(am__objects_12) \ $(am__objects_13) $(am__objects_14) $(am__objects_15) \ @@ -307,14 +312,14 @@ am__objects_35 = $(am__objects_4) $(am__objects_5) $(am__objects_6) \ $(am__objects_25) $(am__objects_26) $(am__objects_27) \ $(am__objects_28) $(am__objects_29) $(am__objects_30) \ $(am__objects_31) $(am__objects_32) $(am__objects_33) \ - $(am__objects_34) -@LIBGFOR_MINIMAL_FALSE@am__objects_36 = close.lo file_pos.lo format.lo \ + $(am__objects_34) $(am__objects_35) +@LIBGFOR_MINIMAL_FALSE@am__objects_37 = close.lo file_pos.lo format.lo \ @LIBGFOR_MINIMAL_FALSE@ inquire.lo intrinsics.lo list_read.lo \ @LIBGFOR_MINIMAL_FALSE@ lock.lo open.lo read.lo transfer.lo \ @LIBGFOR_MINIMAL_FALSE@ transfer128.lo unit.lo unix.lo write.lo \ @LIBGFOR_MINIMAL_FALSE@ fbuf.lo -am__objects_37 = size_from_kind.lo $(am__objects_36) -@LIBGFOR_MINIMAL_FALSE@am__objects_38 = access.lo c99_functions.lo \ +am__objects_38 = size_from_kind.lo $(am__objects_37) +@LIBGFOR_MINIMAL_FALSE@am__objects_39 = access.lo c99_functions.lo \ @LIBGFOR_MINIMAL_FALSE@ chdir.lo chmod.lo clock.lo cpu_time.lo \ @LIBGFOR_MINIMAL_FALSE@ ctime.lo date_and_time.lo dtime.lo \ @LIBGFOR_MINIMAL_FALSE@ env.lo etime.lo execute_command_line.lo \ @@ -324,19 +329,19 @@ am__objects_37 = size_from_kind.lo $(am__objects_36) @LIBGFOR_MINIMAL_FALSE@ rename.lo stat.lo symlnk.lo \ @LIBGFOR_MINIMAL_FALSE@ system_clock.lo time.lo umask.lo \ @LIBGFOR_MINIMAL_FALSE@ unlink.lo -@IEEE_SUPPORT_TRUE@am__objects_39 = ieee_helper.lo -am__objects_40 = associated.lo abort.lo args.lo cshift0.lo eoshift0.lo \ +@IEEE_SUPPORT_TRUE@am__objects_40 = ieee_helper.lo +am__objects_41 = associated.lo abort.lo args.lo cshift0.lo eoshift0.lo \ eoshift2.lo erfc_scaled.lo extends_type_of.lo fnum.lo \ ierrno.lo ishftc.lo mvbits.lo move_alloc.lo pack_generic.lo \ selected_char_kind.lo size.lo spread_generic.lo \ string_intrinsics.lo rand.lo random.lo reshape_generic.lo \ reshape_packed.lo selected_int_kind.lo selected_real_kind.lo \ unpack_generic.lo in_pack_generic.lo in_unpack_generic.lo \ - $(am__objects_38) $(am__objects_39) -@IEEE_SUPPORT_TRUE@am__objects_41 = ieee_arithmetic.lo \ + $(am__objects_39) $(am__objects_40) +@IEEE_SUPPORT_TRUE@am__objects_42 = ieee_arithmetic.lo \ @IEEE_SUPPORT_TRUE@ ieee_exceptions.lo ieee_features.lo -am__objects_42 = -am__objects_43 = _abs_c4.lo _abs_c8.lo _abs_c10.lo _abs_c16.lo \ +am__objects_43 = +am__objects_44 = _abs_c4.lo _abs_c8.lo _abs_c10.lo _abs_c16.lo \ _abs_i4.lo _abs_i8.lo _abs_i16.lo _abs_r4.lo _abs_r8.lo \ _abs_r10.lo _abs_r16.lo _aimag_c4.lo _aimag_c8.lo \ _aimag_c10.lo _aimag_c16.lo _exp_r4.lo _exp_r8.lo _exp_r10.lo \ @@ -360,19 +365,19 @@ am__objects_43 = _abs_c4.lo _abs_c8.lo _abs_c10.lo _abs_c16.lo \ _conjg_c4.lo _conjg_c8.lo _conjg_c10.lo _conjg_c16.lo \ _aint_r4.lo _aint_r8.lo _aint_r10.lo _aint_r16.lo _anint_r4.lo \ _anint_r8.lo _anint_r10.lo _anint_r16.lo -am__objects_44 = _sign_i4.lo _sign_i8.lo _sign_i16.lo _sign_r4.lo \ +am__objects_45 = _sign_i4.lo _sign_i8.lo _sign_i16.lo _sign_r4.lo \ _sign_r8.lo _sign_r10.lo _sign_r16.lo _dim_i4.lo _dim_i8.lo \ _dim_i16.lo _dim_r4.lo _dim_r8.lo _dim_r10.lo _dim_r16.lo \ _atan2_r4.lo _atan2_r8.lo _atan2_r10.lo _atan2_r16.lo \ _mod_i4.lo _mod_i8.lo _mod_i16.lo _mod_r4.lo _mod_r8.lo \ _mod_r10.lo _mod_r16.lo -am__objects_45 = misc_specifics.lo -am__objects_46 = $(am__objects_43) $(am__objects_44) $(am__objects_45) \ +am__objects_46 = misc_specifics.lo +am__objects_47 = $(am__objects_44) $(am__objects_45) $(am__objects_46) \ dprod_r8.lo f2c_specifics.lo -am__objects_47 = $(am__objects_3) $(am__objects_35) $(am__objects_37) \ - $(am__objects_40) $(am__objects_41) $(am__objects_42) \ - $(am__objects_46) -@onestep_FALSE@am_libgfortran_la_OBJECTS = $(am__objects_47) +am__objects_48 = $(am__objects_3) $(am__objects_36) $(am__objects_38) \ + $(am__objects_41) $(am__objects_42) $(am__objects_43) \ + $(am__objects_47) +@onestep_FALSE@am_libgfortran_la_OBJECTS = $(am__objects_48) @onestep_TRUE@am_libgfortran_la_OBJECTS = libgfortran_c.lo libgfortran_la_OBJECTS = $(am_libgfortran_la_OBJECTS) DEFAULT_INCLUDES = -I.@am__isrc@ @@ -879,6 +884,21 @@ $(srcdir)/generated/matmul_c8.c \ $(srcdir)/generated/matmul_c10.c \ $(srcdir)/generated/matmul_c16.c +i_matmulavx128_c = \ +$(srcdir)/generated/matmulavx128_i1.c \ +$(srcdir)/generated/matmulavx128_i2.c \ +$(srcdir)/generated/matmulavx128_i4.c \ +$(srcdir)/generated/matmulavx128_i8.c \ +$(srcdir)/generated/matmulavx128_i16.c \ +$(srcdir)/generated/matmulavx128_r4.c \ +$(srcdir)/generated/matmulavx128_r8.c \ +$(srcdir)/generated/matmulavx128_r10.c \ +$(srcdir)/generated/matmulavx128_r16.c \ +$(srcdir)/generated/matmulavx128_c4.c \ +$(srcdir)/generated/matmulavx128_c8.c \ +$(srcdir)/generated/matmulavx128_c10.c \ +$(srcdir)/generated/matmulavx128_c16.c + i_matmull_c = \ $(srcdir)/generated/matmul_l4.c \ $(srcdir)/generated/matmul_l8.c \ @@ -1059,7 +1079,7 @@ gfor_built_src = $(i_all_c) $(i_any_c) $(i_count_c) $(i_maxloc0_c) \ $(i_iparity_c) $(i_norm2_c) $(i_parity_c) \ $(i_matmul_c) $(i_matmull_c) $(i_shape_c) $(i_eoshift1_c) \ $(i_eoshift3_c) $(i_cshift1_c) $(i_reshape_c) $(in_pack_c) $(in_unpack_c) \ - $(i_pow_c) $(i_pack_c) $(i_unpack_c) \ + $(i_pow_c) $(i_pack_c) $(i_unpack_c) $(i_matmulavx128_c) \ $(i_spread_c) selected_int_kind.inc selected_real_kind.inc kinds.h \ $(i_cshift0_c) kinds.inc c99_protos.inc fpu-target.h fpu-target.inc @@ -1518,6 +1538,19 @@ distclean-compile: @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmul_r16.Plo@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmul_r4.Plo@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmul_r8.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_c10.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_c16.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_c4.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_c8.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_i1.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_i16.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_i2.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_i4.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_i8.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_r10.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_r16.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_r4.Plo@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/matmulavx128_r8.Plo@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/maxloc0_16_i1.Plo@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/maxloc0_16_i16.Plo@am__quote@ @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/maxloc0_16_i2.Plo@am__quote@ @@ -4584,6 +4617,97 @@ unpack_c16.lo: $(srcdir)/generated/unpack_c16.c @AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ @am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o unpack_c16.lo `test -f '$(srcdir)/generated/unpack_c16.c' || echo '$(srcdir)/'`$(srcdir)/generated/unpack_c16.c +matmulavx128_i1.lo: $(srcdir)/generated/matmulavx128_i1.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_i1.lo -MD -MP -MF $(DEPDIR)/matmulavx128_i1.Tpo -c -o matmulavx128_i1.lo `test -f '$(srcdir)/generated/matmulavx128_i1.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i1.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_i1.Tpo $(DEPDIR)/matmulavx128_i1.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_i1.c' object='matmulavx128_i1.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_i1.lo `test -f '$(srcdir)/generated/matmulavx128_i1.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i1.c + +matmulavx128_i2.lo: $(srcdir)/generated/matmulavx128_i2.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_i2.lo -MD -MP -MF $(DEPDIR)/matmulavx128_i2.Tpo -c -o matmulavx128_i2.lo `test -f '$(srcdir)/generated/matmulavx128_i2.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i2.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_i2.Tpo $(DEPDIR)/matmulavx128_i2.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_i2.c' object='matmulavx128_i2.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_i2.lo `test -f '$(srcdir)/generated/matmulavx128_i2.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i2.c + +matmulavx128_i4.lo: $(srcdir)/generated/matmulavx128_i4.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_i4.lo -MD -MP -MF $(DEPDIR)/matmulavx128_i4.Tpo -c -o matmulavx128_i4.lo `test -f '$(srcdir)/generated/matmulavx128_i4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i4.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_i4.Tpo $(DEPDIR)/matmulavx128_i4.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_i4.c' object='matmulavx128_i4.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_i4.lo `test -f '$(srcdir)/generated/matmulavx128_i4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i4.c + +matmulavx128_i8.lo: $(srcdir)/generated/matmulavx128_i8.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_i8.lo -MD -MP -MF $(DEPDIR)/matmulavx128_i8.Tpo -c -o matmulavx128_i8.lo `test -f '$(srcdir)/generated/matmulavx128_i8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i8.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_i8.Tpo $(DEPDIR)/matmulavx128_i8.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_i8.c' object='matmulavx128_i8.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_i8.lo `test -f '$(srcdir)/generated/matmulavx128_i8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i8.c + +matmulavx128_i16.lo: $(srcdir)/generated/matmulavx128_i16.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_i16.lo -MD -MP -MF $(DEPDIR)/matmulavx128_i16.Tpo -c -o matmulavx128_i16.lo `test -f '$(srcdir)/generated/matmulavx128_i16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i16.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_i16.Tpo $(DEPDIR)/matmulavx128_i16.