diff mbox

PR target/66144, PowerPC improve vector compare

Message ID 20170203212459.GA5647@ibm-tiger.the-meissners.org
State New
Headers show

Commit Message

Michael Meissner Feb. 3, 2017, 9:25 p.m. UTC
This patch improves the code generation for:

	vector char
	compare_vc (vector char a, vector char b)
	{
	  return a == b;
	}

Previously it generated:

        vcmpequb 2,2,3
        vspltisw 1,-1
        vspltisw 0,0
        xxsel 34,32,33,34

and now it generates:

        vcmpequb 2,2,3

I changed the vector conditional support for integer vectors to allow constant
all 0's or all 1's in addition to registers, and it knows that after a vector
comparison, the vector element will be all 1's if the comparison was true, and
all 0's if the comparison was false.

I did the normal bootstrap build and make check.  There were no regressions.

I built Spec 2006 with the compiler, and I noticed that 15 out of the 29
benchmarks had code changes with this fix (bzip2, gcc, gamess,milc, gromacs,
gobmk, dealII, calculix, hmmer, sjeng, libquantum, h264ref, tonto, wrf,
sphinx3).  I did a quick run (one pass instead of the normal 3) of the
benchmarks that had code changes.  There were no regressions in performance,
and 2 runs that were faster (gcc and tonto) by 3-4%.

Can I check this fix into the GCC trunk?

Note, I will be on starting on vacation on Wednesday (February 8th), and I will
return to work on Friday February 24th.  If I don't get the approval by Monday,
I will likely hold the patches until I return to work.

[gcc]
2017-02-03  Michael Meissner  <meissner@linux.vnet.ibm.com>

	PR target/66144
	* config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
	false values to be constant vectors with all 0 or all 1 bits set.
	(vcondu<mode><mode>): Likewise.
	* config/rs6000/predicates.md (vector_int_same_bit): New
	predicate.
	(vector_int_reg_or_same_bit): Likewise.
	(fpmask_comparison_operator): Update comment.
	(vecint_comparison_operator): New predicate.
	* config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
	vector conditionals when the true and false values are constant
	vectors with all 0 bits or all 1 bits set.

[gcc/testsuite]
2017-02-03  Michael Meissner  <meissner@linux.vnet.ibm.com>

	PR target/66144
	* gcc.target/powerpc/pr66144-1.c: New test.
	* gcc.target/powerpc/pr66144-2.c: Likewise.
	* gcc.target/powerpc/pr66144-3.c: Likewise.

Comments

Segher Boessenkool Feb. 4, 2017, 12:07 a.m. UTC | #1
On Fri, Feb 03, 2017 at 04:25:00PM -0500, Michael Meissner wrote:
> +;; Return 1 if operand is either a vector constant of all 0 bits of a vector
> +;; constant of all 1 bits.
> +(define_predicate "vector_int_same_bit"
> +  (match_code "const_vector")
> +{
> +  if (GET_MODE_CLASS (mode) != MODE_VECTOR_INT)
> +    return 0;
> +
> +  else
> +    return op == CONST0_RTX (mode) || op == CONSTM1_RTX (mode);
> +})

This predicate is unused as far as I see?

> +  /* Optimize vec1 == vec2, to know the mask generates -1/0.  */
> +  if (GET_MODE_CLASS (dest_mode) == MODE_VECTOR_INT)
>      {
> -      tmp = op_true;
> -      op_true = op_false;
> -      op_false = tmp;
> +      if (op_true == constant_m1 && op_false == constant_0)
> +	{
> +	  emit_move_insn (dest, mask);
> +	  return 1;
> +	}
> +
> +      else if (op_true == constant_0 && op_false == constant_m1)
> +	{
> +	  emit_insn (gen_rtx_SET (dest, gen_rtx_NOT (dest_mode, mask)));
> +	  return 1;
> +	}
>      }

Do you need to test for dest_mode == mask_mode here, like below?

