diff options
author | Uros Bizjak <uros@gcc.gnu.org> | 2007-07-10 21:26:58 +0200 |
---|---|---|
committer | Uros Bizjak <uros@gcc.gnu.org> | 2007-07-10 21:26:58 +0200 |
commit | 558be301bddc36ccee1daf43d74c0a6296d3eb13 (patch) | |
tree | ee7ed21f874f952e037c7236575aadae2ed40fa9 /gcc | |
parent | 1ef8ecdd2e46e4b291cf7f3f1514497ba8bade5f (diff) | |
download | gcc-558be301bddc36ccee1daf43d74c0a6296d3eb13.zip gcc-558be301bddc36ccee1daf43d74c0a6296d3eb13.tar.gz gcc-558be301bddc36ccee1daf43d74c0a6296d3eb13.tar.bz2 |
re PR target/32708 (_mm_cvtsi64x_si128() and _mm_cvtsi128_si64x() inefficient)
PR target/32708
* config/i386/sse.md (vec_concatv2di): Disable for TARGET_64BIT.
(*vec_concatv2di_rex): New insn pattern.
testsuite/ChangeLog:
PR target/32708
* gcc.target/i386/pr32708-1.c: New test.
* gcc.target/i386/pr32708-2.c: Ditto.
* gcc.target/i386/pr32708-3.c: Ditto.
From-SVN: r126523
Diffstat (limited to 'gcc')
-rw-r--r-- | gcc/ChangeLog | 10 | ||||
-rw-r--r-- | gcc/config/i386/sse.md | 19 | ||||
-rw-r--r-- | gcc/testsuite/ChangeLog | 7 | ||||
-rw-r--r-- | gcc/testsuite/gcc.target/i386/pr32708-1.c | 18 | ||||
-rw-r--r-- | gcc/testsuite/gcc.target/i386/pr32708-2.c | 18 | ||||
-rw-r--r-- | gcc/testsuite/gcc.target/i386/pr32708-3.c | 18 |
6 files changed, 87 insertions, 3 deletions
diff --git a/gcc/ChangeLog b/gcc/ChangeLog index 8970931..6ebd9b0 100644 --- a/gcc/ChangeLog +++ b/gcc/ChangeLog @@ -1,3 +1,9 @@ +2007-07-10 Uros Bizjak <ubizjak@gmail.com> + + PR target/32708 + * config/i386/sse.md (vec_concatv2di): Disable for TARGET_64BIT. + (*vec_concatv2di_rex): New insn pattern. + 2007-07-10 Rainer Orth <ro@TechFak.Uni-Bielefeld.DE> PR target/32538 @@ -132,8 +138,8 @@ * optabs.h: Added declaration for signbit_optab. * optabs.c: (init_optabs): Added initialization for signbit_optab. * genoptinit.c (optabs): Added entry for signbit insns. - * builtins.c (expand_builtin_signbit): Added code to use a signbit insn, - if available. + * builtins.c (expand_builtin_signbit): Added code to use a signbit + insn, if available. * config/s390/s390.h (S390_TDC_SIGNBIT_SET): New constant. * config/s390/s390.md (signbit<mode>2): New expander. diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md index 12b8cc8..027d43b 100644 --- a/gcc/config/i386/sse.md +++ b/gcc/config/i386/sse.md @@ -4717,7 +4717,7 @@ (vec_concat:V2DI (match_operand:DI 1 "nonimmediate_operand" " m,*y ,0 ,0,0,m") (match_operand:DI 2 "vector_move_operand" " C, C,Yt,x,m,0")))] - "TARGET_SSE" + "!TARGET_64BIT && TARGET_SSE" "@ movq\t{%1, %0|%0, %1} movq2dq\t{%1, %0|%0, %1} @@ -4728,6 +4728,23 @@ [(set_attr "type" "ssemov,ssemov,sselog,ssemov,ssemov,ssemov") (set_attr "mode" "TI,TI,TI,V4SF,V2SF,V2SF")]) +(define_insn "*vec_concatv2di_rex" + [(set (match_operand:V2DI 0 "register_operand" "=Yt,Yi,!Yt,Yt,x,x,x") + (vec_concat:V2DI + (match_operand:DI 1 "nonimmediate_operand" " m,r ,*y ,0 ,0,0,m") + (match_operand:DI 2 "vector_move_operand" " C,C ,C ,Yt,x,m,0")))] + "TARGET_64BIT" + "@ + movq\t{%1, %0|%0, %1} + movq\t{%1, %0|%0, %1} + movq2dq\t{%1, %0|%0, %1} + punpcklqdq\t{%2, %0|%0, %2} + movlhps\t{%2, %0|%0, %2} + movhps\t{%2, %0|%0, %2} + movlps\t{%1, %0|%0, %1}" + [(set_attr "type" "ssemov,ssemov,ssemov,sselog,ssemov,ssemov,ssemov") + (set_attr "mode" "TI,TI,TI,TI,V4SF,V2SF,V2SF")]) + (define_expand "vec_setv2di" [(match_operand:V2DI 0 "register_operand" "") (match_operand:DI 1 "register_operand" "") diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 78fc492..a473461 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,10 @@ +2007-07-10 Uros Bizjak <ubizjak@gmail.com> + + PR target/32708 + * gcc.target/i386/pr32708-1.c: New test. + * gcc.target/i386/pr32708-2.c: Ditto. + * gcc.target/i386/pr32708-3.c: Ditto. + 2007-07-10 Mark Mitchell <mark@codesourcery.com> * g++.dg/init/new20.C: Remove svn:executable flag. diff --git a/gcc/testsuite/gcc.target/i386/pr32708-1.c b/gcc/testsuite/gcc.target/i386/pr32708-1.c new file mode 100644 index 0000000..4c2b50c --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr32708-1.c @@ -0,0 +1,18 @@ +/* { dg-do compile { target i?86-*-* x86_64-*-* } } */ +/* { dg-options "-O2 -msse2" } */ + +typedef long long __v2di __attribute__ ((__vector_size__ (16))); +typedef long long __m128i __attribute__ ((__vector_size__ (16))); + +static __inline __m128i __attribute__((__always_inline__)) +_mm_set_epi64x (long long __q1, long long __q0) +{ + return __extension__ (__m128i)(__v2di){ __q0, __q1 }; +} + +__m128i long2vector(long long __i) +{ + return _mm_set_epi64x (0, __i); +} + +/* { dg-final { scan-assembler-not "movq2dq" } } */ diff --git a/gcc/testsuite/gcc.target/i386/pr32708-2.c b/gcc/testsuite/gcc.target/i386/pr32708-2.c new file mode 100644 index 0000000..d8560f7 --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr32708-2.c @@ -0,0 +1,18 @@ +/* { dg-do compile { target { { i?86-*-* x86_64-*-* } && lp64 } } } */ +/* { dg-options "-O2 -mtune=k8" } */ + +typedef long long __v2di __attribute__ ((__vector_size__ (16))); +typedef long long __m128i __attribute__ ((__vector_size__ (16))); + +static __inline __m128i __attribute__((__always_inline__)) +_mm_set_epi64x (long long __q1, long long __q0) +{ + return __extension__ (__m128i)(__v2di){ __q0, __q1 }; +} + +__m128i long2vector(long long __i) +{ + return _mm_set_epi64x (0, __i); +} + +/* { dg-final { scan-assembler-not "movq2dq" } } */ diff --git a/gcc/testsuite/gcc.target/i386/pr32708-3.c b/gcc/testsuite/gcc.target/i386/pr32708-3.c new file mode 100644 index 0000000..b61b82a --- /dev/null +++ b/gcc/testsuite/gcc.target/i386/pr32708-3.c @@ -0,0 +1,18 @@ +/* { dg-do compile { target { { i?86-*-* x86_64-*-* } && lp64 } } } */ +/* { dg-options "-O2 -mtune=core2" } */ + +typedef long long __v2di __attribute__ ((__vector_size__ (16))); +typedef long long __m128i __attribute__ ((__vector_size__ (16))); + +static __inline __m128i __attribute__((__always_inline__)) +_mm_set_epi64x (long long __q1, long long __q0) +{ + return __extension__ (__m128i)(__v2di){ __q0, __q1 }; +} + +__m128i long2vector(long long __i) +{ + return _mm_set_epi64x (0, __i); +} + +/* { dg-final { scan-assembler-not "movq2dq" } } */ |