summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86')
-rw-r--r--llvm/test/CodeGen/X86/avx2-shift.ll10
-rw-r--r--llvm/test/CodeGen/X86/avx2-vector-shifts.ll10
-rw-r--r--llvm/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll30
-rw-r--r--llvm/test/CodeGen/X86/vector-fshl-128.ll19
-rw-r--r--llvm/test/CodeGen/X86/vector-fshl-rot-128.ll21
-rw-r--r--llvm/test/CodeGen/X86/vector-fshr-128.ll15
-rw-r--r--llvm/test/CodeGen/X86/vector-fshr-rot-128.ll19
-rw-r--r--llvm/test/CodeGen/X86/vector-rotate-128.ll21
-rw-r--r--llvm/test/CodeGen/X86/vector-shift-lshr-128.ll5
-rw-r--r--llvm/test/CodeGen/X86/vector-shift-lshr-sub128.ll10
10 files changed, 67 insertions, 93 deletions
diff --git a/llvm/test/CodeGen/X86/avx2-shift.ll b/llvm/test/CodeGen/X86/avx2-shift.ll
index f4924b3e2fb..1a0993bd23d 100644
--- a/llvm/test/CodeGen/X86/avx2-shift.ll
+++ b/llvm/test/CodeGen/X86/avx2-shift.ll
@@ -580,9 +580,8 @@ define <8 x i16> @variable_lshr16(<8 x i16> %lhs, <8 x i16> %rhs) {
; X32-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; X32-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X32-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; X32-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; X32-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X32-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X32-NEXT: vextracti128 $1, %ymm0, %xmm1
+; X32-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; X32-NEXT: vzeroupper
; X32-NEXT: retl
;
@@ -591,9 +590,8 @@ define <8 x i16> @variable_lshr16(<8 x i16> %lhs, <8 x i16> %rhs) {
; X64-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; X64-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X64-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; X64-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; X64-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X64-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X64-NEXT: vextracti128 $1, %ymm0, %xmm1
+; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
%res = lshr <8 x i16> %lhs, %rhs
diff --git a/llvm/test/CodeGen/X86/avx2-vector-shifts.ll b/llvm/test/CodeGen/X86/avx2-vector-shifts.ll
index d95f8a01451..905101cd1fc 100644
--- a/llvm/test/CodeGen/X86/avx2-vector-shifts.ll
+++ b/llvm/test/CodeGen/X86/avx2-vector-shifts.ll
@@ -637,9 +637,8 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; X32-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; X32-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X32-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; X32-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; X32-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X32-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X32-NEXT: vextracti128 $1, %ymm0, %xmm1
+; X32-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; X32-NEXT: vzeroupper
; X32-NEXT: retl
;
@@ -648,9 +647,8 @@ define <8 x i16> @lshr_8i16(<8 x i16> %r, <8 x i16> %a) nounwind {
; X64-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; X64-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X64-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; X64-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; X64-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; X64-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; X64-NEXT: vextracti128 $1, %ymm0, %xmm1
+; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
%lshr = lshr <8 x i16> %r, %a
diff --git a/llvm/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll b/llvm/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
index 363df5335ef..6663459f49d 100644
--- a/llvm/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
+++ b/llvm/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
@@ -762,7 +762,7 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
; X86-BMI1-NEXT: movb {{[0-9]+}}(%esp), %cl
; X86-BMI1-NEXT: movl $1, %eax
; X86-BMI1-NEXT: shrl %cl, %eax
-; X86-BMI1-NEXT: testl $-1437226411, %eax # imm = 0xAA55AA55
+; X86-BMI1-NEXT: testb %al, %al
