[llvm] r369434 - [X86] Add a DAG combine to transform (i8 (bitcast (v8i1 (extract_subvector (v16i1 X), 0)))) -> (i8 (trunc (i16 (bitcast (v16i1 X))))) on KNL target
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Tue Aug 20 13:20:04 PDT 2019
Author: ctopper
Date: Tue Aug 20 13:20:04 2019
New Revision: 369434
URL: http://llvm.org/viewvc/llvm-project?rev=369434&view=rev
Log:
[X86] Add a DAG combine to transform (i8 (bitcast (v8i1 (extract_subvector (v16i1 X), 0)))) -> (i8 (trunc (i16 (bitcast (v16i1 X))))) on KNL target
Without AVX512DQ we don't have KMOVB so we can't really copy 8-bits of a k-register to a GPR. We have to copy 16 bits instead. We do this even if the DAG copy is from v8i1->v16i1. If we detect the (i8 (bitcast (v8i1 (extract_subvector (v16i1 X), 0)))) we should rewrite the types to match the copy we do support. By doing this, we can help known bits to propagate without losing the upper 8 bits of the input to the extract_subvector. This allows some zero extends to be removed since we have an isel pattern to use kmovw for (zero_extend (i16 (bitcast (v16i1 X))).
Differential Revision: https://reviews.llvm.org/D66489
Modified:
llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll
llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
llvm/trunk/test/CodeGen/X86/avx512vlcd-intrinsics-fast-isel.ll
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=369434&r1=369433&r2=369434&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Aug 20 13:20:04 2019
@@ -35402,6 +35402,18 @@ static SDValue combineBitcast(SDNode *N,
}
}
+ // Look for (i8 (bitcast (v8i1 (extract_subvector (v16i1 X), 0)))) and
+ // replace with (i8 (trunc (i16 (bitcast (v16i1 X))))). This can occur
+ // due to insert_subvector legalization on KNL. By promoting the copy to i16
+ // we can help with known bits propagation from the vXi1 domain to the
+ // scalar domain.
+ if (VT == MVT::i8 && SrcVT == MVT::v8i1 && Subtarget.hasAVX512() &&
+ !Subtarget.hasDQI() && N0.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
+ N0.getOperand(0).getValueType() == MVT::v16i1 &&
+ isNullConstant(N0.getOperand(1)))
+ return DAG.getNode(ISD::TRUNCATE, SDLoc(N), VT,
+ DAG.getBitcast(MVT::i16, N0.getOperand(0)));
+
// Since MMX types are special and don't usually play with other vector types,
// it's better to handle them early to be sure we emit efficient code by
// avoiding store-load conversions.
Modified: llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll?rev=369434&r1=369433&r2=369434&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll Tue Aug 20 13:20:04 2019
@@ -4557,17 +4557,17 @@ define i8 at test_int_x86_avx512_mask_cmp_s
; X64-NEXT: kmovw %k0, %esi
; X64-NEXT: vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT: kmovw %k0, %eax
-; X64-NEXT: orb %sil, %al
-; X64-NEXT: orb %dl, %al
-; X64-NEXT: orb %cl, %al
+; X64-NEXT: orl %esi, %eax
+; X64-NEXT: orl %edx, %eax
+; X64-NEXT: orl %ecx, %eax
; X64-NEXT: # kill: def $al killed $al killed $eax
; X64-NEXT: retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_sd_all:
; X86: # %bb.0:
-; X86-NEXT: pushl %ebx
+; X86-NEXT: pushl %esi
; X86-NEXT: .cfi_def_cfa_offset 8
-; X86-NEXT: .cfi_offset %ebx, -8
+; X86-NEXT: .cfi_offset %esi, -8
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT: kmovw %eax, %k1
; X86-NEXT: vcmplesd %xmm1, %xmm0, %k0
@@ -4575,14 +4575,14 @@ define i8 at test_int_x86_avx512_mask_cmp_s
; X86-NEXT: vcmpunordsd {sae}, %xmm1, %xmm0, %k0
; X86-NEXT: kmovw %k0, %edx
; X86-NEXT: vcmpneqsd %xmm1, %xmm0, %k0 {%k1}
-; X86-NEXT: kmovw %k0, %ebx
+; X86-NEXT: kmovw %k0, %esi
; X86-NEXT: vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT: kmovw %k0, %eax
-; X86-NEXT: orb %bl, %al
-; X86-NEXT: orb %dl, %al
-; X86-NEXT: orb %cl, %al
+; X86-NEXT: orl %esi, %eax
+; X86-NEXT: orl %edx, %eax
+; X86-NEXT: orl %ecx, %eax
; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %ebx
+; X86-NEXT: popl %esi
; X86-NEXT: .cfi_def_cfa_offset 4
; X86-NEXT: retl
@@ -4634,17 +4634,17 @@ define i8 at test_int_x86_avx512_mask_cmp_s
; X64-NEXT: kmovw %k0, %esi
; X64-NEXT: vcmpnltss {sae}, %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT: kmovw %k0, %eax
-; X64-NEXT: andb %sil, %al
-; X64-NEXT: andb %dl, %al
-; X64-NEXT: andb %cl, %al
+; X64-NEXT: andl %esi, %eax
+; X64-NEXT: andl %edx, %eax
+; X64-NEXT: andl %ecx, %eax
; X64-NEXT: # kill: def $al killed $al killed $eax
; X64-NEXT: retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_ss_all:
; X86: # %bb.0:
-; X86-NEXT: pushl %ebx
+; X86-NEXT: pushl %esi
; X86-NEXT: .cfi_def_cfa_offset 8
-; X86-NEXT: .cfi_offset %ebx, -8
+; X86-NEXT: .cfi_offset %esi, -8
