summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll')
-rw-r--r--llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll245
1 files changed, 60 insertions, 185 deletions
diff --git a/llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll b/llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll
index 2817d9e6229..1d1f10f4d13 100644
--- a/llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll
+++ b/llvm/test/CodeGen/X86/srem-seteq-vec-splat.ll
@@ -9,105 +9,55 @@
define <4 x i32> @test_srem_odd_25(<4 x i32> %X) nounwind {
; CHECK-SSE2-LABEL: test_srem_odd_25:
; CHECK-SSE2: # %bb.0:
-; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-SSE2-NEXT: movdqa %xmm0, %xmm2
-; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm3
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,3,2,3]
-; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; CHECK-SSE2-NEXT: pxor %xmm3, %xmm3
-; CHECK-SSE2-NEXT: pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT: pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT: pand %xmm1, %xmm4
-; CHECK-SSE2-NEXT: psubd %xmm4, %xmm2
-; CHECK-SSE2-NEXT: movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT: psrld $31, %xmm1
-; CHECK-SSE2-NEXT: psrad $3, %xmm2
-; CHECK-SSE2-NEXT: paddd %xmm1, %xmm2
-; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [25,25,25,25]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [3264175145,3264175145,3264175145,3264175145]
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm0
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm4
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
-; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT: psubd %xmm2, %xmm0
-; CHECK-SSE2-NEXT: pcmpeqd %xmm3, %xmm0
-; CHECK-SSE2-NEXT: psrld $31, %xmm0
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT: paddd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pandn {{.*}}(%rip), %xmm0
; CHECK-SSE2-NEXT: retq
;
; CHECK-SSE41-LABEL: test_srem_odd_25:
; CHECK-SSE41: # %bb.0:
-; CHECK-SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT: movdqa {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-SSE41-NEXT: pmuldq %xmm2, %xmm1
-; CHECK-SSE41-NEXT: pmuldq %xmm0, %xmm2
-; CHECK-SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-SSE41-NEXT: movdqa %xmm2, %xmm1
-; CHECK-SSE41-NEXT: psrld $31, %xmm1
-; CHECK-SSE41-NEXT: psrad $3, %xmm2
-; CHECK-SSE41-NEXT: paddd %xmm1, %xmm2
-; CHECK-SSE41-NEXT: pmulld {{.*}}(%rip), %xmm2
-; CHECK-SSE41-NEXT: psubd %xmm2, %xmm0
-; CHECK-SSE41-NEXT: pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT: pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT: paddd {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT: movdqa {{.*#+}} xmm1 = [171798690,171798690,171798690,171798690]
+; CHECK-SSE41-NEXT: pminud %xmm0, %xmm1
; CHECK-SSE41-NEXT: pcmpeqd %xmm1, %xmm0
; CHECK-SSE41-NEXT: psrld $31, %xmm0
; CHECK-SSE41-NEXT: retq
;
; CHECK-AVX1-LABEL: test_srem_odd_25:
; CHECK-AVX1: # %bb.0:
-; CHECK-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-AVX1-NEXT: vpmuldq %xmm2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpmuldq %xmm2, %xmm0, %xmm2
-; CHECK-AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
-; CHECK-AVX1-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX1-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX1-NEXT: vpminud {{.*}}(%rip), %xmm0, %xmm1
; CHECK-AVX1-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX1-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX1-NEXT: retq
;
; CHECK-AVX2-LABEL: test_srem_odd_25:
; CHECK-AVX2: # %bb.0:
-; CHECK-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-AVX2-NEXT: vpmuldq %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpmuldq %xmm2, %xmm0, %xmm2
-; CHECK-AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX2-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX2-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [25,25,25,25]
-; CHECK-AVX2-NEXT: vpmulld %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [3264175145,3264175145,3264175145,3264175145]
