2018-02-23 23:55:27 +08:00
|
|
|
; 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-NOBMI
|
|
|
|
|
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+bmi2 | FileCheck %s --check-prefixes=X86,X86-BMI
|
|
|
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=X64,X64-NOBMI
|
|
|
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+bmi2 | FileCheck %s --check-prefixes=X64,X64-BMI
|
|
|
|
|
|
2007-02-14 07:41:24 +08:00
|
|
|
; PR1198
|
|
|
|
|
|
2018-02-23 23:55:27 +08:00
|
|
|
define i64 @foo(i64 %x, i64 %y) nounwind {
|
|
|
|
|
; X86-NOBMI-LABEL: foo:
|
|
|
|
|
; X86-NOBMI: # %bb.0:
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %edi
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %esi
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ecx, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %esi, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebx, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: adcl $0, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ecx, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %edi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: addl %esi, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %ebp, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: setb %al
|
|
|
|
|
; X86-NOBMI-NEXT: movzbl %al, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %edi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebx, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %ecx, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %ecx, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %edi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebx, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %edi, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebp, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %esi, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: popl %esi
|
|
|
|
|
; X86-NOBMI-NEXT: popl %edi
|
|
|
|
|
; X86-NOBMI-NEXT: popl %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: popl %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: retl
|
|
|
|
|
;
|
|
|
|
|
; X86-BMI-LABEL: foo:
|
|
|
|
|
; X86-BMI: # %bb.0:
|
|
|
|
|
; X86-BMI-NEXT: pushl %ebp
|
|
|
|
|
; X86-BMI-NEXT: pushl %ebx
|
|
|
|
|
; X86-BMI-NEXT: pushl %edi
|
|
|
|
|
; X86-BMI-NEXT: pushl %esi
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %esi
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %edi
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %esi, %edx, %ebx
|
|
|
|
|
; X86-BMI-NEXT: movl %eax, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %esi, %ebp, %eax
|
|
|
|
|
; X86-BMI-NEXT: addl %ebx, %ebp
|
|
|
|
|
; X86-BMI-NEXT: adcl $0, %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %edi, %edx, %ebx
|
|
|
|
|
; X86-BMI-NEXT: addl %ebp, %edx
|
|
|
|
|
; X86-BMI-NEXT: adcl %eax, %ebx
|
|
|
|
|
; X86-BMI-NEXT: setb %al
|
|
|
|
|
; X86-BMI-NEXT: movzbl %al, %eax
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %edi, %edi, %ebp
|
|
|
|
|
; X86-BMI-NEXT: addl %ebx, %edi
|
|
|
|
|
; X86-BMI-NEXT: adcl %eax, %ebp
|
|
|
|
|
; X86-BMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %esi, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %eax, %ebx, %esi
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %eax, %eax, %edx
|
|
|
|
|
; X86-BMI-NEXT: addl %ebx, %eax
|
|
|
|
|
; X86-BMI-NEXT: adcl %esi, %edx
|
|
|
|
|
; X86-BMI-NEXT: addl %edi, %eax
|
|
|
|
|
; X86-BMI-NEXT: adcl %ebp, %edx
|
|
|
|
|
; X86-BMI-NEXT: popl %esi
|
|
|
|
|
; X86-BMI-NEXT: popl %edi
|
|
|
|
|
; X86-BMI-NEXT: popl %ebx
|
|
|
|
|
; X86-BMI-NEXT: popl %ebp
|
|
|
|
|
; X86-BMI-NEXT: retl
|
|
|
|
|
;
|
|
|
|
|
; X64-NOBMI-LABEL: foo:
|
|
|
|
|
; X64-NOBMI: # %bb.0:
|
|
|
|
|
; X64-NOBMI-NEXT: movq %rdi, %rax
|
|
|
|
|
; X64-NOBMI-NEXT: mulq %rsi
|
|
|
|
|
; X64-NOBMI-NEXT: movq %rdx, %rax
|
|
|
|
|
; X64-NOBMI-NEXT: retq
|
|
|
|
|
;
|
|
|
|
|
; X64-BMI-LABEL: foo:
|
|
|
|
|
; X64-BMI: # %bb.0:
|
|
|
|
|
; X64-BMI-NEXT: movq %rdi, %rdx
|
|
|
|
|
; X64-BMI-NEXT: mulxq %rsi, %rcx, %rax
|
|
|
|
|
; X64-BMI-NEXT: retq
|
|
|
|
|
%tmp0 = zext i64 %x to i128
|
|
|
|
|
%tmp1 = zext i64 %y to i128
|
|
|
|
|
%tmp2 = mul i128 %tmp0, %tmp1
|
|
|
|
|
%tmp7 = zext i32 64 to i128
|
|
|
|
|
%tmp3 = lshr i128 %tmp2, %tmp7
|
|
|
|
|
%tmp4 = trunc i128 %tmp3 to i64
|
|
|
|
|
ret i64 %tmp4
|
2007-02-14 07:41:24 +08:00
|
|
|
}
|
2013-06-19 04:14:39 +08:00
|
|
|
|
|
|
|
|
; <rdar://problem/14096009> superfluous multiply by high part of
|
|
|
|
|
; zero-extended value.
