[NFC][Codegen][X86] Add AVX2 runline for '(X & (C l>> Y)) ==/!= 0' tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@362606 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll b/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
index 02ded696861cc345a0512791b48b97f319587ffd..88c939f52bee852e48f17ab898adcff16fdfb3c4 100644 (file)
--- a/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
+++ b/test/CodeGen/X86/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll
@@ -1,10 +1,12 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2            < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-NOBMI
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi       < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-BMI,X86-BMI1
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2 < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-BMI,X86-BMI12
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2            < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-NOBMI
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi       < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-BMI,X64-BMI1
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2 < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-BMI,X64-BMI12
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2                  < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI1,X86-BMI1,V0,X86-V0
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi             < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI1,X86-BMI1,V1,X86-V1
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2       < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI2,X86-BMI2,V2,X86-V2
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2,+avx2 < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,BMI2,X86-BMI2,AVX2,X86-AVX2
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2                  < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI1,X64-BMI1,V0,X64-V0
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi             < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI1,X64-BMI1,V1,X64-V1
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2       < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI2,X64-BMI2,V2,X64-V2
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2,+avx2 < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,BMI2,X64-BMI2,AVX2,X64-AVX2
 
 ; We are looking for the following pattern here:
 ;   (X & (C l>> Y)) ==/!= 0
@@ -95,15 +97,6 @@ define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind {
 ; i16 scalar
 
 define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i16_signbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $32768, %eax # imm = 0x8000
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i16_signbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -113,24 +106,14 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i16_signbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $32768, %ecx # imm = 0x8000
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i16_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $32768, %eax # imm = 0x8000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testw %di, %ax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i16_signbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $32768, %ecx # imm = 0x8000
+; X86-BMI2-NEXT:    shrxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testw %ax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i16_signbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -142,13 +125,13 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i16_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $32768, %eax # imm = 0x8000
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testw %di, %ax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i16_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $32768, %eax # imm = 0x8000
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testw %di, %ax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i16 32768, %y
   %t1 = and i16 %t0, %x
   %res = icmp eq i16 %t1, 0
@@ -156,15 +139,6 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
 }
 
 define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i16_lowestbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $1, %eax
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i16_lowestbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -174,24 +148,14 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i16_lowestbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $1, %ecx
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i16_lowestbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $1, %eax
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testw %di, %ax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i16_lowestbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $1, %ecx
+; X86-BMI2-NEXT:    shrxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testw %ax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i16_lowestbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -203,13 +167,13 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i16_lowestbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $1, %eax
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testw %di, %ax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i16_lowestbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $1, %eax
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testw %di, %ax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i16 1, %y
   %t1 = and i16 %t0, %x
   %res = icmp eq i16 %t1, 0
@@ -217,15 +181,6 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind {
 }
 
 define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i16_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i16_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -235,24 +190,14 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i16_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $4080, %ecx # imm = 0xFF0
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i16_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testw %di, %ax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i16_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $4080, %ecx # imm = 0xFF0
+; X86-BMI2-NEXT:    shrxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testw %ax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i16_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -264,13 +209,13 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i16_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testw %di, %ax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i16_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $4080, %eax # imm = 0xFF0
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testw %di, %ax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i16 4080, %y
   %t1 = and i16 %t0, %x
   %res = icmp eq i16 %t1, 0
@@ -280,15 +225,6 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; i32 scalar
 
 define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_signbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_signbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -298,24 +234,14 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_signbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $-2147483648, %ecx # imm = 0x80000000
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testl %edi, %eax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_signbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $-2147483648, %ecx # imm = 0x80000000
+; X86-BMI2-NEXT:    shrxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testl %eax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_signbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -327,13 +253,13 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testl %edi, %eax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testl %edi, %eax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i32 2147483648, %y
   %t1 = and i32 %t0, %x
   %res = icmp eq i32 %t1, 0
@@ -341,15 +267,6 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
 }
 
 define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_lowestbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $1, %eax
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_lowestbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -359,24 +276,14 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_lowestbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $1, %ecx
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_lowestbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $1, %eax
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testl %edi, %eax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_lowestbit_eq:
+; X86-BMI2:       # %bb.0:
+; 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 %eax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_lowestbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -388,13 +295,13 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_lowestbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $1, %eax
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testl %edi, %eax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_lowestbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $1, %eax
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testl %edi, %eax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i32 1, %y
   %t1 = and i32 %t0, %x
   %res = icmp eq i32 %t1, 0
@@ -402,15 +309,6 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind {
 }
 
 define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -420,24 +318,14 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $16776960, %ecx # imm = 0xFFFF00
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testl %edi, %eax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $16776960, %ecx # imm = 0xFFFF00
+; X86-BMI2-NEXT:    shrxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testl %eax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -449,13 +337,13 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X64-BMI12-NEXT:    shrxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testl %edi, %eax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
+; X64-BMI2-NEXT:    shrxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testl %edi, %eax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i32 16776960, %y
   %t1 = and i32 %t0, %x
   %res = icmp eq i32 %t1, 0
@@ -465,25 +353,6 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
 ; i64 scalar
 
