summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vector-reduce-smax.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86/vector-reduce-smax.ll')
-rw-r--r--llvm/test/CodeGen/X86/vector-reduce-smax.ll613
1 files changed, 472 insertions, 141 deletions
diff --git a/llvm/test/CodeGen/X86/vector-reduce-smax.ll b/llvm/test/CodeGen/X86/vector-reduce-smax.ll
index 8a184cd78e4..4d421d08f61 100644
--- a/llvm/test/CodeGen/X86/vector-reduce-smax.ll
+++ b/llvm/test/CodeGen/X86/vector-reduce-smax.ll
@@ -670,35 +670,108 @@ define i64 @test_v16i64(<16 x i64> %a0) {
define i32 @test_v2i32(<2 x i32> %a0) {
; SSE2-LABEL: test_v2i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE2-NEXT: movdqa %xmm0, %xmm2
-; SSE2-NEXT: pcmpgtd %xmm1, %xmm2
-; SSE2-NEXT: pand %xmm2, %xmm0
-; SSE2-NEXT: pandn %xmm1, %xmm2
-; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: pxor %xmm1, %xmm1
+; SSE2-NEXT: punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE2-NEXT: psllq $32, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,3,2,3]
+; SSE2-NEXT: psrad $31, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,3,2,3]
+; SSE2-NEXT: psrad $31, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm0, %xmm3
+; SSE2-NEXT: pxor %xmm1, %xmm3
+; SSE2-NEXT: pxor %xmm2, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm4
+; SSE2-NEXT: pcmpgtd %xmm3, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE2-NEXT: pand %xmm5, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm3
+; SSE2-NEXT: pand %xmm3, %xmm2
+; SSE2-NEXT: pandn %xmm0, %xmm3
+; SSE2-NEXT: por %xmm2, %xmm3
+; SSE2-NEXT: movd %xmm3, %eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v2i32:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE41-NEXT: pmaxsd %xmm0, %xmm1
-; SSE41-NEXT: movd %xmm1, %eax
+; SSE41-NEXT: movdqa %xmm0, %xmm1
+; SSE41-NEXT: psllq $32, %xmm1
+; SSE41-NEXT: psrad $31, %xmm1
+; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
+; SSE41-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
+; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
+; SSE41-NEXT: psrad $31, %xmm3
+; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3],xmm2[4,5],xmm3[6,7]
+; SSE41-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE41-NEXT: movdqa %xmm3, %xmm0
+; SSE41-NEXT: pxor %xmm2, %xmm0
+; SSE41-NEXT: pxor %xmm1, %xmm2
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
+; SSE41-NEXT: pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
+; SSE41-NEXT: pand %xmm4, %xmm0
+; SSE41-NEXT: por %xmm2, %xmm0
+; SSE41-NEXT: blendvpd %xmm0, %xmm1, %xmm3
+; SSE41-NEXT: movd %xmm3, %eax
; SSE41-NEXT: retq
;
-; AVX-LABEL: test_v2i32:
-; AVX: # %bb.0:
-; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vmovd %xmm0, %eax
-; AVX-NEXT: retq
+; AVX1-LABEL: test_v2i32:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vpsllq $32, %xmm0, %xmm1
+; AVX1-NEXT: vpsrad $31, %xmm1, %xmm1
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT: vpsrad $31, %xmm0, %xmm0
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3],xmm2[4,5],xmm0[6,7]
+; AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX1-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vmovd %xmm0, %eax
+; AVX1-NEXT: retq
;
-; AVX512-LABEL: test_v2i32:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX512-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vmovd %xmm0, %eax
-; AVX512-NEXT: retq