> +  if (op_true == constant_m1 && dest_mode == mask_mode)
> +    op_true = mask;
> +  else if (!REG_P (op_true) && !SUBREG_P (op_true))
> +    op_true = force_reg (dest_mode, op_true);
> +
> +  if (op_false == constant_0 && dest_mode == mask_mode)
> +    op_false = mask;
> +  else if (!REG_P (op_false) && !SUBREG_P (op_false))
> +    op_false = force_reg (dest_mode, op_false);

Another thing you could try is, if either op_true or op_false is 0
or -1, let the result be
	(mask & op_true) | (~mask & op_false)

and let the rest of the optimisers sort it out (it's a single vor/vand
or vorc/vandc, or a vnot, or nothing).  A later improvement perhaps.
Or does it already handle all cases now :-)

Okay for trunk with the unused predicate removed, and the dest_mode ==
mask_mode thing looked at.  Thanks!


Segher
Michael Meissner Feb. 4, 2017, 12:18 a.m. UTC | #2
On Fri, Feb 03, 2017 at 06:07:56PM -0600, Segher Boessenkool wrote:
> On Fri, Feb 03, 2017 at 04:25:00PM -0500, Michael Meissner wrote:
> > +;; Return 1 if operand is either a vector constant of all 0 bits of a vector
> > +;; constant of all 1 bits.
> > +(define_predicate "vector_int_same_bit"
> > +  (match_code "const_vector")
> > +{
> > +  if (GET_MODE_CLASS (mode) != MODE_VECTOR_INT)
> > +    return 0;
> > +
> > +  else
> > +    return op == CONST0_RTX (mode) || op == CONSTM1_RTX (mode);
> > +})
> 
> This predicate is unused as far as I see?

Right.  It was used in my first attempt when I had a peephole2 to
eliminate the extra loads.  Since I moved the processing to when we create the
conditional vector assignment and deleted the peephole2, I missed deleting the
predicate.  Thanks for catching it.

> > +  /* Optimize vec1 == vec2, to know the mask generates -1/0.  */
> > +  if (GET_MODE_CLASS (dest_mode) == MODE_VECTOR_INT)
> >      {
> > -      tmp = op_true;
> > -      op_true = op_false;
> > -      op_false = tmp;
> > +      if (op_true == constant_m1 && op_false == constant_0)
> > +	{
> > +	  emit_move_insn (dest, mask);
> > +	  return 1;
> > +	}
> > +
> > +      else if (op_true == constant_0 && op_false == constant_m1)
> > +	{
> > +	  emit_insn (gen_rtx_SET (dest, gen_rtx_NOT (dest_mode, mask)));
> > +	  return 1;
> > +	}
> >      }
> 
> Do you need to test for dest_mode == mask_mode here, like below?

Yes, because there is support for vcondv4siv4sf and vcondv4sfv4si where the
mask is one of V4SI or V4SF and the destination is the other.  The dest_mode ==
mask_mode checks for that, and ignore that case.

I.e. something like:

static float af[1024], bf[1024], cf[1024];
static int di[1024], ei[1024];

// ..

for (i = 0; i < 1024; i++)
  af[i] = (di[i] == ei[i]) ? bf[i] : cf[i];

> 
> > +  if (op_true == constant_m1 && dest_mode == mask_mode)
> > +    op_true = mask;
> > +  else if (!REG_P (op_true) && !SUBREG_P (op_true))
> > +    op_true = force_reg (dest_mode, op_true);
> > +
> > +  if (op_false == constant_0 && dest_mode == mask_mode)
> > +    op_false = mask;
> > +  else if (!REG_P (op_false) && !SUBREG_P (op_false))
> > +    op_false = force_reg (dest_mode, op_false);
> 
> Another thing you could try is, if either op_true or op_false is 0
> or -1, let the result be
> 	(mask & op_true) | (~mask & op_false)
> 
> and let the rest of the optimisers sort it out (it's a single vor/vand
> or vorc/vandc, or a vnot, or nothing).  A later improvement perhaps.
> Or does it already handle all cases now :-)

I don't know if it would handle it.