 define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i64_signbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    pushl %esi
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X86-NOBMI-NEXT:    xorl %edx, %edx
-; X86-NOBMI-NEXT:    xorl %esi, %esi
-; X86-NOBMI-NEXT:    shrdl %cl, %eax, %esi
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testb $32, %cl
-; X86-NOBMI-NEXT:    cmovnel %eax, %esi
-; X86-NOBMI-NEXT:    cmovnel %edx, %eax
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI-NEXT:    orl %esi, %eax
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    popl %esi
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i64_signbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    pushl %esi
@@ -503,34 +372,24 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
 ; X86-BMI1-NEXT:    popl %esi
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i64_signbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    pushl %esi
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI12-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X86-BMI12-NEXT:    xorl %edx, %edx
-; X86-BMI12-NEXT:    xorl %esi, %esi
-; X86-BMI12-NEXT:    shrdl %cl, %eax, %esi
-; X86-BMI12-NEXT:    shrxl %ecx, %eax, %eax
-; X86-BMI12-NEXT:    testb $32, %cl
-; X86-BMI12-NEXT:    cmovnel %eax, %esi
-; X86-BMI12-NEXT:    cmovnel %edx, %eax
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X86-BMI12-NEXT:    orl %esi, %eax
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    popl %esi
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i64_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movq %rsi, %rcx
-; X64-NOBMI-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-NOBMI-NEXT:    shrq %cl, %rax
-; X64-NOBMI-NEXT:    testq %rdi, %rax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i64_signbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    pushl %esi
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI2-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
+; X86-BMI2-NEXT:    xorl %edx, %edx
+; X86-BMI2-NEXT:    xorl %esi, %esi
+; X86-BMI2-NEXT:    shrdl %cl, %eax, %esi
+; X86-BMI2-NEXT:    shrxl %ecx, %eax, %eax
+; X86-BMI2-NEXT:    testb $32, %cl
+; X86-BMI2-NEXT:    cmovnel %eax, %esi
+; X86-BMI2-NEXT:    cmovnel %edx, %eax
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %esi
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-BMI2-NEXT:    orl %esi, %eax
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    popl %esi
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i64_signbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -542,13 +401,13 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i64_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
-; X64-BMI12-NEXT:    shrxq %rsi, %rax, %rax
-; X64-BMI12-NEXT:    testq %rdi, %rax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i64_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
+; X64-BMI2-NEXT:    shrxq %rsi, %rax, %rax
+; X64-BMI2-NEXT:    testq %rdi, %rax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i64 9223372036854775808, %y
   %t1 = and i64 %t0, %x
   %res = icmp eq i64 %t1, 0
@@ -569,16 +428,6 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
 ; X86-NEXT:    sete %al
 ; X86-NEXT:    retl
 ;
-; X64-NOBMI-LABEL: scalar_i64_lowestbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movq %rsi, %rcx
-; X64-NOBMI-NEXT:    movl $1, %eax
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-NOBMI-NEXT:    shrq %cl, %rax
-; X64-NOBMI-NEXT:    testq %rdi, %rax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
-;
 ; X64-BMI1-LABEL: scalar_i64_lowestbit_eq:
 ; X64-BMI1:       # %bb.0:
 ; X64-BMI1-NEXT:    movq %rsi, %rcx
@@ -589,13 +438,13 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i64_lowestbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $1, %eax
-; X64-BMI12-NEXT:    shrxq %rsi, %rax, %rax
-; X64-BMI12-NEXT:    testq %rdi, %rax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i64_lowestbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $1, %eax
+; X64-BMI2-NEXT:    shrxq %rsi, %rax, %rax
+; X64-BMI2-NEXT:    testq %rdi, %rax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i64 1, %y
   %t1 = and i64 %t0, %x
   %res = icmp eq i64 %t1, 0
@@ -603,25 +452,6 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
 }
 
 define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i64_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    pushl %esi
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $65535, %eax # imm = 0xFFFF
-; X86-NOBMI-NEXT:    movl $-65536, %edx # imm = 0xFFFF0000
-; X86-NOBMI-NEXT:    shrdl %cl, %eax, %edx
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    xorl %esi, %esi
-; X86-NOBMI-NEXT:    testb $32, %cl
-; X86-NOBMI-NEXT:    cmovnel %eax, %edx
-; X86-NOBMI-NEXT:    cmovel %eax, %esi
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %edx
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-NOBMI-NEXT:    orl %edx, %esi
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    popl %esi
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i64_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    pushl %esi
@@ -641,34 +471,24 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ; X86-BMI1-NEXT:    popl %esi
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i64_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    pushl %esi
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI12-NEXT:    movl $65535, %eax # imm = 0xFFFF
-; X86-BMI12-NEXT:    movl $-65536, %edx # imm = 0xFFFF0000
-; X86-BMI12-NEXT:    shrdl %cl, %eax, %edx
-; X86-BMI12-NEXT:    shrxl %ecx, %eax, %eax
-; X86-BMI12-NEXT:    xorl %esi, %esi
-; X86-BMI12-NEXT:    testb $32, %cl
-; X86-BMI12-NEXT:    cmovnel %eax, %edx
-; X86-BMI12-NEXT:    cmovel %eax, %esi
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %edx
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-BMI12-NEXT:    orl %edx, %esi
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    popl %esi
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i64_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movq %rsi, %rcx
-; X64-NOBMI-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-NOBMI-NEXT:    shrq %cl, %rax
-; X64-NOBMI-NEXT:    testq %rdi, %rax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i64_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    pushl %esi
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI2-NEXT:    movl $65535, %eax # imm = 0xFFFF
+; X86-BMI2-NEXT:    movl $-65536, %edx # imm = 0xFFFF0000
+; X86-BMI2-NEXT:    shrdl %cl, %eax, %edx
+; X86-BMI2-NEXT:    shrxl %ecx, %eax, %eax
+; X86-BMI2-NEXT:    xorl %esi, %esi
+; X86-BMI2-NEXT:    testb $32, %cl
+; X86-BMI2-NEXT:    cmovnel %eax, %edx
+; X86-BMI2-NEXT:    cmovel %eax, %esi
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %esi
+; X86-BMI2-NEXT:    orl %edx, %esi
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    popl %esi
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i64_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -680,13 +500,13 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i64_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
-; X64-BMI12-NEXT:    shrxq %rsi, %rax, %rax
-; X64-BMI12-NEXT:    testq %rdi, %rax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i64_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
+; X64-BMI2-NEXT:    shrxq %rsi, %rax, %rax
+; X64-BMI2-NEXT:    testq %rdi, %rax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i64 281474976645120, %y
   %t1 = and i64 %t0, %x
   %res = icmp eq i64 %t1, 0
@@ -698,28 +518,37 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ;------------------------------------------------------------------------------;
 