|
2018-02-23 23:55:27 +08:00
|
|
|
|
|
|
|
|
define i64 @mul1(i64 %n, i64* nocapture %z, i64* nocapture %x, i64 %y) nounwind {
|
|
|
|
|
; X86-NOBMI-LABEL: mul1:
|
|
|
|
|
; X86-NOBMI: # %bb.0: # %entry
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %edi
|
|
|
|
|
; X86-NOBMI-NEXT: pushl %esi
|
|
|
|
|
; X86-NOBMI-NEXT: subl $28, %esp
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: orl %ecx, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: je .LBB1_3
|
|
|
|
|
; X86-NOBMI-NEXT: # %bb.1: # %for.body.preheader
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %ebx, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: movl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
|
|
|
|
|
; X86-NOBMI-NEXT: .p2align 4, 0x90
|
|
|
|
|
; X86-NOBMI-NEXT: .LBB1_2: # %for.body
|
|
|
|
|
; X86-NOBMI-NEXT: # =>This Inner Loop Header: Depth=1
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %ecx
|
2018-02-28 00:59:10 +08:00
|
|
|
; X86-NOBMI-NEXT: movl (%eax,%ebx,8), %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl 4(%eax,%ebx,8), %esi
|
2018-02-23 23:55:27 +08:00
|
|
|
; X86-NOBMI-NEXT: movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ebp, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ebp, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %edi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl %esi, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: mull %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: addl %edi, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: adcl $0, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ebp, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
|
|
|
; X86-NOBMI-NEXT: mull %edx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %edi
|
|
|
|
|
; X86-NOBMI-NEXT: addl %esi, %edi
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %ecx, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: setb {{[0-9]+}}(%esp) # 1-byte Folded Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
|
|
|
|
|
; X86-NOBMI-NEXT: mull {{[0-9]+}}(%esp)
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebp, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: movzbl {{[0-9]+}}(%esp), %eax # 1-byte Folded Reload
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %eax, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: mull %edx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: movl %eax, %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %eax # 4-byte Reload
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: mull %edx
|
|
|
|
|
; X86-NOBMI-NEXT: addl %ebp, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: adcl {{[0-9]+}}(%esp), %edx # 4-byte Folded Reload
|
|
|
|
|
; X86-NOBMI-NEXT: addl %esi, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: adcl %ecx, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
|
|
|
|
|
; X86-NOBMI-NEXT: addl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
|
|
|
|
|
; X86-NOBMI-NEXT: adcl {{[0-9]+}}(%esp), %edi # 4-byte Folded Reload
|
|
|
|
|
; X86-NOBMI-NEXT: adcl $0, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: adcl $0, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %esi, (%ecx,%ebx,8)
