[X86] X86DAGToDAGISel::matchBitExtract(): extract 'lshr' from `X`

Summary:
As discussed in previous review, and noted in the FIXME, if `X` is actually an `lshr Y, Z` (logical!),
we can fold the `Z` into 'control`, and let the `BEXTR` do this too.
We could just insert those 8 bits of shift amount into control,
but it is better to instead zero-extend them, and 'or' them in place.

We can only do this for `lshr`, not `ashr`, because we do not know that the mask cover only the bits of `Y`,
and not any of the sign-extended bits.

The obvious question is, is this actually legal to do?
I believe it is. Relevant quotes, from `Intel® 64 and IA-32 Architectures Software Developer’s Manual`, `BEXTR — Bit Field Extract`:
* `Bit 7:0 of the second source operand specifies the starting bit position of bit extraction.`
* `A START value exceeding the operand size will not extract any bits from the second source operand.`
* `Only bit positions up to (OperandSize -1) of the first source operand are extracted.`
* `All higher order bits in the destination operand (starting at bit position LENGTH) are zeroed.`
* `The destination register is cleared if no bits are extracted.`

FIXME: if we can do this, i wonder if we should prefer `BEXTR` over `BZHI` in such cases.

Reviewers: RKSimon, craig.topper, spatel, andreadb

Reviewed By: RKSimon, craig.topper, andreadb

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D54095

llvm-svn: 347048

1 files changed, 176 insertions, 230 deletions

diff --git a/llvm/test/CodeGen/X86/extract-bits.ll b/llvm/test/CodeGen/X86/extract-bits.ll
index cfe6ba571df..2f1b28627a4 100644
--- a/llvm/test/CodeGen/X86/extract-bits.ll
+++ b/llvm/test/CodeGen/X86/extract-bits.ll
@@ -49,11 +49,10 @@ define i32 @bextr32_a0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ; X86-BMI1NOTBM-LABEL: bextr32_a0:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_a0:
@@ -78,11 +77,10 @@ define i32 @bextr32_a0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_a0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_a0:
@@ -183,11 +181,10 @@ define i32 @bextr32_a1_indexzext(i32 %val, i8 zeroext %numskipbits, i8 zeroext %
 ; X86-BMI1NOTBM-LABEL: bextr32_a1_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_a1_indexzext:
@@ -212,11 +209,10 @@ define i32 @bextr32_a1_indexzext(i32 %val, i8 zeroext %numskipbits, i8 zeroext %
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_a1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_a1_indexzext:
@@ -252,13 +248,12 @@ define i32 @bextr32_a2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ;
 ; X86-BMI1NOTBM-LABEL: bextr32_a2_load:
 ; X86-BMI1NOTBM:       # %bb.0:
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_a2_load:
@@ -285,12 +280,10 @@ define i32 @bextr32_a2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_a2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_a2_load:
@@ -325,13 +318,12 @@ define i32 @bextr32_a3_load_indexzext(i32* %w, i8 zeroext %numskipbits, i8 zeroe
 ;
 ; X86-BMI1NOTBM-LABEL: bextr32_a3_load_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_a3_load_indexzext:
@@ -358,12 +350,10 @@ define i32 @bextr32_a3_load_indexzext(i32* %w, i8 zeroext %numskipbits, i8 zeroe
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_a3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_a3_load_indexzext:
@@ -400,11 +390,10 @@ define i32 @bextr32_a4_commutative(i32 %val, i32 %numskipbits, i32 %numlowbits)
 ; X86-BMI1NOTBM-LABEL: bextr32_a4_commutative:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_a4_commutative:
@@ -429,11 +418,10 @@ define i32 @bextr32_a4_commutative(i32 %val, i32 %numskipbits, i32 %numlowbits)
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_a4_commutative:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_a4_commutative:
@@ -476,13 +464,13 @@ define i32 @bextr32_a5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    pushl %esi
 ; X86-BMI1NOTBM-NEXT:    subl $8, %esp
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %esi
-; X86-BMI1NOTBM-NEXT:    movl %ecx, (%esp)
+; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl %al, %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, {{[0-9]+}}(%esp), %esi