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT: kmovw %eax, %k1
; X86-NEXT: vcmpless %xmm1, %xmm0, %k0
@@ -4652,14 +4652,14 @@ define i8 at test_int_x86_avx512_mask_cmp_s
; X86-NEXT: vcmpunordss {sae}, %xmm1, %xmm0, %k0
; X86-NEXT: kmovw %k0, %edx
; X86-NEXT: vcmpneqss %xmm1, %xmm0, %k0 {%k1}
-; X86-NEXT: kmovw %k0, %ebx
+; X86-NEXT: kmovw %k0, %esi
; X86-NEXT: vcmpnltss {sae}, %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT: kmovw %k0, %eax
-; X86-NEXT: andb %bl, %al
-; X86-NEXT: andb %dl, %al
-; X86-NEXT: andb %cl, %al
+; X86-NEXT: andl %esi, %eax
+; X86-NEXT: andl %edx, %eax
+; X86-NEXT: andl %ecx, %eax
; X86-NEXT: # kill: def $al killed $al killed $eax
-; X86-NEXT: popl %ebx
+; X86-NEXT: popl %esi
; X86-NEXT: .cfi_def_cfa_offset 4
; X86-NEXT: retl
%res1 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 2, i8 -1, i32 4)
Modified: llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll?rev=369434&r1=369433&r2=369434&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll Tue Aug 20 13:20:04 2019
@@ -8345,19 +8345,19 @@ define <8 x i8> @test_mask_cmp_q_256(<4
; X86-LABEL: test_mask_cmp_q_256:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
-; X86-NEXT: vpcmpgtq %ymm0, %ymm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x2a,0x37,0xc8]
-; X86-NEXT: vpcmpleq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xd9,0x02]
-; X86-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe9,0x05]
-; X86-NEXT: vpcmpgtq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x37,0xf1]
-; X86-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X86-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; X86-NEXT: vpcmpgtq %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x37,0xd0]
+; X86-NEXT: vpcmpleq %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd9,0x02]
+; X86-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltq %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x05]
+; X86-NEXT: vpcmpgtq %ymm1, %ymm0, %k6 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x37,0xf1]
+; X86-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X86-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8367,26 +8367,26 @@ define <8 x i8> @test_mask_cmp_q_256(<4
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_q_256:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
-; X64-NEXT: vpcmpgtq %ymm0, %ymm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x2a,0x37,0xc8]
-; X64-NEXT: vpcmpleq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xd9,0x02]
-; X64-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe9,0x05]
-; X64-NEXT: vpcmpgtq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x37,0xf1]
-; X64-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X64-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; X64-NEXT: vpcmpgtq %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x37,0xd0]
+; X64-NEXT: vpcmpleq %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xd9,0x02]
+; X64-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltq %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe9,0x05]
+; X64-NEXT: vpcmpgtq %ymm1, %ymm0, %k6 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x37,0xf1]
+; X64-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X64-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8396,7 +8396,7 @@ define <8 x i8> @test_mask_cmp_q_256(<4
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT: retq # encoding: [0xc3]
@@ -8470,19 +8470,19 @@ define <8 x i8> @test_mask_ucmp_q_256(<4
; X86-LABEL: test_mask_ucmp_q_256:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
-; X86-NEXT: vpcmpltuq %ymm1, %ymm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xc9,0x01]
-; X86-NEXT: vpcmpleuq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xd9,0x02]
-; X86-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltuq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xe9,0x05]
-; X86-NEXT: vpcmpnleuq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xf1,0x06]
-; X86-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X86-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; X86-NEXT: vpcmpltuq %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd1,0x01]
+; X86-NEXT: vpcmpleuq %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd9,0x02]
+; X86-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltuq %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xe9,0x05]
+; X86-NEXT: vpcmpnleuq %ymm1, %ymm0, %k6 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf1,0x06]
+; X86-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X86-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8492,26 +8492,26 @@ define <8 x i8> @test_mask_ucmp_q_256(<4