 define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; CHECK-LABEL: vec_4xi32_splat_eq:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [1,1,1,1]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm5
-; CHECK-NEXT:    psrld %xmm2, %xmm5
-; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    psrld %xmm1, %xmm3
-; CHECK-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; CHECK-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; CHECK-NEXT:    andps %xmm5, %xmm0
-; CHECK-NEXT:    pxor %xmm1, %xmm1
-; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-NEXT:    ret{{[l|q]}}
+; SSE2-LABEL: vec_4xi32_splat_eq:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,1,1,1]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    psrld %xmm2, %xmm5
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    psrld %xmm1, %xmm3
+; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE2-NEXT:    andps %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: vec_4xi32_splat_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsrlvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
   %t0 = lshr <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
@@ -727,28 +556,37 @@ define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 }
 
 define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; CHECK-LABEL: vec_4xi32_nonsplat_eq:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,16776960,2147483648]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm5
-; CHECK-NEXT:    psrld %xmm2, %xmm5
-; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    psrld %xmm1, %xmm3
-; CHECK-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; CHECK-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; CHECK-NEXT:    andps %xmm5, %xmm0
-; CHECK-NEXT:    pxor %xmm1, %xmm1
-; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-NEXT:    ret{{[l|q]}}
+; SSE2-LABEL: vec_4xi32_nonsplat_eq:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,16776960,2147483648]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    psrld %xmm2, %xmm5
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    psrld %xmm1, %xmm3
+; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE2-NEXT:    andps %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
+; AVX2-NEXT:    vpsrlvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
   %t0 = lshr <4 x i32> <i32 0, i32 1, i32 16776960, i32 2147483648>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
@@ -756,84 +594,111 @@ define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 }
 
 define <4 x i1> @vec_4xi32_nonsplat_undef0_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; CHECK-LABEL: vec_4xi32_nonsplat_undef0_eq:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = <1,1,u,1>
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm5
-; CHECK-NEXT:    psrld %xmm2, %xmm5
-; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    psrld %xmm1, %xmm3
-; CHECK-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; CHECK-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; CHECK-NEXT:    andps %xmm5, %xmm0
-; CHECK-NEXT:    pxor %xmm1, %xmm1
-; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-NEXT:    ret{{[l|q]}}
+; SSE2-LABEL: vec_4xi32_nonsplat_undef0_eq:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = <1,1,u,1>
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    psrld %xmm2, %xmm5
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    psrld %xmm1, %xmm3
+; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE2-NEXT:    andps %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef0_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsrlvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
   %t0 = lshr <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %res
 }
 define <4 x i1> @vec_4xi32_nonsplat_undef1_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; CHECK-LABEL: vec_4xi32_nonsplat_undef1_eq:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = [1,1,1,1]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm5
-; CHECK-NEXT:    psrld %xmm2, %xmm5
-; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    psrld %xmm1, %xmm3
-; CHECK-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; CHECK-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; CHECK-NEXT:    andps %xmm5, %xmm0
-; CHECK-NEXT:    pxor %xmm1, %xmm1
-; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-NEXT:    ret{{[l|q]}}
+; SSE2-LABEL: vec_4xi32_nonsplat_undef1_eq:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,1,1,1]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    psrld %xmm2, %xmm5
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    psrld %xmm1, %xmm3
+; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE2-NEXT:    andps %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef1_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsrlvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
   %t0 = lshr <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
   ret <4 x i1> %res
 }
 define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; CHECK-LABEL: vec_4xi32_nonsplat_undef2_eq:
-; CHECK:       # %bb.0:
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa {{.*#+}} xmm3 = <1,1,u,1>
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm5
-; CHECK-NEXT:    psrld %xmm2, %xmm5
-; CHECK-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
-; CHECK-NEXT:    movdqa %xmm3, %xmm4
-; CHECK-NEXT:    psrld %xmm2, %xmm4
-; CHECK-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
-; CHECK-NEXT:    psrld %xmm1, %xmm3
-; CHECK-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
-; CHECK-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; CHECK-NEXT:    andps %xmm5, %xmm0
-; CHECK-NEXT:    pxor %xmm1, %xmm1
-; CHECK-NEXT:    pcmpeqd %xmm1, %xmm0
-; CHECK-NEXT:    ret{{[l|q]}}
+; SSE2-LABEL: vec_4xi32_nonsplat_undef2_eq:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = <1,1,u,1>
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    psrld %xmm2, %xmm5
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm4[0]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
+; SSE2-NEXT:    movdqa %xmm3, %xmm4
+; SSE2-NEXT:    psrld %xmm2, %xmm4
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
+; SSE2-NEXT:    psrld %xmm1, %xmm3
+; SSE2-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
+; SSE2-NEXT:    andps %xmm5, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    ret{{[l|q]}}
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef2_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsrlvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
   %t0 = lshr <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
@@ -894,15 +759,6 @@ define i1 @scalar_i32_x_is_const_eq(i32 %y) nounwind {
   ret i1 %res
 }
 define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_x_is_const2_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $1, %eax
-; X86-NOBMI-NEXT:    shrl %cl, %eax
-; X86-NOBMI-NEXT:    testl $-1437226411, %eax # imm = 0xAA55AA55
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_x_is_const2_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -912,24 +768,14 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_x_is_const2_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $1, %ecx
-; X86-BMI12-NEXT:    shrxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl $-1437226411, %eax # imm = 0xAA55AA55
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_x_is_const2_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %edi, %ecx
-; X64-NOBMI-NEXT:    movl $1, %eax
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shrl %cl, %eax
-; X64-NOBMI-NEXT:    testl $-1437226411, %eax # imm = 0xAA55AA55
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_x_is_const2_eq:
+; X86-BMI2:       # %bb.0:
+; 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:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_x_is_const2_eq:
 ; X64-BMI1:       # %bb.0:
@@ -941,13 +787,13 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_x_is_const2_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $1, %eax
-; X64-BMI12-NEXT:    shrxl %edi, %eax, %eax
-; X64-BMI12-NEXT:    testl $-1437226411, %eax # imm = 0xAA55AA55
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_x_is_const2_eq:
+; 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:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = lshr i32 1, %y
   %t1 = and i32 %t0, 2857740885
   %res = icmp eq i32 %t1, 0

