2013-12-11 Sriraman Tallam <tmsriram@google.com>

	PR target/59390
	* config/i386/i386.c (get_builtin): New function.
	(ix86_builtin_vectorized_function): Replace all instances of
	ix86_builtins[...] with get_builtin(...).
	(ix86_builtin_reciprocal): Ditto.

	PR target/59390
	* gcc.target/i386/pr59390.c: New test.
	* gcc.target/i386/pr59390_1.c: New test.
	* gcc.target/i386/pr59390_2.c: New test.

Tested boot-strap on all languages and test parity on x86_64 and with -m32.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@205904 138bc75d-0d04-0410-961f-82ee72b054a4

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 3e41401..d95ce52 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,12 @@
+2013-12-11  Sriraman Tallam  <tmsriram@google.com>
+
+       PR target/59390
+       * config/i386/i386.c (get_builtin): New function.
+       (ix86_builtin_vectorized_function): Replace all instances of
+       ix86_builtins[...] with get_builtin(...).
+       (ix86_builtin_reciprocal): Ditto.
+
+
 2013-12-11  Balaji V. Iyer  <balaji.v.iyer@intel.com>
 
        * langhooks.h (lang_hooks_for_decls): Remove lang_hooks_for_cilkplus.

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index cdd63e5..8f79209 100644 (file)
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -33777,6 +33777,31 @@ addcarryx:
   gcc_unreachable ();
 }
 
+/* This returns the target-specific builtin with code CODE if
+   current_function_decl has visibility on this builtin, which is checked
+   using isa flags.  Returns NULL_TREE otherwise.  */
+
+static tree ix86_get_builtin (enum ix86_builtins code)
+{
+  struct cl_target_option *opts;
+  tree target_tree = NULL_TREE;
+
+  /* Determine the isa flags of current_function_decl.  */
+
+  if (current_function_decl)
+    target_tree = DECL_FUNCTION_SPECIFIC_TARGET (current_function_decl);
+
+  if (target_tree == NULL)
+    target_tree = target_option_default_node;
+
+  opts = TREE_TARGET_OPTION (target_tree);
+
+  if (ix86_builtins_isa[(int) code].isa & opts->x_ix86_isa_flags)
+    return ix86_builtin_decl (code, true);
+  else
+    return NULL_TREE;
+}
+
 /* Returns a function decl for a vectorized version of the builtin function
    with builtin function code FN and the result vector type TYPE, or NULL_TREE
    if it is not available.  */
@@ -33805,9 +33830,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_SQRTPD];
+           return ix86_get_builtin (IX86_BUILTIN_SQRTPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_SQRTPD256];
+           return ix86_get_builtin (IX86_BUILTIN_SQRTPD256);
        }
       break;
 
@@ -33815,9 +33840,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_SQRTPS_NR];
+           return ix86_get_builtin (IX86_BUILTIN_SQRTPS_NR);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_SQRTPS_NR256];
+           return ix86_get_builtin (IX86_BUILTIN_SQRTPS_NR256);
        }
       break;
 
@@ -33831,9 +33856,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == DFmode)
        {
          if (out_n == 4 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX);
          else if (out_n == 8 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPD_VEC_PACK_SFIX256);
        }
       break;
 
@@ -33847,9 +33872,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_FLOORPS_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPS_SFIX);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_FLOORPS_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPS_SFIX256);
        }
       break;
 
@@ -33863,9 +33888,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == DFmode)
        {
          if (out_n == 4 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_CEILPD_VEC_PACK_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPD_VEC_PACK_SFIX);
          else if (out_n == 8 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CEILPD_VEC_PACK_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPD_VEC_PACK_SFIX256);
        }
       break;
 
@@ -33879,9 +33904,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CEILPS_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPS_SFIX);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_CEILPS_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPS_SFIX256);
        }
       break;
 
@@ -33891,9 +33916,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == DFmode)
        {
          if (out_n == 4 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_VEC_PACK_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_VEC_PACK_SFIX);
          else if (out_n == 8 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_VEC_PACK_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_VEC_PACK_SFIX256);
        }
       break;
 
