[ARC] Rework code for profiling.

gcc/
2016-12-16  Claudiu Zissulescu  <claziss@synopsys.com>

	* config/arc/arc.h (LINK_SPEC): Tidy up.
	(ENDFILE_SPEC): Likewise.
	(LIB_SPEC): Likewise.
	(STARTFILE_SPEC): Include gcrt0 when profiling.
	(FUNCTION_PROFILER): Use __mcount.
	* config/arc/arc.opt (mucb-mcount): Remove.
	* doc/invoke.texi (ARC): Remove mucb-mcount doc.
	* arc/arc-protos.h (arc_profile_call): Remove.
	* config/arc/arc.c (write_profile_sections): Likewise.
	(arc_profile_call): Likewise.
	(unspec_prof_hash): Likewise.
	(unspec_prof_htab_eq): Likewise.
	(arc_legitimate_constant_p): Remove UNSPEC_PROF.
	(arc_reorg): Remove call to write_profile_sections.
	* config/arc/arc.md (call): Remove call to arc_profile_call.
	(call_value): Likewise.
	(sibcall): Likewise.
	(sibcall_value): Likewise.
	(define_constants): Remove UNSPEC_PROF.

libgcc/
	* config.host (arc*-*-linux-uclibc*): Remove libgmon, crtg, and
	crtgend.
	(arc*-*-elf*): Likewise.
	* config/arc/t-arc: Remove old gmon lib targets.
	* config/arc/crtg.S: Remove.
	* config/arc/crtgend.S: Likewise.
	* config/arc/gmon/atomic.h: Likewise.
	* config/arc/gmon/auxreg.h: Likewise.
	* config/arc/gmon/dcache_linesz.S: Likewise.
	* config/arc/gmon/gmon.c: Likewise.
	* config/arc/gmon/machine-gmon.h: Likewise.
	* config/arc/gmon/mcount.c: Likewise.
	* config/arc/gmon/prof-freq-stub.S: Likewise.
	* config/arc/gmon/prof-freq.c: Likewise.
	* config/arc/gmon/profil.S: Likewise.
	* config/arc/gmon/sys/gmon.h: Likewise.
	* config/arc/gmon/sys/gmon_out.h: Likewise.
	* config/arc/t-arc-newlib: Likewise.
	* config/arc/t-arc700-uClibc: Renamed to t-arc-uClibc.

From-SVN: r243742

7 files changed, 39 insertions, 211 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index fe59c99..e5f92a4 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,5 +1,27 @@
 2016-12-16  Claudiu Zissulescu  <claziss@synopsys.com>
 
+	* config/arc/arc.h (LINK_SPEC): Tidy up.
+	(ENDFILE_SPEC): Likewise.
+	(LIB_SPEC): Likewise.
+	(STARTFILE_SPEC): Include gcrt0 when profiling.
+	(FUNCTION_PROFILER): Use __mcount.
+	* config/arc/arc.opt (mucb-mcount): Remove.
+	* doc/invoke.texi (ARC): Remove mucb-mcount doc.
+	* arc/arc-protos.h (arc_profile_call): Remove.
+	* config/arc/arc.c (write_profile_sections): Likewise.
+	(arc_profile_call): Likewise.
+	(unspec_prof_hash): Likewise.
+	(unspec_prof_htab_eq): Likewise.
+	(arc_legitimate_constant_p): Remove UNSPEC_PROF.
+	(arc_reorg): Remove call to write_profile_sections.
+	* config/arc/arc.md (call): Remove call to arc_profile_call.
+	(call_value): Likewise.
+	(sibcall): Likewise.
+	(sibcall_value): Likewise.
+	(define_constants): Remove UNSPEC_PROF.
+
+2016-12-16  Claudiu Zissulescu  <claziss@synopsys.com>
+
 	* config/arc/arc.md (mulsidi_600): Change to insn_and_split,
 	generate new mul64 insn for core multiplication work.
 	(umulsidi_600): Likewise, but use mulu64 insn.
diff --git a/gcc/config/arc/arc-protos.h b/gcc/config/arc/arc-protos.h
index bdbf7ce..d896fe9 100644
--- a/gcc/config/arc/arc-protos.h
+++ b/gcc/config/arc/arc-protos.h
@@ -70,7 +70,6 @@ extern bool arc_raw_symbolic_reference_mentioned_p (rtx, bool);
 extern bool arc_legitimate_pic_operand_p (rtx);
 extern bool arc_is_longcall_p (rtx);
 extern bool arc_is_shortcall_p (rtx);
-extern bool arc_profile_call (rtx callee);
 extern bool valid_brcc_with_delay_p (rtx *);
 extern bool small_data_pattern (rtx , machine_mode);
 extern rtx arc_rewrite_small_data (rtx);
diff --git a/gcc/config/arc/arc.c b/gcc/config/arc/arc.c
index a0aa16e..9a72786 100644
--- a/gcc/config/arc/arc.c
+++ b/gcc/config/arc/arc.c
@@ -3610,97 +3610,6 @@ arc_print_operand_address (FILE *file , rtx addr)
     }
 }
 