> 
> Okay for trunk with the unused predicate removed, and the dest_mode ==
> mask_mode thing looked at.  Thanks!
diff mbox

Patch

Index: gcc/config/rs6000/vector.md
===================================================================
--- gcc/config/rs6000/vector.md	(revision 245101)
+++ gcc/config/rs6000/vector.md	(working copy)
@@ -390,13 +390,13 @@  (define_expand "vcond<mode><mode>"
 }")
 
 (define_expand "vcond<mode><mode>"
-  [(set (match_operand:VEC_I 0 "vint_operand" "")
+  [(set (match_operand:VEC_I 0 "vint_operand")
 	(if_then_else:VEC_I
 	 (match_operator 3 "comparison_operator"
-			 [(match_operand:VEC_I 4 "vint_operand" "")
-			  (match_operand:VEC_I 5 "vint_operand" "")])
-	 (match_operand:VEC_I 1 "vint_operand" "")
-	 (match_operand:VEC_I 2 "vint_operand" "")))]
+			 [(match_operand:VEC_I 4 "vint_operand")
+			  (match_operand:VEC_I 5 "vint_operand")])
+	 (match_operand:VEC_I 1 "vector_int_reg_or_same_bit")
+	 (match_operand:VEC_I 2 "vector_int_reg_or_same_bit")))]
   "VECTOR_UNIT_ALTIVEC_OR_VSX_P (<MODE>mode)"
   "
 {
@@ -446,13 +446,13 @@  (define_expand "vcondv4siv4sf"
 }")
 
 (define_expand "vcondu<mode><mode>"
-  [(set (match_operand:VEC_I 0 "vint_operand" "")
+  [(set (match_operand:VEC_I 0 "vint_operand")
 	(if_then_else:VEC_I
 	 (match_operator 3 "comparison_operator"
-			 [(match_operand:VEC_I 4 "vint_operand" "")
-			  (match_operand:VEC_I 5 "vint_operand" "")])
-	 (match_operand:VEC_I 1 "vint_operand" "")
-	 (match_operand:VEC_I 2 "vint_operand" "")))]
+			 [(match_operand:VEC_I 4 "vint_operand")
+			  (match_operand:VEC_I 5 "vint_operand")])
+	 (match_operand:VEC_I 1 "vector_int_reg_or_same_bit")
+	 (match_operand:VEC_I 2 "vector_int_reg_or_same_bit")))]
   "VECTOR_UNIT_ALTIVEC_OR_VSX_P (<MODE>mode)"
   "
 {
Index: gcc/config/rs6000/predicates.md
===================================================================
--- gcc/config/rs6000/predicates.md	(revision 245101)
+++ gcc/config/rs6000/predicates.md	(working copy)
@@ -808,6 +808,33 @@  (define_predicate "all_ones_constant"
   (and (match_code "const_int,const_double,const_wide_int,const_vector")
        (match_test "op == CONSTM1_RTX (mode) && !FLOAT_MODE_P (mode)")))
 
+;; Return 1 if operand is either a vector constant of all 0 bits of a vector
+;; constant of all 1 bits.
+(define_predicate "vector_int_same_bit"
+  (match_code "const_vector")
+{
+  if (GET_MODE_CLASS (mode) != MODE_VECTOR_INT)
+    return 0;
+
+  else
+    return op == CONST0_RTX (mode) || op == CONSTM1_RTX (mode);
+})
+
+;; Return 1 if operand is a vector int register or is either a vector constant
+;; of all 0 bits of a vector constant of all 1 bits.
+(define_predicate "vector_int_reg_or_same_bit"
+  (match_code "reg,subreg,const_vector")
+{
+  if (GET_MODE_CLASS (mode) != MODE_VECTOR_INT)
+    return 0;
+
+  else if (REG_P (op) || SUBREG_P (op))
+    return vint_operand (op, mode);
+
+  else
+    return op == CONST0_RTX (mode) || op == CONSTM1_RTX (mode);
+})
+
 ;; Return 1 if operand is 0.0.
 (define_predicate "zero_fp_constant"
   (and (match_code "const_double")
@@ -1260,8 +1287,8 @@  (define_predicate "scc_rev_comparison_op
   (and (match_operand 0 "branch_comparison_operator")
        (match_code "ne,le,ge,leu,geu,ordered")))
 