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_q_256:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
-; X64-NEXT: vpcmpltuq %ymm1, %ymm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xc9,0x01]
-; X64-NEXT: vpcmpleuq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xd9,0x02]
-; X64-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltuq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xe9,0x05]
-; X64-NEXT: vpcmpnleuq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xf1,0x06]
-; X64-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X64-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
+; X64-NEXT: vpcmpltuq %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd1,0x01]
+; X64-NEXT: vpcmpleuq %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xd9,0x02]
+; X64-NEXT: vpcmpneqq %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltuq %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xe9,0x05]
+; X64-NEXT: vpcmpnleuq %ymm1, %ymm0, %k6 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x1e,0xf1,0x06]
+; X64-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X64-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8521,7 +8521,7 @@ define <8 x i8> @test_mask_ucmp_q_256(<4
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT: retq # encoding: [0xc3]
@@ -8594,19 +8594,19 @@ define <8 x i8> @test_mask_cmp_d_128(<4
; X86-LABEL: test_mask_cmp_d_128:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
-; X86-NEXT: vpcmpgtd %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf1,0x75,0x0a,0x66,0xc8]
-; X86-NEXT: vpcmpled %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xd9,0x02]
-; X86-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltd %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe9,0x05]
-; X86-NEXT: vpcmpgtd %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x66,0xf1]
-; X86-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X86-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; X86-NEXT: vpcmpgtd %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x66,0xd0]
+; X86-NEXT: vpcmpled %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd9,0x02]
+; X86-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltd %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x05]
+; X86-NEXT: vpcmpgtd %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x66,0xf1]
+; X86-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X86-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8616,25 +8616,25 @@ define <8 x i8> @test_mask_cmp_d_128(<4
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_d_128:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
-; X64-NEXT: vpcmpgtd %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf1,0x75,0x0a,0x66,0xc8]
-; X64-NEXT: vpcmpled %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xd9,0x02]
-; X64-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltd %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe9,0x05]
-; X64-NEXT: vpcmpgtd %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x66,0xf1]
-; X64-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X64-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; X64-NEXT: vpcmpgtd %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x66,0xd0]
+; X64-NEXT: vpcmpled %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xd9,0x02]
+; X64-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltd %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe9,0x05]
+; X64-NEXT: vpcmpgtd %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x66,0xf1]
+; X64-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X64-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8644,7 +8644,7 @@ define <8 x i8> @test_mask_cmp_d_128(<4
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: retq # encoding: [0xc3]
%res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
@@ -8716,19 +8716,19 @@ define <8 x i8> @test_mask_ucmp_d_128(<4
; X86-LABEL: test_mask_ucmp_d_128:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
-; X86-NEXT: vpcmpltud %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xc9,0x01]
-; X86-NEXT: vpcmpleud %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xd9,0x02]
-; X86-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltud %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xe9,0x05]
-; X86-NEXT: vpcmpnleud %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xf1,0x06]
-; X86-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X86-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; X86-NEXT: vpcmpltud %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd1,0x01]
+; X86-NEXT: vpcmpleud %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd9,0x02]