-/* Called via walk_stores.  DATA points to a hash table we can use to
-   establish a unique SYMBOL_REF for each counter, which corresponds to
-   a caller-callee pair.
-   X is a store which we want to examine for an UNSPEC_PROF, which
-   would be an address loaded into a register, or directly used in a MEM.
-   If we found an UNSPEC_PROF, if we encounter a new counter the first time,
-   write out a description and a data allocation for a 32 bit counter.
-   Also, fill in the appropriate symbol_ref into each UNSPEC_PROF instance.  */
-
-static void
-write_profile_sections (rtx dest ATTRIBUTE_UNUSED, rtx x, void *data)
-{
-  rtx *srcp, src;
-  htab_t htab = (htab_t) data;
-  rtx *slot;
-
-  if (GET_CODE (x) != SET)
-    return;
-  srcp = &SET_SRC (x);
-  if (MEM_P (*srcp))
-    srcp = &XEXP (*srcp, 0);
-  else if (MEM_P (SET_DEST (x)))
-    srcp = &XEXP (SET_DEST (x), 0);
-  src = *srcp;
-  if (GET_CODE (src) != CONST)
-    return;
-  src = XEXP (src, 0);
-  if (GET_CODE (src) != UNSPEC || XINT (src, 1) != UNSPEC_PROF)
-    return;
-
-  gcc_assert (XVECLEN (src, 0) == 3);
-  if (!htab_elements (htab))
-    {
-      output_asm_insn (".section .__arc_profile_desc, \"a\"\n"
-		       "\t.long %0 + 1\n",
-		       &XVECEXP (src, 0, 0));
-    }
-  slot = (rtx *) htab_find_slot (htab, src, INSERT);
-  if (*slot == HTAB_EMPTY_ENTRY)
-    {
-      static int count_nr;
-      char buf[24];
-      rtx count;
-
-      *slot = src;
-      sprintf (buf, "__prof_count%d", count_nr++);
-      count = gen_rtx_SYMBOL_REF (Pmode, xstrdup (buf));
-      XVECEXP (src, 0, 2) = count;
-      output_asm_insn (".section\t.__arc_profile_desc, \"a\"\n"
-		       "\t.long\t%1\n"
-		       "\t.section\t.__arc_profile_counters, \"aw\"\n"
-		       "\t.type\t%o2, @object\n"
-		       "\t.size\t%o2, 4\n"
-		       "%o2:\t.zero 4",
-		       &XVECEXP (src, 0, 0));
-      *srcp = count;
-    }
-  else
-    *srcp = XVECEXP (*slot, 0, 2);
-}
-
-/* Hash function for UNSPEC_PROF htab.  Use both the caller's name and
-   the callee's name (if known).  */
-
-static hashval_t
-unspec_prof_hash (const void *x)
-{
-  const_rtx u = (const_rtx) x;
-  const_rtx s1 = XVECEXP (u, 0, 1);
-
-  return (htab_hash_string (XSTR (XVECEXP (u, 0, 0), 0))
-	  ^ (s1->code == SYMBOL_REF ? htab_hash_string (XSTR (s1, 0)) : 0));
-}
-
-/* Equality function for UNSPEC_PROF htab.  Two pieces of UNSPEC_PROF rtl
-   shall refer to the same counter if both caller name and callee rtl
-   are identical.  */
-
-static int
-unspec_prof_htab_eq (const void *x, const void *y)
-{
-  const_rtx u0 = (const_rtx) x;
-  const_rtx u1 = (const_rtx) y;
-  const_rtx s01 = XVECEXP (u0, 0, 1);
-  const_rtx s11 = XVECEXP (u1, 0, 1);
-
-  return (!strcmp (XSTR (XVECEXP (u0, 0, 0), 0),
-		   XSTR (XVECEXP (u1, 0, 0), 0))
-	  && rtx_equal_p (s01, s11));
-}
-
 /* Conditional execution support.
 