-;; Return 1 if OP is a comparison operator suitable for vector/scalar
-;; comparisons that generate a -1/0 mask.
+;; Return 1 if OP is a comparison operator suitable for floating point
+;; vector/scalar comparisons that generate a -1/0 mask.
 (define_predicate "fpmask_comparison_operator"
   (match_code "eq,gt,ge"))
 
@@ -1271,6 +1298,11 @@  (define_predicate "fpmask_comparison_ope
 (define_predicate "invert_fpmask_comparison_operator"
   (match_code "ne,unlt,unle"))
 
+;; Return 1 if OP is a comparison operation suitable for integer vector/scalar
+;; comparisons that generate a -1/0 mask.
+(define_predicate "vecint_comparison_operator"
+  (match_code "eq,gt,gtu"))
+
 ;; Return 1 if OP is a comparison operation that is valid for a branch
 ;; insn, which is true if the corresponding bit in the CC register is set.
 (define_predicate "branch_positive_comparison_operator"
Index: gcc/config/rs6000/rs6000.c
===================================================================
--- gcc/config/rs6000/rs6000.c	(revision 245101)
+++ gcc/config/rs6000/rs6000.c	(working copy)
@@ -25122,8 +25122,9 @@  rs6000_emit_vector_cond_expr (rtx dest, 
   machine_mode cc_mode = CCmode;
   rtx mask;
   rtx cond2;
-  rtx tmp;
   bool invert_move = false;
+  rtx constant_0;
+  rtx constant_m1;
 
   if (VECTOR_UNIT_NONE_P (dest_mode))
     return 0;
@@ -25166,15 +25167,40 @@  rs6000_emit_vector_cond_expr (rtx dest, 
   if (!mask)
     return 0;
 
+  constant_0 = CONST0_RTX (dest_mode);
+  constant_m1 = CONSTM1_RTX (dest_mode);
+
   if (invert_move)
+    std::swap (op_true, op_false);
+
+  /* Optimize vec1 == vec2, to know the mask generates -1/0.  */
+  if (GET_MODE_CLASS (dest_mode) == MODE_VECTOR_INT)
     {
-      tmp = op_true;
-      op_true = op_false;
-      op_false = tmp;
+      if (op_true == constant_m1 && op_false == constant_0)
+	{
+	  emit_move_insn (dest, mask);
+	  return 1;
+	}
+
+      else if (op_true == constant_0 && op_false == constant_m1)
+	{
+	  emit_insn (gen_rtx_SET (dest, gen_rtx_NOT (dest_mode, mask)));
+	  return 1;
+	}
     }
 