; X86-BMI1-NEXT: sete %al
; X86-BMI1-NEXT: retl
;
@@ -771,7 +771,7 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
; X86-BMI2-NEXT: movb {{[0-9]+}}(%esp), %al
; X86-BMI2-NEXT: movl $1, %ecx
; X86-BMI2-NEXT: shrxl %eax, %ecx, %eax
-; X86-BMI2-NEXT: testl $-1437226411, %eax # imm = 0xAA55AA55
+; X86-BMI2-NEXT: testb %al, %al
; X86-BMI2-NEXT: sete %al
; X86-BMI2-NEXT: retl
;
@@ -781,7 +781,7 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
; X64-BMI1-NEXT: movl $1, %eax
; X64-BMI1-NEXT: # kill: def $cl killed $cl killed $ecx
; X64-BMI1-NEXT: shrl %cl, %eax
-; X64-BMI1-NEXT: testl $-1437226411, %eax # imm = 0xAA55AA55
+; X64-BMI1-NEXT: testb %al, %al
; X64-BMI1-NEXT: sete %al
; X64-BMI1-NEXT: retq
;
@@ -789,7 +789,7 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
; X64-BMI2: # %bb.0:
; X64-BMI2-NEXT: movl $1, %eax
; X64-BMI2-NEXT: shrxl %edi, %eax, %eax
-; X64-BMI2-NEXT: testl $-1437226411, %eax # imm = 0xAA55AA55
+; X64-BMI2-NEXT: testb %al, %al
; X64-BMI2-NEXT: sete %al
; X64-BMI2-NEXT: retq
%t0 = lshr i32 1, %y
@@ -803,24 +803,10 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
;------------------------------------------------------------------------------;
define i1 @negative_scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind {
-; X86-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; X86: # %bb.0:
-; X86-NEXT: movb {{[0-9]+}}(%esp), %cl
-; X86-NEXT: movb $24, %al
-; X86-NEXT: shrb %cl, %al
-; X86-NEXT: andb {{[0-9]+}}(%esp), %al
-; X86-NEXT: shrb $7, %al
-; X86-NEXT: retl
-;
-; X64-LABEL: negative_scalar_i8_bitsinmiddle_slt:
-; X64: # %bb.0:
-; X64-NEXT: movl %esi, %ecx
-; X64-NEXT: movb $24, %al
-; X64-NEXT: # kill: def $cl killed $cl killed $ecx
-; X64-NEXT: shrb %cl, %al
-; X64-NEXT: andb %dil, %al
-; X64-NEXT: shrb $7, %al
-; X64-NEXT: retq
+; CHECK-LABEL: negative_scalar_i8_bitsinmiddle_slt:
+; CHECK: # %bb.0:
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: ret{{[l|q]}}
%t0 = lshr i8 24, %y
%t1 = and i8 %t0, %x
%res = icmp slt i8 %t1, 0
diff --git a/llvm/test/CodeGen/X86/vector-fshl-128.ll b/llvm/test/CodeGen/X86/vector-fshl-128.ll
index 12a5f2bc2cc..5fb1c6fb088 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-128.ll
@@ -643,20 +643,19 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt)
;
; AVX2-LABEL: var_funnnel_v8i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm3 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm2, %xmm3, %xmm3
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX2-NEXT: vpsrlvd %ymm3, %ymm1, %ymm1
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm1, %xmm1
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT: vpsllvd %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3
+; AVX2-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm5 = [16,16,16,16,16,16,16,16]
-; AVX2-NEXT: vpsubw %xmm2, %xmm5, %xmm5
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsrlvd %ymm5, %ymm1, %ymm1
-; AVX2-NEXT: vpshufb %ymm4, %ymm1, %ymm1
-; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
; AVX2-NEXT: vpor %xmm1, %xmm3, %xmm1
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vpcmpeqw %xmm3, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-fshl-rot-128.ll b/llvm/test/CodeGen/X86/vector-fshl-rot-128.ll
index ce521ad8896..7ef1ed243a6 100644
--- a/llvm/test/CodeGen/X86/vector-fshl-rot-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshl-rot-128.ll
@@ -338,20 +338,19 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %amt) nounwind {
;
; AVX2-LABEL: var_funnnel_v8i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsllvd %ymm2, %ymm0, %ymm2
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16]
-; AVX2-NEXT: vpsubw %xmm1, %xmm4, %xmm1
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT: vpsrlvd %ymm2, %ymm0, %ymm2
+; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vpsllvd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX2-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-fshr-128.ll b/llvm/test/CodeGen/X86/vector-fshr-128.ll