+; X86-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltud %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xe9,0x05]
+; X86-NEXT: vpcmpnleud %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf1,0x06]
+; X86-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X86-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8738,25 +8738,25 @@ define <8 x i8> @test_mask_ucmp_d_128(<4
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_d_128:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
-; X64-NEXT: vpcmpltud %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xc9,0x01]
-; X64-NEXT: vpcmpleud %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xd9,0x02]
-; X64-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltud %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xe9,0x05]
-; X64-NEXT: vpcmpnleud %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xf1,0x06]
-; X64-NEXT: kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
-; X64-NEXT: kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
+; X64-NEXT: vpcmpltud %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd1,0x01]
+; X64-NEXT: vpcmpleud %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xd9,0x02]
+; X64-NEXT: vpcmpneqd %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltud %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xe9,0x05]
+; X64-NEXT: vpcmpnleud %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1e,0xf1,0x06]
+; X64-NEXT: kshiftlw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0c]
+; X64-NEXT: kshiftrw $12, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0c]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8766,7 +8766,7 @@ define <8 x i8> @test_mask_ucmp_d_128(<4
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: retq # encoding: [0xc3]
%res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
@@ -8838,19 +8838,19 @@ define <8 x i8> @test_mask_cmp_q_128(<2
; X86-LABEL: test_mask_cmp_q_128:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
-; X86-NEXT: vpcmpgtq %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x0a,0x37,0xc8]
-; X86-NEXT: vpcmpleq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xd9,0x02]
-; X86-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe9,0x05]
-; X86-NEXT: vpcmpgtq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x37,0xf1]
-; X86-NEXT: kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
-; X86-NEXT: kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; X86-NEXT: vpcmpgtq %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x37,0xd0]
+; X86-NEXT: vpcmpleq %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd9,0x02]
+; X86-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltq %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x05]
+; X86-NEXT: vpcmpgtq %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x37,0xf1]
+; X86-NEXT: kshiftlw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; X86-NEXT: kshiftrw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8860,25 +8860,25 @@ define <8 x i8> @test_mask_cmp_q_128(<2
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_q_128:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
-; X64-NEXT: vpcmpgtq %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x0a,0x37,0xc8]
-; X64-NEXT: vpcmpleq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xd9,0x02]
-; X64-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe9,0x05]
-; X64-NEXT: vpcmpgtq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x37,0xf1]
-; X64-NEXT: kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
-; X64-NEXT: kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; X64-NEXT: vpcmpgtq %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x37,0xd0]
+; X64-NEXT: vpcmpleq %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xd9,0x02]
+; X64-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltq %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe9,0x05]
+; X64-NEXT: vpcmpgtq %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x37,0xf1]
+; X64-NEXT: kshiftlw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; X64-NEXT: kshiftrw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8888,7 +8888,7 @@ define <8 x i8> @test_mask_cmp_q_128(<2
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: retq # encoding: [0xc3]
%res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
@@ -8960,19 +8960,19 @@ define <8 x i8> @test_mask_ucmp_q_128(<2
; X86-LABEL: test_mask_ucmp_q_128:
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
-; X86-NEXT: kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
-; X86-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