|
|
|
|
|
; X86-NOBMI-NEXT: movl %edi, 4(%ecx,%ebx,8)
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ecx, %edi
|
|
|
|
|
; X86-NOBMI-NEXT: addl $1, %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
|
|
|
|
|
; X86-NOBMI-NEXT: adcl $0, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: movl %ebx, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %ebp, %ecx
|
|
|
|
|
; X86-NOBMI-NEXT: movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %edi, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: orl %ecx, %esi
|
|
|
|
|
; X86-NOBMI-NEXT: jne .LBB1_2
|
|
|
|
|
; X86-NOBMI-NEXT: .LBB1_3: # %for.end
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X86-NOBMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-NOBMI-NEXT: addl $28, %esp
|
|
|
|
|
; X86-NOBMI-NEXT: popl %esi
|
|
|
|
|
; X86-NOBMI-NEXT: popl %edi
|
|
|
|
|
; X86-NOBMI-NEXT: popl %ebx
|
|
|
|
|
; X86-NOBMI-NEXT: popl %ebp
|
|
|
|
|
; X86-NOBMI-NEXT: retl
|
|
|
|
|
;
|
|
|
|
|
; X86-BMI-LABEL: mul1:
|
|
|
|
|
; X86-BMI: # %bb.0: # %entry
|
|
|
|
|
; X86-BMI-NEXT: pushl %ebp
|
|
|
|
|
; X86-BMI-NEXT: pushl %ebx
|
|
|
|
|
; X86-BMI-NEXT: pushl %edi
|
|
|
|
|
; X86-BMI-NEXT: pushl %esi
|
|
|
|
|
; X86-BMI-NEXT: subl $20, %esp
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %ecx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: orl %ecx, %eax
|
|
|
|
|
; X86-BMI-NEXT: je .LBB1_3
|
|
|
|
|
; X86-BMI-NEXT: # %bb.1: # %for.body.preheader
|
|
|
|
|
; X86-BMI-NEXT: xorl %ecx, %ecx
|
|
|
|
|
; X86-BMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-BMI-NEXT: xorl %edi, %edi
|
|
|
|
|
; X86-BMI-NEXT: movl $0, {{[0-9]+}}(%esp) # 4-byte Folded Spill
|
|
|
|
|
; X86-BMI-NEXT: .p2align 4, 0x90
|
|
|
|
|
; X86-BMI-NEXT: .LBB1_2: # %for.body
|
|
|
|
|
; X86-BMI-NEXT: # =>This Inner Loop Header: Depth=1
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: movl (%eax,%edi,8), %ecx
|
|
|
|
|
; X86-BMI-NEXT: movl 4(%eax,%edi,8), %ebx
|
|
|
|
|
; X86-BMI-NEXT: movl %ebx, (%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %eax, %esi
|
2018-02-28 00:59:10 +08:00
|
|
|
; X86-BMI-NEXT: mulxl %eax, %eax, %ebp
|
2018-02-23 23:55:27 +08:00
|
|
|
; X86-BMI-NEXT: movl %eax, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl %ebx, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %esi, %eax, %esi
|
|
|
|
|
; X86-BMI-NEXT: addl %ebp, %eax
|
|
|
|
|
; X86-BMI-NEXT: adcl $0, %esi
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: mulxl {{[0-9]+}}(%esp), %ebp, %ebx
|
|
|
|
|
; X86-BMI-NEXT: addl %eax, %ebp
|
|
|
|
|
; X86-BMI-NEXT: adcl %esi, %ebx
|
|
|
|
|
; X86-BMI-NEXT: movl (%esp), %edx # 4-byte Reload
|
|
|
|
|
; X86-BMI-NEXT: mulxl {{[0-9]+}}(%esp), %eax, %esi
|
|
|
|
|
; X86-BMI-NEXT: setb %dl
|
|
|
|
|
; X86-BMI-NEXT: addl %ebx, %eax
|
|
|
|
|
; X86-BMI-NEXT: movzbl %dl, %edx
|
|
|
|
|
; X86-BMI-NEXT: adcl %edx, %esi
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %edx
|
|
|
|
|
; X86-BMI-NEXT: xorl %ebx, %ebx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %ebx, %ebx, %edx
|
|
|
|
|
