Support AVX for cmpss/cmpsd.

gcc/

2010-05-12  H.J. Lu  <hongjiu.lu@intel.com>

	PR target/44088
	* config/i386/sse.md (*avx_vmmaskcmp<mode>3): New.

gcc/testsuite/

2010-05-12  H.J. Lu  <hongjiu.lu@intel.com>

	PR target/44088
	* gcc.target/i386/avx-cmpsd-1.c: New.
	* gcc.target/i386/avx-cmpsd-2.c: Likewise.
	* gcc.target/i386/avx-cmpss-1.c: Likewise.
	* gcc.target/i386/avx-cmpss-2.c: Likewise.
	* gcc.target/i386/sse-cmpss-1.c: Likewise.
	* gcc.target/i386/sse2-cmpsd-1.c: Likewise.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@159319 138bc75d-0d04-0410-961f-82ee72b054a4

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 52f8319..09043fd 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,8 @@
+2010-05-12  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR target/44088
+       * config/i386/sse.md (*avx_vmmaskcmp<mode>3): New.
+
 2010-05-12  Jakub Jelinek  <jakub@redhat.com>
 
        PR middle-end/44085

diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index 0baefd8..efe55e8 100644 (file)
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -1483,6 +1483,20 @@
    (set_attr "length_immediate" "1")
    (set_attr "mode" "<MODE>")])
 
+(define_insn "*avx_vmmaskcmp<mode>3"
+  [(set (match_operand:SSEMODEF2P 0 "register_operand" "=x")
+       (vec_merge:SSEMODEF2P
+        (match_operator:SSEMODEF2P 3 "sse_comparison_operator"
+               [(match_operand:SSEMODEF2P 1 "register_operand" "x")
+                (match_operand:SSEMODEF2P 2 "nonimmediate_operand" "xm")])
+        (match_dup 1)
+        (const_int 1)))]
+  "AVX_VEC_FLOAT_MODE_P (<MODE>mode)"
+  "vcmp%D3s<ssemodesuffixf2c>\t{%2, %1, %0|%0, %1, %2}"
+  [(set_attr "type" "ssecmp")
+   (set_attr "prefix" "vex")
+   (set_attr "mode" "<ssescalarmode>")])
+
 (define_insn "<sse>_vmmaskcmp<mode>3"
   [(set (match_operand:SSEMODEF2P 0 "register_operand" "=x")
        (vec_merge:SSEMODEF2P

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index d2628e4..e5f3a50 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,13 @@
+2010-05-12  H.J. Lu  <hongjiu.lu@intel.com>
+
+       PR target/44088
+       * gcc.target/i386/avx-cmpsd-1.c: New.
+       * gcc.target/i386/avx-cmpsd-2.c: Likewise.
+       * gcc.target/i386/avx-cmpss-1.c: Likewise.
+       * gcc.target/i386/avx-cmpss-2.c: Likewise.
+       * gcc.target/i386/sse-cmpss-1.c: Likewise.
+       * gcc.target/i386/sse2-cmpsd-1.c: Likewise.
+
 2010-05-12  Jakub Jelinek  <jakub@redhat.com>
 
        PR middle-end/44085

diff --git a/gcc/testsuite/gcc.target/i386/avx-cmpsd-1.c b/gcc/testsuite/gcc.target/i386/avx-cmpsd-1.c
new file mode 100644 (file)
index 0000000..ad59f26
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx-cmpsd-1.c
@@ -0,0 +1,8 @@
+/* { dg-do run } */
+/* { dg-require-effective-target avx } */
+/* { dg-options "-O2 -mavx" } */
+
+#define CHECK_H "avx-check.h"
+#define TEST avx_test
+
+#include "sse2-cmpsd-1.c"

diff --git a/gcc/testsuite/gcc.target/i386/avx-cmpsd-2.c b/gcc/testsuite/gcc.target/i386/avx-cmpsd-2.c
new file mode 100644 (file)
index 0000000..3162912
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx-cmpsd-2.c
@@ -0,0 +1,13 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx" } */
+
+#include <emmintrin.h>
+
+__m128d
+foo (__m128d x, __m128d y)
+{
+  return _mm_cmpeq_sd (x, y);
+}
+
+
+/* { dg-final { scan-assembler "vcmpeqsd" } } */

diff --git a/gcc/testsuite/gcc.target/i386/avx-cmpss-1.c b/gcc/testsuite/gcc.target/i386/avx-cmpss-1.c
new file mode 100644 (file)
index 0000000..c0cdd68
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx-cmpss-1.c
@@ -0,0 +1,8 @@
+/* { dg-do run } */
+/* { dg-require-effective-target avx } */
+/* { dg-options "-O2 -mavx" } */
+
+#define CHECK_H "avx-check.h"
+#define TEST avx_test
+
+#include "sse-cmpss-1.c"

diff --git a/gcc/testsuite/gcc.target/i386/avx-cmpss-2.c b/gcc/testsuite/gcc.target/i386/avx-cmpss-2.c
new file mode 100644 (file)
index 0000000..0fcc620
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx-cmpss-2.c
@@ -0,0 +1,13 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx" } */
+
+#include <xmmintrin.h>
+
+__m128
+foo (__m128 x, __m128 y)
+{
+  return _mm_cmpeq_ss (x, y);
+}
+
+
+/* { dg-final { scan-assembler "vcmpeqss" } } */