diff --git a/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll b/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll
index a1b01be7cf6b64bdc578ca5f9721c947999fa8d2..aa4a135fd681f571ac909448d77b7f6cd3c16807 100644 (file)
--- a/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll
+++ b/test/CodeGen/X86/hoist-and-by-const-from-shl-in-eqcmp-zero.ll
@@ -1,10 +1,12 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2            < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-NOBMI
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi       < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-BMI,X86-BMI1
-; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2 < %s | FileCheck %s --check-prefixes=CHECK,X86,X86-BMI,X86-BMI12
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2            < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-NOBMI
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi       < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-BMI,X64-BMI1
-; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2 < %s | FileCheck %s --check-prefixes=CHECK,X64,X64-BMI,X64-BMI12
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2                  < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI1,X86-BMI1,V0,X86-V0
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi             < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI1,X86-BMI1,V1,X86-V1
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2       < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,SSE2,X86-SSE2,BMI2,X86-BMI2,V2,X86-V2
+; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2,+avx2 < %s | FileCheck %s --check-prefixes=CHECK,X86,V0123,X86-V0123,BMI2,X86-BMI2,AVX2,X86-AVX2
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2                  < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI1,X64-BMI1,V0,X64-V0
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi             < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI1,X64-BMI1,V1,X64-V1
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2       < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,SSE2,X64-SSE2,BMI2,X64-BMI2,V2,X64-V2
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+sse,sse2,+bmi,+bmi2,+avx2 < %s | FileCheck %s --check-prefixes=CHECK,X64,V0123,X64-V0123,BMI2,X64-BMI2,AVX2,X64-AVX2
 
 ; We are looking for the following pattern here:
 ;   (X & (C << Y)) ==/!= 0
@@ -90,15 +92,6 @@ define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind {
 ; i16 scalar
 
 define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i16_signbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-32768, %eax # imm = 0x8000
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i16_signbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -108,24 +101,14 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i16_signbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $-32768, %ecx # imm = 0x8000
-; X86-BMI12-NEXT:    shlxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i16_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $-32768, %eax # imm = 0x8000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shll %cl, %eax
-; X64-NOBMI-NEXT:    testw %di, %ax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i16_signbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $-32768, %ecx # imm = 0x8000
+; X86-BMI2-NEXT:    shlxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testw %ax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i16_signbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -137,13 +120,13 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i16_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $-32768, %eax # imm = 0x8000
-; X64-BMI12-NEXT:    shlxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testw %di, %ax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i16_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $-32768, %eax # imm = 0x8000
+; X64-BMI2-NEXT:    shlxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testw %di, %ax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i16 32768, %y
   %t1 = and i16 %t0, %x
   %res = icmp eq i16 %t1, 0
@@ -171,15 +154,6 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind {
 }
 
 define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i16_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i16_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -189,24 +163,14 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i16_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $4080, %ecx # imm = 0xFF0
-; X86-BMI12-NEXT:    shlxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testw %ax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i16_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shll %cl, %eax
-; X64-NOBMI-NEXT:    testw %di, %ax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i16_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $4080, %ecx # imm = 0xFF0
+; X86-BMI2-NEXT:    shlxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testw %ax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i16_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -218,13 +182,13 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i16_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $4080, %eax # imm = 0xFF0
-; X64-BMI12-NEXT:    shlxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testw %di, %ax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i16_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $4080, %eax # imm = 0xFF0
+; X64-BMI2-NEXT:    shlxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testw %di, %ax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i16 4080, %y
   %t1 = and i16 %t0, %x
   %res = icmp eq i16 %t1, 0
@@ -234,15 +198,6 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind {
 ; i32 scalar
 
 define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_signbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_signbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -252,24 +207,14 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_signbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $-2147483648, %ecx # imm = 0x80000000
-; X86-BMI12-NEXT:    shlxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shll %cl, %eax
-; X64-NOBMI-NEXT:    testl %edi, %eax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_signbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $-2147483648, %ecx # imm = 0x80000000
+; X86-BMI2-NEXT:    shlxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testl %eax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_signbit_eq:
 ; X64-BMI1:       # %bb.0:
@@ -281,13 +226,13 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
-; X64-BMI12-NEXT:    shlxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testl %edi, %eax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $-2147483648, %eax # imm = 0x80000000
+; X64-BMI2-NEXT:    shlxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testl %edi, %eax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i32 2147483648, %y
   %t1 = and i32 %t0, %x
   %res = icmp eq i32 %t1, 0
@@ -315,15 +260,6 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind {
 }
 
