rs6000: Refine small loop unroll in loop_unroll_adjust hook

In this patch, loop unroll adjust hook is introduced for powerpc.  We
can do target related heuristic adjustment in this hook.  In this patch,
-funroll-loops is enabled for small loops at O2 and above with an option
-munroll-small-loops to guard the small loops unrolling, and it works
fine with -flto.


gcc/
2019-11-11  Jiufu Guo  <guojiufu@linux.ibm.com>

	PR tree-optimization/88760
	* gcc/config/rs6000/rs6000.opt (-munroll-only-small-loops): New option.
	* gcc/common/config/rs6000/rs6000-common.c
	(rs6000_option_optimization_table) [OPT_LEVELS_2_PLUS_SPEED_ONLY]:
	Turn on -funroll-loops and -munroll-only-small-loops.
	[OPT_LEVELS_ALL]: Turn off -fweb and -frename-registers.
	* config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
	set of PARAM_MAX_UNROLL_TIMES and PARAM_MAX_UNROLLED_INSNS.
	Turn off -munroll-only-small-loops for explicit -funroll-loops.
	(TARGET_LOOP_UNROLL_ADJUST): Add loop unroll adjust hook.
	(rs6000_loop_unroll_adjust): Define it.  Use -munroll-only-small-loops.

gcc.testsuite/
2019-11-11  Jiufu Guo  <guojiufu@linux.ibm.com>

	PR tree-optimization/88760
	* gcc.dg/pr59643.c: Update back to r277550.

From-SVN: r278034

gcc/ChangeLog

+2019-11-11  Jiufu Guo  <guojiufu@linux.ibm.com>
+
+	PR tree-optimization/88760
+	* gcc/config/rs6000/rs6000.opt (-munroll-only-small-loops): New option.
+	* gcc/common/config/rs6000/rs6000-common.c
+	(rs6000_option_optimization_table) [OPT_LEVELS_2_PLUS_SPEED_ONLY]:
+	Turn on -funroll-loops and -munroll-only-small-loops.
+	[OPT_LEVELS_ALL]: Turn off -fweb and -frename-registers.
+	* config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
+	set of PARAM_MAX_UNROLL_TIMES and PARAM_MAX_UNROLLED_INSNS.
+	Turn off -munroll-only-small-loops for explicit -funroll-loops.
+	(TARGET_LOOP_UNROLL_ADJUST): Add loop unroll adjust hook.
+	(rs6000_loop_unroll_adjust): Define it.  Use -munroll-only-small-loops.
+
 2019-11-11  Kewen Lin  <linkw@gcc.gnu.org>
 
 	* config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):

gcc/common/config/rs6000/rs6000-common.c

@@ -35,7 +35,14 @@ static const struct default_options rs6000_option_optimization_table[] =
     { OPT_LEVELS_ALL, OPT_fsplit_wide_types_early, NULL, 1 },
     /* Enable -fsched-pressure for first pass instruction scheduling.  */
     { OPT_LEVELS_1_PLUS, OPT_fsched_pressure, NULL, 1 },
-    { OPT_LEVELS_2_PLUS, OPT_funroll_loops, NULL, 1 },
+    /* Enable -munroll-only-small-loops with -funroll-loops to unroll small
+       loops at -O2 and above by default.  */
+    { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_funroll_loops, NULL, 1 },
+    { OPT_LEVELS_2_PLUS_SPEED_ONLY, OPT_munroll_only_small_loops, NULL, 1 },
+    /* -fweb and -frename-registers are useless in general for rs6000,
+       turn them off.  */
+    { OPT_LEVELS_ALL, OPT_fweb, NULL, 0 },
+    { OPT_LEVELS_ALL, OPT_frename_registers, NULL, 0 },
     { OPT_LEVELS_NONE, 0, NULL, 0 }
   };
 

gcc/config/rs6000/rs6000.c

@@ -1428,6 +1428,9 @@ static const struct attribute_spec rs6000_attribute_table[] =
 #undef TARGET_VECTORIZE_DESTROY_COST_DATA
 #define TARGET_VECTORIZE_DESTROY_COST_DATA rs6000_destroy_cost_data
 
