re PR target/32961 (Gcc has different requirements for x86 shift xmm intrinsics)

	PR target/32961
	* config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLLWI128,
	IX86_BUILTIN_PSLLDI128, BUILTIN_PSLLQI128, IX86_BUILTIN_PSRAWI128,
	IX86_BUILTIN_PSRADI128, IX86_BUILTIN_PSRLWI128,
	IX86_BUILTIN_PSRLDI128, IX86_BUILTIN_PSRLQI128]: Do not require
	immediate shift value.
	* config/i386/emmintrin.h (_mm_slli_epi16, _mm_slli_epi32,
	_mm_slli_epi64, _mm_srai_epi16, _mm_srai_epi32, _mm_srli_epi16,
	_mm_srli_epi32, _mm_srli_epi64):  Enable disabled functions and
	remove equivalent macro definitions.

testsuite/ChangeLog:

	PR target/32961
	* gcc.target/i386/pr32961.c: New testcase.

From-SVN: r129433

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f766783a44aef7f35e0b532a21fae9ecdb67fbf2..7acfc843ec3ec4ca64ca8229010d23dd6a9463b9 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,16 @@
+2007-10-18  Uros Bizjak  <ubizjak@gmail.com>
+
+       PR target/32961
+       * config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLLWI128,
+       IX86_BUILTIN_PSLLDI128, BUILTIN_PSLLQI128, IX86_BUILTIN_PSRAWI128,
+       IX86_BUILTIN_PSRADI128, IX86_BUILTIN_PSRLWI128,
+       IX86_BUILTIN_PSRLDI128, IX86_BUILTIN_PSRLQI128]: Do not require
+       immediate shift value.
+       * config/i386/emmintrin.h (_mm_slli_epi16, _mm_slli_epi32,
+       _mm_slli_epi64, _mm_srai_epi16, _mm_srai_epi32, _mm_srli_epi16,
+       _mm_srli_epi32, _mm_srli_epi64):  Enable disabled functions and
+       remove equivalent macro definitions.
+
 2007-10-18  Chen liqin  <liqin@sunnorth.com.cn>
 
         * config.gcc : update score-*-elf(extra_objs).

diff --git a/gcc/config/i386/emmintrin.h b/gcc/config/i386/emmintrin.h
index d30184bc8057258f99f63e026dce7831fbbd97a3..0d82f28d6c399737056e154d5e5c52b958ece49b 100644 (file)
--- a/gcc/config/i386/emmintrin.h
+++ b/gcc/config/i386/emmintrin.h
@@ -1100,7 +1100,6 @@ _mm_mul_epu32 (__m128i __A, __m128i __B)
   return (__m128i)__builtin_ia32_pmuludq128 ((__v4si)__A, (__v4si)__B);
 }
 
-#if 0
 static __inline __m128i __attribute__((__always_inline__))
 _mm_slli_epi16 (__m128i __A, int __B)
 {
@@ -1118,16 +1117,7 @@ _mm_slli_epi64 (__m128i __A, int __B)
 {
   return (__m128i)__builtin_ia32_psllqi128 ((__v2di)__A, __B);
 }
-#else
-#define _mm_slli_epi16(__A, __B) \
-  ((__m128i)__builtin_ia32_psllwi128 ((__v8hi)(__A), __B))
-#define _mm_slli_epi32(__A, __B) \
-  ((__m128i)__builtin_ia32_pslldi128 ((__v8hi)(__A), __B))
-#define _mm_slli_epi64(__A, __B) \
-  ((__m128i)__builtin_ia32_psllqi128 ((__v8hi)(__A), __B))
-#endif
 
-#if 0
 static __inline __m128i __attribute__((__always_inline__))
 _mm_srai_epi16 (__m128i __A, int __B)
 {
@@ -1139,12 +1129,6 @@ _mm_srai_epi32 (__m128i __A, int __B)
 {
   return (__m128i)__builtin_ia32_psradi128 ((__v4si)__A, __B);
 }
-#else
-#define _mm_srai_epi16(__A, __B) \
-  ((__m128i)__builtin_ia32_psrawi128 ((__v8hi)(__A), __B))
-#define _mm_srai_epi32(__A, __B) \
-  ((__m128i)__builtin_ia32_psradi128 ((__v8hi)(__A), __B))
-#endif
 
 #if 0
 static __m128i __attribute__((__always_inline__))
@@ -1165,7 +1149,6 @@ _mm_srli_si128 (__m128i __A, int __B)
   ((__m128i)__builtin_ia32_pslldqi128 (__A, (__B) * 8))
 #endif
 
-#if 0
 static __inline __m128i __attribute__((__always_inline__))
 _mm_srli_epi16 (__m128i __A, int __B)
 {
@@ -1183,14 +1166,6 @@ _mm_srli_epi64 (__m128i __A, int __B)
 {
   return (__m128i)__builtin_ia32_psrlqi128 ((__v2di)__A, __B);
 }
-#else
-#define _mm_srli_epi16(__A, __B) \
-  ((__m128i)__builtin_ia32_psrlwi128 ((__v8hi)(__A), __B))
-#define _mm_srli_epi32(__A, __B) \
-  ((__m128i)__builtin_ia32_psrldi128 ((__v4si)(__A), __B))
-#define _mm_srli_epi64(__A, __B) \
-  ((__m128i)__builtin_ia32_psrlqi128 ((__v4si)(__A), __B))
-#endif
 
 static __inline __m128i __attribute__((__always_inline__))
 _mm_sll_epi16 (__m128i __A, __m128i __B)

diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c
index 6da9fc0185074a64ca7298500c8da5bb9d2ff2bc..ddc02b5c673a754279a51a9e827f6d41444cdafd 100644 (file)
--- a/gcc/config/i386/i386.c
+++ b/gcc/config/i386/i386.c
@@ -16209,80 +16209,38 @@ ix86_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED,
       emit_insn (pat);
       return target;
 