+  if (op_true == constant_m1 && dest_mode == mask_mode)
+    op_true = mask;
+  else if (!REG_P (op_true) && !SUBREG_P (op_true))
+    op_true = force_reg (dest_mode, op_true);
+
+  if (op_false == constant_0 && dest_mode == mask_mode)
+    op_false = mask;
+  else if (!REG_P (op_false) && !SUBREG_P (op_false))
+    op_false = force_reg (dest_mode, op_false);
+
   cond2 = gen_rtx_fmt_ee (NE, cc_mode, gen_lowpart (dest_mode, mask),
-			  CONST0_RTX (dest_mode));
+			  constant_0);
   emit_insn (gen_rtx_SET (dest,
 			  gen_rtx_IF_THEN_ELSE (dest_mode,
 						cond2,
Index: gcc/testsuite/gcc.target/powerpc/pr66144-1.c
===================================================================
--- gcc/testsuite/gcc.target/powerpc/pr66144-1.c	(revision 0)
+++ gcc/testsuite/gcc.target/powerpc/pr66144-1.c	(revision 0)
@@ -0,0 +1,20 @@ 
+/* { dg-do compile { target { powerpc64*-*-* } } } */
+/* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power9" } } */
+/* { dg-require-effective-target powerpc_p9vector_ok } */
+/* { dg-options "-mcpu=power9 -O2" } */
+
+/* Verify that we optimize vector1 = (vector2 == vector3) by not loading up
+   0/-1.  */
+
+vector int
+test (vector int a, vector int b)
+{
+  return a == b;
+}
+
+/* { dg-final { scan-assembler     {\mvcmpequw\M} } } */
+/* { dg-final { scan-assembler-not {\mxxspltib\M} } } */
+/* { dg-final { scan-assembler-not {\mvspltisw\M} } } */
+/* { dg-final { scan-assembler-not {\mxxlxor\M}   } } */
+/* { dg-final { scan-assembler-not {\mxxlxorc\M}  } } */
+/* { dg-final { scan-assembler-not {\mxxsel\M}    } } */
Index: gcc/testsuite/gcc.target/powerpc/pr66144-2.c
===================================================================
--- gcc/testsuite/gcc.target/powerpc/pr66144-2.c	(revision 0)
+++ gcc/testsuite/gcc.target/powerpc/pr66144-2.c	(revision 0)
@@ -0,0 +1,21 @@ 
+/* { dg-do compile { target { powerpc64*-*-* } } } */
+/* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power8" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2" } */
+
+/* Verify that we optimize vector1 = (vector2 != vector3) by not loading up
+   0/-1.  */
+
+vector unsigned char
+test (vector unsigned char a, vector unsigned char b)
+{
+  return a != b;
+}
+
+/* { dg-final { scan-assembler     {\mvcmpequb\M} } } */
+/* { dg-final { scan-assembler     {\mxxlnor\M}   } } */
+/* { dg-final { scan-assembler-not {\mxxspltib\M} } } */
+/* { dg-final { scan-assembler-not {\mvspltisw\M} } } */
+/* { dg-final { scan-assembler-not {\mxxlxor\M}   } } */
+/* { dg-final { scan-assembler-not {\mxxlxorc\M}  } } */
+/* { dg-final { scan-assembler-not {\mxxsel\M}    } } */
Index: gcc/testsuite/gcc.target/powerpc/pr66144-3.c
===================================================================
--- gcc/testsuite/gcc.target/powerpc/pr66144-3.c	(revision 0)
+++ gcc/testsuite/gcc.target/powerpc/pr66144-3.c	(revision 0)
@@ -0,0 +1,27 @@ 
+/* { dg-do compile { target { powerpc64*-*-* } } } */
+/* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power8" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize" } */
+
+/* Verify that we can optimize a vector conditional move, where one of the arms
+   is all 1's into using the mask as one of the inputs to XXSEL.  */
+
+#include <altivec.h>
+
+static int a[1024], b[1024], c[1024];
+
+int *p_a = a, *p_b = b, *p_c = c;
+
+void
+test (void)
+{
+  unsigned long i;
+
+  for (i = 0; i < 1024; i++)
+    a[i] = (b[i] == c[i]) ? -1 : a[i];
+}
+
+/* { dg-final { scan-assembler     {\mvcmpequw\M} } } */
+/* { dg-final { scan-assembler     {\mxxsel\M}    } } */
+/* { dg-final { scan-assembler-not {\mvspltisw\M} } } */
+/* { dg-final { scan-assembler-not {\mxxlorc\M}   } } */