; X86-BMI-NEXT: movl %edx, (%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: xorl %ecx, %ecx
|
|
|
|
|
; X86-BMI-NEXT: mulxl %ecx, %ecx, %edx
|
|
|
|
|
; X86-BMI-NEXT: addl %ebx, %ecx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %ebx
|
|
|
|
|
; X86-BMI-NEXT: adcl (%esp), %edx # 4-byte Folded Reload
|
|
|
|
|
; X86-BMI-NEXT: addl %eax, %ecx
|
|
|
|
|
; X86-BMI-NEXT: adcl %esi, %edx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
|
|
|
|
|
; X86-BMI-NEXT: addl {{[0-9]+}}(%esp), %esi # 4-byte Folded Reload
|
|
|
|
|
; X86-BMI-NEXT: adcl {{[0-9]+}}(%esp), %ebp # 4-byte Folded Reload
|
|
|
|
|
; X86-BMI-NEXT: adcl $0, %ecx
|
|
|
|
|
; X86-BMI-NEXT: adcl $0, %edx
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %esi, (%eax,%edi,8)
|
|
|
|
|
; X86-BMI-NEXT: movl %ebp, 4(%eax,%edi,8)
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %eax, %esi
|
|
|
|
|
; X86-BMI-NEXT: addl $1, %edi
|
|
|
|
|
; X86-BMI-NEXT: movl {{[0-9]+}}(%esp), %ebp # 4-byte Reload
|
|
|
|
|
; X86-BMI-NEXT: adcl $0, %ebp
|
|
|
|
|
; X86-BMI-NEXT: movl %edi, %eax
|
|
|
|
|
; X86-BMI-NEXT: xorl %esi, %eax
|
|
|
|
|
; X86-BMI-NEXT: movl %ebp, {{[0-9]+}}(%esp) # 4-byte Spill
|
|
|
|
|
; X86-BMI-NEXT: movl %ebp, %esi
|
|
|
|
|
; X86-BMI-NEXT: xorl %ebx, %esi
|
|
|
|
|
; X86-BMI-NEXT: orl %eax, %esi
|
|
|
|
|
; X86-BMI-NEXT: jne .LBB1_2
|
|
|
|
|
; X86-BMI-NEXT: .LBB1_3: # %for.end
|
|
|
|
|
; X86-BMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X86-BMI-NEXT: xorl %edx, %edx
|
|
|
|
|
; X86-BMI-NEXT: addl $20, %esp
|
|
|
|
|
; X86-BMI-NEXT: popl %esi
|
|
|
|
|
; X86-BMI-NEXT: popl %edi
|
|
|
|
|
; X86-BMI-NEXT: popl %ebx
|
|
|
|
|
; X86-BMI-NEXT: popl %ebp
|
|
|
|
|
; X86-BMI-NEXT: retl
|
|
|
|
|
;
|
|
|
|
|
; X64-NOBMI-LABEL: mul1:
|
|
|
|
|
; X64-NOBMI: # %bb.0: # %entry
|
|
|
|
|
; X64-NOBMI-NEXT: testq %rdi, %rdi
|
|
|
|
|
; X64-NOBMI-NEXT: je .LBB1_3
|
|
|
|
|
; X64-NOBMI-NEXT: # %bb.1: # %for.body.preheader
|
[CodeGen] Add a new pass for PostRA sink
Summary:
This pass sinks COPY instructions into a successor block, if the COPY is not
used in the current block and the COPY is live-in to a single successor
(i.e., doesn't require the COPY to be duplicated). This avoids executing the
the copy on paths where their results aren't needed. This also exposes
additional opportunites for dead copy elimination and shrink wrapping.
These copies were either not handled by or are inserted after the MachineSink
pass. As an example of the former case, the MachineSink pass cannot sink
COPY instructions with allocatable source registers; for AArch64 these type
of copy instructions are frequently used to move function parameters (PhyReg)
into virtual registers in the entry block..
For the machine IR below, this pass will sink %w19 in the entry into its
successor (%bb.1) because %w19 is only live-in in %bb.1.
```
%bb.0:
%wzr = SUBSWri %w1, 1
%w19 = COPY %w0
Bcc 11, %bb.2
%bb.1:
Live Ins: %w19
BL @fun
%w0 = ADDWrr %w0, %w19
RET %w0
%bb.2:
%w0 = COPY %wzr
RET %w0
```
As we sink %w19 (CSR in AArch64) into %bb.1, the shrink-wrapping pass will be
able to see %bb.0 as a candidate.