+; AVX2-LABEL: test_v2i32:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpsllq $32, %xmm0, %xmm1
+; AVX2-NEXT: vpsrad $31, %xmm1, %xmm1
+; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
+; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[2,2,3,3]
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-NEXT: vpsrad $31, %xmm0, %xmm0
+; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0],xmm0[1],xmm2[2],xmm0[3]
+; AVX2-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX2-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT: vmovd %xmm0, %eax
+; AVX2-NEXT: retq
+;
+; AVX512BW-LABEL: test_v2i32:
+; AVX512BW: # %bb.0:
+; AVX512BW-NEXT: vpsllq $32, %xmm0, %xmm1
+; AVX512BW-NEXT: vpsraq $32, %zmm1, %zmm1
+; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX512BW-NEXT: vpsraq $32, %zmm0, %zmm0
+; AVX512BW-NEXT: vpmaxsq %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT: vmovd %xmm0, %eax
+; AVX512BW-NEXT: vzeroupper
+; AVX512BW-NEXT: retq
+;
+; AVX512VL-LABEL: test_v2i32:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpsllq $32, %xmm0, %xmm1
+; AVX512VL-NEXT: vpsraq $32, %xmm1, %xmm1
+; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512VL-NEXT: vpsllq $32, %xmm0, %xmm0
+; AVX512VL-NEXT: vpsraq $32, %xmm0, %xmm0
+; AVX512VL-NEXT: vpmaxsq %xmm0, %xmm1, %xmm0
+; AVX512VL-NEXT: vmovd %xmm0, %eax
+; AVX512VL-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.smax.v2i32(<2 x i32> %a0)
ret i32 %1
}
@@ -1041,62 +1114,199 @@ define i32 @test_v32i32(<32 x i32> %a0) {
;
define i16 @test_v2i16(<2 x i16> %a0) {
-; SSE-LABEL: test_v2i16:
-; SSE: # %bb.0:
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: pmaxsw %xmm0, %xmm1
-; SSE-NEXT: movd %xmm1, %eax
-; SSE-NEXT: # kill: def $ax killed $ax killed $eax
-; SSE-NEXT: retq
-;
-; AVX-LABEL: test_v2i16:
-; AVX: # %bb.0:
-; AVX-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vmovd %xmm0, %eax
-; AVX-NEXT: # kill: def $ax killed $ax killed $eax
-; AVX-NEXT: retq
+; SSE2-LABEL: test_v2i16:
+; SSE2: # %bb.0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT: psllq $48, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: psrad $31, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; SSE2-NEXT: psrad $16, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT: psllq $48, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm2
+; SSE2-NEXT: psrad $31, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; SSE2-NEXT: psrad $16, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm0, %xmm3
+; SSE2-NEXT: pxor %xmm2, %xmm3
+; SSE2-NEXT: pxor %xmm1, %xmm2
+; SSE2-NEXT: movdqa %xmm3, %xmm4
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm5, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE2-NEXT: por %xmm2, %xmm3
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: pandn %xmm1, %xmm3
+; SSE2-NEXT: por %xmm0, %xmm3
+; SSE2-NEXT: movd %xmm3, %eax
+; SSE2-NEXT: # kill: def $ax killed $ax killed $eax
+; SSE2-NEXT: retq
;
-; AVX512-LABEL: test_v2i16:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vmovd %xmm0, %eax
-; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
-; AVX512-NEXT: retq
+; SSE41-LABEL: test_v2i16:
+; SSE41: # %bb.0:
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE41-NEXT: psllq $48, %xmm0
+; SSE41-NEXT: movdqa %xmm0, %xmm2
+; SSE41-NEXT: psrad $31, %xmm2
+; SSE41-NEXT: psrad $16, %xmm0
+; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3],xmm3[4,5],xmm2[6,7]
+; SSE41-NEXT: psllq $48, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: psrad $31, %xmm0
+; SSE41-NEXT: psrad $16, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3],xmm1[4,5],xmm0[6,7]
+; SSE41-NEXT: movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
+; SSE41-NEXT: movdqa %xmm3, %xmm2
+; SSE41-NEXT: pxor %xmm0, %xmm2
+; SSE41-NEXT: pxor %xmm1, %xmm0
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
+; SSE41-NEXT: pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
+; SSE41-NEXT: pand %xmm4, %xmm0
+; SSE41-NEXT: por %xmm2, %xmm0
+; SSE41-NEXT: blendvpd %xmm0, %xmm3, %xmm1
+; SSE41-NEXT: movd %xmm1, %eax
+; SSE41-NEXT: # kill: def $ax killed $ax killed $eax
+; SSE41-NEXT: retq
+;
+; AVX1-LABEL: test_v2i16:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vpsllq $48, %xmm0, %xmm1
+; AVX1-NEXT: vpsrad $31, %xmm1, %xmm2
+; AVX1-NEXT: vpsrad $16, %xmm1, %xmm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX1-NEXT: vpsllq $48, %xmm0, %xmm0
+; AVX1-NEXT: vpsrad $31, %xmm0, %xmm2
+; AVX1-NEXT: vpsrad $16, %xmm0, %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
+; AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX1-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vmovd %xmm0, %eax
+; AVX1-NEXT: # kill: def $ax killed $ax killed $eax
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: test_v2i16:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpsllq $48, %xmm0, %xmm1
+; AVX2-NEXT: vpsrad $31, %xmm1, %xmm2
+; AVX2-NEXT: vpsrad $16, %xmm1, %xmm1
+; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX2-NEXT: vpsllq $48, %xmm0, %xmm0
+; AVX2-NEXT: vpsrad $31, %xmm0, %xmm2
+; AVX2-NEXT: vpsrad $16, %xmm0, %xmm0
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX2-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX2-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT: vmovd %xmm0, %eax
+; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
+; AVX2-NEXT: retq
+;
+; AVX512BW-LABEL: test_v2i16:
+; AVX512BW: # %bb.0:
+; AVX512BW-NEXT: vpsllq $48, %xmm0, %xmm1
+; AVX512BW-NEXT: vpsraq $48, %zmm1, %zmm1
+; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512BW-NEXT: vpsllq $48, %xmm0, %xmm0
+; AVX512BW-NEXT: vpsraq $48, %zmm0, %zmm0
+; AVX512BW-NEXT: vpmaxsq %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT: vmovd %xmm0, %eax
+; AVX512BW-NEXT: # kill: def $ax killed $ax killed $eax
+; AVX512BW-NEXT: vzeroupper
+; AVX512BW-NEXT: retq
+;
+; AVX512VL-LABEL: test_v2i16:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpsllq $48, %xmm0, %xmm1
+; AVX512VL-NEXT: vpsraq $48, %xmm1, %xmm1
+; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512VL-NEXT: vpsllq $48, %xmm0, %xmm0
+; AVX512VL-NEXT: vpsraq $48, %xmm0, %xmm0
+; AVX512VL-NEXT: vpmaxsq %xmm0, %xmm1, %xmm0
+; AVX512VL-NEXT: vmovd %xmm0, %eax
+; AVX512VL-NEXT: # kill: def $ax killed $ax killed $eax
+; AVX512VL-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.smax.v2i16(<2 x i16> %a0)
ret i16 %1
}
define i16 @test_v4i16(<4 x i16> %a0) {
-; SSE-LABEL: test_v4i16:
-; SSE: # %bb.0:
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE-NEXT: pmaxsw %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: pmaxsw %xmm1, %xmm0
-; SSE-NEXT: movd %xmm0, %eax
-; SSE-NEXT: # kill: def $ax killed $ax killed $eax
-; SSE-NEXT: retq