+#undef TARGET_LOOP_UNROLL_ADJUST
+#define TARGET_LOOP_UNROLL_ADJUST rs6000_loop_unroll_adjust
+
 #undef TARGET_INIT_BUILTINS
 #define TARGET_INIT_BUILTINS rs6000_init_builtins
 #undef TARGET_BUILTIN_DECL
@@ -4540,25 +4543,12 @@ rs6000_option_override_internal (bool global_init_p)
 			     global_options.x_param_values,
 			     global_options_set.x_param_values);
 
-      /* unroll very small loops 2 time if no -funroll-loops.  */
-      if (!global_options_set.x_flag_unroll_loops
-	  && !global_options_set.x_flag_unroll_all_loops)
-	{
-	  maybe_set_param_value (PARAM_MAX_UNROLL_TIMES, 2,
-				 global_options.x_param_values,
-				 global_options_set.x_param_values);
-
-	  maybe_set_param_value (PARAM_MAX_UNROLLED_INSNS, 20,
-				 global_options.x_param_values,
-				 global_options_set.x_param_values);
-
-	  /* If fweb or frename-registers are not specificed in command-line,
-	     do not turn them on implicitly.  */
-	  if (!global_options_set.x_flag_web)
-	    global_options.x_flag_web = 0;
-	  if (!global_options_set.x_flag_rename_registers)
-	    global_options.x_flag_rename_registers = 0;
-	}
+      /* Explicit -funroll-loops turns -munroll-only-small-loops off.  */
+      if (((global_options_set.x_flag_unroll_loops && flag_unroll_loops)
+	   || (global_options_set.x_flag_unroll_all_loops
+	       && flag_unroll_all_loops))
+	  && !global_options_set.x_unroll_only_small_loops)
+	unroll_only_small_loops = 0;
 
       /* If using typedef char *va_list, signal that
 	 __builtin_va_start (&ap, 0) can be optimized to
@@ -5105,6 +5095,25 @@ rs6000_destroy_cost_data (void *data)
   free (data);
 }
 
+/* Implement targetm.loop_unroll_adjust.  */
+
+static unsigned
+rs6000_loop_unroll_adjust (unsigned nunroll, struct loop *loop)
+{
+   if (unroll_only_small_loops)
+    {
+      /* TODO: This is hardcoded to 10 right now.  It can be refined, for
+	 example we may want to unroll very small loops more times (4 perhaps).
+	 We also should use a PARAM for this.  */
+      if (loop->ninsns <= 10)
+	return MIN (2, nunroll);
+      else
+	return 0;
+    }
+
+  return nunroll;
+}
+
 /* Handler for the Mathematical Acceleration Subsystem (mass) interface to a
    library with vectorized intrinsics.  */
 

gcc/config/rs6000/rs6000.opt

@@ -501,6 +501,10 @@ moptimize-swaps
 Target Undocumented Var(rs6000_optimize_swaps) Init(1) Save
 Analyze and remove doubleword swaps from VSX computations.
 
+munroll-only-small-loops
+Target Undocumented Var(unroll_only_small_loops) Init(0) Save
+; Use conservative small loop unrolling.
+
 mpower9-misc
 Target Undocumented Report Mask(P9_MISC) Var(rs6000_isa_flags)
 Use certain scalar instructions added in ISA 3.0.

gcc/testsuite/ChangeLog

+2019-11-11  Jiufu Guo  <guojiufu@linux.ibm.com>
+
+	PR tree-optimization/88760
+	* gcc.dg/pr59643.c: Update back to r277550.
+
 2019-11-10  Paul Thomas  <pault@gcc.gnu.org>
 
 	PR fortran/92123

gcc/testsuite/gcc.dg/pr59643.c

 /* PR tree-optimization/59643 */
 /* { dg-do compile } */
 /* { dg-options "-O3 -fdump-tree-pcom-details" } */
-/* { dg-additional-options "--param max-unrolled-insns=400" { target { powerpc*-*-* } } } */
-/* Implicit threashold of max-unrolled-insn on ppc at O3 is too small for the
-   loop of this case.  */
 
 void
 foo (double *a, double *b, double *c, double d, double e, int n)