mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 19:23:23 +01:00
5fca27122d
As shown in PR46237: https://bugs.llvm.org/show_bug.cgi?id=46237 The size-savings win for hoisting an 8-bit ALU immediate (intentionally excluding store constants) requires extreme conditions; it may not even be possible when including REX prefix bytes on x86-64. I did draft a version of this patch that included use counts after the loop, but I suspect that accounting is not working as expected. I think that is because the number of constant uses are changing as we select instructions (for example as we transform shl/add into LEA). Differential Revision: https://reviews.llvm.org/D81468
55 lines
1.5 KiB
LLVM
55 lines
1.5 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
|
|
|
|
define i1 @foo(i32 %i) optsize {
|
|
; CHECK-LABEL: foo:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl $305419896, %eax # imm = 0x12345678
|
|
; CHECK-NEXT: andl %eax, %edi
|
|
; CHECK-NEXT: cmpl %eax, %edi
|
|
; CHECK-NEXT: sete %al
|
|
; CHECK-NEXT: retq
|
|
%and = and i32 %i, 305419896
|
|
%cmp = icmp eq i32 %and, 305419896
|
|
ret i1 %cmp
|
|
}
|
|
|
|
; 8-bit ALU immediates probably have small encodings.
|
|
; We do not want to hoist the constant into a register here.
|
|
|
|
define zeroext i1 @g(i32 %x) optsize {
|
|
; CHECK-LABEL: g:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: orl $1, %edi
|
|
; CHECK-NEXT: cmpl $1, %edi
|
|
; CHECK-NEXT: sete %al
|
|
; CHECK-NEXT: retq
|
|
%t0 = or i32 %x, 1
|
|
%t1 = icmp eq i32 %t0, 1
|
|
ret i1 %t1
|
|
}
|
|
|
|
; 8-bit ALU immediates probably have small encodings.
|
|
; We do not want to hoist the constant into a register here.
|
|
|
|
define i64 @PR46237(i64 %x, i64 %y, i64 %z) optsize {
|
|
; CHECK-LABEL: PR46237:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: movl %edx, %eax
|
|
; CHECK-NEXT: shll $6, %eax
|
|
; CHECK-NEXT: movzbl %al, %ecx
|
|
; CHECK-NEXT: andl $7, %esi
|
|
; CHECK-NEXT: andl $7, %edx
|
|
; CHECK-NEXT: leaq (%rdx,%rsi,8), %rax
|
|
; CHECK-NEXT: orq %rcx, %rax
|
|
; CHECK-NEXT: retq
|
|
%and = shl i64 %z, 6
|
|
%shl = and i64 %and, 192
|
|
%and1 = shl i64 %y, 3
|
|
%shl2 = and i64 %and1, 56
|
|
%and3 = and i64 %z, 7
|
|
%or = or i64 %and3, %shl2
|
|
%or4 = or i64 %or, %shl
|
|
ret i64 %or4
|
|
}
|