    This is based on the ARM port but for now is much simpler.
@@ -5438,7 +5347,6 @@ arc_legitimate_constant_p (machine_mode mode, rtx x)
 	  case UNSPEC_TLS_GD:
 	  case UNSPEC_TLS_IE:
 	  case UNSPEC_TLS_OFF:
-	  case UNSPEC_PROF:
 	    return true;
 
 	  default:
@@ -6359,47 +6267,6 @@ arc_is_shortcall_p (rtx sym_ref)
 
 }
 
-/* Emit profiling code for calling CALLEE.  Return true if a special
-   call pattern needs to be generated.  */
-
-bool
-arc_profile_call (rtx callee)
-{
-  rtx from = XEXP (DECL_RTL (current_function_decl), 0);
-
-  if (TARGET_UCB_MCOUNT)
-    /* Profiling is done by instrumenting the callee.  */
-    return false;
-
-  if (CONSTANT_P (callee))
-    {
-      rtx count_ptr
-	= gen_rtx_CONST (Pmode,
-			 gen_rtx_UNSPEC (Pmode,
-					 gen_rtvec (3, from, callee,
-						    CONST0_RTX (Pmode)),
-					 UNSPEC_PROF));
-      rtx counter = gen_rtx_MEM (SImode, count_ptr);
-      /* ??? The increment would better be done atomically, but as there is
-	 no proper hardware support, that would be too expensive.  */
-      emit_move_insn (counter,
-		      force_reg (SImode, plus_constant (SImode, counter, 1)));
-      return false;
-    }
-  else
-    {
-      rtx count_list_ptr
-	= gen_rtx_CONST (Pmode,
-			 gen_rtx_UNSPEC (Pmode,
-					 gen_rtvec (3, from, CONST0_RTX (Pmode),
-						    CONST0_RTX (Pmode)),
-					 UNSPEC_PROF));
-      emit_move_insn (gen_rtx_REG (Pmode, 8), count_list_ptr);
-      emit_move_insn (gen_rtx_REG (Pmode, 9), callee);
-      return true;
-    }
-}
-
 /* Worker function for TARGET_RETURN_IN_MEMORY.  */
 
 static bool
@@ -6621,25 +6488,6 @@ arc_reorg (void)
   cfun->machine->arc_reorg_started = 1;
   arc_reorg_in_progress = 1;
 
-  /* Emit special sections for profiling.  */
-  if (crtl->profile)
-    {
-      section *save_text_section;
-      rtx_insn *insn;
-      int size = get_max_uid () >> 4;
-      htab_t htab = htab_create (size, unspec_prof_hash, unspec_prof_htab_eq,
-				 NULL);
-
-      save_text_section = in_section;
-      for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
-	if (NONJUMP_INSN_P (insn))
-	  walk_stores (PATTERN (insn), write_profile_sections, htab);
-      if (htab_elements (htab))
-	in_section = 0;
-      switch_to_section (save_text_section);
-      htab_delete (htab);
-    }
-
   /* Link up loop ends with their loop start.  */
   {
     for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
diff --git a/gcc/config/arc/arc.h b/gcc/config/arc/arc.h
index f9512c4..87a2e11 100644
--- a/gcc/config/arc/arc.h
+++ b/gcc/config/arc/arc.h
@@ -121,12 +121,11 @@ extern const char *arc_cpu_to_as (int argc, const char **argv);
 		   -X %{mbig-endian:-EB} \
 		   %{EB} %{EL} \
 		   %{marclinux*} \
-		   %{!marclinux*: %{pg|p|profile:-marclinux_prof;: -marclinux}} \
+		   %{!marclinux*: -marclinux} \
 		   %{!z:-z max-page-size=0x2000 -z common-page-size=0x2000} \
 		   %{shared:-shared}"
 #else
-#define LINK_SPEC "%{mbig-endian:-EB} %{EB} %{EL}\
-  %{pg|p:-marcelf_prof;mA7|mARC700|mcpu=arc700|mcpu=ARC700: -marcelf}"
+#define LINK_SPEC "%{mbig-endian:-EB} %{EB} %{EL}"
 #endif
 