-; X86-NEXT: vpcmpltuq %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xc9,0x01]
-; X86-NEXT: vpcmpleuq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xd9,0x02]
-; X86-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
-; X86-NEXT: vpcmpnltuq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xe9,0x05]
-; X86-NEXT: vpcmpnleuq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xf1,0x06]
-; X86-NEXT: kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
-; X86-NEXT: kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
+; X86-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; X86-NEXT: vpcmpltuq %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd1,0x01]
+; X86-NEXT: vpcmpleuq %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd9,0x02]
+; X86-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe1,0x04]
+; X86-NEXT: vpcmpnltuq %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xe9,0x05]
+; X86-NEXT: vpcmpnleuq %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf1,0x06]
+; X86-NEXT: kshiftlw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; X86-NEXT: kshiftrw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
; X86-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X86-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -8982,25 +8982,25 @@ define <8 x i8> @test_mask_ucmp_q_128(<2
; X86-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X86-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X86-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X86-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_q_128:
; X64: # %bb.0:
-; X64-NEXT: kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
-; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
-; X64-NEXT: vpcmpltuq %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xc9,0x01]
-; X64-NEXT: vpcmpleuq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xd9,0x02]
-; X64-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
-; X64-NEXT: vpcmpnltuq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xe9,0x05]
-; X64-NEXT: vpcmpnleuq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xf1,0x06]
-; X64-NEXT: kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
-; X64-NEXT: kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
+; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
+; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
+; X64-NEXT: vpcmpltuq %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd1,0x01]
+; X64-NEXT: vpcmpleuq %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xd9,0x02]
+; X64-NEXT: vpcmpneqq %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1f,0xe1,0x04]
+; X64-NEXT: vpcmpnltuq %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xe9,0x05]
+; X64-NEXT: vpcmpnleuq %xmm1, %xmm0, %k6 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x1e,0xf1,0x06]
+; X64-NEXT: kshiftlw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x32,0xc9,0x0e]
+; X64-NEXT: kshiftrw $14, %k1, %k1 # encoding: [0xc4,0xe3,0xf9,0x30,0xc9,0x0e]
; X64-NEXT: kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
-; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
+; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
; X64-NEXT: kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
@@ -9010,7 +9010,7 @@ define <8 x i8> @test_mask_ucmp_q_128(<2
; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
; X64-NEXT: kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
-; X64-NEXT: kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
+; X64-NEXT: kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
; X64-NEXT: retq # encoding: [0xc3]
%res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
Modified: llvm/trunk/test/CodeGen/X86/avx512vlcd-intrinsics-fast-isel.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512vlcd-intrinsics-fast-isel.ll?rev=369434&r1=369433&r2=369434&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx512vlcd-intrinsics-fast-isel.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx512vlcd-intrinsics-fast-isel.ll Tue Aug 20 13:20:04 2019
@@ -7,7 +7,6 @@ define <2 x i64> @test_mm_broadcastmb_ep
; X86: # %bb.0: # %entry
; X86-NEXT: vpcmpeqd %xmm1, %xmm0, %k0
; X86-NEXT: kmovw %k0, %eax
-; X86-NEXT: movzbl %al, %eax
; X86-NEXT: vmovd %eax, %xmm0
; X86-NEXT: vpbroadcastq %xmm0, %xmm0
; X86-NEXT: retl
@@ -34,7 +33,6 @@ define <4 x i64> @test_mm256_broadcastmb
; X86: # %bb.0: # %entry
; X86-NEXT: vpcmpeqq %ymm1, %ymm0, %k0
; X86-NEXT: kmovw %k0, %eax
-; X86-NEXT: movzbl %al, %eax
; X86-NEXT: vmovd %eax, %xmm0
; X86-NEXT: vpbroadcastq %xmm0, %ymm0
; X86-NEXT: retl
More information about the llvm-commits
mailing list