+; SSE2-LABEL: test_v4i16:
+; SSE2: # %bb.0:
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT: pslld $16, %xmm0
+; SSE2-NEXT: psrad $16, %xmm0
+; SSE2-NEXT: pslld $16, %xmm1
+; SSE2-NEXT: psrad $16, %xmm1
+; SSE2-NEXT: movdqa %xmm0, %xmm2
+; SSE2-NEXT: pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT: pand %xmm2, %xmm0
+; SSE2-NEXT: pandn %xmm1, %xmm2
+; SSE2-NEXT: por %xmm0, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
+; SSE2-NEXT: movdqa %xmm2, %xmm1
+; SSE2-NEXT: pcmpgtd %xmm0, %xmm1
+; SSE2-NEXT: pand %xmm1, %xmm2
+; SSE2-NEXT: pandn %xmm0, %xmm1
+; SSE2-NEXT: por %xmm2, %xmm1
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: # kill: def $ax killed $ax killed $eax
+; SSE2-NEXT: retq
+;
+; SSE41-LABEL: test_v4i16:
+; SSE41: # %bb.0:
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE41-NEXT: pslld $16, %xmm0
+; SSE41-NEXT: psrad $16, %xmm0
+; SSE41-NEXT: pslld $16, %xmm1
+; SSE41-NEXT: psrad $16, %xmm1
+; SSE41-NEXT: pmaxsd %xmm0, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE41-NEXT: pmaxsd %xmm1, %xmm0
+; SSE41-NEXT: movd %xmm0, %eax
+; SSE41-NEXT: # kill: def $ax killed $ax killed $eax
+; SSE41-NEXT: retq
;
; AVX-LABEL: test_v4i16:
; AVX: # %bb.0:
+; AVX-NEXT: vpslld $16, %xmm0, %xmm1
+; AVX-NEXT: vpsrad $16, %xmm1, %xmm1
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT: vpslld $16, %xmm0, %xmm0
+; AVX-NEXT: vpsrad $16, %xmm0, %xmm0
+; AVX-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $ax killed $ax killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4i16:
; AVX512: # %bb.0:
+; AVX512-NEXT: vpslld $16, %xmm0, %xmm1
+; AVX512-NEXT: vpsrad $16, %xmm1, %xmm1
+; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512-NEXT: vpslld $16, %xmm0, %xmm0
+; AVX512-NEXT: vpsrad $16, %xmm0, %xmm0
+; AVX512-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX512-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq
@@ -1384,41 +1594,132 @@ define i16 @test_v64i16(<64 x i16> %a0) {
define i8 @test_v2i8(<2 x i8> %a0) {
; SSE2-LABEL: test_v2i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: psrlw $8, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT: psllq $56, %xmm0
; SSE2-NEXT: movdqa %xmm0, %xmm2
-; SSE2-NEXT: pcmpgtb %xmm1, %xmm2
-; SSE2-NEXT: pand %xmm2, %xmm0
-; SSE2-NEXT: pandn %xmm1, %xmm2
-; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: psrad $31, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; SSE2-NEXT: psrad $24, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE2-NEXT: psllq $56, %xmm1
+; SSE2-NEXT: movdqa %xmm1, %xmm2
+; SSE2-NEXT: psrad $31, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,2,3]
+; SSE2-NEXT: psrad $24, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,3,2,3]
+; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm0, %xmm3
+; SSE2-NEXT: pxor %xmm2, %xmm3
+; SSE2-NEXT: pxor %xmm1, %xmm2
+; SSE2-NEXT: movdqa %xmm3, %xmm4
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm5, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,1,3,3]
+; SSE2-NEXT: por %xmm2, %xmm3
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: pandn %xmm1, %xmm3
+; SSE2-NEXT: por %xmm0, %xmm3
+; SSE2-NEXT: movd %xmm3, %eax
; SSE2-NEXT: # kill: def $al killed $al killed $eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v2i8:
; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm1
-; SSE41-NEXT: psrlw $8, %xmm1