diff --git a/gcc/testsuite/gcc.target/i386/sse-cmpss-1.c b/gcc/testsuite/gcc.target/i386/sse-cmpss-1.c
new file mode 100644 (file)
index 0000000..e4be731
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/sse-cmpss-1.c
@@ -0,0 +1,59 @@
+/* { dg-do run } */
+/* { dg-options "-O2 -msse -std=c99" } */
+
+#ifndef CHECK_H
+#define CHECK_H "sse-check.h"
+#endif
+
+#ifndef TEST
+#define TEST sse_test
+#endif
+
+#include CHECK_H
+
+#include <math.h>
+
+float s1[]={2134.3343, 6678.346, 453.345635, 54646.464356};
+float s2[]={41124.234, 6678.346, 8653.65635, 856.43576};
+int dd[] = {1, 2, 3, 4};
+float d[4];
+union{int i[4]; float f[4];} e;
+
+void check(char *id)
+{
+    if(checkVi((int*)d, e.i, 4)){
+       printf("mm_cmp%s_ss FAILED\n", id);
+    }
+}
+
+static void
+TEST ()
+{
+    __m128 source1, source2, dest;
+    int i;
+
+#define CMP(cmp, rel)                                  \
+    e.i[0] = rel ? -1 : 0;                             \
+    dest = _mm_loadu_ps((float*)dd);                   \
+    source1 = _mm_loadu_ps(s1);                                \
+    source2 = _mm_loadu_ps(s2);                                \
+    dest = _mm_cmp##cmp##_ss(source1, source2);                \
+    _mm_storeu_ps(d, dest);                            \
+    check("" #cmp "");
+
+    for(i = 1; i < 4; i++) e.f[i] = s1[i];
+    
+    CMP(eq, !isunordered(s1[0], s2[0]) && s1[0] == s2[0]);
+    CMP(lt, !isunordered(s1[0], s2[0]) && s1[0] < s2[0]);
+    CMP(le, !isunordered(s1[0], s2[0]) && s1[0] <= s2[0]);
+    CMP(unord, isunordered(s1[0], s2[0]));
+    CMP(neq, isunordered(s1[0], s2[0]) || s1[0] != s2[0]);
+    CMP(nlt, isunordered(s1[0], s2[0]) || s1[0] >= s2[0]);
+    CMP(nle, isunordered(s1[0], s2[0]) || s1[0] > s2[0]);
+    CMP(ord, !isunordered(s1[0], s2[0]));
+
+    CMP(ge, isunordered(s1[0], s2[0]) || s1[0] >= s2[0]);
+    CMP(gt, isunordered(s1[0], s2[0]) || s1[0] > s2[0]);
+    CMP(nge, !isunordered(s1[0], s2[0]) && s1[0] < s2[0]);
+    CMP(ngt, !isunordered(s1[0], s2[0]) && s1[0] <= s2[0]);
+}

diff --git a/gcc/testsuite/gcc.target/i386/sse2-cmpsd-1.c b/gcc/testsuite/gcc.target/i386/sse2-cmpsd-1.c
new file mode 100644 (file)
index 0000000..224a609
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/sse2-cmpsd-1.c
@@ -0,0 +1,57 @@
+/* { dg-do run } */
+/* { dg-options "-O2 -msse2 -std=c99" } */
+
+#ifndef CHECK_H
+#define CHECK_H "sse2-check.h"
+#endif
+
+#ifndef TEST
+#define TEST sse2_test
+#endif
+
+#include CHECK_H
+
+#include <math.h>
+
+double s1[] = {2134.3343, 6678.346};
+double s2[] = {41124.234, 6678.346};
+long long dd[] =  {1, 2}, d[2];
+union{long long l[2]; double d[2];} e;
+
+void check(char *id)
+{
+    if(checkVl(d, e.l, 2)){
+       printf("mm_cmp%s_sd FAILED\n", id);
+    }
+}
+
+#define CMP(cmp, rel)                                  \
+    e.l[0] = rel ? -1 : 0;                             \
+    dest = _mm_loadu_pd((double*)dd);                  \
+    source1 = _mm_loadu_pd(s1);                                \
+    source2 = _mm_loadu_pd(s2);                                \
+    dest = _mm_cmp##cmp##_sd(source1, source2);                \
+    _mm_storeu_pd((double*) d, dest);                  \
+    check("" #cmp "");
+
+static void
+TEST ()
+{
+    __m128d source1, source2, dest;
+
+    e.d[1] = s1[1];
+    
+    CMP(eq, !isunordered(s1[0], s2[0]) && s1[0] == s2[0]);
+    CMP(lt, !isunordered(s1[0], s2[0]) && s1[0] < s2[0]);
+    CMP(le, !isunordered(s1[0], s2[0]) && s1[0] <= s2[0]);
+    CMP(unord, isunordered(s1[0], s2[0]));
+    CMP(neq, isunordered(s1[0], s2[0]) || s1[0] != s2[0]);
+    CMP(nlt, isunordered(s1[0], s2[0]) || s1[0] >= s2[0]);
+    CMP(nle, isunordered(s1[0], s2[0]) || s1[0] > s2[0]);
+    CMP(ord, !isunordered(s1[0], s2[0]));
+
+    CMP(ge, isunordered(s1[0], s2[0]) || s1[0] >= s2[0]);
+    CMP(gt, isunordered(s1[0], s2[0]) || s1[0] > s2[0]);
+    CMP(nge, !isunordered(s1[0], s2[0]) && s1[0] < s2[0]);
+    CMP(ngt, !isunordered(s1[0], s2[0]) && s1[0] <= s2[0]);
+}