@@ -33903,9 +33928,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CVTPS2DQ];
+           return ix86_get_builtin (IX86_BUILTIN_CVTPS2DQ);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_CVTPS2DQ256];
+           return ix86_get_builtin (IX86_BUILTIN_CVTPS2DQ256);
        }
       break;
 
@@ -33919,9 +33944,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == DFmode)
        {
          if (out_n == 4 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX);
          else if (out_n == 8 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPD_AZ_VEC_PACK_SFIX256);
        }
       break;
 
@@ -33935,9 +33960,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SImode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPS_AZ_SFIX];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ_SFIX);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPS_AZ_SFIX256];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ_SFIX256);
        }
       break;
 
@@ -33945,9 +33970,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_CPYSGNPD];
+           return ix86_get_builtin (IX86_BUILTIN_CPYSGNPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CPYSGNPD256];
+           return ix86_get_builtin (IX86_BUILTIN_CPYSGNPD256);
        }
       break;
 
@@ -33955,9 +33980,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CPYSGNPS];
+           return ix86_get_builtin (IX86_BUILTIN_CPYSGNPS);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_CPYSGNPS256];
+           return ix86_get_builtin (IX86_BUILTIN_CPYSGNPS256);
        }
       break;
 
@@ -33969,9 +33994,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_FLOORPD];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_FLOORPD256];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPD256);
        }
       break;
 
@@ -33983,9 +34008,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_FLOORPS];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPS);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_FLOORPS256];
+           return ix86_get_builtin (IX86_BUILTIN_FLOORPS256);
        }
       break;
 
@@ -33997,9 +34022,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_CEILPD];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CEILPD256];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPD256);
        }
       break;
 
@@ -34011,9 +34036,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_CEILPS];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPS);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_CEILPS256];
+           return ix86_get_builtin (IX86_BUILTIN_CEILPS256);
        }
       break;
 
@@ -34025,9 +34050,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_TRUNCPD];
+           return ix86_get_builtin (IX86_BUILTIN_TRUNCPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_TRUNCPD256];
+           return ix86_get_builtin (IX86_BUILTIN_TRUNCPD256);
        }
       break;
 
@@ -34039,9 +34064,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_TRUNCPS];
+           return ix86_get_builtin (IX86_BUILTIN_TRUNCPS);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_TRUNCPS256];
+           return ix86_get_builtin (IX86_BUILTIN_TRUNCPS256);
        }
       break;
 
@@ -34053,9 +34078,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_RINTPD];
+           return ix86_get_builtin (IX86_BUILTIN_RINTPD);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_RINTPD256];
+           return ix86_get_builtin (IX86_BUILTIN_RINTPD256);
        }
       break;
 
@@ -34067,9 +34092,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_RINTPS];
+           return ix86_get_builtin (IX86_BUILTIN_RINTPS);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_RINTPS256];
+           return ix86_get_builtin (IX86_BUILTIN_RINTPS256);
        }
       break;
 
@@ -34081,9 +34106,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPD_AZ];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPD_AZ);
          else if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPD_AZ256];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPD_AZ256);
        }
       break;
 
@@ -34095,9 +34120,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPS_AZ];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ);
          else if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_ROUNDPS_AZ256];
+           return ix86_get_builtin (IX86_BUILTIN_ROUNDPS_AZ256);
        }
       break;
 
@@ -34105,9 +34130,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == DFmode && in_mode == DFmode)
        {
          if (out_n == 2 && in_n == 2)
-           return ix86_builtins[IX86_BUILTIN_VFMADDPD];
+           return ix86_get_builtin (IX86_BUILTIN_VFMADDPD);
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_VFMADDPD256];
+           return ix86_get_builtin (IX86_BUILTIN_VFMADDPD256);
        }
       break;
 
@@ -34115,9 +34140,9 @@ ix86_builtin_vectorized_function (tree fndecl, tree type_out,
       if (out_mode == SFmode && in_mode == SFmode)
        {
          if (out_n == 4 && in_n == 4)
-           return ix86_builtins[IX86_BUILTIN_VFMADDPS];
+           return ix86_get_builtin (IX86_BUILTIN_VFMADDPS);
          if (out_n == 8 && in_n == 8)
-           return ix86_builtins[IX86_BUILTIN_VFMADDPS256];
+           return ix86_get_builtin (IX86_BUILTIN_VFMADDPS256);
        }
       break;
 
