summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vector-reduce-mul.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-reduce-mul.ll')
-rw-r--r--llvm/test/CodeGen/X86/vector-reduce-mul.ll115
1 files changed, 41 insertions, 74 deletions
diff --git a/llvm/test/CodeGen/X86/vector-reduce-mul.ll b/llvm/test/CodeGen/X86/vector-reduce-mul.ll
index 5f1455daab9..b474cc0b75f 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-mul.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-mul.ll
@@ -807,12 +807,8 @@ define i32 @test_v4i32(<4 x i32> %a0) {
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
+; SSE2-NEXT: pmuludq %xmm3, %xmm1
; SSE2-NEXT: movd %xmm1, %eax
; SSE2-NEXT: retq
;
@@ -849,23 +845,16 @@ define i32 @test_v4i32(<4 x i32> %a0) {
define i32 @test_v8i32(<8 x i32> %a0) {
; SSE2-LABEL: test_v8i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm2, %xmm3
; SSE2-NEXT: pmuludq %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
-; SSE2-NEXT: pmuludq %xmm0, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,2,0,0]
-; SSE2-NEXT: pmuludq %xmm1, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[2,2,0,0]
+; SSE2-NEXT: pmuludq %xmm3, %xmm0
+; SSE2-NEXT: pmuludq %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v8i32:
@@ -920,29 +909,22 @@ define i32 @test_v8i32(<8 x i32> %a0) {
define i32 @test_v16i32(<16 x i32> %a0) {
; SSE2-LABEL: test_v16i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm4, %xmm5
-; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm0[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm4, %xmm6
-; SSE2-NEXT: pmuludq %xmm5, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm3, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm0
; SSE2-NEXT: pmuludq %xmm1, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm4, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm5, %xmm2
+; SSE2-NEXT: pmuludq %xmm1, %xmm2
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm6[2,2,0,0]
-; SSE2-NEXT: pmuludq %xmm6, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,2,0,0]
+; SSE2-NEXT: pmuludq %xmm2, %xmm0
+; SSE2-NEXT: pmuludq %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v16i32:
@@ -1005,39 +987,32 @@ define i32 @test_v16i32(<16 x i32> %a0) {
define i32 @test_v32i32(<32 x i32> %a0) {
; SSE2-LABEL: test_v32i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm2[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm6[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm2[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm8, %xmm9
+; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm4[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm0[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm8, %xmm10
+; SSE2-NEXT: pmuludq %xmm9, %xmm10
+; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm7[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm3[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm8, %xmm9
+; SSE2-NEXT: pshufd {{.*#+}} xmm8 = xmm5[1,1,3,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm1[1,1,3,3]
+; SSE2-NEXT: pmuludq %xmm8, %xmm11
+; SSE2-NEXT: pmuludq %xmm9, %xmm11
+; SSE2-NEXT: pmuludq %xmm10, %xmm11
; SSE2-NEXT: pmuludq %xmm6, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm4, %xmm0
; SSE2-NEXT: pmuludq %xmm2, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm7, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm1[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm5, %xmm1
; SSE2-NEXT: pmuludq %xmm3, %xmm1
; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm6[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm8, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm9, %xmm3
-; SSE2-NEXT: pmuludq %xmm0, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm7[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm2, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
-; SSE2-NEXT: pmuludq %xmm10, %xmm2
-; SSE2-NEXT: pmuludq %xmm0, %xmm2
-; SSE2-NEXT: pmuludq %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,2,0,0]
-; SSE2-NEXT: pmuludq %xmm2, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE2-NEXT: pmuludq %xmm1, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm11[2,2,0,0]
+; SSE2-NEXT: pmuludq %xmm11, %xmm1
; SSE2-NEXT: pmuludq %xmm0, %xmm1
; SSE2-NEXT: movd %xmm1, %eax
; SSE2-NEXT: retq
@@ -1151,12 +1126,8 @@ define i16 @test_v4i16(<4 x i16> %a0) {
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
+; SSE2-NEXT: pmuludq %xmm3, %xmm1
; SSE2-NEXT: pextrw $0, %xmm1, %eax
; SSE2-NEXT: # kill: def $ax killed $ax killed $eax
; SSE2-NEXT: retq
@@ -1603,12 +1574,8 @@ define i8 @test_v4i8(<4 x i8> %a0) {
; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,1,1]
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; SSE2-NEXT: pmuludq %xmm2, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; SSE2-NEXT: pmuludq %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[0,2,2,3]
-; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
; SSE2-NEXT: pmuludq %xmm0, %xmm1
+; SSE2-NEXT: pmuludq %xmm3, %xmm1
; SSE2-NEXT: movd %xmm1, %eax
; SSE2-NEXT: # kill: def $al killed $al killed $eax
; SSE2-NEXT: retq
OpenPOWER on IntegriCloud