 define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -333,24 +269,14 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $16776960, %ecx # imm = 0xFFFF00
-; X86-BMI12-NEXT:    shlxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testl %eax, {{[0-9]+}}(%esp)
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %esi, %ecx
-; X64-NOBMI-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shll %cl, %eax
-; X64-NOBMI-NEXT:    testl %edi, %eax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $16776960, %ecx # imm = 0xFFFF00
+; X86-BMI2-NEXT:    shlxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testl %eax, {{[0-9]+}}(%esp)
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -362,13 +288,13 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
-; X64-BMI12-NEXT:    shlxl %esi, %eax, %eax
-; X64-BMI12-NEXT:    testl %edi, %eax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $16776960, %eax # imm = 0xFFFF00
+; X64-BMI2-NEXT:    shlxl %esi, %eax, %eax
+; X64-BMI2-NEXT:    testl %edi, %eax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i32 16776960, %y
   %t1 = and i32 %t0, %x
   %res = icmp eq i32 %t1, 0
@@ -391,16 +317,6 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
 ; X86-NEXT:    sete %al
 ; X86-NEXT:    retl
 ;
-; X64-NOBMI-LABEL: scalar_i64_signbit_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movq %rsi, %rcx
-; X64-NOBMI-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-NOBMI-NEXT:    shlq %cl, %rax
-; X64-NOBMI-NEXT:    testq %rdi, %rax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
-;
 ; X64-BMI1-LABEL: scalar_i64_signbit_eq:
 ; X64-BMI1:       # %bb.0:
 ; X64-BMI1-NEXT:    movq %rsi, %rcx
@@ -411,13 +327,13 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i64_signbit_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
-; X64-BMI12-NEXT:    shlxq %rsi, %rax, %rax
-; X64-BMI12-NEXT:    testq %rdi, %rax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i64_signbit_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movabsq $-9223372036854775808, %rax # imm = 0x8000000000000000
+; X64-BMI2-NEXT:    shlxq %rsi, %rax, %rax
+; X64-BMI2-NEXT:    testq %rdi, %rax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i64 9223372036854775808, %y
   %t1 = and i64 %t0, %x
   %res = icmp eq i64 %t1, 0
@@ -425,25 +341,6 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind {
 }
 
 define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i64_lowestbit_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    pushl %esi
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $1, %eax
-; X86-NOBMI-NEXT:    xorl %edx, %edx
-; X86-NOBMI-NEXT:    xorl %esi, %esi
-; X86-NOBMI-NEXT:    shldl %cl, %eax, %esi
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testb $32, %cl
-; X86-NOBMI-NEXT:    cmovnel %eax, %esi
-; X86-NOBMI-NEXT:    cmovnel %edx, %eax
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X86-NOBMI-NEXT:    orl %esi, %eax
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    popl %esi
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i64_lowestbit_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    pushl %esi
@@ -463,24 +360,24 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
 ; X86-BMI1-NEXT:    popl %esi
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i64_lowestbit_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    pushl %esi
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI12-NEXT:    movl $1, %eax
-; X86-BMI12-NEXT:    xorl %edx, %edx
-; X86-BMI12-NEXT:    xorl %esi, %esi
-; X86-BMI12-NEXT:    shldl %cl, %eax, %esi
-; X86-BMI12-NEXT:    shlxl %ecx, %eax, %eax
-; X86-BMI12-NEXT:    testb $32, %cl
-; X86-BMI12-NEXT:    cmovnel %eax, %esi
-; X86-BMI12-NEXT:    cmovnel %edx, %eax
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %eax
-; X86-BMI12-NEXT:    orl %esi, %eax
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    popl %esi
-; X86-BMI12-NEXT:    retl
+; X86-BMI2-LABEL: scalar_i64_lowestbit_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    pushl %esi
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI2-NEXT:    movl $1, %eax
+; X86-BMI2-NEXT:    xorl %edx, %edx
+; X86-BMI2-NEXT:    xorl %esi, %esi
+; X86-BMI2-NEXT:    shldl %cl, %eax, %esi
+; X86-BMI2-NEXT:    shlxl %ecx, %eax, %eax
+; X86-BMI2-NEXT:    testb $32, %cl
+; X86-BMI2-NEXT:    cmovnel %eax, %esi
+; X86-BMI2-NEXT:    cmovnel %edx, %eax
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %esi
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-BMI2-NEXT:    orl %esi, %eax
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    popl %esi
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-LABEL: scalar_i64_lowestbit_eq:
 ; X64:       # %bb.0:
@@ -494,25 +391,6 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind {
 }
 
 define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i64_bitsinmiddle_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    pushl %esi
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-65536, %eax # imm = 0xFFFF0000
-; X86-NOBMI-NEXT:    movl $65535, %edx # imm = 0xFFFF
-; X86-NOBMI-NEXT:    shldl %cl, %eax, %edx
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    xorl %esi, %esi
-; X86-NOBMI-NEXT:    testb $32, %cl
-; X86-NOBMI-NEXT:    cmovnel %eax, %edx
-; X86-NOBMI-NEXT:    cmovel %eax, %esi
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %edx
-; X86-NOBMI-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-NOBMI-NEXT:    orl %edx, %esi
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    popl %esi
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i64_bitsinmiddle_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    pushl %esi
@@ -532,34 +410,24 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ; X86-BMI1-NEXT:    popl %esi
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i64_bitsinmiddle_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    pushl %esi
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI12-NEXT:    movl $-65536, %eax # imm = 0xFFFF0000
-; X86-BMI12-NEXT:    movl $65535, %edx # imm = 0xFFFF
-; X86-BMI12-NEXT:    shldl %cl, %eax, %edx
-; X86-BMI12-NEXT:    shlxl %ecx, %eax, %eax
-; X86-BMI12-NEXT:    xorl %esi, %esi
-; X86-BMI12-NEXT:    testb $32, %cl
-; X86-BMI12-NEXT:    cmovnel %eax, %edx
-; X86-BMI12-NEXT:    cmovel %eax, %esi
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %edx
-; X86-BMI12-NEXT:    andl {{[0-9]+}}(%esp), %esi
-; X86-BMI12-NEXT:    orl %edx, %esi
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    popl %esi
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i64_bitsinmiddle_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movq %rsi, %rcx
-; X64-NOBMI-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-NOBMI-NEXT:    shlq %cl, %rax
-; X64-NOBMI-NEXT:    testq %rdi, %rax
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i64_bitsinmiddle_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    pushl %esi
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI2-NEXT:    movl $-65536, %eax # imm = 0xFFFF0000
+; X86-BMI2-NEXT:    movl $65535, %edx # imm = 0xFFFF
+; X86-BMI2-NEXT:    shldl %cl, %eax, %edx
+; X86-BMI2-NEXT:    shlxl %ecx, %eax, %eax
+; X86-BMI2-NEXT:    xorl %esi, %esi
+; X86-BMI2-NEXT:    testb $32, %cl
+; X86-BMI2-NEXT:    cmovnel %eax, %edx
+; X86-BMI2-NEXT:    cmovel %eax, %esi
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-BMI2-NEXT:    andl {{[0-9]+}}(%esp), %esi
+; X86-BMI2-NEXT:    orl %edx, %esi
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    popl %esi
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i64_bitsinmiddle_eq:
 ; X64-BMI1:       # %bb.0:
@@ -571,13 +439,13 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i64_bitsinmiddle_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
-; X64-BMI12-NEXT:    shlxq %rsi, %rax, %rax
-; X64-BMI12-NEXT:    testq %rdi, %rax
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i64_bitsinmiddle_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movabsq $281474976645120, %rax # imm = 0xFFFFFFFF0000
+; X64-BMI2-NEXT:    shlxq %rsi, %rax, %rax
+; X64-BMI2-NEXT:    testq %rdi, %rax
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i64 281474976645120, %y
   %t1 = and i64 %t0, %x
   %res = icmp eq i64 %t1, 0
@@ -589,39 +457,48 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind {
 ;------------------------------------------------------------------------------;
 