+; X86-BMI1NOTBM-NEXT:    movl %eax, (%esp)
 ; X86-BMI1NOTBM-NEXT:    calll use32
 ; X86-BMI1NOTBM-NEXT:    movl %esi, %eax
 ; X86-BMI1NOTBM-NEXT:    addl $8, %esp
@@ -523,10 +511,10 @@ define i32 @bextr32_a5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr32_a5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %ebx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %ebx
 ; X64-BMI1NOTBM-NEXT:    movl %esi, %edi
 ; X64-BMI1NOTBM-NEXT:    callq use32
 ; X64-BMI1NOTBM-NEXT:    movl %ebx, %eax
@@ -677,11 +665,10 @@ define i64 @bextr64_a0(i64 %val, i64 %numskipbits, i64 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_a0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_a0:
@@ -967,11 +954,10 @@ define i64 @bextr64_a1_indexzext(i64 %val, i8 zeroext %numskipbits, i8 zeroext %
 ; X64-BMI1NOTBM-LABEL: bextr64_a1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_a1_indexzext:
@@ -1118,12 +1104,10 @@ define i64 @bextr64_a2_load(i64* %w, i64 %numskipbits, i64 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_a2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_a2_load:
@@ -1268,12 +1252,10 @@ define i64 @bextr64_a3_load_indexzext(i64* %w, i8 zeroext %numskipbits, i8 zeroe
 ; X64-BMI1NOTBM-LABEL: bextr64_a3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_a3_load_indexzext:
@@ -1417,11 +1399,10 @@ define i64 @bextr64_a4_commutative(i64 %val, i64 %numskipbits, i64 %numlowbits)
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_a4_commutative:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_a4_commutative:
@@ -1605,10 +1586,10 @@ define i64 @bextr64_a5_skipextrauses(i64 %val, i64 %numskipbits, i64 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr64_a5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rbx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rbx
 ; X64-BMI1NOTBM-NEXT:    movq %rsi, %rdi
 ; X64-BMI1NOTBM-NEXT:    callq use64
 ; X64-BMI1NOTBM-NEXT:    movq %rbx, %rax
@@ -1656,11 +1637,10 @@ define i32 @bextr32_b0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ; X86-BMI1NOTBM-LABEL: bextr32_b0:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_b0:
@@ -1685,11 +1665,10 @@ define i32 @bextr32_b0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_b0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_b0:
@@ -1723,11 +1702,10 @@ define i32 @bextr32_b1_indexzext(i32 %val, i8 zeroext %numskipbits, i8 zeroext %
 ; X86-BMI1NOTBM-LABEL: bextr32_b1_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_b1_indexzext:
@@ -1752,11 +1730,10 @@ define i32 @bextr32_b1_indexzext(i32 %val, i8 zeroext %numskipbits, i8 zeroext %
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_b1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_b1_indexzext:
@@ -1792,13 +1769,12 @@ define i32 @bextr32_b2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ;
 ; X86-BMI1NOTBM-LABEL: bextr32_b2_load:
 ; X86-BMI1NOTBM:       # %bb.0:
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_b2_load:
@@ -1825,12 +1801,10 @@ define i32 @bextr32_b2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_b2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_b2_load:
@@ -1865,13 +1839,12 @@ define i32 @bextr32_b3_load_indexzext(i32* %w, i8 zeroext %numskipbits, i8 zeroe
 ;
 ; X86-BMI1NOTBM-LABEL: bextr32_b3_load_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_b3_load_indexzext:
@@ -1898,12 +1871,10 @@ define i32 @bextr32_b3_load_indexzext(i32* %w, i8 zeroext %numskipbits, i8 zeroe
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_b3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_b3_load_indexzext:
@@ -1940,11 +1911,10 @@ define i32 @bextr32_b4_commutative(i32 %val, i32 %numskipbits, i32 %numlowbits)
 ; X86-BMI1NOTBM-LABEL: bextr32_b4_commutative:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_b4_commutative:
@@ -1969,11 +1939,10 @@ define i32 @bextr32_b4_commutative(i32 %val, i32 %numskipbits, i32 %numlowbits)
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_b4_commutative:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_b4_commutative:
@@ -2016,13 +1985,13 @@ define i32 @bextr32_b5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    pushl %esi