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_i16.c' object='matmulavx128_i16.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_i16.lo `test -f '$(srcdir)/generated/matmulavx128_i16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_i16.c + +matmulavx128_r4.lo: $(srcdir)/generated/matmulavx128_r4.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_r4.lo -MD -MP -MF $(DEPDIR)/matmulavx128_r4.Tpo -c -o matmulavx128_r4.lo `test -f '$(srcdir)/generated/matmulavx128_r4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r4.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_r4.Tpo $(DEPDIR)/matmulavx128_r4.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_r4.c' object='matmulavx128_r4.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_r4.lo `test -f '$(srcdir)/generated/matmulavx128_r4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r4.c + +matmulavx128_r8.lo: $(srcdir)/generated/matmulavx128_r8.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_r8.lo -MD -MP -MF $(DEPDIR)/matmulavx128_r8.Tpo -c -o matmulavx128_r8.lo `test -f '$(srcdir)/generated/matmulavx128_r8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r8.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_r8.Tpo $(DEPDIR)/matmulavx128_r8.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_r8.c' object='matmulavx128_r8.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_r8.lo `test -f '$(srcdir)/generated/matmulavx128_r8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r8.c + +matmulavx128_r10.lo: $(srcdir)/generated/matmulavx128_r10.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_r10.lo -MD -MP -MF $(DEPDIR)/matmulavx128_r10.Tpo -c -o matmulavx128_r10.lo `test -f '$(srcdir)/generated/matmulavx128_r10.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r10.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_r10.Tpo $(DEPDIR)/matmulavx128_r10.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_r10.c' object='matmulavx128_r10.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_r10.lo `test -f '$(srcdir)/generated/matmulavx128_r10.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r10.c + +matmulavx128_r16.lo: $(srcdir)/generated/matmulavx128_r16.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_r16.lo -MD -MP -MF $(DEPDIR)/matmulavx128_r16.Tpo -c -o matmulavx128_r16.lo `test -f '$(srcdir)/generated/matmulavx128_r16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r16.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_r16.Tpo $(DEPDIR)/matmulavx128_r16.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_r16.c' object='matmulavx128_r16.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_r16.lo `test -f '$(srcdir)/generated/matmulavx128_r16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_r16.c + +matmulavx128_c4.lo: $(srcdir)/generated/matmulavx128_c4.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_c4.lo -MD -MP -MF $(DEPDIR)/matmulavx128_c4.Tpo -c -o matmulavx128_c4.lo `test -f '$(srcdir)/generated/matmulavx128_c4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c4.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_c4.Tpo $(DEPDIR)/matmulavx128_c4.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_c4.c' object='matmulavx128_c4.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_c4.lo `test -f '$(srcdir)/generated/matmulavx128_c4.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c4.c + +matmulavx128_c8.lo: $(srcdir)/generated/matmulavx128_c8.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_c8.lo -MD -MP -MF $(DEPDIR)/matmulavx128_c8.Tpo -c -o matmulavx128_c8.lo `test -f '$(srcdir)/generated/matmulavx128_c8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c8.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_c8.Tpo $(DEPDIR)/matmulavx128_c8.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_c8.c' object='matmulavx128_c8.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_c8.lo `test -f '$(srcdir)/generated/matmulavx128_c8.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c8.c + +matmulavx128_c10.lo: $(srcdir)/generated/matmulavx128_c10.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_c10.lo -MD -MP -MF $(DEPDIR)/matmulavx128_c10.Tpo -c -o matmulavx128_c10.lo `test -f '$(srcdir)/generated/matmulavx128_c10.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c10.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_c10.Tpo $(DEPDIR)/matmulavx128_c10.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_c10.c' object='matmulavx128_c10.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_c10.lo `test -f '$(srcdir)/generated/matmulavx128_c10.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c10.c + +matmulavx128_c16.lo: $(srcdir)/generated/matmulavx128_c16.c +@am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT matmulavx128_c16.lo -MD -MP -MF $(DEPDIR)/matmulavx128_c16.Tpo -c -o matmulavx128_c16.lo `test -f '$(srcdir)/generated/matmulavx128_c16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c16.c +@am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/matmulavx128_c16.Tpo $(DEPDIR)/matmulavx128_c16.Plo +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$(srcdir)/generated/matmulavx128_c16.c' object='matmulavx128_c16.lo' libtool=yes @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -c -o matmulavx128_c16.lo `test -f '$(srcdir)/generated/matmulavx128_c16.c' || echo '$(srcdir)/'`$(srcdir)/generated/matmulavx128_c16.c + spread_i1.lo: $(srcdir)/generated/spread_i1.c @am__fastdepCC_TRUE@ $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) $(LIBTOOLFLAGS) --mode=compile $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) -MT spread_i1.lo -MD -MP -MF $(DEPDIR)/spread_i1.Tpo -c -o spread_i1.lo `test -f '$(srcdir)/generated/spread_i1.c' || echo '$(srcdir)/'`$(srcdir)/generated/spread_i1.c @am__fastdepCC_TRUE@ $(am__mv) $(DEPDIR)/spread_i1.Tpo $(DEPDIR)/spread_i1.Plo @@ -5567,7 +5691,10 @@ uninstall-am: uninstall-cafexeclibLTLIBRARIES \ @LIBGFOR_USE_SYMVER_SUN_TRUE@@LIBGFOR_USE_SYMVER_TRUE@ > $@ || (rm -f $@ ; exit 1) # Turn on vectorization and loop unrolling for matmul. -$(patsubst %.c,%.lo,$(notdir $(i_matmul_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 +$(patsubst %.c,%.lo,$(notdir $(i_matmul_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 + +# Turn on AVX128 for AMD-specific matmul, but only if the compiler understands -mprefer-avx128 +@HAVE_AVX128_TRUE@$(patsubst %.c,%.lo,$(notdir $(i_matmulavx128_c))): AM_CFLAGS += -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 -mprefer-avx128 # Logical matmul doesn't vectorize. $(patsubst %.c,%.lo,$(notdir $(i_matmull_c))): AM_CFLAGS += -funroll-loops @@ -5667,6 +5794,9 @@ fpu-target.inc: fpu-target.h $(srcdir)/libgfortran.h @MAINTAINER_MODE_TRUE@$(i_matmul_c): m4/matmul.m4 m4/matmul_internal.m4 $(I_M4_DEPS) @MAINTAINER_MODE_TRUE@ $(M4) -Dfile=$@ -I$(srcdir)/m4 matmul.m4 > $@ +@MAINTAINER_MODE_TRUE@$(i_matmulavx128_c): m4/matmulavx128.m4 m4/matmul_internal.m4 $(I_M4_DEPS) +@MAINTAINER_MODE_TRUE@ $(M4) -Dfile=$@ -I$(srcdir)/m4 matmulavx128.m4 > $@ + @MAINTAINER_MODE_TRUE@$(i_matmull_c): m4/matmull.m4 $(I_M4_DEPS) @MAINTAINER_MODE_TRUE@ $(M4) -Dfile=$@ -I$(srcdir)/m4 matmull.m4 > $@ diff --git a/libgfortran/acinclude.m4 b/libgfortran/acinclude.m4 index ceb80a1..cd8dfab 100644 --- a/libgfortran/acinclude.m4 +++ b/libgfortran/acinclude.m4 @@ -452,3 +452,53 @@ AC_DEFUN([LIBGFOR_CHECK_AVX512F], [ []) CFLAGS="$ac_save_CFLAGS" ]) + +dnl Check for FMA3 +dnl +AC_DEFUN([LIBGFOR_CHECK_FMA3], [ + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mfma -mno-fma4" + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ + float + flt_mul_add (float a, float b, float c) + { + return __builtin_fmaf (a, b, c); + }]], [[]])], + AC_DEFINE(HAVE_FMA3, 1, + [Define if FMA3 instructions can be compiled.]), + []) + CFLAGS="$ac_save_CFLAGS" +]) + +dnl Check for FMA4 +dnl +AC_DEFUN([LIBGFOR_CHECK_FMA4], [ + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mfma4 -mno-fma" + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ + float + flt_mul_add (float a, float b, float c) + { + return __builtin_fmaf (a, b, c); + }]], [[]])], + AC_DEFINE(HAVE_FMA4, 1, + [Define if FMA4 instructions can be compiled.]), + []) + CFLAGS="$ac_save_CFLAGS" +]) + +dnl Check for -mprefer-avx128 +dnl This also defines an automake conditional. +AC_DEFUN([LIBGFOR_CHECK_AVX128], [ + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mavx -mprefer-avx128" + AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ + void foo() + { + }]], [[]])], + AC_DEFINE(HAVE_AVX128, 1, + [Define if -mprefer-avx128 is supported.]) + AM_CONDITIONAL([HAVE_AVX128],true), + []) + CFLAGS="$ac_save_CFLAGS" +]) diff --git a/libgfortran/config.h.in b/libgfortran/config.h.in index b762d099..0274e5d 100644 --- a/libgfortran/config.h.in +++ b/libgfortran/config.h.in @@ -81,6 +81,9 @@ /* Define if AVX instructions can be compiled. */ #undef HAVE_AVX +/* Define if -mprefer-avx128 is supported. */ +#undef HAVE_AVX128 + /* Define if AVX2 instructions can be compiled. */ #undef HAVE_AVX2 @@ -375,6 +378,12 @@ /* Define to 1 if you have the `floorl' function. */ #undef HAVE_FLOORL +/* Define if FMA3 instructions can be compiled. */ +#undef HAVE_FMA3 + +/* Define if FMA4 instructions can be compiled. */ +#undef HAVE_FMA4 + /* Define to 1 if you have the `fmod' function. */ #undef HAVE_FMOD diff --git a/libgfortran/configure b/libgfortran/configure index 81238fc..36e0155 100755 --- a/libgfortran/configure +++ b/libgfortran/configure @@ -606,6 +606,8 @@ am__EXEEXT_TRUE LTLIBOBJS LIBOBJS get_gcc_base_ver +HAVE_AVX128_FALSE +HAVE_AVX128_TRUE IEEE_FLAGS IEEE_SUPPORT IEEE_SUPPORT_FALSE @@ -12421,7 +12423,7 @@ else lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2 lt_status=$lt_dlunknown cat > conftest.