 define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; X86-LABEL: vec_4xi32_splat_eq:
-; X86:       # %bb.0:
-; X86-NEXT:    pslld $23, %xmm1
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    cvttps2dq %xmm1, %xmm1
-; X86-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X86-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm2, %xmm3
-; X86-NEXT:    pmuludq %xmm1, %xmm2
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X86-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X86-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X86-NEXT:    pand %xmm1, %xmm0
-; X86-NEXT:    pxor %xmm1, %xmm1
-; X86-NEXT:    pcmpeqd %xmm1, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: vec_4xi32_splat_eq:
-; X64:       # %bb.0:
-; X64-NEXT:    pslld $23, %xmm1
-; X64-NEXT:    paddd {{.*}}(%rip), %xmm1
-; X64-NEXT:    cvttps2dq %xmm1, %xmm1
-; X64-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm2, %xmm3
-; X64-NEXT:    pmuludq %xmm1, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X64-NEXT:    pand %xmm1, %xmm0
-; X64-NEXT:    pxor %xmm1, %xmm1
-; X64-NEXT:    pcmpeqd %xmm1, %xmm0
-; X64-NEXT:    retq
+; X86-SSE2-LABEL: vec_4xi32_splat_eq:
+; X86-SSE2:       # %bb.0:
+; X86-SSE2-NEXT:    pslld $23, %xmm1
+; X86-SSE2-NEXT:    paddd {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X86-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X86-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE2-NEXT:    pand %xmm1, %xmm0
+; X86-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X86-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X86-SSE2-NEXT:    retl
+;
+; AVX2-LABEL: vec_4xi32_splat_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsllvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; X64-SSE2-LABEL: vec_4xi32_splat_eq:
+; X64-SSE2:       # %bb.0:
+; X64-SSE2-NEXT:    pslld $23, %xmm1
+; X64-SSE2-NEXT:    paddd {{.*}}(%rip), %xmm1
+; X64-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X64-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X64-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X64-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-SSE2-NEXT:    pand %xmm1, %xmm0
+; X64-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X64-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X64-SSE2-NEXT:    retq
   %t0 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
@@ -629,41 +506,50 @@ define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 }
 
 define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; X86-LABEL: vec_4xi32_nonsplat_eq:
-; X86:       # %bb.0:
-; X86-NEXT:    pslld $23, %xmm1
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    cvttps2dq %xmm1, %xmm1
-; X86-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
-; X86-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm1, %xmm2
-; X86-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm3, %xmm1
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; X86-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; X86-NEXT:    pand %xmm2, %xmm0
-; X86-NEXT:    pxor %xmm1, %xmm1
-; X86-NEXT:    pcmpeqd %xmm1, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: vec_4xi32_nonsplat_eq:
-; X64:       # %bb.0:
-; X64-NEXT:    pslld $23, %xmm1
-; X64-NEXT:    paddd {{.*}}(%rip), %xmm1
-; X64-NEXT:    cvttps2dq %xmm1, %xmm1
-; X64-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm1, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm3, %xmm1
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; X64-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; X64-NEXT:    pand %xmm2, %xmm0
-; X64-NEXT:    pxor %xmm1, %xmm1
-; X64-NEXT:    pcmpeqd %xmm1, %xmm0
-; X64-NEXT:    retq
+; X86-SSE2-LABEL: vec_4xi32_nonsplat_eq:
+; X86-SSE2:       # %bb.0:
+; X86-SSE2-NEXT:    pslld $23, %xmm1
+; X86-SSE2-NEXT:    paddd {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm3, %xmm1
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X86-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X86-SSE2-NEXT:    pand %xmm2, %xmm0
+; X86-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X86-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X86-SSE2-NEXT:    retl
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
+; AVX2-NEXT:    vpsllvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; X64-SSE2-LABEL: vec_4xi32_nonsplat_eq:
+; X64-SSE2:       # %bb.0:
+; X64-SSE2-NEXT:    pslld $23, %xmm1
+; X64-SSE2-NEXT:    paddd {{.*}}(%rip), %xmm1
+; X64-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X64-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [0,1,16776960,2147483648]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm3, %xmm1
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; X64-SSE2-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X64-SSE2-NEXT:    pand %xmm2, %xmm0
+; X64-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X64-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X64-SSE2-NEXT:    retq
   %t0 = shl <4 x i32> <i32 0, i32 1, i32 16776960, i32 2147483648>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
@@ -671,117 +557,144 @@ define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
 }
 
