Enable TARGET_LOOP_UNROLL_ADJUST for bdver3/bdver4

From-SVN: r206110

gcc/ChangeLog

+2013-12-19  Ganesh Gopalasubramanian  <Ganesh.Gopalasubramanian@amd.com>
+
+	* config/i386/i386.c: Include cfgloop.h. 
+	(ix86_loop_memcount): New function.
+	(ix86_loop_unroll_adjust): New function.
+	(TARGET_LOOP_UNROLL_ADJUST): Define.
+	* config/i386/i386.h
+	(TARGET_ADJUST_UNROLL): Define.
+	* config/i386/x86-tune.def
+	(X86_TUNE_ADJUST_UNROLL): Define.
+
 2013-12-19  Marek Polacek  <polacek@redhat.com>
 
 	* config/i386/i386.c (ix86_parse_stringop_strategy_string): Remove

gcc/config/i386/i386.c

@@ -64,6 +64,7 @@ along with GCC; see the file COPYING3.  If not see
 #include "is-a.h"
 #include "gimple.h"
 #include "gimplify.h"
+#include "cfgloop.h"
 #include "dwarf2.h"
 #include "df.h"
 #include "tm-constrs.h"
@@ -44014,6 +44015,64 @@ ix86_simd_clone_usable (struct cgraph_node *node)
     }
 }
 
+/* This function gives out the number of memory references.
+   This value determines the unrolling factor for
+   bdver3 and bdver4 architectures. */
+
+static int
+ix86_loop_memcount (rtx *x, unsigned *mem_count)
+{
+  if (*x != NULL_RTX && MEM_P (*x))
+   {
+     enum machine_mode mode;
+     unsigned int n_words;
+
+     mode = GET_MODE (*x);
+     n_words = GET_MODE_SIZE (mode) / UNITS_PER_WORD;
+
+    if (n_words > 4)
+       (*mem_count)+=2;
+    else
+       (*mem_count)+=1;
+   }
+  return 0;
+}
+
+/* This function adjusts the unroll factor based on
+   the hardware capabilities. For ex, bdver3 has
+   a loop buffer which makes unrolling of smaller
+   loops less important. This function decides the
+   unroll factor using number of memory references
+   (value 32 is used) as a heuristic. */
+
+static unsigned
+ix86_loop_unroll_adjust (unsigned nunroll, struct loop *loop)
+{
+  basic_block *bbs;
+  rtx insn;
+  unsigned i;
+  unsigned mem_count = 0;
+
+  if (!TARGET_ADJUST_UNROLL)
+     return nunroll;
+
+  /* Count the number of memory references within the loop body.  */
+  bbs = get_loop_body (loop);
+  for (i = 0; i < loop->num_nodes; i++)
+    {
+      for (insn = BB_HEAD (bbs[i]); insn != BB_END (bbs[i]); insn = NEXT_INSN (insn))
+        if (NONDEBUG_INSN_P (insn))
+            for_each_rtx (&insn, (rtx_function) ix86_loop_memcount, &mem_count);
+    }
+  free (bbs);
+
+  if (mem_count && mem_count <=32)
+    return 32/mem_count;
+
+  return nunroll;
+}
+
+
 /* Implement TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P.  */
 
 static bool
@@ -44499,6 +44558,9 @@ ix86_atomic_assign_expand_fenv (tree *hold, tree *clear, tree *update)
 #define TARGET_INIT_LIBFUNCS darwin_rename_builtins
 #endif
 
+#undef TARGET_LOOP_UNROLL_ADJUST
+#define TARGET_LOOP_UNROLL_ADJUST ix86_loop_unroll_adjust
+
 #undef TARGET_SPILL_CLASS
 #define TARGET_SPILL_CLASS ix86_spill_class
 

gcc/config/i386/i386.h

@@ -443,6 +443,8 @@ extern unsigned char ix86_tune_features[X86_TUNE_LAST];
 	ix86_tune_features[X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE]
 #define TARGET_SPLIT_MEM_OPND_FOR_FP_CONVERTS \
 	ix86_tune_features[X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS]
+#define TARGET_ADJUST_UNROLL \
+    ix86_tune_features[X86_TUNE_ADJUST_UNROLL]
 
 /* Feature tests against the various architecture variations.  */
 enum ix86_arch_indices {

gcc/config/i386/x86-tune.def

@@ -503,3 +503,9 @@ DEF_TUNE (X86_TUNE_QIMODE_MATH, "qimode_math", ~0)
    arithmetic to 32bit via PROMOTE_MODE macro.  This code generation scheme
    is usually used for RISC targets.  */
 DEF_TUNE (X86_TUNE_PROMOTE_QI_REGS, "promote_qi_regs", 0)
+
+/* X86_TUNE_ADJUST_UNROLL: This enables adjusting the unroll factor based
+   on hardware capabilities. Bdver3 hardware has a loop buffer which makes
+   unrolling small loop less important. For, such architectures we adjust
+   the unroll factor so that the unrolled loop fits the loop buffer.  */
+DEF_TUNE (X86_TUNE_ADJUST_UNROLL, "adjust_unroll_factor", m_BDVER3 | m_BDVER4)