$ac_ext <<_LT_EOF -#line 12424 "configure" +#line 12426 "configure" #include "confdefs.h" #if HAVE_DLFCN_H @@ -12527,7 +12529,7 @@ else lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2 lt_status=$lt_dlunknown cat > conftest.$ac_ext <<_LT_EOF -#line 12530 "configure" +#line 12532 "configure" #include "confdefs.h" #if HAVE_DLFCN_H @@ -26363,6 +26365,99 @@ rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext CFLAGS="$ac_save_CFLAGS" +# Check for FMA3 extensions + + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mfma -mno-fma4" + cat confdefs.h - <<_ACEOF >conftest.$ac_ext +/* end confdefs.h. */ + + float + flt_mul_add (float a, float b, float c) + { + return __builtin_fmaf (a, b, c); + } +int +main () +{ + + ; + return 0; +} +_ACEOF +if ac_fn_c_try_compile "$LINENO"; then : + +$as_echo "#define HAVE_FMA3 1" >>confdefs.h + +fi +rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext + CFLAGS="$ac_save_CFLAGS" + + +# Check for FMA4 extensions + + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mfma4 -mno-fma" + cat confdefs.h - <<_ACEOF >conftest.$ac_ext +/* end confdefs.h. */ + + float + flt_mul_add (float a, float b, float c) + { + return __builtin_fmaf (a, b, c); + } +int +main () +{ + + ; + return 0; +} +_ACEOF +if ac_fn_c_try_compile "$LINENO"; then : + +$as_echo "#define HAVE_FMA4 1" >>confdefs.h + +fi +rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext + CFLAGS="$ac_save_CFLAGS" + + +# Check if AVX128 works + + ac_save_CFLAGS="$CFLAGS" + CFLAGS="-O2 -mavx -mprefer-avx128" + cat confdefs.h - <<_ACEOF >conftest.$ac_ext +/* end confdefs.h. */ + + void foo() + { + } +int +main () +{ + + ; + return 0; +} +_ACEOF +if ac_fn_c_try_compile "$LINENO"; then : + +$as_echo "#define HAVE_AVX128 1" >>confdefs.h + + if true; then + HAVE_AVX128_TRUE= + HAVE_AVX128_FALSE='#' +else + HAVE_AVX128_TRUE='#' + HAVE_AVX128_FALSE= +fi + +fi +rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext + CFLAGS="$ac_save_CFLAGS" + + # Determine what GCC version number to use in filesystem paths. get_gcc_base_ver="cat" @@ -26615,6 +26710,10 @@ if test -z "${IEEE_SUPPORT_TRUE}" && test -z "${IEEE_SUPPORT_FALSE}"; then as_fn_error "conditional \"IEEE_SUPPORT\" was never defined. Usually this means the macro was only invoked conditionally." "$LINENO" 5 fi +if test -z "${HAVE_AVX128_TRUE}" && test -z "${HAVE_AVX128_FALSE}"; then + as_fn_error "conditional \"HAVE_AVX128\" was never defined. +Usually this means the macro was only invoked conditionally." "$LINENO" 5 +fi : ${CONFIG_STATUS=./config.status} ac_write_fail=0 diff --git a/libgfortran/configure.ac b/libgfortran/configure.ac index 37b12d2..78214ac 100644 --- a/libgfortran/configure.ac +++ b/libgfortran/configure.ac @@ -624,6 +624,15 @@ LIBGFOR_CHECK_AVX2 # Check wether we support AVX512f extensions LIBGFOR_CHECK_AVX512F +# Check for FMA3 extensions +LIBGFOR_CHECK_FMA3 + +# Check for FMA4 extensions +LIBGFOR_CHECK_FMA4 + +# Check if AVX128 works +LIBGFOR_CHECK_AVX128 + # Determine what GCC version number to use in filesystem paths. GCC_BASE_VER diff --git a/libgfortran/generated/matmul_c10.c b/libgfortran/generated/matmul_c10.c index dbe3d3a..54e2714 100644 --- a/libgfortran/generated/matmul_c10.c +++ b/libgfortran/generated/matmul_c10.c @@ -1734,6 +1734,24 @@ matmul_c10_avx512f (gfc_array_c10 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c10_avx128_fma3 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c10_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c10_avx128_fma4 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c10_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_c10_vanilla (gfc_array_c10 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_c10 (gfc_array_c10 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_c10_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_c10_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_c16.c b/libgfortran/generated/matmul_c16.c index 36c8747..cd8aacd 100644 --- a/libgfortran/generated/matmul_c16.c +++ b/libgfortran/generated/matmul_c16.c @@ -1734,6 +1734,24 @@ matmul_c16_avx512f (gfc_array_c16 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c16_avx128_fma3 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c16_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c16_avx128_fma4 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c16_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_c16_vanilla (gfc_array_c16 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_c16 (gfc_array_c16 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_c16_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_c16_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_c4.c b/libgfortran/generated/matmul_c4.c index 79343dd..ead22ef 100644 --- a/libgfortran/generated/matmul_c4.c +++ b/libgfortran/generated/matmul_c4.c @@ -1734,6 +1734,24 @@ matmul_c4_avx512f (gfc_array_c4 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c4_avx128_fma3 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c4_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c4_avx128_fma4 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c4_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_c4_vanilla (gfc_array_c4 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_c4 (gfc_array_c4 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_c4_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_c4_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_c8.c b/libgfortran/generated/matmul_c8.c index 0307d3f..a52e4bd 100644 --- a/libgfortran/generated/matmul_c8.c +++ b/libgfortran/generated/matmul_c8.c @@ -1734,6 +1734,24 @@ matmul_c8_avx512f (gfc_array_c8 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c8_avx128_fma3 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c8_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c8_avx128_fma4 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c8_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_c8_vanilla (gfc_array_c8 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_c8 (gfc_array_c8 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_c8_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_c8_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_i1.c b/libgfortran/generated/matmul_i1.c index 8cda1c32..dfd47e1 100644 --- a/libgfortran/generated/matmul_i1.c +++ b/libgfortran/generated/matmul_i1.c @@ -1734,6 +1734,24 @@ matmul_i1_avx512f (gfc_array_i1 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i1_avx128_fma3 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i1_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i1_avx128_fma4 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i1_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_i1_vanilla (gfc_array_i1 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_i1 (gfc_array_i1 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_i1_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_i1_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_i16.c b/libgfortran/generated/matmul_i16.c index 0603522..a7bdcb5 100644 --- a/libgfortran/generated/matmul_i16.c +++ b/libgfortran/generated/matmul_i16.c @@ -1734,6 +1734,24 @@ matmul_i16_avx512f (gfc_array_i16 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i16_avx128_fma3 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i16_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i16_avx128_fma4 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i16_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_i16_vanilla (gfc_array_i16 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_i16 (gfc_array_i16 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_i16_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_i16_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_i2.c b/libgfortran/generated/matmul_i2.c index 93316cb..d541fa3 100644 --- a/libgfortran/generated/matmul_i2.c +++ b/libgfortran/generated/matmul_i2.c @@ -1734,6 +1734,24 @@ matmul_i2_avx512f (gfc_array_i2 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i2_avx128_fma3 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i2_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i2_avx128_fma4 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i2_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_i2_vanilla (gfc_array_i2 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_i2 (gfc_array_i2 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_i2_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_i2_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_i4.c b/libgfortran/generated/matmul_i4.c index 475b214..f8f0cdb 100644 --- a/libgfortran/generated/matmul_i4.c +++ b/libgfortran/generated/matmul_i4.c @@ -1734,6 +1734,24 @@ matmul_i4_avx512f (gfc_array_i4 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i4_avx128_fma3 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i4_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i4_avx128_fma4 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i4_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_i4_vanilla (gfc_array_i4 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_i4 (gfc_array_i4 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_i4_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_i4_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_i8.c b/libgfortran/generated/matmul_i8.c index d29a4b4..2aac1d7 100644 --- a/libgfortran/generated/matmul_i8.c +++ b/libgfortran/generated/matmul_i8.c @@ -1734,6 +1734,24 @@ matmul_i8_avx512f (gfc_array_i8 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i8_avx128_fma3 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i8_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i8_avx128_fma4 