[X86] Add tests for BMI patterns across non-adjacent and assosiative instructions.

[goldstein.w.n]

Summary

I.e for blsi match (and (sub 0, x), x) but we currently miss valid
patterns like (and (and (sub 0, x), y), x).

[goldstein.w.n]

N/A

[goldstein.w.n]

ping.

[craig.topper]

Can you add a description of what "indirect" means in this context?

[goldstein.w.n]

Update commit message to be more clear

[goldstein.w.n]

In D141178#4048341, @craig.topper wrote:

Can you add a description of what "indirect" means in this context?

Done.

[pengfei]

LGTM except for a nit.

llvm/test/CodeGen/X86/bmi-out-of-order.ll

5–7

These are not necessary. You can just remove them.

[goldstein.w.n]

Remove unneeded meta data

[goldstein.w.n]

In D141178#4051146, @pengfei wrote:

LGTM except for a nit.

Fixed nit, will wait 24 hrs to push.

llvm/test/CodeGen/X86/bmi-out-of-order.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+cmov,+bmi | FileCheck %s --check-prefix=X86
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+cmov,+bmi | FileCheck %s --check-prefix=X64
+
+source_filename = "/home/noah/.tmp/objdumps/test.c-2023-01-06-T18-38-57.o.c"
+target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+define i32 @blsmsk_used2(i32 %a) nounwind {
+; X86-LABEL: blsmsk_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    blsmskl %eax, %ecx
+; X86-NEXT:    decl %eax
+; X86-NEXT:    imull %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmsk_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    blsmskl %edi, %ecx
+; X64-NEXT:    leal -1(%rdi), %eax
+; X64-NEXT:    imull %ecx, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add i32 %a, -1
+  %xor = xor i32 %sub, %a
+  %mul = mul i32 %xor, %sub
+  ret i32 %mul
+}
+
+define i64 @blsmask_through1(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsmask_through1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    addl $-1, %eax
+; X86-NEXT:    movl %esi, %edx
+; X86-NEXT:    adcl $-1, %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl %ecx, %eax
+; X86-NEXT:    xorl %esi, %edx
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmask_through1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    xorq %rsi, %rdi
+; X64-NEXT:    leaq -1(%rsi), %rax
+; X64-NEXT:    xorq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = add i64 %b, -1
+  %0 = xor i64 %sub, %a
+  %xor1 = xor i64 %0, %b
+  ret i64 %xor1
+}
+
+define i32 @blsmask_through2(i32 %a, i32 %b, i32 %c) nounwind {
+; X86-LABEL: blsmask_through2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    leal -1(%ecx), %eax
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmask_through2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $esi killed $esi def $rsi
+; X64-NEXT:    xorl %edx, %edi
+; X64-NEXT:    xorl %esi, %edi
+; X64-NEXT:    leal -1(%rsi), %eax
+; X64-NEXT:    xorl %edi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add nsw i32 %b, -1
+  %0 = xor i32 %sub, %a
+  %1 = xor i32 %0, %c
+  %xor2 = xor i32 %1, %b
+  ret i32 %xor2
+}
+
+define i64 @blsmask_through3(i64 %a, i64 %b, i64 %c, i64 %d) nounwind {
+; X86-LABEL: blsmask_through3:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %esi, %eax
+; X86-NEXT:    addl $-1, %eax
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    adcl $-1, %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl %esi, %eax
+; X86-NEXT:    xorl %ecx, %edx
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmask_through3:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    xorq %rdx, %rdi
+; X64-NEXT:    xorq %rcx, %rdi
+; X64-NEXT:    xorq %rsi, %rdi
+; X64-NEXT:    leaq -1(%rsi), %rax
+; X64-NEXT:    xorq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = add nsw i64 %b, -1
+  %0 = xor i64 %sub, %a
+  %1 = xor i64 %0, %c
+  %2 = xor i64 %1, %d
+  %xor3 = xor i64 %2, %b
+  ret i64 %xor3
+}
+
+define i32 @blsmask_through1_used1(i32 %a, i32 %b) nounwind {
+; X86-LABEL: blsmask_through1_used1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    leal -1(%ecx), %eax
+; X86-NEXT:    xorl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl %eax, %ecx
+; X86-NEXT:    incl %eax
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmask_through1_used1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $esi killed $esi def $rsi
+; X64-NEXT:    leal -1(%rsi), %eax
+; X64-NEXT:    xorl %edi, %eax
+; X64-NEXT:    xorl %eax, %esi
+; X64-NEXT:    incl %eax
+; X64-NEXT:    orl %esi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add i32 %b, -1
+  %xor = xor i32 %sub, %a
+  %xor1 = xor i32 %xor, %b
+  %add = add i32 %xor, 1
+  %or = or i32 %add, %xor1
+  ret i32 %or
+}
+
+define i64 @blsmask_through1_used2(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsmask_through1_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %ecx, %edi