-; SSE41-NEXT: pmaxsb %xmm0, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE41-NEXT: psllq $56, %xmm0
+; SSE41-NEXT: movdqa %xmm0, %xmm2
+; SSE41-NEXT: psrad $31, %xmm2
+; SSE41-NEXT: psrad $24, %xmm0
+; SSE41-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
+; SSE41-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3],xmm3[4,5],xmm2[6,7]
+; SSE41-NEXT: psllq $56, %xmm1
+; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: psrad $31, %xmm0
+; SSE41-NEXT: psrad $24, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3],xmm1[4,5],xmm0[6,7]
+; SSE41-NEXT: movdqa {{.*#+}} xmm0 = [2147483648,2147483648]
+; SSE41-NEXT: movdqa %xmm3, %xmm2
+; SSE41-NEXT: pxor %xmm0, %xmm2
+; SSE41-NEXT: pxor %xmm1, %xmm0
+; SSE41-NEXT: movdqa %xmm2, %xmm4
+; SSE41-NEXT: pcmpeqd %xmm0, %xmm4
+; SSE41-NEXT: pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,0,2,2]
+; SSE41-NEXT: pand %xmm4, %xmm0
+; SSE41-NEXT: por %xmm2, %xmm0
+; SSE41-NEXT: blendvpd %xmm0, %xmm3, %xmm1
; SSE41-NEXT: pextrb $0, %xmm1, %eax
; SSE41-NEXT: # kill: def $al killed $al killed $eax
; SSE41-NEXT: retq
;
-; AVX-LABEL: test_v2i8:
-; AVX: # %bb.0:
-; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpextrb $0, %xmm0, %eax
-; AVX-NEXT: # kill: def $al killed $al killed $eax
-; AVX-NEXT: retq
+; AVX1-LABEL: test_v2i8:
+; AVX1: # %bb.0:
+; AVX1-NEXT: vpsllq $56, %xmm0, %xmm1
+; AVX1-NEXT: vpsrad $31, %xmm1, %xmm2
+; AVX1-NEXT: vpsrad $24, %xmm1, %xmm1
+; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX1-NEXT: vpsllq $56, %xmm0, %xmm0
+; AVX1-NEXT: vpsrad $31, %xmm0, %xmm2
+; AVX1-NEXT: vpsrad $24, %xmm0, %xmm0
+; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
+; AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX1-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vpextrb $0, %xmm0, %eax
+; AVX1-NEXT: # kill: def $al killed $al killed $eax
+; AVX1-NEXT: retq
;
-; AVX512-LABEL: test_v2i8:
-; AVX512: # %bb.0:
-; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpextrb $0, %xmm0, %eax
-; AVX512-NEXT: # kill: def $al killed $al killed $eax
-; AVX512-NEXT: retq
+; AVX2-LABEL: test_v2i8:
+; AVX2: # %bb.0:
+; AVX2-NEXT: vpsllq $56, %xmm0, %xmm1
+; AVX2-NEXT: vpsrad $31, %xmm1, %xmm2
+; AVX2-NEXT: vpsrad $24, %xmm1, %xmm1
+; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX2-NEXT: vpsllq $56, %xmm0, %xmm0
+; AVX2-NEXT: vpsrad $31, %xmm0, %xmm2
+; AVX2-NEXT: vpsrad $24, %xmm0, %xmm0
+; AVX2-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3]
+; AVX2-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm2
+; AVX2-NEXT: vblendvpd %xmm2, %xmm1, %xmm0, %xmm0
+; AVX2-NEXT: vpextrb $0, %xmm0, %eax
+; AVX2-NEXT: # kill: def $al killed $al killed $eax
+; AVX2-NEXT: retq
+;
+; AVX512BW-LABEL: test_v2i8:
+; AVX512BW: # %bb.0:
+; AVX512BW-NEXT: vpsllq $56, %xmm0, %xmm1
+; AVX512BW-NEXT: vpsraq $56, %zmm1, %zmm1
+; AVX512BW-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512BW-NEXT: vpsllq $56, %xmm0, %xmm0
+; AVX512BW-NEXT: vpsraq $56, %zmm0, %zmm0
+; AVX512BW-NEXT: vpmaxsq %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT: vpextrb $0, %xmm0, %eax
+; AVX512BW-NEXT: # kill: def $al killed $al killed $eax
+; AVX512BW-NEXT: vzeroupper
+; AVX512BW-NEXT: retq
+;
+; AVX512VL-LABEL: test_v2i8:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpsllq $56, %xmm0, %xmm1
+; AVX512VL-NEXT: vpsraq $56, %xmm1, %xmm1