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i8_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_i8_vanilla (gfc_array_i8 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_i8 (gfc_array_i8 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_i8_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_i8_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_r10.c b/libgfortran/generated/matmul_r10.c index f1885a4..448c96b 100644 --- a/libgfortran/generated/matmul_r10.c +++ b/libgfortran/generated/matmul_r10.c @@ -1734,6 +1734,24 @@ matmul_r10_avx512f (gfc_array_r10 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r10_avx128_fma3 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r10_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r10_avx128_fma4 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r10_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_r10_vanilla (gfc_array_r10 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_r10 (gfc_array_r10 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_r10_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_r10_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_r16.c b/libgfortran/generated/matmul_r16.c index a7aa44f..57a47dc 100644 --- a/libgfortran/generated/matmul_r16.c +++ b/libgfortran/generated/matmul_r16.c @@ -1734,6 +1734,24 @@ matmul_r16_avx512f (gfc_array_r16 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r16_avx128_fma3 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r16_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r16_avx128_fma4 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r16_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_r16_vanilla (gfc_array_r16 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_r16 (gfc_array_r16 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_r16_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_r16_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_r4.c b/libgfortran/generated/matmul_r4.c index 1353025..52eea53 100644 --- a/libgfortran/generated/matmul_r4.c +++ b/libgfortran/generated/matmul_r4.c @@ -1734,6 +1734,24 @@ matmul_r4_avx512f (gfc_array_r4 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r4_avx128_fma3 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r4_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r4_avx128_fma4 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r4_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_r4_vanilla (gfc_array_r4 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_r4 (gfc_array_r4 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_r4_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_r4_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmul_r8.c b/libgfortran/generated/matmul_r8.c index a4af000..074697d 100644 --- a/libgfortran/generated/matmul_r8.c +++ b/libgfortran/generated/matmul_r8.c @@ -1734,6 +1734,24 @@ matmul_r8_avx512f (gfc_array_r8 * const restrict retarray, #endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r8_avx128_fma3 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r8_avx128_fma3); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r8_avx128_fma4 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r8_avx128_fma4); +#endif + /* Function to fall back to if there is no special processor-specific version. */ static void matmul_r8_vanilla (gfc_array_r8 * const restrict retarray, @@ -2332,6 +2350,26 @@ void matmul_r8 (gfc_array_r8 * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_r8_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_r8_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/generated/matmulavx128_c10.c b/libgfortran/generated/matmulavx128_c10.c new file mode 100644 index 0000000..53cdecb --- /dev/null +++ b/libgfortran/generated/matmulavx128_c10.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_COMPLEX_10) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_COMPLEX_10 *, const GFC_COMPLEX_10 *, + const int *, const GFC_COMPLEX_10 *, const int *, + const GFC_COMPLEX_10 *, GFC_COMPLEX_10 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c10_avx128_fma3 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c10_avx128_fma3); +void +matmul_c10_avx128_fma3 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_10 * restrict abase; + const GFC_COMPLEX_10 * restrict bbase; + GFC_COMPLEX_10 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_10)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_10 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_10 *a, *b; + GFC_COMPLEX_10 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_10 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_10 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_10)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_10 *restrict abase_x; + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 *restrict dest_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_10)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_10 *restrict abase_x; + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 *restrict dest_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c10_avx128_fma4 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c10_avx128_fma4); +void +matmul_c10_avx128_fma4 (gfc_array_c10 * const restrict retarray, + gfc_array_c10 * const restrict a, gfc_array_c10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_10 * restrict abase; + const GFC_COMPLEX_10 * restrict bbase; + GFC_COMPLEX_10 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_10)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_10 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_10 *a, *b; + GFC_COMPLEX_10 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_10 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_10 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_10)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_10 *restrict abase_x; + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 *restrict dest_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_10)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_10 *restrict abase_x; + const GFC_COMPLEX_10 *restrict bbase_y; + GFC_COMPLEX_10 *restrict dest_y; + GFC_COMPLEX_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_c16.c b/libgfortran/generated/matmulavx128_c16.c new file mode 100644 index 0000000..e7657a0 --- /dev/null +++ b/libgfortran/generated/matmulavx128_c16.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_COMPLEX_16) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_COMPLEX_16 *, const GFC_COMPLEX_16 *, + const int *, const GFC_COMPLEX_16 *, const int *, + const GFC_COMPLEX_16 *, GFC_COMPLEX_16 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c16_avx128_fma3 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c16_avx128_fma3); +void +matmul_c16_avx128_fma3 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_16 * restrict abase; + const GFC_COMPLEX_16 * restrict bbase; + GFC_COMPLEX_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_16 *a, *b; + GFC_COMPLEX_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_16 *restrict abase_x; + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 *restrict dest_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_16 *restrict abase_x; + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 *restrict dest_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c16_avx128_fma4 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c16_avx128_fma4); +void +matmul_c16_avx128_fma4 (gfc_array_c16 * const restrict retarray, + gfc_array_c16 * const restrict a, gfc_array_c16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_16 * restrict abase; + const GFC_COMPLEX_16 * restrict bbase; + GFC_COMPLEX_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_16 *a, *b; + GFC_COMPLEX_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_16 *restrict abase_x; + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 *restrict dest_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_16 *restrict abase_x; + const GFC_COMPLEX_16 *restrict bbase_y; + GFC_COMPLEX_16 *restrict dest_y; + GFC_COMPLEX_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_c4.c b/libgfortran/generated/matmulavx128_c4.c new file mode 100644 index 0000000..950f1eb --- /dev/null +++ b/libgfortran/generated/matmulavx128_c4.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_COMPLEX_4) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_COMPLEX_4 *, const GFC_COMPLEX_4 *, + const int *, const GFC_COMPLEX_4 *, const int *, + const GFC_COMPLEX_4 *, GFC_COMPLEX_4 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c4_avx128_fma3 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c4_avx128_fma3); +void +matmul_c4_avx128_fma3 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_4 * restrict abase; + const GFC_COMPLEX_4 * restrict bbase; + GFC_COMPLEX_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_4 *a, *b; + GFC_COMPLEX_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_4 *restrict abase_x; + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 *restrict dest_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_4 *restrict abase_x; + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 *restrict dest_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c4_avx128_fma4 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c4_avx128_fma4); +void +matmul_c4_avx128_fma4 (gfc_array_c4 * const restrict retarray, + gfc_array_c4 * const restrict a, gfc_array_c4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_4 * restrict abase; + const GFC_COMPLEX_4 * restrict bbase; + GFC_COMPLEX_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_4 *a, *b; + GFC_COMPLEX_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_4 *restrict abase_x; + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 *restrict dest_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_4 *restrict abase_x; + const GFC_COMPLEX_4 *restrict bbase_y; + GFC_COMPLEX_4 *restrict dest_y; + GFC_COMPLEX_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_c8.c b/libgfortran/generated/matmulavx128_c8.c new file mode 100644 index 0000000..a41c160 --- /dev/null +++ b/libgfortran/generated/matmulavx128_c8.