[AARCH64] Add support for floating-point vcond.

gcc/
	* config/aarch64/aarch64-simd.md
	(aarch64_simd_bsl<mode>_internal): Add floating-point modes.
	(aarch64_simd_bsl): Likewise.
	(aarch64_vcond_internal<mode>): Likewise.
	(vcond<mode><mode>): Likewise.
	(aarch64_cm<cmp><mode>): Fix constraints, add new modes.
	* config/aarch64/iterators.md (V_cmp_result): Add V2DF.

gcc/testsuite/
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-d.c: New.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-f.c: Likewise.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-d.c: Likewise.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-f.c: Likewise.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-d.c: Likewise.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-f.c: Likewise.
	* gcc/testsuite/gcc.target/aarch64/vect-fcm.x: Likewise.
	* gcc/testsuite/lib/target-supports.exp
	(check_effective_target_vect_cond): Enable for AArch64.

From-SVN: r195018

gcc/ChangeLog

 2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
 
+	* config/aarch64/aarch64-simd.md
+	(aarch64_simd_bsl<mode>_internal): Add floating-point modes.
+	(aarch64_simd_bsl): Likewise.
+	(aarch64_vcond_internal<mode>): Likewise.
+	(vcond<mode><mode>): Likewise.
+	(aarch64_cm<cmp><mode>): Fix constraints, add new modes.
+	* config/aarch64/iterators.md (V_cmp_result): Add V2DF.
+
+2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
+
 	* config/aarch64/aarch64-builtins.c
 	(aarch64_builtin_vectorized_function): Handle sqrt, sqrtf.
 

gcc/config/aarch64/aarch64-simd.md

@@ -1463,7 +1463,7 @@
    (set_attr "simd_mode" "V2SI")]
 )
 
-;; vbsl_* intrinsics may compile to any of vbsl/vbif/vbit depending on register
+;; vbsl_* intrinsics may compile to any of bsl/bif/bit depending on register
 ;; allocation.  For an intrinsic of form:
 ;;   vD = bsl_* (vS, vN, vM)
 ;; We can use any of:
@@ -1472,11 +1472,12 @@
 ;;   bif vD, vM, vS  (if D = N, so 0-bits in vS choose bits from vM, else vN)
 
 (define_insn "aarch64_simd_bsl<mode>_internal"
-  [(set (match_operand:VDQ 0 "register_operand"		     "=w,w,w")
-	(unspec:VDQ [(match_operand:VDQ 1 "register_operand" " 0,w,w")
-		     (match_operand:VDQ 2 "register_operand" " w,w,0")
-		     (match_operand:VDQ 3 "register_operand" " w,0,w")]
-		    UNSPEC_BSL))]
+  [(set (match_operand:VALL 0 "register_operand"		"=w,w,w")
+	(unspec:VALL
+	 [(match_operand:<V_cmp_result> 1 "register_operand"	" 0,w,w")
+	  (match_operand:VALL 2 "register_operand"		" w,w,0")
+	  (match_operand:VALL 3 "register_operand"		" w,0,w")]
+	 UNSPEC_BSL))]
   "TARGET_SIMD"
   "@
   bsl\\t%0.<Vbtype>, %2.<Vbtype>, %3.<Vbtype>
@@ -1485,15 +1486,15 @@
 )
 
 (define_expand "aarch64_simd_bsl<mode>"
-  [(set (match_operand:VDQ 0 "register_operand")
-	(unspec:VDQ [(match_operand:<V_cmp_result> 1 "register_operand")
-		     (match_operand:VDQ 2 "register_operand")
-		     (match_operand:VDQ 3 "register_operand")]
-		    UNSPEC_BSL))]
+  [(set (match_operand:VALL 0 "register_operand")
+	(unspec:VALL [(match_operand:<V_cmp_result> 1 "register_operand")
+		      (match_operand:VALL 2 "register_operand")
+		      (match_operand:VALL 3 "register_operand")]
+		     UNSPEC_BSL))]
   "TARGET_SIMD"
 {
   /* We can't alias operands together if they have different modes.  */
-  operands[1] = gen_lowpart (<MODE>mode, operands[1]);
+  operands[1] = gen_lowpart (<V_cmp_result>mode, operands[1]);
 })
 