 define <4 x i1> @vec_4xi32_nonsplat_undef0_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; X86-LABEL: vec_4xi32_nonsplat_undef0_eq:
-; X86:       # %bb.0:
-; X86-NEXT:    pslld $23, %xmm1
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    cvttps2dq %xmm1, %xmm1
-; X86-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X86-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm2, %xmm3
-; X86-NEXT:    pmuludq %xmm1, %xmm2
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X86-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X86-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X86-NEXT:    pand %xmm1, %xmm0
-; X86-NEXT:    pxor %xmm1, %xmm1
-; X86-NEXT:    pcmpeqd %xmm1, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: vec_4xi32_nonsplat_undef0_eq:
-; X64:       # %bb.0:
-; X64-NEXT:    pslld $23, %xmm1
-; X64-NEXT:    paddd {{.*}}(%rip), %xmm1
-; X64-NEXT:    cvttps2dq %xmm1, %xmm1
-; X64-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm2, %xmm3
-; X64-NEXT:    pmuludq %xmm1, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X64-NEXT:    pand %xmm1, %xmm0
-; X64-NEXT:    pxor %xmm1, %xmm1
-; X64-NEXT:    pcmpeqd %xmm1, %xmm0
-; X64-NEXT:    retq
+; X86-SSE2-LABEL: vec_4xi32_nonsplat_undef0_eq:
+; X86-SSE2:       # %bb.0:
+; X86-SSE2-NEXT:    pslld $23, %xmm1
+; X86-SSE2-NEXT:    paddd {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X86-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X86-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE2-NEXT:    pand %xmm1, %xmm0
+; X86-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X86-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X86-SSE2-NEXT:    retl
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef0_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsllvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; X64-SSE2-LABEL: vec_4xi32_nonsplat_undef0_eq:
+; X64-SSE2:       # %bb.0:
+; X64-SSE2-NEXT:    pslld $23, %xmm1
+; X64-SSE2-NEXT:    paddd {{.*}}(%rip), %xmm1
+; X64-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X64-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X64-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X64-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-SSE2-NEXT:    pand %xmm1, %xmm0
+; X64-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X64-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X64-SSE2-NEXT:    retq
   %t0 = shl <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0>
   ret <4 x i1> %res
 }
 define <4 x i1> @vec_4xi32_nonsplat_undef1_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; X86-LABEL: vec_4xi32_nonsplat_undef1_eq:
-; X86:       # %bb.0:
-; X86-NEXT:    pslld $23, %xmm1
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    cvttps2dq %xmm1, %xmm1
-; X86-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X86-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm2, %xmm3
-; X86-NEXT:    pmuludq %xmm1, %xmm2
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X86-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X86-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X86-NEXT:    pand %xmm1, %xmm0
-; X86-NEXT:    pxor %xmm1, %xmm1
-; X86-NEXT:    pcmpeqd %xmm1, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: vec_4xi32_nonsplat_undef1_eq:
-; X64:       # %bb.0:
-; X64-NEXT:    pslld $23, %xmm1
-; X64-NEXT:    paddd {{.*}}(%rip), %xmm1
-; X64-NEXT:    cvttps2dq %xmm1, %xmm1
-; X64-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm2, %xmm3
-; X64-NEXT:    pmuludq %xmm1, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X64-NEXT:    pand %xmm1, %xmm0
-; X64-NEXT:    pxor %xmm1, %xmm1
-; X64-NEXT:    pcmpeqd %xmm1, %xmm0
-; X64-NEXT:    retq
+; X86-SSE2-LABEL: vec_4xi32_nonsplat_undef1_eq:
+; X86-SSE2:       # %bb.0:
+; X86-SSE2-NEXT:    pslld $23, %xmm1
+; X86-SSE2-NEXT:    paddd {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X86-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X86-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE2-NEXT:    pand %xmm1, %xmm0
+; X86-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X86-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X86-SSE2-NEXT:    retl
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef1_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsllvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; X64-SSE2-LABEL: vec_4xi32_nonsplat_undef1_eq:
+; X64-SSE2:       # %bb.0:
+; X64-SSE2-NEXT:    pslld $23, %xmm1
+; X64-SSE2-NEXT:    paddd {{.*}}(%rip), %xmm1
+; X64-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X64-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X64-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X64-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-SSE2-NEXT:    pand %xmm1, %xmm0
+; X64-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X64-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X64-SSE2-NEXT:    retq
   %t0 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
   ret <4 x i1> %res
 }
 define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwind {
-; X86-LABEL: vec_4xi32_nonsplat_undef2_eq:
-; X86:       # %bb.0:
-; X86-NEXT:    pslld $23, %xmm1
-; X86-NEXT:    paddd {{\.LCPI.*}}, %xmm1
-; X86-NEXT:    cvttps2dq %xmm1, %xmm1
-; X86-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X86-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X86-NEXT:    pmuludq %xmm2, %xmm3
-; X86-NEXT:    pmuludq %xmm1, %xmm2
-; X86-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X86-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X86-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X86-NEXT:    pand %xmm1, %xmm0
-; X86-NEXT:    pxor %xmm1, %xmm1
-; X86-NEXT:    pcmpeqd %xmm1, %xmm0
-; X86-NEXT:    retl
-;
-; X64-LABEL: vec_4xi32_nonsplat_undef2_eq:
-; X64:       # %bb.0:
-; X64-NEXT:    pslld $23, %xmm1
-; X64-NEXT:    paddd {{.*}}(%rip), %xmm1
-; X64-NEXT:    cvttps2dq %xmm1, %xmm1
-; X64-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
-; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
-; X64-NEXT:    pmuludq %xmm2, %xmm3
-; X64-NEXT:    pmuludq %xmm1, %xmm2
-; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
-; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; X64-NEXT:    pand %xmm1, %xmm0
-; X64-NEXT:    pxor %xmm1, %xmm1
-; X64-NEXT:    pcmpeqd %xmm1, %xmm0
-; X64-NEXT:    retq
+; X86-SSE2-LABEL: vec_4xi32_nonsplat_undef2_eq:
+; X86-SSE2:       # %bb.0:
+; X86-SSE2-NEXT:    pslld $23, %xmm1
+; X86-SSE2-NEXT:    paddd {{\.LCPI.*}}, %xmm1
+; X86-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X86-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X86-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X86-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE2-NEXT:    pand %xmm1, %xmm0
+; X86-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X86-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X86-SSE2-NEXT:    retl
+;
+; AVX2-LABEL: vec_4xi32_nonsplat_undef2_eq:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
+; AVX2-NEXT:    vpsllvd %xmm1, %xmm2, %xmm1
+; AVX2-NEXT:    vpand %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    ret{{[l|q]}}
+;
+; X64-SSE2-LABEL: vec_4xi32_nonsplat_undef2_eq:
+; X64-SSE2:       # %bb.0:
+; X64-SSE2-NEXT:    pslld $23, %xmm1
+; X64-SSE2-NEXT:    paddd {{.*}}(%rip), %xmm1
+; X64-SSE2-NEXT:    cvttps2dq %xmm1, %xmm1
+; X64-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [1,1,1,1]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,3,3]
+; X64-SSE2-NEXT:    pmuludq %xmm2, %xmm3
+; X64-SSE2-NEXT:    pmuludq %xmm1, %xmm2
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
+; X64-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; X64-SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X64-SSE2-NEXT:    pand %xmm1, %xmm0
+; X64-SSE2-NEXT:    pxor %xmm1, %xmm1
+; X64-SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; X64-SSE2-NEXT:    retq
   %t0 = shl <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y
   %t1 = and <4 x i32> %t0, %x
   %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0>
@@ -822,15 +735,6 @@ define i1 @scalar_i8_signbit_ne(i8 %x, i8 %y) nounwind {
 ;------------------------------------------------------------------------------;
 