 #if DEFAULT_LIBC != LIBC_UCLIBC
@@ -135,7 +134,7 @@ extern const char *arc_cpu_to_as (int argc, const char **argv);
 #define EXTRA_SPECS \
   { "arc_tls_extra_start_spec", ARC_TLS_EXTRA_START_SPEC }, \
 
-#define STARTFILE_SPEC "%{!shared:crt0.o%s} crti%O%s %{pg|p:crtg.o%s} " \
+#define STARTFILE_SPEC "%{pg|p:gcrt0.o%s}%{!pg:%{!p:crt0.o%s}} crti%O%s " \
   "%(arc_tls_extra_start_spec) crtbegin.o%s"
 #else
 #define STARTFILE_SPEC							\
@@ -143,7 +142,7 @@ extern const char *arc_cpu_to_as (int argc, const char **argv);
 #endif
 
 #if DEFAULT_LIBC != LIBC_UCLIBC
-#define ENDFILE_SPEC "%{pg|p:crtgend.o%s} crtend.o%s crtn%O%s"
+#define ENDFILE_SPEC "crtend.o%s crtn%O%s"
 #else
 #define ENDFILE_SPEC							\
   LINUX_OR_ANDROID_LD (GNU_USER_TARGET_ENDFILE_SPEC, ANDROID_ENDFILE_SPEC)
@@ -154,12 +153,11 @@ extern const char *arc_cpu_to_as (int argc, const char **argv);
 #define LIB_SPEC  \
   "%{pthread:-lpthread} \
    %{shared:-lc} \
-   %{!shared:%{pg|p|profile:-lgmon -u profil --defsym __profil=profil} -lc}"
+   %{!shared:%{profile:-lc_p}%{!profile:-lc}}"
 #define TARGET_ASM_FILE_END file_end_indicate_exec_stack
 #else
 #undef LIB_SPEC
-/* -lc_p not present for arc-elf32-* : ashwin */
-#define LIB_SPEC "%{!shared:%{g*:-lg} %{pg|p:-lgmon} -lc}"
+#define LIB_SPEC "%{!shared:%{g*:-lg} -lc}"
 #endif
 
 #ifndef DRIVER_ENDIAN_SELF_SPECS
@@ -916,12 +914,14 @@ extern int arc_initial_elimination_offset(int from, int to);
   (OFFSET) = arc_initial_elimination_offset ((FROM), (TO))
 
 /* Output assembler code to FILE to increment profiler label # LABELNO
-   for profiling a function entry.
-   We actually emit the profiler code at the call site, so leave this one
-   empty.  */
-#define FUNCTION_PROFILER(FILE, LABELNO) \
-  if (TARGET_UCB_MCOUNT) \
-    fprintf (FILE, "\t%s\n", arc_output_libcall ("__mcount"))
+   for profiling a function entry.  */
+#define FUNCTION_PROFILER(FILE, LABELNO)			\
+  do {								\
+  if (flag_pic)							\
+    fprintf (FILE, "\tbl\t__mcount@plt\n");			\
+  else								\
+    fprintf (FILE, "\tbl\t__mcount\n");				\
+  } while (0);
 
 #define NO_PROFILE_COUNTERS  1
 
diff --git a/gcc/config/arc/arc.md b/gcc/config/arc/arc.md
index 6fe237e..2d3e19b 100644
--- a/gcc/config/arc/arc.md
+++ b/gcc/config/arc/arc.md
@@ -9,7 +9,7 @@
 ;;    Saurabh Verma (saurabh.verma@codito.com)
 ;;    Ramana Radhakrishnan(ramana.radhakrishnan@codito.com)
 ;;
-;;    Profiling support and performance improvements by
+;;    Performance improvements by
 ;;    Joern Rennecke (joern.rennecke@embecosm.com)
 ;;
 
@@ -165,9 +165,7 @@
   ])
 