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_COMPLEX_8) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_COMPLEX_8 *, const GFC_COMPLEX_8 *, + const int *, const GFC_COMPLEX_8 *, const int *, + const GFC_COMPLEX_8 *, GFC_COMPLEX_8 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_c8_avx128_fma3 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_c8_avx128_fma3); +void +matmul_c8_avx128_fma3 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_8 * restrict abase; + const GFC_COMPLEX_8 * restrict bbase; + GFC_COMPLEX_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_8 *a, *b; + GFC_COMPLEX_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_8 *restrict abase_x; + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 *restrict dest_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_8 *restrict abase_x; + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 *restrict dest_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_c8_avx128_fma4 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_c8_avx128_fma4); +void +matmul_c8_avx128_fma4 (gfc_array_c8 * const restrict retarray, + gfc_array_c8 * const restrict a, gfc_array_c8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_COMPLEX_8 * restrict abase; + const GFC_COMPLEX_8 * restrict bbase; + GFC_COMPLEX_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_COMPLEX_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_COMPLEX_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_COMPLEX_8 *a, *b; + GFC_COMPLEX_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_COMPLEX_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_COMPLEX_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_COMPLEX_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_COMPLEX_8 *restrict abase_x; + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 *restrict dest_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_COMPLEX_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_COMPLEX_8 *restrict abase_x; + const GFC_COMPLEX_8 *restrict bbase_y; + GFC_COMPLEX_8 *restrict dest_y; + GFC_COMPLEX_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_COMPLEX_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_i1.c b/libgfortran/generated/matmulavx128_i1.c new file mode 100644 index 0000000..e187157 --- /dev/null +++ b/libgfortran/generated/matmulavx128_i1.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_INTEGER_1) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_INTEGER_1 *, const GFC_INTEGER_1 *, + const int *, const GFC_INTEGER_1 *, const int *, + const GFC_INTEGER_1 *, GFC_INTEGER_1 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i1_avx128_fma3 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i1_avx128_fma3); +void +matmul_i1_avx128_fma3 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_1 * restrict abase; + const GFC_INTEGER_1 * restrict bbase; + GFC_INTEGER_1 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_1)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_1 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_1 *a, *b; + GFC_INTEGER_1 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_1 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_1 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_1)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_1 *restrict abase_x; + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 *restrict dest_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_1)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_1 *restrict abase_x; + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 *restrict dest_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i1_avx128_fma4 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i1_avx128_fma4); +void +matmul_i1_avx128_fma4 (gfc_array_i1 * const restrict retarray, + gfc_array_i1 * const restrict a, gfc_array_i1 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_1 * restrict abase; + const GFC_INTEGER_1 * restrict bbase; + GFC_INTEGER_1 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_1)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_1 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_1 *a, *b; + GFC_INTEGER_1 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_1 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_1 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_1)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_1 *restrict abase_x; + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 *restrict dest_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_1)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_1 *restrict abase_x; + const GFC_INTEGER_1 *restrict bbase_y; + GFC_INTEGER_1 *restrict dest_y; + GFC_INTEGER_1 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_1) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_i16.c b/libgfortran/generated/matmulavx128_i16.c new file mode 100644 index 0000000..1a7b733 --- /dev/null +++ b/libgfortran/generated/matmulavx128_i16.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_INTEGER_16) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_INTEGER_16 *, const GFC_INTEGER_16 *, + const int *, const GFC_INTEGER_16 *, const int *, + const GFC_INTEGER_16 *, GFC_INTEGER_16 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i16_avx128_fma3 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i16_avx128_fma3); +void +matmul_i16_avx128_fma3 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_16 * restrict abase; + const GFC_INTEGER_16 * restrict bbase; + GFC_INTEGER_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_16 *a, *b; + GFC_INTEGER_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_16 *restrict abase_x; + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 *restrict dest_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_16 *restrict abase_x; + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 *restrict dest_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i16_avx128_fma4 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i16_avx128_fma4); +void +matmul_i16_avx128_fma4 (gfc_array_i16 * const restrict retarray, + gfc_array_i16 * const restrict a, gfc_array_i16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_16 * restrict abase; + const GFC_INTEGER_16 * restrict bbase; + GFC_INTEGER_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_16 *a, *b; + GFC_INTEGER_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_16 *restrict abase_x; + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 *restrict dest_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_16 *restrict abase_x; + const GFC_INTEGER_16 *restrict bbase_y; + GFC_INTEGER_16 *restrict dest_y; + GFC_INTEGER_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_i2.c b/libgfortran/generated/matmulavx128_i2.c new file mode 100644 index 0000000..a095c58 --- /dev/null +++ b/libgfortran/generated/matmulavx128_i2.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_INTEGER_2) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_INTEGER_2 *, const GFC_INTEGER_2 *, + const int *, const GFC_INTEGER_2 *, const int *, + const GFC_INTEGER_2 *, GFC_INTEGER_2 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i2_avx128_fma3 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i2_avx128_fma3); +void +matmul_i2_avx128_fma3 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_2 * restrict abase; + const GFC_INTEGER_2 * restrict bbase; + GFC_INTEGER_2 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_2)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_2 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_2 *a, *b; + GFC_INTEGER_2 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_2 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_2 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_2)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_2 *restrict abase_x; + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 *restrict dest_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_2)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_2 *restrict abase_x; + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 *restrict dest_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i2_avx128_fma4 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i2_avx128_fma4); +void +matmul_i2_avx128_fma4 (gfc_array_i2 * const restrict retarray, + gfc_array_i2 * const restrict a, gfc_array_i2 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_2 * restrict abase; + const GFC_INTEGER_2 * restrict bbase; + GFC_INTEGER_2 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_2)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_2 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_2 *a, *b; + GFC_INTEGER_2 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_2 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_2 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_2)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_2 *restrict abase_x; + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 *restrict dest_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_2)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_2 *restrict abase_x; + const GFC_INTEGER_2 *restrict bbase_y; + GFC_INTEGER_2 *restrict dest_y; + GFC_INTEGER_2 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_2) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_i4.c b/libgfortran/generated/matmulavx128_i4.c new file mode 100644 index 0000000..a01c56f --- /dev/null +++ b/libgfortran/generated/matmulavx128_i4.