+; AVX512VL-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512VL-NEXT: vpsllq $56, %xmm0, %xmm0
+; AVX512VL-NEXT: vpsraq $56, %xmm0, %xmm0
+; AVX512VL-NEXT: vpmaxsq %xmm0, %xmm1, %xmm0
+; AVX512VL-NEXT: vpextrb $0, %xmm0, %eax
+; AVX512VL-NEXT: # kill: def $al killed $al killed $eax
+; AVX512VL-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.smax.v2i8(<2 x i8> %a0)
ret i8 %1
}
@@ -1426,17 +1727,19 @@ define i8 @test_v2i8(<2 x i8> %a0) {
define i8 @test_v4i8(<4 x i8> %a0) {
; SSE2-LABEL: test_v4i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: psrld $16, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT: pslld $24, %xmm0
+; SSE2-NEXT: psrad $24, %xmm0
+; SSE2-NEXT: pslld $24, %xmm1
+; SSE2-NEXT: psrad $24, %xmm1
; SSE2-NEXT: movdqa %xmm0, %xmm2
-; SSE2-NEXT: pcmpgtb %xmm1, %xmm2
+; SSE2-NEXT: pcmpgtd %xmm1, %xmm2
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn %xmm1, %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movdqa %xmm2, %xmm0
-; SSE2-NEXT: psrlw $8, %xmm0
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
; SSE2-NEXT: movdqa %xmm2, %xmm1
-; SSE2-NEXT: pcmpgtb %xmm0, %xmm1
+; SSE2-NEXT: pcmpgtd %xmm0, %xmm1
; SSE2-NEXT: pand %xmm1, %xmm2
; SSE2-NEXT: pandn %xmm0, %xmm1
; SSE2-NEXT: por %xmm2, %xmm1
@@ -1446,32 +1749,42 @@ define i8 @test_v4i8(<4 x i8> %a0) {
;
; SSE41-LABEL: test_v4i8:
; SSE41: # %bb.0:
-; SSE41-NEXT: movdqa %xmm0, %xmm1
-; SSE41-NEXT: psrld $16, %xmm1
-; SSE41-NEXT: pmaxsb %xmm0, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm0
-; SSE41-NEXT: psrlw $8, %xmm0
-; SSE41-NEXT: pmaxsb %xmm1, %xmm0
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE41-NEXT: pslld $24, %xmm0
+; SSE41-NEXT: psrad $24, %xmm0
+; SSE41-NEXT: pslld $24, %xmm1
+; SSE41-NEXT: psrad $24, %xmm1
+; SSE41-NEXT: pmaxsd %xmm0, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE41-NEXT: pmaxsd %xmm1, %xmm0
; SSE41-NEXT: pextrb $0, %xmm0, %eax
; SSE41-NEXT: # kill: def $al killed $al killed $eax
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v4i8:
; AVX: # %bb.0:
-; AVX-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpslld $24, %xmm0, %xmm1
+; AVX-NEXT: vpsrad $24, %xmm1, %xmm1
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT: vpslld $24, %xmm0, %xmm0
+; AVX-NEXT: vpsrad $24, %xmm0, %xmm0
+; AVX-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0
+; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; AVX-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpextrb $0, %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4i8:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vpslld $24, %xmm0, %xmm1
+; AVX512-NEXT: vpsrad $24, %xmm1, %xmm1
+; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512-NEXT: vpslld $24, %xmm0, %xmm0
+; AVX512-NEXT: vpsrad $24, %xmm0, %xmm0
+; AVX512-NEXT: vpmaxsd %xmm0, %xmm1, %xmm0
+; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
+; AVX512-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
@@ -1482,64 +1795,82 @@ define i8 @test_v4i8(<4 x i8> %a0) {
define i8 @test_v8i8(<8 x i8> %a0) {
; SSE2-LABEL: test_v8i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE2-NEXT: movdqa %xmm0, %xmm2
-; SSE2-NEXT: pcmpgtb %xmm1, %xmm2
-; SSE2-NEXT: pand %xmm2, %xmm0
-; SSE2-NEXT: pandn %xmm1, %xmm2
-; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movdqa %xmm2, %xmm0
-; SSE2-NEXT: psrld $16, %xmm0