 (define_constants
-  [(UNSPEC_PROF 18) ; profile callgraph counter
-
-   (R0_REG 0)
+  [(R0_REG 0)
    (R1_REG 1)
    (R2_REG 2)
    (R3_REG 3)
@@ -4108,13 +4106,6 @@
 
     gcc_assert (MEM_P (operands[0]));
     callee  = XEXP (operands[0], 0);
-    if (crtl->profile && arc_profile_call (callee))
-      {
-	emit_call_insn (gen_call_prof (gen_rtx_SYMBOL_REF (Pmode,
-							   \"_mcount_call\"),
-				       operands[1]));
-	DONE;
-      }
     /* This is to decide if we should generate indirect calls by loading the
        32 bit address of the callee into a register before performing the
        branch and link - this exposes cse opportunities.
@@ -4177,14 +4168,6 @@
 
     gcc_assert (MEM_P (operands[1]));
     callee = XEXP (operands[1], 0);
-    if (crtl->profile && arc_profile_call (callee))
-      {
-	emit_call_insn (gen_call_value_prof (operands[0],
-					     gen_rtx_SYMBOL_REF (Pmode,
-							    \"_mcount_call\"),
-					     operands[2]));
-	DONE;
-      }
      /* See the comment in define_expand \"call\".  */
     if (GET_CODE (callee) != REG
 	&& (GET_CODE (callee) == PLUS || arc_is_longcall_p (callee)))
@@ -4667,13 +4650,6 @@
 
     if (operands[2] == NULL_RTX)
       operands[2] = const0_rtx;
-    if (crtl->profile && arc_profile_call (callee))
-      {
-	emit_insn (gen_sibcall_prof
-		    (gen_rtx_SYMBOL_REF (Pmode, \"_mcount_call\"),
-		     operands[1], operands[2]));
-	DONE;
-      }
     if (GET_CODE (callee) != REG
 	&& (GET_CODE (callee) == PLUS || arc_is_longcall_p (callee)))
       XEXP (operands[0], 0) = force_reg (Pmode, callee);
@@ -4693,13 +4669,6 @@
 
     if (operands[3] == NULL_RTX)
       operands[3] = const0_rtx;
-    if (crtl->profile && arc_profile_call (XEXP (operands[1], 0)))
-      {
-	emit_insn (gen_sibcall_value_prof
-		    (operands[0], gen_rtx_SYMBOL_REF (Pmode, \"_mcount_call\"),
-		     operands[2], operands[3]));
-	DONE;
-      }
     if (GET_CODE (callee) != REG && arc_is_longcall_p (callee))
       XEXP (operands[1], 0) = force_reg (Pmode, callee);
   }"
diff --git a/gcc/config/arc/arc.opt b/gcc/config/arc/arc.opt
index 31b305bb..ec4f055 100644
--- a/gcc/config/arc/arc.opt
+++ b/gcc/config/arc/arc.opt
@@ -409,10 +409,6 @@ mlra-priority-noncompact
 Target RejectNegative Var(arc_lra_prioritytag, ARC_LRA_PRIORITY_NONCOMPACT)
 Reduce priority for r0..r3 / r12..r15 with TARGET_REGISTER_PRIORITY.
 
-mucb-mcount
-Target Report Var(TARGET_UCB_MCOUNT)
-instrument with mcount calls as in the ucb code.
-
 ; backward-compatibility aliases, translated by DRIVER_SELF_SPECS
 
 mEA
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index 9eda687..b729964 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -611,7 +611,7 @@ Objective-C and Objective-C++ Dialects}.
 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
 -mlong-calls -mmedium-calls -msdata @gol
--mucb-mcount -mvolatile-cache -mtp-regno=@var{regno} @gol
+-mvolatile-cache -mtp-regno=@var{regno} @gol
 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
@@ -14726,12 +14726,6 @@ Do not generate sdata references.  This is the default for tool chains
 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
 targets.
 
-@item -mucb-mcount
-@opindex mucb-mcount
-Instrument with mcount calls as used in UCB code.  I.e. do the
-counting in the callee, not the caller.  By default ARC instrumentation
-counts in the caller.
-
 @item -mvolatile-cache
 @opindex mvolatile-cache
 Use ordinarily cached memory accesses for volatile references.  This is the