With this change I observed 12% more shrink-wrapping candidate and 13% more dead copies deleted in spec2000/2006/2017 on AArch64.
Reviewers: qcolombet, MatzeB, thegameg, mcrosier, gberry, hfinkel, john.brawn, twoh, RKSimon, sebpop, kparzysz
Reviewed By: sebpop
Subscribers: evandro, sebpop, sfertile, aemerson, mgorny, javed.absar, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D41463
llvm-svn: 328237
2018-03-23 04:06:47 +08:00
|
|
|
; X64-NOBMI-NEXT: movq %rcx, %r8
|
|
|
|
|
; X64-NOBMI-NEXT: movq %rdx, %r9
|
2018-02-23 23:55:27 +08:00
|
|
|
; X64-NOBMI-NEXT: xorl %r10d, %r10d
|
|
|
|
|
; X64-NOBMI-NEXT: xorl %ecx, %ecx
|
|
|
|
|
; X64-NOBMI-NEXT: .p2align 4, 0x90
|
|
|
|
|
; X64-NOBMI-NEXT: .LBB1_2: # %for.body
|
|
|
|
|
; X64-NOBMI-NEXT: # =>This Inner Loop Header: Depth=1
|
|
|
|
|
; X64-NOBMI-NEXT: movq %r8, %rax
|
|
|
|
|
; X64-NOBMI-NEXT: mulq (%r9,%rcx,8)
|
|
|
|
|
; X64-NOBMI-NEXT: addq %r10, %rax
|
|
|
|
|
; X64-NOBMI-NEXT: adcq $0, %rdx
|
|
|
|
|
; X64-NOBMI-NEXT: movq %rax, (%rsi,%rcx,8)
|
|
|
|
|
; X64-NOBMI-NEXT: incq %rcx
|
|
|
|
|
; X64-NOBMI-NEXT: cmpq %rcx, %rdi
|
|
|
|
|
; X64-NOBMI-NEXT: movq %rdx, %r10
|
|
|
|
|
; X64-NOBMI-NEXT: jne .LBB1_2
|
|
|
|
|
; X64-NOBMI-NEXT: .LBB1_3: # %for.end
|
|
|
|
|
; X64-NOBMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X64-NOBMI-NEXT: retq
|
|
|
|
|
;
|
|
|
|
|
; X64-BMI-LABEL: mul1:
|
|
|
|
|
; X64-BMI: # %bb.0: # %entry
|
|
|
|
|
; X64-BMI-NEXT: testq %rdi, %rdi
|
|
|
|
|
; X64-BMI-NEXT: je .LBB1_3
|
|
|
|
|
; X64-BMI-NEXT: # %bb.1: # %for.body.preheader
|
[CodeGen] Add a new pass for PostRA sink
Summary:
This pass sinks COPY instructions into a successor block, if the COPY is not
used in the current block and the COPY is live-in to a single successor
(i.e., doesn't require the COPY to be duplicated). This avoids executing the
the copy on paths where their results aren't needed. This also exposes
additional opportunites for dead copy elimination and shrink wrapping.
These copies were either not handled by or are inserted after the MachineSink
pass. As an example of the former case, the MachineSink pass cannot sink
COPY instructions with allocatable source registers; for AArch64 these type
of copy instructions are frequently used to move function parameters (PhyReg)
into virtual registers in the entry block..
For the machine IR below, this pass will sink %w19 in the entry into its
successor (%bb.1) because %w19 is only live-in in %bb.1.
```
%bb.0:
%wzr = SUBSWri %w1, 1
%w19 = COPY %w0
Bcc 11, %bb.2
%bb.1:
Live Ins: %w19
BL @fun
%w0 = ADDWrr %w0, %w19
RET %w0
%bb.2:
%w0 = COPY %wzr
RET %w0
```
As we sink %w19 (CSR in AArch64) into %bb.1, the shrink-wrapping pass will be
able to see %bb.0 as a candidate.