 ; X86-BMI1NOTBM-NEXT:    subl $8, %esp
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %esi
-; X86-BMI1NOTBM-NEXT:    movl %ecx, (%esp)
+; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl %al, %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, {{[0-9]+}}(%esp), %esi
+; X86-BMI1NOTBM-NEXT:    movl %eax, (%esp)
 ; X86-BMI1NOTBM-NEXT:    calll use32
 ; X86-BMI1NOTBM-NEXT:    movl %esi, %eax
 ; X86-BMI1NOTBM-NEXT:    addl $8, %esp
@@ -2063,10 +2032,10 @@ define i32 @bextr32_b5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr32_b5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %ebx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %ebx
 ; X64-BMI1NOTBM-NEXT:    movl %esi, %edi
 ; X64-BMI1NOTBM-NEXT:    callq use32
 ; X64-BMI1NOTBM-NEXT:    movl %ebx, %eax
@@ -2214,11 +2183,10 @@ define i64 @bextr64_b0(i64 %val, i64 %numskipbits, i64 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_b0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_b0:
@@ -2355,11 +2323,10 @@ define i64 @bextr64_b1_indexzext(i64 %val, i8 zeroext %numskipbits, i8 zeroext %
 ; X64-BMI1NOTBM-LABEL: bextr64_b1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_b1_indexzext:
@@ -2503,12 +2470,10 @@ define i64 @bextr64_b2_load(i64* %w, i64 %numskipbits, i64 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_b2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_b2_load:
@@ -2650,12 +2615,10 @@ define i64 @bextr64_b3_load_indexzext(i64* %w, i8 zeroext %numskipbits, i8 zeroe
 ; X64-BMI1NOTBM-LABEL: bextr64_b3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_b3_load_indexzext:
@@ -2796,11 +2759,10 @@ define i64 @bextr64_b4_commutative(i64 %val, i64 %numskipbits, i64 %numlowbits)
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_b4_commutative:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_b4_commutative:
@@ -2979,10 +2941,10 @@ define i64 @bextr64_b5_skipextrauses(i64 %val, i64 %numskipbits, i64 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr64_b5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rbx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rbx
 ; X64-BMI1NOTBM-NEXT:    movq %rsi, %rdi
 ; X64-BMI1NOTBM-NEXT:    callq use64
 ; X64-BMI1NOTBM-NEXT:    movq %rbx, %rax
@@ -5165,11 +5127,10 @@ define i32 @bextr32_d0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ; X86-BMI1NOTBM-LABEL: bextr32_d0:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %eax
+; X86-BMI1NOTBM-NEXT:    bextrl %eax, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_d0:
@@ -5194,11 +5155,10 @@ define i32 @bextr32_d0(i32 %val, i32 %numskipbits, i32 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_d0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_d0:
@@ -5229,11 +5189,10 @@ define i32 @bextr32_d1_indexzext(i32 %val, i8 %numskipbits, i8 %numlowbits) noun
 ; X86-BMI1NOTBM-LABEL: bextr32_d1_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
 ; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    orl %eax, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_d1_indexzext:
@@ -5258,11 +5217,10 @@ define i32 @bextr32_d1_indexzext(i32 %val, i8 %numskipbits, i8 %numlowbits) noun
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_d1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_d1_indexzext:
@@ -5296,12 +5254,11 @@ define i32 @bextr32_d2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ; X86-BMI1NOTBM-LABEL: bextr32_d2_load:
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    orl %edx, %ecx
+; X86-BMI1NOTBM-NEXT:    bextrl %ecx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_d2_load:
@@ -5327,12 +5284,10 @@ define i32 @bextr32_d2_load(i32* %w, i32 %numskipbits, i32 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_d2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_d2_load:
@@ -5364,13 +5319,12 @@ define i32 @bextr32_d3_load_indexzext(i32* %w, i8 %numskipbits, i8 %numlowbits)
 ;
 ; X86-BMI1NOTBM-LABEL: bextr32_d3_load_indexzext:
 ; X86-BMI1NOTBM:       # %bb.0:
-; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %al
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-BMI1NOTBM-NEXT:    movb {{[0-9]+}}(%esp), %cl
