This PR is related to #99591. In this PR, instead of modifying how the legalisation occurs depending on surrounding instructions, we refine after legalisation. This PR has two parts: * `SDPatternMatch/MatchContext`: Modify a little bit the code to match Operands (used by `m_Node(...)`) and Unary/Binary/Ternary Patterns to make it compatible with `VPMatchContext`, instead of only `m_Opc` supported. Some tests were added to ensure no regressions. * `DAGCombiner`: Add a `foldSubCtlzNot` which detect and rewrite the patterns using matching context. Remaining Tasks: - [ ] GlobalISel - [ ] Currently the pattern matching will occur even before legalisation. Should I restrict it to specific stages instead ? - [ ] Style: Add a visitVP_SUB ?? Move `foldSubCtlzNot` in another location for style consistency purpose ? @topperc --------- Co-authored-by: v01dxyz <v01dxyz@v01d.xyz>
443 lines
13 KiB
LLVM
443 lines
13 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown | FileCheck %s --check-prefixes=X86,X86-NOCMOV
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+cmov | FileCheck %s --check-prefixes=X86,X86-CMOV
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=X64
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+bmi,+lzcnt | FileCheck %s --check-prefix=X86-CLZ
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi,+lzcnt | FileCheck %s --check-prefix=X64-CLZ
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+bmi,+lzcnt,+fast-lzcnt | FileCheck %s --check-prefix=X86-CLZ
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi,+lzcnt,+fast-lzcnt | FileCheck %s --check-prefix=X64-CLZ
|
|
|
|
declare i8 @llvm.ctlz.i8(i8, i1)
|
|
declare i16 @llvm.ctlz.i16(i16, i1)
|
|
declare i32 @llvm.ctlz.i32(i32, i1)
|
|
declare i64 @llvm.ctlz.i64(i64, i1)
|
|
|
|
define i8 @ctlo_i8(i8 %x) {
|
|
; X86-NOCMOV-LABEL: ctlo_i8:
|
|
; X86-NOCMOV: # %bb.0:
|
|
; X86-NOCMOV-NEXT: movzbl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: xorb $-1, %al
|
|
; X86-NOCMOV-NEXT: je .LBB0_1
|
|
; X86-NOCMOV-NEXT: # %bb.2: # %cond.false
|
|
; X86-NOCMOV-NEXT: movzbl %al, %eax
|
|
; X86-NOCMOV-NEXT: bsrl %eax, %eax
|
|
; X86-NOCMOV-NEXT: xorl $7, %eax
|
|
; X86-NOCMOV-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
; X86-NOCMOV-NEXT: .LBB0_1:
|
|
; X86-NOCMOV-NEXT: movb $8, %al
|
|
; X86-NOCMOV-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
;
|
|
; X86-CMOV-LABEL: ctlo_i8:
|
|
; X86-CMOV: # %bb.0:
|
|
; X86-CMOV-NEXT: movzbl {{[0-9]+}}(%esp), %eax
|
|
; X86-CMOV-NEXT: notb %al
|
|
; X86-CMOV-NEXT: movzbl %al, %eax
|
|
; X86-CMOV-NEXT: bsrl %eax, %ecx
|
|
; X86-CMOV-NEXT: movl $15, %eax
|
|
; X86-CMOV-NEXT: cmovnel %ecx, %eax
|
|
; X86-CMOV-NEXT: xorl $7, %eax
|
|
; X86-CMOV-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-CMOV-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i8:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notb %dil
|
|
; X64-NEXT: movzbl %dil, %eax
|
|
; X64-NEXT: bsrl %eax, %ecx
|
|
; X64-NEXT: movl $15, %eax
|
|
; X64-NEXT: cmovnel %ecx, %eax
|
|
; X64-NEXT: xorl $7, %eax
|
|
; X64-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i8:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: shll $24, %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i8:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: shll $24, %edi
|
|
; X64-CLZ-NEXT: notl %edi
|
|
; X64-CLZ-NEXT: lzcntl %edi, %eax
|
|
; X64-CLZ-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i8 %x, -1
|
|
%tmp2 = call i8 @llvm.ctlz.i8( i8 %tmp1, i1 false )
|
|
ret i8 %tmp2
|
|
}
|
|
|
|
