summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-reduce-fmax.ll')
-rw-r--r--llvm/test/CodeGen/X86/vector-reduce-fmax.ll66
1 files changed, 28 insertions, 38 deletions
diff --git a/llvm/test/CodeGen/X86/vector-reduce-fmax.ll b/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
index c2a06cfb7fe..d3c1ca256e8 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-fmax.ll
@@ -15,25 +15,25 @@ define float @test_v2f32(<2 x float> %a0) {
; SSE2: # %bb.0:
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
-; SSE2-NEXT: maxps %xmm1, %xmm0
+; SSE2-NEXT: maxss %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v2f32:
; SSE41: # %bb.0:
; SSE41-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSE41-NEXT: maxps %xmm1, %xmm0
+; SSE41-NEXT: maxss %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v2f32:
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f32:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call float @llvm.experimental.vector.reduce.fmax.f32.v2f32(<2 x float> %a0)
ret float %1
@@ -47,7 +47,7 @@ define float @test_v4f32(<4 x float> %a0) {
; SSE2-NEXT: maxps %xmm1, %xmm0
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
-; SSE2-NEXT: maxps %xmm1, %xmm0
+; SSE2-NEXT: maxss %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v4f32:
@@ -56,7 +56,7 @@ define float @test_v4f32(<4 x float> %a0) {
; SSE41-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE41-NEXT: maxps %xmm1, %xmm0
; SSE41-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSE41-NEXT: maxps %xmm1, %xmm0
+; SSE41-NEXT: maxss %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v4f32:
@@ -64,7 +64,7 @@ define float @test_v4f32(<4 x float> %a0) {
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4f32:
@@ -72,7 +72,7 @@ define float @test_v4f32(<4 x float> %a0) {
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call float @llvm.experimental.vector.reduce.fmax.f32.v4f32(<4 x float> %a0)
ret float %1
@@ -87,7 +87,7 @@ define float @test_v8f32(<8 x float> %a0) {
; SSE2-NEXT: maxps %xmm1, %xmm0
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
-; SSE2-NEXT: maxps %xmm1, %xmm0
+; SSE2-NEXT: maxss %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v8f32:
@@ -97,7 +97,7 @@ define float @test_v8f32(<8 x float> %a0) {
; SSE41-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE41-NEXT: maxps %xmm1, %xmm0
; SSE41-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSE41-NEXT: maxps %xmm1, %xmm0
+; SSE41-NEXT: maxss %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v8f32:
@@ -107,8 +107,7 @@ define float @test_v8f32(<8 x float> %a0) {
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
-; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -119,8 +118,7 @@ define float @test_v8f32(<8 x float> %a0) {
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX512-NEXT: vmaxps %ymm1, %ymm0, %ymm0
-; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call float @llvm.experimental.vector.reduce.fmax.f32.v8f32(<8 x float> %a0)
@@ -138,7 +136,7 @@ define float @test_v16f32(<16 x float> %a0) {
; SSE2-NEXT: maxps %xmm1, %xmm0
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[2,3]
-; SSE2-NEXT: maxps %xmm1, %xmm0
+; SSE2-NEXT: maxss %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v16f32:
@@ -150,7 +148,7 @@ define float @test_v16f32(<16 x float> %a0) {
; SSE41-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE41-NEXT: maxps %xmm1, %xmm0
; SSE41-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSE41-NEXT: maxps %xmm1, %xmm0
+; SSE41-NEXT: maxss %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v16f32:
@@ -161,8 +159,7 @@ define float @test_v16f32(<16 x float> %a0) {
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
-; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -175,8 +172,7 @@ define float @test_v16f32(<16 x float> %a0) {
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX512-NEXT: vmaxps %zmm1, %zmm0, %zmm0
-; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call float @llvm.experimental.vector.reduce.fmax.f32.v16f32(<16 x float> %a0)
@@ -192,19 +188,19 @@ define double @test_v2f64(<2 x double> %a0) {
; SSE: # %bb.0:
; SSE-NEXT: movapd %xmm0, %xmm1
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT: maxpd %xmm1, %xmm0
+; SSE-NEXT: maxsd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.experimental.vector.reduce.fmax.f64.v2f64(<2 x double> %a0)
ret double %1
@@ -216,7 +212,7 @@ define double @test_v4f64(<4 x double> %a0) {
; SSE-NEXT: maxpd %xmm1, %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT: maxpd %xmm1, %xmm0
+; SSE-NEXT: maxsd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_v4f64:
@@ -224,8 +220,7 @@ define double @test_v4f64(<4 x double> %a0) {
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
-; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -234,8 +229,7 @@ define double @test_v4f64(<4 x double> %a0) {
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX512-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
-; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call double @llvm.experimental.vector.reduce.fmax.f64.v4f64(<4 x double> %a0)
@@ -250,7 +244,7 @@ define double @test_v8f64(<8 x double> %a0) {
; SSE-NEXT: maxpd %xmm1, %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT: maxpd %xmm1, %xmm0
+; SSE-NEXT: maxsd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_v8f64:
@@ -259,8 +253,7 @@ define double @test_v8f64(<8 x double> %a0) {
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
-; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -271,8 +264,7 @@ define double @test_v8f64(<8 x double> %a0) {
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
-; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call double @llvm.experimental.vector.reduce.fmax.f64.v8f64(<8 x double> %a0)
@@ -291,7 +283,7 @@ define double @test_v16f64(<16 x double> %a0) {
; SSE-NEXT: maxpd %xmm1, %xmm0
; SSE-NEXT: movapd %xmm0, %xmm1
; SSE-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT: maxpd %xmm1, %xmm0
+; SSE-NEXT: maxsd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: test_v16f64:
@@ -302,8 +294,7 @@ define double @test_v16f64(<16 x double> %a0) {
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
-; AVX-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
@@ -315,8 +306,7 @@ define double @test_v16f64(<16 x double> %a0) {
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
-; AVX512-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call double @llvm.experimental.vector.reduce.fmax.f64.v16f64(<16 x double> %a0)
OpenPOWER on IntegriCloud