-    case IX86_BUILTIN_PSLLWI128:
-      icode = CODE_FOR_ashlv8hi3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSLLDI128:
-      icode = CODE_FOR_ashlv4si3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSLLQI128:
-      icode = CODE_FOR_ashlv2di3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSRAWI128:
-      icode = CODE_FOR_ashrv8hi3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSRADI128:
-      icode = CODE_FOR_ashrv4si3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSRLWI128:
-      icode = CODE_FOR_lshrv8hi3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSRLDI128:
-      icode = CODE_FOR_lshrv4si3;
-      goto do_pshifti;
-    case IX86_BUILTIN_PSRLQI128:
-      icode = CODE_FOR_lshrv2di3;
-      goto do_pshifti;
-    do_pshifti:
-      arg0 = TREE_VALUE (arglist);
-      arg1 = TREE_VALUE (TREE_CHAIN (arglist));
-      op0 = expand_expr (arg0, NULL_RTX, VOIDmode, 0);
-      op1 = expand_expr (arg1, NULL_RTX, VOIDmode, 0);
-
-      if (GET_CODE (op1) != CONST_INT)
-       {
-         error ("shift must be an immediate");
-         return const0_rtx;
-       }
-      if (INTVAL (op1) < 0 || INTVAL (op1) > 255)
-       op1 = GEN_INT (255);
-
-      tmode = insn_data[icode].operand[0].mode;
-      mode1 = insn_data[icode].operand[1].mode;
-      if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
-       op0 = copy_to_reg (op0);
-
-      target = gen_reg_rtx (tmode);
-      pat = GEN_FCN (icode) (target, op0, op1);
-      if (!pat)
-       return 0;
-      emit_insn (pat);
-      return target;
-
     case IX86_BUILTIN_PSLLW128:
+    case IX86_BUILTIN_PSLLWI128:
       icode = CODE_FOR_ashlv8hi3;
       goto do_pshift;
     case IX86_BUILTIN_PSLLD128:
+    case IX86_BUILTIN_PSLLDI128:
       icode = CODE_FOR_ashlv4si3;
       goto do_pshift;
     case IX86_BUILTIN_PSLLQ128:
+    case IX86_BUILTIN_PSLLQI128:
       icode = CODE_FOR_ashlv2di3;
       goto do_pshift;
     case IX86_BUILTIN_PSRAW128:
+    case IX86_BUILTIN_PSRAWI128:
       icode = CODE_FOR_ashrv8hi3;
       goto do_pshift;
     case IX86_BUILTIN_PSRAD128:
+    case IX86_BUILTIN_PSRADI128:
       icode = CODE_FOR_ashrv4si3;
       goto do_pshift;
     case IX86_BUILTIN_PSRLW128:
+    case IX86_BUILTIN_PSRLWI128:
       icode = CODE_FOR_lshrv8hi3;
       goto do_pshift;
     case IX86_BUILTIN_PSRLD128:
+    case IX86_BUILTIN_PSRLDI128:
       icode = CODE_FOR_lshrv4si3;
       goto do_pshift;
     case IX86_BUILTIN_PSRLQ128:
+    case IX86_BUILTIN_PSRLQI128:
       icode = CODE_FOR_lshrv2di3;
-      goto do_pshift;
+
     do_pshift:
       arg0 = TREE_VALUE (arglist);
       arg1 = TREE_VALUE (TREE_CHAIN (arglist));
@@ -16295,7 +16253,9 @@ ix86_expand_builtin (tree exp, rtx target, rtx subtarget ATTRIBUTE_UNUSED,
       if (! (*insn_data[icode].operand[1].predicate) (op0, mode1))
        op0 = copy_to_reg (op0);
 
-      op1 = simplify_gen_subreg (SImode, op1, GET_MODE (op1), 0);
+      if (!CONST_INT_P (op1))
+       op1 = simplify_gen_subreg (SImode, op1, GET_MODE (op1), 0);
+
       if (! (*insn_data[icode].operand[2].predicate) (op1, SImode))
        op1 = copy_to_reg (op1);
 

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 98fc71868aa8245d9de81ca1785614401352fee7..5665b19a38676f87fd703bec2df78a4c098116eb 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2007-10-18  Uros Bizjak  <ubizjak@gmail.com>
+
+       PR target/32961
+       * gcc.target/i386/pr32961.c: New testcase.
+
 2007-10-18  Uros Bizjak  <ubizjak@gmail.com>
 
        * gcc.target/i386/pr31167.c: Add i?86-*-* to target selector.

diff --git a/gcc/testsuite/gcc.target/i386/pr32961.c b/gcc/testsuite/gcc.target/i386/pr32961.c
new file mode 100644 (file)
index 0000000..a232628
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr32961.c
@@ -0,0 +1,10 @@
+/* { dg-do compile } */
+/* { dg-options "-O0 -msse2" } */
+
+#include <xmmintrin.h>
+
+void x (int n)
+{
+  __m128i a;
+  a = _mm_slli_epi32 (a, n);
+}