With this change I observed 12% more shrink-wrapping candidate and 13% more dead copies deleted in spec2000/2006/2017 on AArch64.
Reviewers: qcolombet, MatzeB, thegameg, mcrosier, gberry, hfinkel, john.brawn, twoh, RKSimon, sebpop, kparzysz
Reviewed By: sebpop
Subscribers: evandro, sebpop, sfertile, aemerson, mgorny, javed.absar, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D41463
llvm-svn: 328237
2018-03-23 04:06:47 +08:00
|
|
|
; X64-BMI-NEXT: movq %rcx, %r8
|
|
|
|
|
; X64-BMI-NEXT: movq %rdx, %r9
|
2018-02-23 23:55:27 +08:00
|
|
|
; X64-BMI-NEXT: xorl %r10d, %r10d
|
2018-09-20 02:59:08 +08:00
|
|
|
; X64-BMI-NEXT: xorl %ecx, %ecx
|
2018-02-23 23:55:27 +08:00
|
|
|
; X64-BMI-NEXT: .p2align 4, 0x90
|
|
|
|
|
; X64-BMI-NEXT: .LBB1_2: # %for.body
|
|
|
|
|
; X64-BMI-NEXT: # =>This Inner Loop Header: Depth=1
|
|
|
|
|
; X64-BMI-NEXT: movq %r8, %rdx
|
2018-09-20 02:59:08 +08:00
|
|
|
; X64-BMI-NEXT: mulxq (%r9,%rcx,8), %rax, %rdx
|
|
|
|
|
; X64-BMI-NEXT: addq %r10, %rax
|
2018-02-23 23:55:27 +08:00
|
|
|
; X64-BMI-NEXT: adcq $0, %rdx
|
2018-09-20 02:59:08 +08:00
|
|
|
; X64-BMI-NEXT: movq %rax, (%rsi,%rcx,8)
|
|
|
|
|
; X64-BMI-NEXT: incq %rcx
|
|
|
|
|
; X64-BMI-NEXT: cmpq %rcx, %rdi
|
2018-02-23 23:55:27 +08:00
|
|
|
; X64-BMI-NEXT: movq %rdx, %r10
|
|
|
|
|
; X64-BMI-NEXT: jne .LBB1_2
|
|
|
|
|
; X64-BMI-NEXT: .LBB1_3: # %for.end
|
|
|
|
|
; X64-BMI-NEXT: xorl %eax, %eax
|
|
|
|
|
; X64-BMI-NEXT: retq
|
2013-06-19 04:14:39 +08:00
|
|
|
entry:
|
|
|
|
|
%conv = zext i64 %y to i128
|
|
|
|
|
%cmp11 = icmp eq i64 %n, 0
|
|
|
|
|
br i1 %cmp11, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %entry, %for.body
|
|
|
|
|
%carry.013 = phi i64 [ %conv6, %for.body ], [ 0, %entry ]
|
|
|
|
|
%i.012 = phi i64 [ %inc, %for.body ], [ 0, %entry ]
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx = getelementptr inbounds i64, i64* %x, i64 %i.012
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i64, i64* %arrayidx, align 8
|
2013-06-19 04:14:39 +08:00
|
|
|
%conv2 = zext i64 %0 to i128
|
|
|
|
|
%mul = mul i128 %conv2, %conv
|
|
|
|
|
%conv3 = zext i64 %carry.013 to i128
|
|
|
|
|
%add = add i128 %mul, %conv3
|
|
|
|
|
%conv4 = trunc i128 %add to i64
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx5 = getelementptr inbounds i64, i64* %z, i64 %i.012
|
2013-06-19 04:14:39 +08:00
|
|
|
store i64 %conv4, i64* %arrayidx5, align 8
|
|
|
|
|
%shr = lshr i128 %add, 64
|
|
|
|
|
%conv6 = trunc i128 %shr to i64
|
|
|
|
|
%inc = add i64 %i.012, 1
|
|
|
|
|
%exitcond = icmp eq i64 %inc, %n
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.body, %entry
|
|
|
|
|
ret i64 0
|
|
|
|
|
}
|