+; X86-NEXT:    addl $-1, %edi
+; X86-NEXT:    movl %esi, %ebx
+; X86-NEXT:    adcl $-1, %ebx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    xorl %ebx, %ebp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    xorl %edi, %eax
+; X86-NEXT:    xorl %ebp, %esi
+; X86-NEXT:    xorl %eax, %ecx
+; X86-NEXT:    imull %eax, %ebx
+; X86-NEXT:    mull %edi
+; X86-NEXT:    imull %edi, %ebp
+; X86-NEXT:    addl %ebx, %ebp
+; X86-NEXT:    addl %ebp, %edx
+; X86-NEXT:    orl %esi, %edx
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsmask_through1_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    leaq -1(%rsi), %rax
+; X64-NEXT:    xorq %rax, %rdi
+; X64-NEXT:    xorq %rdi, %rsi
+; X64-NEXT:    imulq %rdi, %rax
+; X64-NEXT:    orq %rsi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = add i64 %b, -1
+  %xor = xor i64 %sub, %a
+  %xor1 = xor i64 %xor, %b
+  %mul = mul i64 %xor, %sub
+  %or = or i64 %mul, %xor1
+  ret i64 %or
+}
+
+define i32 @blsi_used2(i32 %a) nounwind {
+; X86-LABEL: blsi_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    blsil %eax, %ecx
+; X86-NEXT:    negl %eax
+; X86-NEXT:    imull %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    blsil %edi, %eax
+; X64-NEXT:    negl %edi
+; X64-NEXT:    imull %edi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = sub nsw i32 0, %a
+  %and = and i32 %sub, %a
+  %mul = mul nsw i32 %and, %sub
+  ret i32 %mul
+}
+
+define i64 @blsi_through1(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsi_through1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    sbbl %esi, %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %esi, %edx
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_through1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    andq %rsi, %rdi
+; X64-NEXT:    negq %rax
+; X64-NEXT:    andq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = sub nsw i64 0, %b
+  %and = and i64 %sub, %a
+  %and1 = and i64 %and, %b
+  ret i64 %and1
+}
+
+define i32 @blsi_through2(i32 %a, i32 %b, i32 %c) nounwind {
+; X86-LABEL: blsi_through2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_through2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movl %esi, %eax
+; X64-NEXT:    andl %edx, %edi
+; X64-NEXT:    andl %esi, %edi
+; X64-NEXT:    negl %eax
+; X64-NEXT:    andl %edi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = sub i32 0, %b
+  %and = and i32 %sub, %a
+  %0 = and i32 %and, %c
+  %and2 = and i32 %0, %b
+  ret i32 %and2
+}
+
+define i64 @blsi_through3(i64 %a, i64 %b, i64 %c) nounwind {
+; X86-LABEL: blsi_through3:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    sbbl %esi, %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %esi, %edx
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_through3:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    andq %rdx, %rdi
+; X64-NEXT:    andq %rsi, %rdi
+; X64-NEXT:    negq %rax
+; X64-NEXT:    andq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = sub i64 0, %b
+  %and = and i64 %sub, %a
+  %0 = and i64 %and, %c
+  %and3 = and i64 %0, %b
+  ret i64 %and3
+}
+
+define i32 @blsi_through1_used1(i32 %a, i32 %b) nounwind {
+; X86-LABEL: blsi_through1_used1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %eax, %ecx
+; X86-NEXT:    incl %eax
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_through1_used1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movl %esi, %eax
+; X64-NEXT:    negl %eax
+; X64-NEXT:    andl %edi, %eax
+; X64-NEXT:    andl %eax, %esi
+; X64-NEXT:    incl %eax
+; X64-NEXT:    orl %esi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = sub nsw i32 0, %b
+  %and = and i32 %sub, %a
+  %and1 = and i32 %and, %b
+  %add = add nsw i32 %and, 1
+  %or = or i32 %add, %and1
+  ret i32 %or
+}
+
+define i64 @blsi_through1_used2(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsi_through1_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    xorl %edi, %edi
+; X86-NEXT:    movl %ecx, %edx
+; X86-NEXT:    negl %edx
+; X86-NEXT:    sbbl %esi, %edi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
+; X86-NEXT:    andl %edi, %ebx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %edx, %eax
+; X86-NEXT:    andl %ebx, %esi
+; X86-NEXT:    andl %eax, %ecx
+; X86-NEXT:    imull %edx, %ebx
+; X86-NEXT:    imull %eax, %edi
+; X86-NEXT:    addl %ebx, %edi
+; X86-NEXT:    mull %edx
+; X86-NEXT:    addl %edi, %edx
+; X86-NEXT:    orl %esi, %edx
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsi_through1_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    negq %rax
+; X64-NEXT:    andq %rax, %rdi
+; X64-NEXT:    andq %rdi, %rsi
+; X64-NEXT:    imulq %rdi, %rax
+; X64-NEXT:    orq %rsi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = sub nsw i64 0, %b
