c65b4d64d4
Alignment of an alloca in IR can be lower than the preferred alignment on purpose, but this override essentially treats the preferred alignment as the minimum alignment. The patch changes this behavior to always use the specified alignment. If alignment is not set explicitly in LLVM IR, it is set to DL.getPrefTypeAlign(Ty) in computeAllocaDefaultAlign. Tests are changed as well: explicit alignment is increased to match the preferred alignment if it changes output, or omitted when it is hard to determine the right value (e.g. for pointers, some structs, or weird types). Differential Revision: https://reviews.llvm.org/D135462
29 lines
713 B
LLVM
29 lines
713 B
LLVM
; RUN: llc -mtriple=aarch64 -O0 -fast-isel < %s | FileCheck %s
|
|
|
|
; Function Attrs: nounwind
|
|
define i32 @foo() #0 {
|
|
entry:
|
|
%c = alloca i8
|
|
; CHECK: add x0, sp, #12
|
|
%s = alloca i16
|
|
; CHECK-NEXT: add x1, sp, #8
|
|
%i = alloca i32
|
|
; CHECK-NEXT: add x2, sp, #4
|
|
%call = call i32 @bar(ptr %c, ptr %s, ptr %i)
|
|
%0 = load i8, ptr %c, align 1
|
|
%conv = zext i8 %0 to i32
|
|
%add = add nsw i32 %call, %conv
|
|
%1 = load i16, ptr %s, align 2
|
|
%conv1 = sext i16 %1 to i32
|
|
%add2 = add nsw i32 %add, %conv1
|
|
%2 = load i32, ptr %i, align 4
|
|
%add3 = add nsw i32 %add2, %2
|
|
ret i32 %add3
|
|
}
|
|
|
|
declare i32 @bar(ptr, ptr, ptr) #1
|
|
|
|
attributes #0 = { nounwind "frame-pointer"="none" }
|
|
attributes #1 = { "frame-pointer"="none" }
|
|
|