define i8 @ctlo_i8_undef(i8 %x) {
|
|
; X86-LABEL: ctlo_i8_undef:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
|
|
; X86-NEXT: notb %al
|
|
; X86-NEXT: movzbl %al, %eax
|
|
; X86-NEXT: bsrl %eax, %eax
|
|
; X86-NEXT: xorl $7, %eax
|
|
; X86-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i8_undef:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notb %dil
|
|
; X64-NEXT: movzbl %dil, %eax
|
|
; X64-NEXT: bsrl %eax, %eax
|
|
; X64-NEXT: xorl $7, %eax
|
|
; X64-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i8_undef:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movzbl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notb %al
|
|
; X86-CLZ-NEXT: movzbl %al, %eax
|
|
; X86-CLZ-NEXT: shll $24, %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: # kill: def $al killed $al killed $eax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i8_undef:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notb %dil
|
|
; X64-CLZ-NEXT: movzbl %dil, %eax
|
|
; X64-CLZ-NEXT: shll $24, %eax
|
|
; X64-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X64-CLZ-NEXT: # kill: def $al killed $al killed $eax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i8 %x, -1
|
|
%tmp2 = call i8 @llvm.ctlz.i8( i8 %tmp1, i1 true )
|
|
ret i8 %tmp2
|
|
}
|
|
|
|
define i16 @ctlo_i16(i16 %x) {
|
|
; X86-NOCMOV-LABEL: ctlo_i16:
|
|
; X86-NOCMOV: # %bb.0:
|
|
; X86-NOCMOV-NEXT: movzwl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: xorw $-1, %ax
|
|
; X86-NOCMOV-NEXT: je .LBB2_1
|
|
; X86-NOCMOV-NEXT: # %bb.2: # %cond.false
|
|
; X86-NOCMOV-NEXT: bsrw %ax, %ax
|
|
; X86-NOCMOV-NEXT: xorl $15, %eax
|
|
; X86-NOCMOV-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
; X86-NOCMOV-NEXT: .LBB2_1:
|
|
; X86-NOCMOV-NEXT: movw $16, %ax
|
|
; X86-NOCMOV-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
;
|
|
; X86-CMOV-LABEL: ctlo_i16:
|
|
; X86-CMOV: # %bb.0:
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CMOV-NEXT: notl %eax
|
|
; X86-CMOV-NEXT: bsrw %ax, %cx
|
|
; X86-CMOV-NEXT: movw $31, %ax
|
|
; X86-CMOV-NEXT: cmovnew %cx, %ax
|
|
; X86-CMOV-NEXT: xorl $15, %eax
|
|
; X86-CMOV-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-CMOV-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i16:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notl %edi
|
|
; X64-NEXT: bsrw %di, %cx
|
|
; X64-NEXT: movw $31, %ax
|
|
; X64-NEXT: cmovnew %cx, %ax
|
|
; X64-NEXT: xorl $15, %eax
|
|
; X64-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i16:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntw %ax, %ax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i16:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notl %edi
|
|
; X64-CLZ-NEXT: lzcntw %di, %ax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i16 %x, -1
|
|
%tmp2 = call i16 @llvm.ctlz.i16( i16 %tmp1, i1 false )
|
|
ret i16 %tmp2
|
|
}
|
|
|
|
define i16 @ctlo_i16_undef(i16 %x) {
|
|
; X86-LABEL: ctlo_i16_undef:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NEXT: notl %eax
|
|
; X86-NEXT: bsrw %ax, %ax
|
|
; X86-NEXT: xorl $15, %eax
|
|
; X86-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i16_undef:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notl %edi
|
|
; X64-NEXT: bsrw %di, %ax
|
|
; X64-NEXT: xorl $15, %eax
|
|
; X64-NEXT: # kill: def $ax killed $ax killed $eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i16_undef:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntw %ax, %ax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i16_undef:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notl %edi