+; CHECK-AVX2-NEXT: vpmulld %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [85899345,85899345,85899345,85899345]
+; CHECK-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [171798690,171798690,171798690,171798690]
+; CHECK-AVX2-NEXT: vpminud %xmm1, %xmm0, %xmm1
; CHECK-AVX2-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX2-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX2-NEXT: retq
;
; CHECK-AVX512VL-LABEL: test_srem_odd_25:
; CHECK-AVX512VL: # %bb.0:
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX512VL-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpaddd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpminud {{.*}}(%rip){1to4}, %xmm0, %xmm1
; CHECK-AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: retq
@@ -208,18 +158,10 @@ define <4 x i32> @test_srem_even_100(<4 x i32> %X) nounwind {
;
; CHECK-AVX512VL-LABEL: test_srem_even_100:
; CHECK-AVX512VL: # %bb.0:
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX512VL-NEXT: vpsrad $5, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpaddd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vprord $2, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpminud {{.*}}(%rip){1to4}, %xmm0, %xmm1
; CHECK-AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: retq
@@ -235,113 +177,55 @@ define <4 x i32> @test_srem_even_100(<4 x i32> %X) nounwind {
define <4 x i32> @test_srem_odd_neg25(<4 x i32> %X) nounwind {
; CHECK-SSE2-LABEL: test_srem_odd_neg25:
; CHECK-SSE2: # %bb.0:
-; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [1374389535,2920577761,2920577761,1374389535]
-; CHECK-SSE2-NEXT: movdqa %xmm0, %xmm2
+; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [3264175145,3264175145,3264175145,3264175145]
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm0
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; CHECK-SSE2-NEXT: pmuludq %xmm3, %xmm4
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,3,2,3]
-; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; CHECK-SSE2-NEXT: pxor %xmm3, %xmm3
-; CHECK-SSE2-NEXT: pxor %xmm4, %xmm4
-; CHECK-SSE2-NEXT: pcmpgtd %xmm0, %xmm4
-; CHECK-SSE2-NEXT: pand %xmm1, %xmm4
-; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [0,4294967295,4294967295,0]
-; CHECK-SSE2-NEXT: pand %xmm0, %xmm1
-; CHECK-SSE2-NEXT: paddd %xmm4, %xmm1
-; CHECK-SSE2-NEXT: psubd %xmm1, %xmm2
-; CHECK-SSE2-NEXT: movdqa %xmm2, %xmm1
-; CHECK-SSE2-NEXT: psrld $31, %xmm1
-; CHECK-SSE2-NEXT: psrad $3, %xmm2
-; CHECK-SSE2-NEXT: paddd %xmm1, %xmm2
-; CHECK-SSE2-NEXT: movdqa {{.*#+}} xmm1 = [25,4294967271,4294967271,25]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
-; CHECK-SSE2-NEXT: pmuludq %xmm1, %xmm2
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; CHECK-SSE2-NEXT: pmuludq %xmm4, %xmm1
-; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; CHECK-SSE2-NEXT: psubd %xmm2, %xmm0
-; CHECK-SSE2-NEXT: pcmpeqd %xmm3, %xmm0
-; CHECK-SSE2-NEXT: psrld $31, %xmm0
+; CHECK-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; CHECK-SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; CHECK-SSE2-NEXT: paddd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pxor {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pcmpgtd {{.*}}(%rip), %xmm0
+; CHECK-SSE2-NEXT: pandn {{.*}}(%rip), %xmm0
; CHECK-SSE2-NEXT: retq
;
; CHECK-SSE41-LABEL: test_srem_odd_neg25:
; CHECK-SSE41: # %bb.0:
-; CHECK-SSE41-NEXT: movdqa {{.*#+}} xmm1 = [1374389535,2920577761,2920577761,1374389535]
-; CHECK-SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
-; CHECK-SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-SSE41-NEXT: pmuldq %xmm2, %xmm3
-; CHECK-SSE41-NEXT: pmuldq %xmm0, %xmm1
-; CHECK-SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3],xmm1[4,5],xmm3[6,7]
-; CHECK-SSE41-NEXT: movdqa %xmm1, %xmm2
-; CHECK-SSE41-NEXT: psrld $31, %xmm2
-; CHECK-SSE41-NEXT: psrad $3, %xmm1
-; CHECK-SSE41-NEXT: paddd %xmm2, %xmm1
-; CHECK-SSE41-NEXT: pmulld {{.*}}(%rip), %xmm1
-; CHECK-SSE41-NEXT: psubd %xmm1, %xmm0
-; CHECK-SSE41-NEXT: pxor %xmm1, %xmm1
+; CHECK-SSE41-NEXT: pmulld {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT: paddd {{.*}}(%rip), %xmm0