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
-; X86-BMI1NOTBM-NEXT:    movl (%edx), %edx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl {{[0-9]+}}(%esp), %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, (%eax), %eax
 ; X86-BMI1NOTBM-NEXT:    retl
 ;
 ; X86-BMI1BMI2-LABEL: bextr32_d3_load_indexzext:
@@ -5396,12 +5350,10 @@ define i32 @bextr32_d3_load_indexzext(i32* %w, i8 %numskipbits, i8 %numlowbits)
 ;
 ; X64-BMI1NOTBM-LABEL: bextr32_d3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movl (%rdi), %eax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %eax
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %eax, %eax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, (%rdi), %eax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr32_d3_load_indexzext:
@@ -5444,13 +5396,13 @@ define i32 @bextr32_d5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X86-BMI1NOTBM:       # %bb.0:
 ; X86-BMI1NOTBM-NEXT:    pushl %esi
 ; X86-BMI1NOTBM-NEXT:    subl $8, %esp
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-BMI1NOTBM-NEXT:    shrl %cl, %edx
-; X86-BMI1NOTBM-NEXT:    shll $8, %eax
-; X86-BMI1NOTBM-NEXT:    bextrl %eax, %edx, %esi
-; X86-BMI1NOTBM-NEXT:    movl %ecx, (%esp)
+; X86-BMI1NOTBM-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-BMI1NOTBM-NEXT:    shll $8, %ecx
+; X86-BMI1NOTBM-NEXT:    movzbl %al, %edx
+; X86-BMI1NOTBM-NEXT:    orl %ecx, %edx
+; X86-BMI1NOTBM-NEXT:    bextrl %edx, {{[0-9]+}}(%esp), %esi
+; X86-BMI1NOTBM-NEXT:    movl %eax, (%esp)
 ; X86-BMI1NOTBM-NEXT:    calll use32
 ; X86-BMI1NOTBM-NEXT:    movl %esi, %eax
 ; X86-BMI1NOTBM-NEXT:    addl $8, %esp
@@ -5491,10 +5443,10 @@ define i32 @bextr32_d5_skipextrauses(i32 %val, i32 %numskipbits, i32 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr32_d5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    shrl %cl, %edi
 ; X64-BMI1NOTBM-NEXT:    shll $8, %edx
-; X64-BMI1NOTBM-NEXT:    bextrl %edx, %edi, %ebx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orl %edx, %eax
+; X64-BMI1NOTBM-NEXT:    bextrl %eax, %edi, %ebx
 ; X64-BMI1NOTBM-NEXT:    movl %esi, %edi
 ; X64-BMI1NOTBM-NEXT:    callq use32
 ; X64-BMI1NOTBM-NEXT:    movl %ebx, %eax
@@ -5673,11 +5625,10 @@ define i64 @bextr64_d0(i64 %val, i64 %numskipbits, i64 %numlowbits) nounwind {
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_d0:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_d0:
@@ -5845,11 +5796,10 @@ define i64 @bextr64_d1_indexzext(i64 %val, i8 %numskipbits, i8 %numlowbits) noun
 ; X64-BMI1NOTBM-LABEL: bextr64_d1_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_d1_indexzext:
@@ -6023,12 +5973,10 @@ define i64 @bextr64_d2_load(i64* %w, i64 %numskipbits, i64 %numlowbits) nounwind
 ;
 ; X64-BMI1NOTBM-LABEL: bextr64_d2_load:
 ; X64-BMI1NOTBM:       # %bb.0:
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_d2_load:
@@ -6200,12 +6148,10 @@ define i64 @bextr64_d3_load_indexzext(i64* %w, i8 %numskipbits, i8 %numlowbits)
 ; X64-BMI1NOTBM-LABEL: bextr64_d3_load_indexzext:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    # kill: def $edx killed $edx def $rdx
-; X64-BMI1NOTBM-NEXT:    movl %esi, %ecx
-; X64-BMI1NOTBM-NEXT:    movq (%rdi), %rax
-; X64-BMI1NOTBM-NEXT:    # kill: def $cl killed $cl killed $ecx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rax
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rax, %rax
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, (%rdi), %rax
 ; X64-BMI1NOTBM-NEXT:    retq
 ;
 ; X64-BMI1BMI2-LABEL: bextr64_d3_load_indexzext:
@@ -6419,10 +6365,10 @@ define i64 @bextr64_d5_skipextrauses(i64 %val, i64 %numskipbits, i64 %numlowbits
 ; X64-BMI1NOTBM-LABEL: bextr64_d5_skipextrauses:
 ; X64-BMI1NOTBM:       # %bb.0:
 ; X64-BMI1NOTBM-NEXT:    pushq %rbx
-; X64-BMI1NOTBM-NEXT:    movq %rsi, %rcx
-; X64-BMI1NOTBM-NEXT:    shrq %cl, %rdi
 ; X64-BMI1NOTBM-NEXT:    shlq $8, %rdx
-; X64-BMI1NOTBM-NEXT:    bextrq %rdx, %rdi, %rbx
+; X64-BMI1NOTBM-NEXT:    movzbl %sil, %eax
+; X64-BMI1NOTBM-NEXT:    orq %rdx, %rax
+; X64-BMI1NOTBM-NEXT:    bextrq %rax, %rdi, %rbx
 ; X64-BMI1NOTBM-NEXT:    movq %rsi, %rdi
 ; X64-BMI1NOTBM-NEXT:    callq use64
 ; X64-BMI1NOTBM-NEXT:    movq %rbx, %rax