index 00f5d73a4de..ee982383c46 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-128.ll
@@ -656,15 +656,14 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %amt)
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT: vpsrlvd %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm5 = [16,16,16,16,16,16,16,16]
-; AVX2-NEXT: vpsubw %xmm2, %xmm5, %xmm5
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm5 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero,xmm5[4],zero,xmm5[5],zero,xmm5[6],zero,xmm5[7],zero
+; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4
+; AVX2-NEXT: vpackusdw %xmm4, %xmm3, %xmm3
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm2, %xmm4, %xmm4
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm4 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero,xmm4[4],zero,xmm4[5],zero,xmm4[6],zero,xmm4[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX2-NEXT: vpsllvd %ymm5, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm4, %ymm0, %ymm0
+; AVX2-NEXT: vpsllvd %ymm4, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
; AVX2-NEXT: vpor %xmm3, %xmm0, %xmm0
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
diff --git a/llvm/test/CodeGen/X86/vector-fshr-rot-128.ll b/llvm/test/CodeGen/X86/vector-fshr-rot-128.ll
index d88a2a214ca..0439459ac55 100644
--- a/llvm/test/CodeGen/X86/vector-fshr-rot-128.ll
+++ b/llvm/test/CodeGen/X86/vector-fshr-rot-128.ll
@@ -361,19 +361,18 @@ define <8 x i16> @var_funnnel_v8i16(<8 x i16> %x, <8 x i16> %amt) nounwind {
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpsubw %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
-; AVX2-NEXT: vpsllvd %ymm2, %ymm0, %ymm2
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16]
-; AVX2-NEXT: vpsubw %xmm1, %xmm4, %xmm1
+; AVX2-NEXT: vpsrlvd %ymm2, %ymm0, %ymm2
+; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vpsllvd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX2-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-rotate-128.ll b/llvm/test/CodeGen/X86/vector-rotate-128.ll
index 666325b35f7..35dbb109924 100644
--- a/llvm/test/CodeGen/X86/vector-rotate-128.ll
+++ b/llvm/test/CodeGen/X86/vector-rotate-128.ll
@@ -324,20 +324,19 @@ define <8 x i16> @var_rotate_v8i16(<8 x i16> %a, <8 x i16> %b) nounwind {
;
; AVX2-LABEL: var_rotate_v8i16:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpand {{.*}}(%rip), %xmm1, %xmm1
-; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsllvd %ymm2, %ymm0, %ymm2
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[0,2,2,3]
-; AVX2-NEXT: vmovdqa {{.*#+}} xmm4 = [16,16,16,16,16,16,16,16]
-; AVX2-NEXT: vpsubw %xmm1, %xmm4, %xmm1
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT: vpsubw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT: vpsrlvd %ymm2, %ymm0, %ymm2
+; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
-; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vpsllvd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: vpor %xmm0, %xmm2, %xmm0
+; AVX2-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
index 21a011abab0..a587a43f1bb 100644
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-128.ll
@@ -279,9 +279,8 @@ define <8 x i16> @var_shift_v8i16(<8 x i16> %a, <8 x i16> %b) nounwind {
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-sub128.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-sub128.ll
index f351bfb2196..b540421fb3d 100644
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-sub128.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-sub128.ll
@@ -213,9 +213,8 @@ define <4 x i16> @var_shift_v4i16(<4 x i16> %a, <4 x i16> %b) nounwind {
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -387,9 +386,8 @@ define <2 x i16> @var_shift_v2i16(<2 x i16> %a, <2 x i16> %b) nounwind {
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
; AVX2-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; AVX2-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
OpenPOWER on IntegriCloud