+  %and = and i64 %sub, %a
+  %and1 = and i64 %and, %b
+  %mul = mul nsw i64 %and, %sub
+  %or = or i64 %mul, %and1
+  ret i64 %or
+}
+
+define i32 @blsr_used2(i32 %a) nounwind {
+; X86-LABEL: blsr_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    blsrl %eax, %ecx
+; X86-NEXT:    decl %eax
+; X86-NEXT:    imull %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    blsrl %edi, %ecx
+; X64-NEXT:    leal -1(%rdi), %eax
+; X64-NEXT:    imull %ecx, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add i32 %a, -1
+  %and = and i32 %sub, %a
+  %mul = mul i32 %and, %sub
+  ret i32 %mul
+}
+
+define i64 @blsr_through1(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsr_through1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    addl $-1, %eax
+; X86-NEXT:    movl %esi, %edx
+; X86-NEXT:    adcl $-1, %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    andl %esi, %edx
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_through1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    andq %rsi, %rdi
+; X64-NEXT:    leaq -1(%rsi), %rax
+; X64-NEXT:    andq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = add i64 %b, -1
+  %0 = and i64 %sub, %a
+  %and1 = and i64 %0, %b
+  ret i64 %and1
+}
+
+define i32 @blsr_through2(i32 %a, i32 %b, i32 %c) nounwind {
+; X86-LABEL: blsr_through2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    leal -1(%ecx), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_through2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $esi killed $esi def $rsi
+; X64-NEXT:    andl %edx, %edi
+; X64-NEXT:    andl %esi, %edi
+; X64-NEXT:    leal -1(%rsi), %eax
+; X64-NEXT:    andl %edi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add nsw i32 %b, -1
+  %0 = and i32 %sub, %a
+  %1 = and i32 %0, %c
+  %and2 = and i32 %1, %b
+  ret i32 %and2
+}
+
+define i64 @blsr_through3(i64 %a, i64 %b, i64 %c, i64 %d) nounwind {
+; X86-LABEL: blsr_through3:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    sbbl %esi, %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %edx
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %esi, %edx
+; X86-NEXT:    andl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_through3:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movq %rsi, %rax
+; X64-NEXT:    andq %rdx, %rdi
+; X64-NEXT:    andq %rcx, %rdi
+; X64-NEXT:    andq %rsi, %rdi
+; X64-NEXT:    negq %rax
+; X64-NEXT:    andq %rdi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = sub nsw i64 0, %b
+  %and = and i64 %sub, %a
+  %0 = and i64 %and, %c
+  %1 = and i64 %0, %d
+  %and4 = and i64 %1, %b
+  ret i64 %and4
+}
+
+define i32 @blsr_through1_used1(i32 %a, i32 %b) nounwind {
+; X86-LABEL: blsr_through1_used1:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    leal -1(%ecx), %eax
+; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %eax, %ecx
+; X86-NEXT:    incl %eax
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_through1_used1:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $esi killed $esi def $rsi
+; X64-NEXT:    leal -1(%rsi), %eax
+; X64-NEXT:    andl %edi, %eax
+; X64-NEXT:    andl %eax, %esi
+; X64-NEXT:    incl %eax
+; X64-NEXT:    orl %esi, %eax
+; X64-NEXT:    retq
+entry:
+  %sub = add i32 %b, -1
+  %and = and i32 %sub, %a
+  %and1 = and i32 %and, %b
+  %add = add i32 %and, 1
+  %or = or i32 %add, %and1
+  ret i32 %or
+}
+
+define i64 @blsr_through1_used2(i64 %a, i64 %b) nounwind {
+; X86-LABEL: blsr_through1_used2:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    pushl %ebx
+; X86-NEXT:    pushl %edi
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl %ecx, %edi
+; X86-NEXT:    addl $-1, %edi
+; X86-NEXT:    movl %esi, %ebx
+; X86-NEXT:    adcl $-1, %ebx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
+; X86-NEXT:    andl %ebx, %ebp
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    andl %edi, %eax
+; X86-NEXT:    andl %ebp, %esi
+; X86-NEXT:    andl %eax, %ecx
+; X86-NEXT:    imull %eax, %ebx
+; X86-NEXT:    mull %edi
+; X86-NEXT:    imull %edi, %ebp
+; X86-NEXT:    addl %ebx, %ebp
+; X86-NEXT:    addl %ebp, %edx
+; X86-NEXT:    orl %esi, %edx
+; X86-NEXT:    orl %ecx, %eax
+; X86-NEXT:    popl %esi
+; X86-NEXT:    popl %edi
+; X86-NEXT:    popl %ebx
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl
+;
+; X64-LABEL: blsr_through1_used2:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    leaq -1(%rsi), %rax
+; X64-NEXT:    andq %rax, %rdi
+; X64-NEXT:    andq %rdi, %rsi
+; X64-NEXT:    imulq %rdi, %rax
+; X64-NEXT:    orq %rsi, %rax
+; X64-NEXT:    retq
+entry:
+  %sub = add i64 %b, -1
+  %and = and i64 %sub, %a
+  %and1 = and i64 %and, %b
+  %mul = mul i64 %and, %sub
+  %or = or i64 %mul, %and1
+  ret i64 %or
+}