68c1eeb4ba
This rewrites ArgPromotion to be based on offsets rather than GEP structure. We inspect all loads at constant offsets and remember which types are loaded at which offsets. Then we promote based on those types. This generalizes ArgPromotion to work with bitcasted loads, and is compatible with opaque pointers. This patch also fixes incorrect handling of alignment during argument promotion. Previously, the implementation only checked that the pointer is dereferenceable, but was happy to speculate overaligned loads. (I would have fixed this separately in advance, but I found this hard to do with the previous implementation approach). Differential Revision: https://reviews.llvm.org/D118685
67 lines
2.3 KiB
LLVM
67 lines
2.3 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --scrub-attributes
|
|
; RUN: opt -S -argpromotion < %s | FileCheck %s
|
|
|
|
; Test argument promotion involving bitcasts.
|
|
|
|
%opaque = type opaque
|
|
|
|
define internal i32 @callee_basic(i8* %p) {
|
|
; CHECK-LABEL: define {{[^@]+}}@callee_basic
|
|
; CHECK-SAME: (i32 [[P_0_VAL:%.*]], i32 [[P_4_VAL:%.*]]) {
|
|
; CHECK-NEXT: [[Z:%.*]] = add i32 [[P_0_VAL]], [[P_4_VAL]]
|
|
; CHECK-NEXT: ret i32 [[Z]]
|
|
;
|
|
%p.32 = bitcast i8* %p to i32*
|
|
%x = load i32, i32* %p.32
|
|
%p1 = getelementptr i8, i8* %p, i64 4
|
|
%p1.32 = bitcast i8* %p1 to i32*
|
|
%y = load i32, i32* %p1.32
|
|
%z = add i32 %x, %y
|
|
ret i32 %z
|
|
}
|
|
|
|
define void @caller_basic(i8* %p) {
|
|
; CHECK-LABEL: define {{[^@]+}}@caller_basic
|
|
; CHECK-SAME: (i8* [[P:%.*]]) {
|
|
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i8* [[P]] to i32*
|
|
; CHECK-NEXT: [[P_VAL:%.*]] = load i32, i32* [[TMP1]], align 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, i8* [[P]], i64 4
|
|
; CHECK-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
|
|
; CHECK-NEXT: [[P_VAL1:%.*]] = load i32, i32* [[TMP3]], align 4
|
|
; CHECK-NEXT: [[TMP4:%.*]] = call i32 @callee_basic(i32 [[P_VAL]], i32 [[P_VAL1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
call i32 @callee_basic(i8* %p)
|
|
ret void
|
|
}
|
|
|
|
define internal i32 @callee_opaque(%opaque* %p) {
|
|
; CHECK-LABEL: define {{[^@]+}}@callee_opaque
|
|
; CHECK-SAME: (i32 [[P_0_VAL:%.*]], i32 [[P_4_VAL:%.*]]) {
|
|
; CHECK-NEXT: [[Z:%.*]] = add i32 [[P_0_VAL]], [[P_4_VAL]]
|
|
; CHECK-NEXT: ret i32 [[Z]]
|
|
;
|
|
%p.32 = bitcast %opaque* %p to i32*
|
|
%x = load i32, i32* %p.32
|
|
%p1.32 = getelementptr i32, i32* %p.32, i64 1
|
|
%y = load i32, i32* %p1.32
|
|
%z = add i32 %x, %y
|
|
ret i32 %z
|
|
}
|
|
|
|
define void @caller_opaque(%opaque* %p) {
|
|
; CHECK-LABEL: define {{[^@]+}}@caller_opaque
|
|
; CHECK-SAME: (%opaque* [[P:%.*]]) {
|
|
; CHECK-NEXT: [[TMP1:%.*]] = bitcast %opaque* [[P]] to i32*
|
|
; CHECK-NEXT: [[P_VAL:%.*]] = load i32, i32* [[TMP1]], align 4
|
|
; CHECK-NEXT: [[TMP2:%.*]] = bitcast %opaque* [[P]] to i8*
|
|
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, i8* [[TMP2]], i64 4
|
|
; CHECK-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to i32*
|
|
; CHECK-NEXT: [[P_VAL1:%.*]] = load i32, i32* [[TMP4]], align 4
|
|
; CHECK-NEXT: [[TMP5:%.*]] = call i32 @callee_opaque(i32 [[P_VAL]], i32 [[P_VAL1]])
|
|
; CHECK-NEXT: ret void
|
|
;
|
|
call i32 @callee_opaque(%opaque* %p)
|
|
ret void
|
|
}
|