|
|
; X64-CLZ-NEXT: lzcntw %di, %ax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i16 %x, -1
|
|
%tmp2 = call i16 @llvm.ctlz.i16( i16 %tmp1, i1 true )
|
|
ret i16 %tmp2
|
|
}
|
|
|
|
define i32 @ctlo_i32(i32 %x) {
|
|
; X86-NOCMOV-LABEL: ctlo_i32:
|
|
; X86-NOCMOV: # %bb.0:
|
|
; X86-NOCMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: xorl $-1, %eax
|
|
; X86-NOCMOV-NEXT: je .LBB4_1
|
|
; X86-NOCMOV-NEXT: # %bb.2: # %cond.false
|
|
; X86-NOCMOV-NEXT: bsrl %eax, %eax
|
|
; X86-NOCMOV-NEXT: xorl $31, %eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
; X86-NOCMOV-NEXT: .LBB4_1:
|
|
; X86-NOCMOV-NEXT: movl $32, %eax
|
|
; X86-NOCMOV-NEXT: retl
|
|
;
|
|
; X86-CMOV-LABEL: ctlo_i32:
|
|
; X86-CMOV: # %bb.0:
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CMOV-NEXT: notl %eax
|
|
; X86-CMOV-NEXT: bsrl %eax, %ecx
|
|
; X86-CMOV-NEXT: movl $63, %eax
|
|
; X86-CMOV-NEXT: cmovnel %ecx, %eax
|
|
; X86-CMOV-NEXT: xorl $31, %eax
|
|
; X86-CMOV-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i32:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notl %edi
|
|
; X64-NEXT: bsrl %edi, %ecx
|
|
; X64-NEXT: movl $63, %eax
|
|
; X64-NEXT: cmovnel %ecx, %eax
|
|
; X64-NEXT: xorl $31, %eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i32:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i32:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notl %edi
|
|
; X64-CLZ-NEXT: lzcntl %edi, %eax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i32 %x, -1
|
|
%tmp2 = call i32 @llvm.ctlz.i32( i32 %tmp1, i1 false )
|
|
ret i32 %tmp2
|
|
}
|
|
|
|
define i32 @ctlo_i32_undef(i32 %x) {
|
|
; X86-LABEL: ctlo_i32_undef:
|
|
; X86: # %bb.0:
|
|
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NEXT: notl %eax
|
|
; X86-NEXT: bsrl %eax, %eax
|
|
; X86-NEXT: xorl $31, %eax
|
|
; X86-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i32_undef:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notl %edi
|
|
; X64-NEXT: bsrl %edi, %eax
|
|
; X64-NEXT: xorl $31, %eax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i32_undef:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i32_undef:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notl %edi
|
|
; X64-CLZ-NEXT: lzcntl %edi, %eax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i32 %x, -1
|
|
%tmp2 = call i32 @llvm.ctlz.i32( i32 %tmp1, i1 true )
|
|
ret i32 %tmp2
|
|
}
|
|
|
|
define i64 @ctlo_i64(i64 %x) {
|
|
; X86-NOCMOV-LABEL: ctlo_i64:
|
|
; X86-NOCMOV: # %bb.0:
|
|
; X86-NOCMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X86-NOCMOV-NEXT: notl %ecx
|
|
; X86-NOCMOV-NEXT: notl %eax
|
|
; X86-NOCMOV-NEXT: bsrl %eax, %edx
|
|
; X86-NOCMOV-NEXT: movl $63, %eax
|
|
; X86-NOCMOV-NEXT: je .LBB6_2
|
|
; X86-NOCMOV-NEXT: # %bb.1:
|
|
; X86-NOCMOV-NEXT: movl %edx, %eax
|
|
; X86-NOCMOV-NEXT: .LBB6_2:
|
|
; X86-NOCMOV-NEXT: testl %ecx, %ecx
|
|
; X86-NOCMOV-NEXT: jne .LBB6_3
|
|
; X86-NOCMOV-NEXT: # %bb.4:
|
|
; X86-NOCMOV-NEXT: xorl $31, %eax
|
|
; X86-NOCMOV-NEXT: addl $32, %eax
|
|
; X86-NOCMOV-NEXT: xorl %edx, %edx
|
|
; X86-NOCMOV-NEXT: retl
|
|
; X86-NOCMOV-NEXT: .LBB6_3:
|
|
; X86-NOCMOV-NEXT: bsrl %ecx, %eax
|
|
; X86-NOCMOV-NEXT: xorl $31, %eax
|
|
; X86-NOCMOV-NEXT: xorl %edx, %edx
|
|
; X86-NOCMOV-NEXT: retl
|
|
;
|
|
; X86-CMOV-LABEL: ctlo_i64:
|
|
; X86-CMOV: # %bb.0:
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X86-CMOV-NEXT: notl %ecx
|
|
; X86-CMOV-NEXT: notl %eax
|
|
; X86-CMOV-NEXT: bsrl %eax, %eax
|
|
; X86-CMOV-NEXT: movl $63, %edx
|
|
; X86-CMOV-NEXT: cmovnel %eax, %edx