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_INTEGER_4) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_INTEGER_4 *, const GFC_INTEGER_4 *, + const int *, const GFC_INTEGER_4 *, const int *, + const GFC_INTEGER_4 *, GFC_INTEGER_4 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i4_avx128_fma3 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i4_avx128_fma3); +void +matmul_i4_avx128_fma3 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_4 * restrict abase; + const GFC_INTEGER_4 * restrict bbase; + GFC_INTEGER_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_4 *a, *b; + GFC_INTEGER_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_4 *restrict abase_x; + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 *restrict dest_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_4 *restrict abase_x; + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 *restrict dest_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i4_avx128_fma4 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i4_avx128_fma4); +void +matmul_i4_avx128_fma4 (gfc_array_i4 * const restrict retarray, + gfc_array_i4 * const restrict a, gfc_array_i4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_4 * restrict abase; + const GFC_INTEGER_4 * restrict bbase; + GFC_INTEGER_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_4 *a, *b; + GFC_INTEGER_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_4 *restrict abase_x; + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 *restrict dest_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_4 *restrict abase_x; + const GFC_INTEGER_4 *restrict bbase_y; + GFC_INTEGER_4 *restrict dest_y; + GFC_INTEGER_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_i8.c b/libgfortran/generated/matmulavx128_i8.c new file mode 100644 index 0000000..bc78ffe --- /dev/null +++ b/libgfortran/generated/matmulavx128_i8.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_INTEGER_8) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_INTEGER_8 *, const GFC_INTEGER_8 *, + const int *, const GFC_INTEGER_8 *, const int *, + const GFC_INTEGER_8 *, GFC_INTEGER_8 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_i8_avx128_fma3 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_i8_avx128_fma3); +void +matmul_i8_avx128_fma3 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_8 * restrict abase; + const GFC_INTEGER_8 * restrict bbase; + GFC_INTEGER_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_8 *a, *b; + GFC_INTEGER_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_8 *restrict abase_x; + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 *restrict dest_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_8 *restrict abase_x; + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 *restrict dest_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_i8_avx128_fma4 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_i8_avx128_fma4); +void +matmul_i8_avx128_fma4 (gfc_array_i8 * const restrict retarray, + gfc_array_i8 * const restrict a, gfc_array_i8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_INTEGER_8 * restrict abase; + const GFC_INTEGER_8 * restrict bbase; + GFC_INTEGER_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_INTEGER_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_INTEGER_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_INTEGER_8 *a, *b; + GFC_INTEGER_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_INTEGER_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_INTEGER_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_INTEGER_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_INTEGER_8 *restrict abase_x; + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 *restrict dest_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_INTEGER_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_INTEGER_8 *restrict abase_x; + const GFC_INTEGER_8 *restrict bbase_y; + GFC_INTEGER_8 *restrict dest_y; + GFC_INTEGER_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_INTEGER_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_r10.c b/libgfortran/generated/matmulavx128_r10.c new file mode 100644 index 0000000..943678d --- /dev/null +++ b/libgfortran/generated/matmulavx128_r10.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_REAL_10) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_REAL_10 *, const GFC_REAL_10 *, + const int *, const GFC_REAL_10 *, const int *, + const GFC_REAL_10 *, GFC_REAL_10 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r10_avx128_fma3 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r10_avx128_fma3); +void +matmul_r10_avx128_fma3 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_10 * restrict abase; + const GFC_REAL_10 * restrict bbase; + GFC_REAL_10 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_10)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_10 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_10 *a, *b; + GFC_REAL_10 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_10 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_10 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_10)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_10)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_10 *restrict abase_x; + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 *restrict dest_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_10)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_10 *restrict abase_x; + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 *restrict dest_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r10_avx128_fma4 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r10_avx128_fma4); +void +matmul_r10_avx128_fma4 (gfc_array_r10 * const restrict retarray, + gfc_array_r10 * const restrict a, gfc_array_r10 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_10 * restrict abase; + const GFC_REAL_10 * restrict bbase; + GFC_REAL_10 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_10)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_10 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_10 *a, *b; + GFC_REAL_10 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_10 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_10 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_10)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_10)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_10 *restrict abase_x; + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 *restrict dest_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_10)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_10 *restrict abase_x; + const GFC_REAL_10 *restrict bbase_y; + GFC_REAL_10 *restrict dest_y; + GFC_REAL_10 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_10) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_r16.c b/libgfortran/generated/matmulavx128_r16.c new file mode 100644 index 0000000..3d5738b --- /dev/null +++ b/libgfortran/generated/matmulavx128_r16.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_REAL_16) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_REAL_16 *, const GFC_REAL_16 *, + const int *, const GFC_REAL_16 *, const int *, + const GFC_REAL_16 *, GFC_REAL_16 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r16_avx128_fma3 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r16_avx128_fma3); +void +matmul_r16_avx128_fma3 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_16 * restrict abase; + const GFC_REAL_16 * restrict bbase; + GFC_REAL_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_16 *a, *b; + GFC_REAL_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_16 *restrict abase_x; + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 *restrict dest_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_16 *restrict abase_x; + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 *restrict dest_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r16_avx128_fma4 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r16_avx128_fma4); +void +matmul_r16_avx128_fma4 (gfc_array_r16 * const restrict retarray, + gfc_array_r16 * const restrict a, gfc_array_r16 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_16 * restrict abase; + const GFC_REAL_16 * restrict bbase; + GFC_REAL_16 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_16)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_16 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_16 *a, *b; + GFC_REAL_16 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_16 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_16 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_16)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_16)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_16 *restrict abase_x; + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 *restrict dest_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_16)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_16 *restrict abase_x; + const GFC_REAL_16 *restrict bbase_y; + GFC_REAL_16 *restrict dest_y; + GFC_REAL_16 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_16) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_r4.c b/libgfortran/generated/matmulavx128_r4.c new file mode 100644 index 0000000..6c6da39 --- /dev/null +++ b/libgfortran/generated/matmulavx128_r4.