+; CHECK-SSE41-NEXT: movdqa {{.*#+}} xmm1 = [171798690,171798690,171798690,171798690]
+; CHECK-SSE41-NEXT: pminud %xmm0, %xmm1
; CHECK-SSE41-NEXT: pcmpeqd %xmm1, %xmm0
; CHECK-SSE41-NEXT: psrld $31, %xmm0
; CHECK-SSE41-NEXT: retq
;
; CHECK-AVX1-LABEL: test_srem_odd_neg25:
; CHECK-AVX1: # %bb.0:
-; CHECK-AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [1374389535,2920577761,2920577761,1374389535]
-; CHECK-AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
-; CHECK-AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX1-NEXT: vpmuldq %xmm2, %xmm3, %xmm2
-; CHECK-AVX1-NEXT: vpmuldq %xmm1, %xmm0, %xmm1
-; CHECK-AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
-; CHECK-AVX1-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX1-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-AVX1-NEXT: vpminud {{.*}}(%rip), %xmm0, %xmm1
; CHECK-AVX1-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX1-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX1-NEXT: retq
;
; CHECK-AVX2-LABEL: test_srem_odd_neg25:
; CHECK-AVX2: # %bb.0:
-; CHECK-AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [1374389535,2920577761,2920577761,1374389535]
-; CHECK-AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
-; CHECK-AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX2-NEXT: vpmuldq %xmm2, %xmm3, %xmm2
-; CHECK-AVX2-NEXT: vpmuldq %xmm1, %xmm0, %xmm1
-; CHECK-AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX2-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX2-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [3264175145,3264175145,3264175145,3264175145]
+; CHECK-AVX2-NEXT: vpmulld %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [85899345,85899345,85899345,85899345]
+; CHECK-AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
+; CHECK-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [171798690,171798690,171798690,171798690]
+; CHECK-AVX2-NEXT: vpminud %xmm1, %xmm0, %xmm1
; CHECK-AVX2-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX2-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX2-NEXT: retq
;
; CHECK-AVX512VL-LABEL: test_srem_odd_neg25:
; CHECK-AVX512VL: # %bb.0:
-; CHECK-AVX512VL-NEXT: vmovdqa {{.*#+}} xmm1 = [1374389535,2920577761,2920577761,1374389535]
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,2,3,3]
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm3, %xmm2
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm1, %xmm0, %xmm1
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
-; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX512VL-NEXT: vpsrad $3, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpaddd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpminud {{.*}}(%rip){1to4}, %xmm0, %xmm1
; CHECK-AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: retq
@@ -442,19 +326,10 @@ define <4 x i32> @test_srem_even_neg100(<4 x i32> %X) nounwind {
;
; CHECK-AVX512VL-LABEL: test_srem_even_neg100:
; CHECK-AVX512VL: # %bb.0:
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1374389535,1374389535,1374389535,1374389535]
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm2 = [2920577761,2920577761,2920577761,2920577761]
-; CHECK-AVX512VL-NEXT: vpmuldq %xmm2, %xmm0, %xmm2
-; CHECK-AVX512VL-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
-; CHECK-AVX512VL-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3]
-; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm1, %xmm2
-; CHECK-AVX512VL-NEXT: vpsrad $5, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpaddd %xmm2, %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
-; CHECK-AVX512VL-NEXT: vpsubd %xmm1, %xmm0, %xmm0
-; CHECK-AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; CHECK-AVX512VL-NEXT: vpmulld {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpaddd {{.*}}(%rip){1to4}, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vprord $2, %xmm0, %xmm0
+; CHECK-AVX512VL-NEXT: vpminud {{.*}}(%rip){1to4}, %xmm0, %xmm1
; CHECK-AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: vpsrld $31, %xmm0, %xmm0
; CHECK-AVX512VL-NEXT: retq
OpenPOWER on IntegriCloud