 define i1 @scalar_i32_x_is_const_eq(i32 %y) nounwind {
-; X86-NOBMI-LABEL: scalar_i32_x_is_const_eq:
-; X86-NOBMI:       # %bb.0:
-; X86-NOBMI-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-NOBMI-NEXT:    movl $-1437226411, %eax # imm = 0xAA55AA55
-; X86-NOBMI-NEXT:    shll %cl, %eax
-; X86-NOBMI-NEXT:    testb $1, %al
-; X86-NOBMI-NEXT:    sete %al
-; X86-NOBMI-NEXT:    retl
-;
 ; X86-BMI1-LABEL: scalar_i32_x_is_const_eq:
 ; X86-BMI1:       # %bb.0:
 ; X86-BMI1-NEXT:    movb {{[0-9]+}}(%esp), %cl
@@ -840,24 +744,14 @@ define i1 @scalar_i32_x_is_const_eq(i32 %y) nounwind {
 ; X86-BMI1-NEXT:    sete %al
 ; X86-BMI1-NEXT:    retl
 ;
-; X86-BMI12-LABEL: scalar_i32_x_is_const_eq:
-; X86-BMI12:       # %bb.0:
-; X86-BMI12-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI12-NEXT:    movl $-1437226411, %ecx # imm = 0xAA55AA55
-; X86-BMI12-NEXT:    shlxl %eax, %ecx, %eax
-; X86-BMI12-NEXT:    testb $1, %al
-; X86-BMI12-NEXT:    sete %al
-; X86-BMI12-NEXT:    retl
-;
-; X64-NOBMI-LABEL: scalar_i32_x_is_const_eq:
-; X64-NOBMI:       # %bb.0:
-; X64-NOBMI-NEXT:    movl %edi, %ecx
-; X64-NOBMI-NEXT:    movl $-1437226411, %eax # imm = 0xAA55AA55
-; X64-NOBMI-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-NOBMI-NEXT:    shll %cl, %eax
-; X64-NOBMI-NEXT:    testb $1, %al
-; X64-NOBMI-NEXT:    sete %al
-; X64-NOBMI-NEXT:    retq
+; X86-BMI2-LABEL: scalar_i32_x_is_const_eq:
+; X86-BMI2:       # %bb.0:
+; X86-BMI2-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI2-NEXT:    movl $-1437226411, %ecx # imm = 0xAA55AA55
+; X86-BMI2-NEXT:    shlxl %eax, %ecx, %eax
+; X86-BMI2-NEXT:    testb $1, %al
+; X86-BMI2-NEXT:    sete %al
+; X86-BMI2-NEXT:    retl
 ;
 ; X64-BMI1-LABEL: scalar_i32_x_is_const_eq:
 ; X64-BMI1:       # %bb.0:
@@ -869,13 +763,13 @@ define i1 @scalar_i32_x_is_const_eq(i32 %y) nounwind {
 ; X64-BMI1-NEXT:    sete %al
 ; X64-BMI1-NEXT:    retq
 ;
-; X64-BMI12-LABEL: scalar_i32_x_is_const_eq:
-; X64-BMI12:       # %bb.0:
-; X64-BMI12-NEXT:    movl $-1437226411, %eax # imm = 0xAA55AA55
-; X64-BMI12-NEXT:    shlxl %edi, %eax, %eax
-; X64-BMI12-NEXT:    testb $1, %al
-; X64-BMI12-NEXT:    sete %al
-; X64-BMI12-NEXT:    retq
+; X64-BMI2-LABEL: scalar_i32_x_is_const_eq:
+; X64-BMI2:       # %bb.0:
+; X64-BMI2-NEXT:    movl $-1437226411, %eax # imm = 0xAA55AA55
+; X64-BMI2-NEXT:    shlxl %edi, %eax, %eax
+; X64-BMI2-NEXT:    testb $1, %al
+; X64-BMI2-NEXT:    sete %al
+; X64-BMI2-NEXT:    retq
   %t0 = shl i32 2857740885, %y
   %t1 = and i32 %t0, 1
   %res = icmp eq i32 %t1, 0