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_REAL_4) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_REAL_4 *, const GFC_REAL_4 *, + const int *, const GFC_REAL_4 *, const int *, + const GFC_REAL_4 *, GFC_REAL_4 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r4_avx128_fma3 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r4_avx128_fma3); +void +matmul_r4_avx128_fma3 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_4 * restrict abase; + const GFC_REAL_4 * restrict bbase; + GFC_REAL_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_4 *a, *b; + GFC_REAL_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_4 *restrict abase_x; + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 *restrict dest_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_4 *restrict abase_x; + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 *restrict dest_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r4_avx128_fma4 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r4_avx128_fma4); +void +matmul_r4_avx128_fma4 (gfc_array_r4 * const restrict retarray, + gfc_array_r4 * const restrict a, gfc_array_r4 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_4 * restrict abase; + const GFC_REAL_4 * restrict bbase; + GFC_REAL_4 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_4)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_4 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_4 *a, *b; + GFC_REAL_4 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_4 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_4 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_4)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_4)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_4 *restrict abase_x; + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 *restrict dest_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_4)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_4 *restrict abase_x; + const GFC_REAL_4 *restrict bbase_y; + GFC_REAL_4 *restrict dest_y; + GFC_REAL_4 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_4) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/generated/matmulavx128_r8.c b/libgfortran/generated/matmulavx128_r8.c new file mode 100644 index 0000000..d628200 --- /dev/null +++ b/libgfortran/generated/matmulavx128_r8.c @@ -0,0 +1,1152 @@ +/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h> + + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +#if defined (HAVE_GFC_REAL_8) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const GFC_REAL_8 *, const GFC_REAL_8 *, + const int *, const GFC_REAL_8 *, const int *, + const GFC_REAL_8 *, GFC_REAL_8 *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +void +matmul_r8_avx128_fma3 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto(matmul_r8_avx128_fma3); +void +matmul_r8_avx128_fma3 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_8 * restrict abase; + const GFC_REAL_8 * restrict bbase; + GFC_REAL_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_8 *a, *b; + GFC_REAL_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_8 *restrict abase_x; + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 *restrict dest_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_8 *restrict abase_x; + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 *restrict dest_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +void +matmul_r8_avx128_fma4 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto(matmul_r8_avx128_fma4); +void +matmul_r8_avx128_fma4 (gfc_array_r8 * const restrict retarray, + gfc_array_r8 * const restrict a, gfc_array_r8 * const restrict b, int try_blas, + int blas_limit, blas_call gemm) +{ + const GFC_REAL_8 * restrict abase; + const GFC_REAL_8 * restrict bbase; + GFC_REAL_8 * restrict dest; + + index_type rxstride, rystride, axstride, aystride, bxstride, bystride; + index_type x, y, n, count, xcount, ycount; + + assert (GFC_DESCRIPTOR_RANK (a) == 2 + || GFC_DESCRIPTOR_RANK (b) == 2); + +/* C[xcount,ycount] = A[xcount, count] * B[count,ycount] + + Either A or B (but not both) can be rank 1: + + o One-dimensional argument A is implicitly treated as a row matrix + dimensioned [1,count], so xcount=1. + + o One-dimensional argument B is implicitly treated as a column matrix + dimensioned [count, 1], so ycount=1. +*/ + + if (retarray->base_addr == NULL) + { + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + } + else + { + GFC_DIMENSION_SET(retarray->dim[0], 0, + GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); + + GFC_DIMENSION_SET(retarray->dim[1], 0, + GFC_DESCRIPTOR_EXTENT(b,1) - 1, + GFC_DESCRIPTOR_EXTENT(retarray,0)); + } + + retarray->base_addr + = xmallocarray (size0 ((array_t *) retarray), sizeof (GFC_REAL_8)); + retarray->offset = 0; + } + else if (unlikely (compile_options.bounds_check)) + { + index_type ret_extent, arg_extent; + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else if (GFC_DESCRIPTOR_RANK (b) == 1) + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic: is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + else + { + arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 1:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + + arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); + ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); + if (arg_extent != ret_extent) + runtime_error ("Incorrect extent in return array in" + " MATMUL intrinsic for dimension 2:" + " is %ld, should be %ld", + (long int) ret_extent, (long int) arg_extent); + } + } + + + if (GFC_DESCRIPTOR_RANK (retarray) == 1) + { + /* One-dimensional result may be addressed in the code below + either as a row or a column matrix. We want both cases to + work. */ + rxstride = rystride = GFC_DESCRIPTOR_STRIDE(retarray,0); + } + else + { + rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); + rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); + } + + + if (GFC_DESCRIPTOR_RANK (a) == 1) + { + /* Treat it as a a row matrix A[1,count]. */ + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = 1; + + xcount = 1; + count = GFC_DESCRIPTOR_EXTENT(a,0); + } + else + { + axstride = GFC_DESCRIPTOR_STRIDE(a,0); + aystride = GFC_DESCRIPTOR_STRIDE(a,1); + + count = GFC_DESCRIPTOR_EXTENT(a,1); + xcount = GFC_DESCRIPTOR_EXTENT(a,0); + } + + if (count != GFC_DESCRIPTOR_EXTENT(b,0)) + { + if (count > 0 || GFC_DESCRIPTOR_EXTENT(b,0) > 0) + runtime_error ("dimension of array B incorrect in MATMUL intrinsic"); + } + + if (GFC_DESCRIPTOR_RANK (b) == 1) + { + /* Treat it as a column matrix B[count,1] */ + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + + /* bystride should never be used for 1-dimensional b. + The value is only used for calculation of the + memory by the buffer. */ + bystride = 256; + ycount = 1; + } + else + { + bxstride = GFC_DESCRIPTOR_STRIDE(b,0); + bystride = GFC_DESCRIPTOR_STRIDE(b,1); + ycount = GFC_DESCRIPTOR_EXTENT(b,1); + } + + abase = a->base_addr; + bbase = b->base_addr; + dest = retarray->base_addr; + + /* Now that everything is set up, we perform the multiplication + itself. */ + +#define POW3(x) (((float) (x)) * ((float) (x)) * ((float) (x))) +#define min(a,b) ((a) <= (b) ? (a) : (b)) +#define max(a,b) ((a) >= (b) ? (a) : (b)) + + if (try_blas && rxstride == 1 && (axstride == 1 || aystride == 1) + && (bxstride == 1 || bystride == 1) + && (((float) xcount) * ((float) ycount) * ((float) count) + > POW3(blas_limit))) + { + const int m = xcount, n = ycount, k = count, ldc = rystride; + const GFC_REAL_8 one = 1, zero = 0; + const int lda = (axstride == 1) ? aystride : axstride, + ldb = (bxstride == 1) ? bystride : bxstride; + + if (lda > 0 && ldb > 0 && ldc > 0 && m > 1 && n > 1 && k > 1) + { + assert (gemm != NULL); + gemm (axstride == 1 ? "N" : "T", bxstride == 1 ? "N" : "T", &m, + &n, &k, &one, abase, &lda, bbase, &ldb, &zero, dest, + &ldc, 1, 1); + return; + } + } + + if (rxstride == 1 && axstride == 1 && bxstride == 1) + { + /* This block of code implements a tuned matmul, derived from + Superscalar GEMM-based level 3 BLAS, Beta version 0.1 + + Bo Kagstrom and Per Ling + Department of Computing Science + Umea University + S-901 87 Umea, Sweden + + from netlib.org, translated to C, and modified for matmul.m4. */ + + const GFC_REAL_8 *a, *b; + GFC_REAL_8 *c; + const index_type m = xcount, n = ycount, k = count; + + /* System generated locals */ + index_type a_dim1, a_offset, b_dim1, b_offset, c_dim1, c_offset, + i1, i2, i3, i4, i5, i6; + + /* Local variables */ + GFC_REAL_8 f11, f12, f21, f22, f31, f32, f41, f42, + f13, f14, f23, f24, f33, f34, f43, f44; + index_type i, j, l, ii, jj, ll; + index_type isec, jsec, lsec, uisec, ujsec, ulsec; + GFC_REAL_8 *t1; + + a = abase; + b = bbase; + c = retarray->base_addr; + + /* Parameter adjustments */ + c_dim1 = rystride; + c_offset = 1 + c_dim1; + c -= c_offset; + a_dim1 = aystride; + a_offset = 1 + a_dim1; + a -= a_offset; + b_dim1 = bystride; + b_offset = 1 + b_dim1; + b -= b_offset; + + /* Early exit if possible */ + if (m == 0 || n == 0 || k == 0) + return; + + /* Adjust size of t1 to what is needed. */ + index_type t1_dim; + t1_dim = (a_dim1-1) * 256 + b_dim1; + if (t1_dim > 65536) + t1_dim = 65536; + + t1 = malloc (t1_dim * sizeof(GFC_REAL_8)); + + /* Empty c first. */ + for (j=1; j<=n; j++) + for (i=1; i<=m; i++) + c[i + j * c_dim1] = (GFC_REAL_8)0; + + /* Start turning the crank. */ + i1 = n; + for (jj = 1; jj <= i1; jj += 512) + { + /* Computing MIN */ + i2 = 512; + i3 = n - jj + 1; + jsec = min(i2,i3); + ujsec = jsec - jsec % 4; + i2 = k; + for (ll = 1; ll <= i2; ll += 256) + { + /* Computing MIN */ + i3 = 256; + i4 = k - ll + 1; + lsec = min(i3,i4); + ulsec = lsec - lsec % 2; + + i3 = m; + for (ii = 1; ii <= i3; ii += 256) + { + /* Computing MIN */ + i4 = 256; + i5 = m - ii + 1; + isec = min(i4,i5); + uisec = isec - isec % 2; + i4 = ll + ulsec - 1; + for (l = ll; l <= i4; l += 2) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 2) + { + t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] = + a[i + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 1) << 8) - 257] = + a[i + (l + 1) * a_dim1]; + t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + l * a_dim1]; + t1[l - ll + 2 + ((i - ii + 2) << 8) - 257] = + a[i + 1 + (l + 1) * a_dim1]; + } + if (uisec < isec) + { + t1[l - ll + 1 + (isec << 8) - 257] = + a[ii + isec - 1 + l * a_dim1]; + t1[l - ll + 2 + (isec << 8) - 257] = + a[ii + isec - 1 + (l + 1) * a_dim1]; + } + } + if (ulsec < lsec) + { + i4 = ii + isec - 1; + for (i = ii; i<= i4; ++i) + { + t1[lsec + ((i - ii + 1) << 8) - 257] = + a[i + (ll + lsec - 1) * a_dim1]; + } + } + + uisec = isec - isec % 4; + i4 = jj + ujsec - 1; + for (j = jj; j <= i4; j += 4) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f22 = c[i + 1 + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f23 = c[i + 1 + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + f24 = c[i + 1 + (j + 3) * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + f32 = c[i + 2 + (j + 1) * c_dim1]; + f42 = c[i + 3 + (j + 1) * c_dim1]; + f33 = c[i + 2 + (j + 2) * c_dim1]; + f43 = c[i + 3 + (j + 2) * c_dim1]; + f34 = c[i + 2 + (j + 3) * c_dim1]; + f44 = c[i + 3 + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f22 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f23 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f24 += t1[l - ll + 1 + ((i - ii + 2) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + j * b_dim1]; + f32 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f42 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 1) * b_dim1]; + f33 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f43 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 2) * b_dim1]; + f34 += t1[l - ll + 1 + ((i - ii + 3) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + f44 += t1[l - ll + 1 + ((i - ii + 4) << 8) - 257] + * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + (j + 1) * c_dim1] = f12; + c[i + 1 + (j + 1) * c_dim1] = f22; + c[i + (j + 2) * c_dim1] = f13; + c[i + 1 + (j + 2) * c_dim1] = f23; + c[i + (j + 3) * c_dim1] = f14; + c[i + 1 + (j + 3) * c_dim1] = f24; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + c[i + 2 + (j + 1) * c_dim1] = f32; + c[i + 3 + (j + 1) * c_dim1] = f42; + c[i + 2 + (j + 2) * c_dim1] = f33; + c[i + 3 + (j + 2) * c_dim1] = f43; + c[i + 2 + (j + 3) * c_dim1] = f34; + c[i + 3 + (j + 3) * c_dim1] = f44; + } + if (uisec < isec) + { + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + f12 = c[i + (j + 1) * c_dim1]; + f13 = c[i + (j + 2) * c_dim1]; + f14 = c[i + (j + 3) * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f12 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 1) * b_dim1]; + f13 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 2) * b_dim1]; + f14 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + (j + 3) * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + (j + 1) * c_dim1] = f12; + c[i + (j + 2) * c_dim1] = f13; + c[i + (j + 3) * c_dim1] = f14; + } + } + } + if (ujsec < jsec) + { + i4 = jj + jsec - 1; + for (j = jj + ujsec; j <= i4; ++j) + { + i5 = ii + uisec - 1; + for (i = ii; i <= i5; i += 4) + { + f11 = c[i + j * c_dim1]; + f21 = c[i + 1 + j * c_dim1]; + f31 = c[i + 2 + j * c_dim1]; + f41 = c[i + 3 + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + f21 += t1[l - ll + 1 + ((i - ii + 2) << 8) - + 257] * b[l + j * b_dim1]; + f31 += t1[l - ll + 1 + ((i - ii + 3) << 8) - + 257] * b[l + j * b_dim1]; + f41 += t1[l - ll + 1 + ((i - ii + 4) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + c[i + 1 + j * c_dim1] = f21; + c[i + 2 + j * c_dim1] = f31; + c[i + 3 + j * c_dim1] = f41; + } + i5 = ii + isec - 1; + for (i = ii + uisec; i <= i5; ++i) + { + f11 = c[i + j * c_dim1]; + i6 = ll + lsec - 1; + for (l = ll; l <= i6; ++l) + { + f11 += t1[l - ll + 1 + ((i - ii + 1) << 8) - + 257] * b[l + j * b_dim1]; + } + c[i + j * c_dim1] = f11; + } + } + } + } + } + } + free(t1); + return; + } + else if (rxstride == 1 && aystride == 1 && bxstride == 1) + { + if (GFC_DESCRIPTOR_RANK (a) != 1) + { + const GFC_REAL_8 *restrict abase_x; + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 *restrict dest_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n] * bbase_y[n]; + dest_y[x] = s; + } + } + } + else + { + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n]; + dest[y*rystride] = s; + } + } + } + else if (axstride < aystride) + { + for (y = 0; y < ycount; y++) + for (x = 0; x < xcount; x++) + dest[x*rxstride + y*rystride] = (GFC_REAL_8)0; + + for (y = 0; y < ycount; y++) + for (n = 0; n < count; n++) + for (x = 0; x < xcount; x++) + /* dest[x,y] += a[x,n] * b[n,y] */ + dest[x*rxstride + y*rystride] += + abase[x*axstride + n*aystride] * + bbase[n*bxstride + y*bystride]; + } + else if (GFC_DESCRIPTOR_RANK (a) == 1) + { + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase[n*axstride] * bbase_y[n*bxstride]; + dest[y*rxstride] = s; + } + } + else + { + const GFC_REAL_8 *restrict abase_x; + const GFC_REAL_8 *restrict bbase_y; + GFC_REAL_8 *restrict dest_y; + GFC_REAL_8 s; + + for (y = 0; y < ycount; y++) + { + bbase_y = &bbase[y*bystride]; + dest_y = &dest[y*rystride]; + for (x = 0; x < xcount; x++) + { + abase_x = &abase[x*axstride]; + s = (GFC_REAL_8) 0; + for (n = 0; n < count; n++) + s += abase_x[n*aystride] * bbase_y[n*bxstride]; + dest_y[x*rxstride] = s; + } + } + } +} +#undef POW3 +#undef min +#undef max + +#endif + +#endif + diff --git a/libgfortran/m4/matmul.m4 b/libgfortran/m4/matmul.m4 index 7976fda..c2f6415 100644 --- a/libgfortran/m4/matmul.m4 +++ b/libgfortran/m4/matmul.m4 @@ -106,6 +106,26 @@ static' include(matmul_internal.m4)dnl static' include(matmul_internal.m4)dnl `#endif /* HAVE_AVX512F */ +/* AMD-specifix funtions with AVX128 and FMA3/FMA4. */ + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +'define(`matmul_name',`matmul_'rtype_code`_avx128_fma3')dnl +`void +'matmul_name` ('rtype` * const restrict retarray, + 'rtype` * const restrict a, 'rtype` * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto('matmul_name`); +#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +'define(`matmul_name',`matmul_'rtype_code`_avx128_fma4')dnl +`void +'matmul_name` ('rtype` * const restrict retarray, + 'rtype` * const restrict a, 'rtype` * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto('matmul_name`); +#endif + /* Function to fall back to if there is no special processor-specific version. */ 'define(`matmul_name',`matmul_'rtype_code`_vanilla')dnl `static' include(matmul_internal.m4)dnl @@ -161,6 +181,26 @@ void matmul_'rtype_code` ('rtype` * const restrict retarray, } #endif /* HAVE_AVX */ } + else if (__cpu_model.__cpu_vendor == VENDOR_AMD) + { +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA))) + { + matmul_fn = matmul_'rtype_code`_avx128_fma3; + goto store; + } +#endif +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) + if ((__cpu_model.__cpu_features[0] & (1 << FEATURE_AVX)) + && (__cpu_model.__cpu_features[0] & (1 << FEATURE_FMA4))) + { + matmul_fn = matmul_'rtype_code`_avx128_fma4; + goto store; + } +#endif + + } store: __atomic_store_n (&matmul_p, matmul_fn, __ATOMIC_RELAXED); } diff --git a/libgfortran/m4/matmulavx128.m4 b/libgfortran/m4/matmulavx128.m4 new file mode 100644 index 0000000..1417284 --- /dev/null +++ b/libgfortran/m4/matmulavx128.m4 @@ -0,0 +1,67 @@ +`/* Implementation of the MATMUL intrinsic + Copyright (C) 2002-2017 Free Software Foundation, Inc. + Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>. + +This file is part of the GNU Fortran runtime library (libgfortran). + +Libgfortran is free software; you can redistribute it and/or +modify it under the terms of the GNU General Public +License as published by the Free Software Foundation; either +version 3 of the License, or (at your option) any later version. + +Libgfortran is distributed in the hope that it will be useful, +but WITHOUT ANY WARRANTY; without even the implied warranty of +MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +GNU General Public License for more details. + +Under Section 7 of GPL version 3, you are granted additional +permissions described in the GCC Runtime Library Exception, version +3.1, as published by the Free Software Foundation. + +You should have received a copy of the GNU General Public License and +a copy of the GCC Runtime Library Exception along with this program; +see the files COPYING3 and COPYING.RUNTIME respectively. If not, see +<http://www.gnu.org/licenses/>. */ + +#include "libgfortran.h" +#include <string.h> +#include <assert.h>' + +include(iparm.m4)dnl + +/* These are the specific versions of matmul with -mprefer-avx128. */ + +`#if defined (HAVE_'rtype_name`) + +/* Prototype for the BLAS ?gemm subroutine, a pointer to which can be + passed to us by the front-end, in which case we call it for large + matrices. */ + +typedef void (*blas_call)(const char *, const char *, const int *, const int *, + const int *, const 'rtype_name` *, const 'rtype_name` *, + const int *, const 'rtype_name` *, const int *, + const 'rtype_name` *, 'rtype_name` *, const int *, + int, int); + +#if defined(HAVE_AVX) && defined(HAVE_FMA3) && defined(HAVE_AVX128) +'define(`matmul_name',`matmul_'rtype_code`_avx128_fma3')dnl +`void +'matmul_name` ('rtype` * const restrict retarray, + 'rtype` * const restrict a, 'rtype` * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma"))); +internal_proto('matmul_name`); +'include(matmul_internal.m4)dnl +`#endif + +#if defined(HAVE_AVX) && defined(HAVE_FMA4) && defined(HAVE_AVX128) +'define(`matmul_name',`matmul_'rtype_code`_avx128_fma4')dnl +`void +'matmul_name` ('rtype` * const restrict retarray, + 'rtype` * const restrict a, 'rtype` * const restrict b, int try_blas, + int blas_limit, blas_call gemm) __attribute__((__target__("avx,fma4"))); +internal_proto('matmul_name`); +'include(matmul_internal.m4)dnl +`#endif + +#endif +' |