 (define_expand "aarch64_vcond_internal<mode>"
@@ -1574,14 +1575,64 @@
   DONE;
 })
 
+(define_expand "aarch64_vcond_internal<mode>"
+  [(set (match_operand:VDQF 0 "register_operand")
+	(if_then_else:VDQF
+	  (match_operator 3 "comparison_operator"
+	    [(match_operand:VDQF 4 "register_operand")
+	     (match_operand:VDQF 5 "nonmemory_operand")])
+	  (match_operand:VDQF 1 "register_operand")
+	  (match_operand:VDQF 2 "register_operand")))]
+  "TARGET_SIMD"
+{
+  int inverse = 0;
+  rtx mask = gen_reg_rtx (<V_cmp_result>mode);
+
+  if (!REG_P (operands[5])
+      && (operands[5] != CONST0_RTX (<MODE>mode)))
+    operands[5] = force_reg (<MODE>mode, operands[5]);
+
+  switch (GET_CODE (operands[3]))
+    {
+    case LT:
+      inverse = 1;
+      /* Fall through.  */
+    case GE:
+      emit_insn (gen_aarch64_cmge<mode> (mask, operands[4], operands[5]));
+      break;
+    case LE:
+      inverse = 1;
+      /* Fall through.  */
+    case GT:
+      emit_insn (gen_aarch64_cmgt<mode> (mask, operands[4], operands[5]));
+      break;
+    case NE:
+      inverse = 1;
+      /* Fall through.  */
+    case EQ:
+      emit_insn (gen_aarch64_cmeq<mode> (mask, operands[4], operands[5]));
+      break;
+    default:
+      gcc_unreachable ();
+    }
+
+  if (inverse)
+    emit_insn (gen_aarch64_simd_bsl<mode> (operands[0], mask, operands[2],
+				    operands[1]));
+  else
+    emit_insn (gen_aarch64_simd_bsl<mode> (operands[0], mask, operands[1],
+				    operands[2]));
+  DONE;
+})
+
 (define_expand "vcond<mode><mode>"
-  [(set (match_operand:VDQ 0 "register_operand")
-	(if_then_else:VDQ
+  [(set (match_operand:VALL 0 "register_operand")
+	(if_then_else:VALL
 	  (match_operator 3 "comparison_operator"
-	    [(match_operand:VDQ 4 "register_operand")
-	     (match_operand:VDQ 5 "nonmemory_operand")])
-	  (match_operand:VDQ 1 "register_operand")
-	  (match_operand:VDQ 2 "register_operand")))]
+	    [(match_operand:VALL 4 "register_operand")
+	     (match_operand:VALL 5 "nonmemory_operand")])
+	  (match_operand:VALL 1 "register_operand")
+	  (match_operand:VALL 2 "register_operand")))]
   "TARGET_SIMD"
 {
   emit_insn (gen_aarch64_vcond_internal<mode> (operands[0], operands[1],
@@ -2866,6 +2917,22 @@
    (set_attr "simd_mode" "<MODE>")]
 )
 
+;; fcm(eq|ge|le|lt|gt)
+
+(define_insn "aarch64_cm<cmp><mode>"
+  [(set (match_operand:<V_cmp_result> 0 "register_operand" "=w,w")
+	(unspec:<V_cmp_result>
+	  [(match_operand:VDQF 1 "register_operand" "w,w")
+	   (match_operand:VDQF 2 "aarch64_simd_reg_or_zero" "w,Dz")]
+	   VCMP_S))]
+  "TARGET_SIMD"
+  "@
+  fcm<cmp>\t%<v>0<Vmtype>, %<v>1<Vmtype>, %<v>2<Vmtype>
+  fcm<cmp>\t%<v>0<Vmtype>, %<v>1<Vmtype>, 0"
+  [(set_attr "simd_type" "simd_fcmp")
+   (set_attr "simd_mode" "<MODE>")]
+)
+
 ;; addp
 
 (define_insn "aarch64_addp<mode>"

gcc/testsuite/ChangeLog

 2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
 
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-d.c: New.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-f.c: Likewise.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-d.c: Likewise.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-f.c: Likewise.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-d.c: Likewise.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-f.c: Likewise.
+	* gcc/testsuite/gcc.target/aarch64/vect-fcm.x: Likewise.
+	* gcc/testsuite/lib/target-supports.exp
+	(check_effective_target_vect_cond): Enable for AArch64.
+
+2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
+
 	* gcc.target/aarch64/vsqrt.c (test_square_root_v2sf): Use
 	endian-safe float pool loading.
 	(test_square_root_v4sf): Likewise.

gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-d.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE double
+#define OP ==
+#define INV_OP !=
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmeq\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, v\[0-9\]+\.2d" 2 } } */
+/* { dg-final { scan-assembler-times "fcmeq\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm-eq-f.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE float
+#define OP ==
+#define INV_OP !=
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmeq\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s" 2 } } */
+/* { dg-final { scan-assembler-times "fcmeq\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-d.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE double
+#define OP >=
+#define INV_OP <
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmge\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, v\[0-9\]+\.2d" 2 } } */
+/* { dg-final { scan-assembler-times "fcmge\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm-ge-f.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE float
+#define OP >=
+#define INV_OP <
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmge\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s" 2 } } */
+/* { dg-final { scan-assembler-times "fcmge\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-d.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE double
+#define OP >
+#define INV_OP <=
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmgt\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, v\[0-9\]+\.2d" 2 } } */
+/* { dg-final { scan-assembler-times "fcmgt\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm-gt-f.c

+/* { dg-do run } */
+/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-all -fno-unroll-loops --save-temps -fno-inline" } */
+
+#define FTYPE float
+#define OP >
+#define INV_OP <=
+
+#include "vect-fcm.x"
+
+/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 3 "vect" } } */
+/* { dg-final { scan-assembler-times "fcmgt\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s" 2 } } */
+/* { dg-final { scan-assembler-times "fcmgt\\tv\[0-9\]+\.\[24\]s, v\[0-9\]+\.\[24\]s, 0" 1 } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */
+/* { dg-final { cleanup-saved-temps } } */

gcc/testsuite/gcc.target/aarch64/vect-fcm.x

+#include <stdlib.h>
+#define N 16
+
+FTYPE input1[N] =
+{2.0, 4.0, 8.0, 16.0,
+ 2.125, 4.25, 8.5, 17.0,
+ -2.0, -4.0, -8.0, -16.0,
+ -2.125, -4.25, -8.5, -17.0};
+
+FTYPE input2[N] =
+{-2.0, 4.0, -8.0, 16.0,
+ 2.125, -4.25, 8.5, -17.0,
+ 2.0, -4.0, 8.0, -16.0,
+ -2.125, 4.25, -8.5, 17.0};
+
+void
+foo (FTYPE *in1, FTYPE *in2, FTYPE *output)
+{
+  int i = 0;
+  /* Vectorizable.  */
+  for (i = 0; i < N; i++)
+    output[i] = (in1[i] OP in2[i]) ? 2.0 : 4.0;
+}
+
+void
+bar (FTYPE *in1, FTYPE *in2, FTYPE *output)
+{
+  int i = 0;
+  /* Vectorizable.  */
+  for (i = 0; i < N; i++)
+    output[i] = (in1[i] INV_OP in2[i]) ? 4.0 : 2.0;
+}
+
+void
+foobar (FTYPE *in1, FTYPE *in2, FTYPE *output)
+{
+  int i = 0;
+  /* Vectorizable.  */
+  for (i = 0; i < N; i++)
+    output[i] = (in1[i] OP 0.0) ? 4.0 : 2.0;
+}
+
+int
+main (int argc, char **argv)
+{
+  FTYPE out1[N];
+  FTYPE out2[N];
+  int i = 0;
+  foo (input1, input2, out1);
+  bar (input1, input2, out2);
+  for (i = 0; i < N; i++)
+    if (out1[i] != out2[i])
+      abort ();
+  return 0;
+}
+

gcc/testsuite/lib/target-supports.exp

@@ -3699,7 +3699,8 @@ proc check_effective_target_vect_condition { } {
 	verbose "check_effective_target_vect_cond: using cached result" 2
     } else {
 	set et_vect_cond_saved 0
-	if { [istarget powerpc*-*-*]
+	if { [istarget aarch64*-*-*]
+	     || [istarget powerpc*-*-*]
 	     || [istarget ia64-*-*]
 	     || [istarget i?86-*-*]
 	     || [istarget spu-*-*]