-; SSE2-NEXT: movdqa %xmm2, %xmm1
-; SSE2-NEXT: pcmpgtb %xmm0, %xmm1
-; SSE2-NEXT: pand %xmm1, %xmm2
-; SSE2-NEXT: pandn %xmm0, %xmm1
-; SSE2-NEXT: por %xmm2, %xmm1
-; SSE2-NEXT: movdqa %xmm1, %xmm0
-; SSE2-NEXT: psrlw $8, %xmm0
-; SSE2-NEXT: movdqa %xmm1, %xmm2
-; SSE2-NEXT: pcmpgtb %xmm0, %xmm2
-; SSE2-NEXT: pand %xmm2, %xmm1
-; SSE2-NEXT: pandn %xmm0, %xmm2
-; SSE2-NEXT: por %xmm1, %xmm2
-; SSE2-NEXT: movd %xmm2, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE2-NEXT: psllw $8, %xmm0
+; SSE2-NEXT: psraw $8, %xmm0
+; SSE2-NEXT: psllw $8, %xmm1
+; SSE2-NEXT: psraw $8, %xmm1
+; SSE2-NEXT: pmaxsw %xmm0, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE2-NEXT: psllw $8, %xmm0
+; SSE2-NEXT: psraw $8, %xmm0
+; SSE2-NEXT: pmaxsw %xmm1, %xmm0
+; SSE2-NEXT: movdqa %xmm0, %xmm1
+; SSE2-NEXT: psrld $16, %xmm1
+; SSE2-NEXT: psllw $8, %xmm1
+; SSE2-NEXT: psraw $8, %xmm1
+; SSE2-NEXT: pmaxsw %xmm0, %xmm1
+; SSE2-NEXT: movd %xmm1, %eax
; SSE2-NEXT: # kill: def $al killed $al killed $eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v8i8:
; SSE41: # %bb.0:
-; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; SSE41-NEXT: pmaxsb %xmm0, %xmm1
-; SSE41-NEXT: movdqa %xmm1, %xmm0
-; SSE41-NEXT: psrld $16, %xmm0
-; SSE41-NEXT: pmaxsb %xmm1, %xmm0
+; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE41-NEXT: psllw $8, %xmm0
+; SSE41-NEXT: psraw $8, %xmm0
+; SSE41-NEXT: psllw $8, %xmm1
+; SSE41-NEXT: psraw $8, %xmm1
+; SSE41-NEXT: pmaxsw %xmm0, %xmm1
+; SSE41-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE41-NEXT: psllw $8, %xmm0
+; SSE41-NEXT: psraw $8, %xmm0
+; SSE41-NEXT: pmaxsw %xmm1, %xmm0
; SSE41-NEXT: movdqa %xmm0, %xmm1
-; SSE41-NEXT: psrlw $8, %xmm1
-; SSE41-NEXT: pmaxsb %xmm0, %xmm1
+; SSE41-NEXT: psrld $16, %xmm1
+; SSE41-NEXT: psllw $8, %xmm1
+; SSE41-NEXT: psraw $8, %xmm1
+; SSE41-NEXT: pmaxsw %xmm0, %xmm1
; SSE41-NEXT: pextrb $0, %xmm1, %eax
; SSE41-NEXT: # kill: def $al killed $al killed $eax
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v8i8:
; AVX: # %bb.0:
+; AVX-NEXT: vpsllw $8, %xmm0, %xmm1
+; AVX-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX-NEXT: vpsllw $8, %xmm0, %xmm0
+; AVX-NEXT: vpsraw $8, %xmm0, %xmm0
+; AVX-NEXT: vpmaxsw %xmm0, %xmm1, %xmm0
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpsllw $8, %xmm1, %xmm1
+; AVX-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vpsllw $8, %xmm1, %xmm1
+; AVX-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpextrb $0, %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8i8:
; AVX512: # %bb.0:
+; AVX512-NEXT: vpsllw $8, %xmm0, %xmm1
+; AVX512-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX512-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX512-NEXT: vpsllw $8, %xmm0, %xmm0
+; AVX512-NEXT: vpsraw $8, %xmm0, %xmm0
+; AVX512-NEXT: vpmaxsw %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,2,3]
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vpsllw $8, %xmm1, %xmm1
+; AVX512-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX512-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
-; AVX512-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0
+; AVX512-NEXT: vpsllw $8, %xmm1, %xmm1
+; AVX512-NEXT: vpsraw $8, %xmm1, %xmm1
+; AVX512-NEXT: vpmaxsw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpextrb $0, %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
OpenPOWER on IntegriCloud