@@ -34397,7 +34422,7 @@ ix86_vectorize_builtin_gather (const_tree mem_vectype,
       return NULL_TREE;
     }
 
-  return ix86_builtins[code];
+  return ix86_get_builtin (code);
 }
 
 /* Returns a code for a target-specific builtin that implements
@@ -34418,10 +34443,10 @@ ix86_builtin_reciprocal (unsigned int fn, bool md_fn,
       {
        /* Vectorized version of sqrt to rsqrt conversion.  */
       case IX86_BUILTIN_SQRTPS_NR:
-       return ix86_builtins[IX86_BUILTIN_RSQRTPS_NR];
+       return ix86_get_builtin (IX86_BUILTIN_RSQRTPS_NR);
 
       case IX86_BUILTIN_SQRTPS_NR256:
-       return ix86_builtins[IX86_BUILTIN_RSQRTPS_NR256];
+       return ix86_get_builtin (IX86_BUILTIN_RSQRTPS_NR256);
 
       default:
        return NULL_TREE;
@@ -34432,7 +34457,7 @@ ix86_builtin_reciprocal (unsigned int fn, bool md_fn,
       {
        /* Sqrt to rsqrt conversion.  */
       case BUILT_IN_SQRTF:
-       return ix86_builtins[IX86_BUILTIN_RSQRTF];
+       return ix86_get_builtin (IX86_BUILTIN_RSQRTF);
 
       default:
        return NULL_TREE;

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index e57eaf0..248928e 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,10 @@
+2013-12-11  Sriraman Tallam  <tmsriram@google.com>
+
+       PR target/59390
+       * gcc.target/i386/pr59390.c: New test.
+       * gcc.target/i386/pr59390_1.c: New test.
+       * gcc.target/i386/pr59390_2.c: New test.
+
 2013-12-11  Balaji V. Iyer  <balaji.v.iyer@intel.com>
 
        * g++.dg/cilk-plus/CK/catch_exc.cc: New test case.

diff --git a/gcc/testsuite/gcc.target/i386/pr59390.c b/gcc/testsuite/gcc.target/i386/pr59390.c
new file mode 100644 (file)
index 0000000..49ce02d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr59390.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-std=c99 -O3" } */
+
+#include "math.h"
+void fun() __attribute__((target("fma")));
+
+void 
+other_fun(double *restrict out, double * restrict a, double * restrict b, double * restrict c, int n)
+{
+    int i;
+    for (i = 0; i < n; i++) {
+        out[i] = fma(a[i], b[i], c[i]);
+    }   
+}
+
+/* { dg-final { scan-assembler-not "vfmadd" } } */

diff --git a/gcc/testsuite/gcc.target/i386/pr59390_1.c b/gcc/testsuite/gcc.target/i386/pr59390_1.c
new file mode 100644 (file)
index 0000000..2bd32a4
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr59390_1.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+/* { dg-options "-std=c99 -O3" } */
+
+#include "math.h"
+void fun() __attribute__((target("fma")));
+
+__attribute__((target("fma")))
+void 
+other_fun(double *restrict out, double * restrict a, double * restrict b, double * restrict c, int n)
+{
+    int i;
+    for (i = 0; i < n; i++) {
+        out[i] = fma(a[i], b[i], c[i]);
+    }   
+}
+
+/* { dg-final { scan-assembler "vfmadd" } } */

diff --git a/gcc/testsuite/gcc.target/i386/pr59390_2.c b/gcc/testsuite/gcc.target/i386/pr59390_2.c
new file mode 100644 (file)
index 0000000..55a181a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr59390_2.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-options "-std=c99 -O3 -mfma" } */
+
+#include "math.h"
+void fun() __attribute__((target("fma")));
+
+void 
+other_fun(double *restrict out, double * restrict a, double * restrict b, double * restrict c, int n)
+{
+    int i;
+    for (i = 0; i < n; i++) {
+        out[i] = fma(a[i], b[i], c[i]);
+    }   
+}
+
+/* { dg-final { scan-assembler "vfmadd" } } */