|
|
; X86-CMOV-NEXT: xorl $31, %edx
|
|
; X86-CMOV-NEXT: addl $32, %edx
|
|
; X86-CMOV-NEXT: bsrl %ecx, %eax
|
|
; X86-CMOV-NEXT: xorl $31, %eax
|
|
; X86-CMOV-NEXT: testl %ecx, %ecx
|
|
; X86-CMOV-NEXT: cmovel %edx, %eax
|
|
; X86-CMOV-NEXT: xorl %edx, %edx
|
|
; X86-CMOV-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i64:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notq %rdi
|
|
; X64-NEXT: bsrq %rdi, %rcx
|
|
; X64-NEXT: movl $127, %eax
|
|
; X64-NEXT: cmovneq %rcx, %rax
|
|
; X64-NEXT: xorq $63, %rax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i64:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: testl %eax, %eax
|
|
; X86-CLZ-NEXT: jne .LBB6_1
|
|
; X86-CLZ-NEXT: # %bb.2:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: addl $32, %eax
|
|
; X86-CLZ-NEXT: xorl %edx, %edx
|
|
; X86-CLZ-NEXT: retl
|
|
; X86-CLZ-NEXT: .LBB6_1:
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: xorl %edx, %edx
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i64:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notq %rdi
|
|
; X64-CLZ-NEXT: lzcntq %rdi, %rax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i64 %x, -1
|
|
%tmp2 = call i64 @llvm.ctlz.i64( i64 %tmp1, i1 false )
|
|
ret i64 %tmp2
|
|
}
|
|
|
|
define i64 @ctlo_i64_undef(i64 %x) {
|
|
; X86-NOCMOV-LABEL: ctlo_i64_undef:
|
|
; X86-NOCMOV: # %bb.0:
|
|
; X86-NOCMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: notl %eax
|
|
; X86-NOCMOV-NEXT: testl %eax, %eax
|
|
; X86-NOCMOV-NEXT: jne .LBB7_1
|
|
; X86-NOCMOV-NEXT: # %bb.2:
|
|
; X86-NOCMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-NOCMOV-NEXT: notl %eax
|
|
; X86-NOCMOV-NEXT: bsrl %eax, %eax
|
|
; X86-NOCMOV-NEXT: xorl $31, %eax
|
|
; X86-NOCMOV-NEXT: orl $32, %eax
|
|
; X86-NOCMOV-NEXT: xorl %edx, %edx
|
|
; X86-NOCMOV-NEXT: retl
|
|
; X86-NOCMOV-NEXT: .LBB7_1:
|
|
; X86-NOCMOV-NEXT: bsrl %eax, %eax
|
|
; X86-NOCMOV-NEXT: xorl $31, %eax
|
|
; X86-NOCMOV-NEXT: xorl %edx, %edx
|
|
; X86-NOCMOV-NEXT: retl
|
|
;
|
|
; X86-CMOV-LABEL: ctlo_i64_undef:
|
|
; X86-CMOV: # %bb.0:
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CMOV-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
; X86-CMOV-NEXT: notl %eax
|
|
; X86-CMOV-NEXT: notl %ecx
|
|
; X86-CMOV-NEXT: bsrl %ecx, %edx
|
|
; X86-CMOV-NEXT: xorl $31, %edx
|
|
; X86-CMOV-NEXT: bsrl %eax, %eax
|
|
; X86-CMOV-NEXT: xorl $31, %eax
|
|
; X86-CMOV-NEXT: orl $32, %eax
|
|
; X86-CMOV-NEXT: testl %ecx, %ecx
|
|
; X86-CMOV-NEXT: cmovnel %edx, %eax
|
|
; X86-CMOV-NEXT: xorl %edx, %edx
|
|
; X86-CMOV-NEXT: retl
|
|
;
|
|
; X64-LABEL: ctlo_i64_undef:
|
|
; X64: # %bb.0:
|
|
; X64-NEXT: notq %rdi
|
|
; X64-NEXT: bsrq %rdi, %rax
|
|
; X64-NEXT: xorq $63, %rax
|
|
; X64-NEXT: retq
|
|
;
|
|
; X86-CLZ-LABEL: ctlo_i64_undef:
|
|
; X86-CLZ: # %bb.0:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: testl %eax, %eax
|
|
; X86-CLZ-NEXT: jne .LBB7_1
|
|
; X86-CLZ-NEXT: # %bb.2:
|
|
; X86-CLZ-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
; X86-CLZ-NEXT: notl %eax
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: addl $32, %eax
|
|
; X86-CLZ-NEXT: xorl %edx, %edx
|
|
; X86-CLZ-NEXT: retl
|
|
; X86-CLZ-NEXT: .LBB7_1:
|
|
; X86-CLZ-NEXT: lzcntl %eax, %eax
|
|
; X86-CLZ-NEXT: xorl %edx, %edx
|
|
; X86-CLZ-NEXT: retl
|
|
;
|
|
; X64-CLZ-LABEL: ctlo_i64_undef:
|
|
; X64-CLZ: # %bb.0:
|
|
; X64-CLZ-NEXT: notq %rdi
|
|
; X64-CLZ-NEXT: lzcntq %rdi, %rax
|
|
; X64-CLZ-NEXT: retq
|
|
%tmp1 = xor i64 %x, -1
|
|
%tmp2 = call i64 @llvm.ctlz.i64( i64 %tmp1, i1 true )
|
|
ret i64 %tmp2
|
|
}
|