bf789b1957
For the longest time we used `AAValueSimplify` and
`genericValueTraversal` to determine "potential values". This was
problematic for many reasons:
- We recomputed the result a lot as there was no caching for the 9
locations calling `genericValueTraversal`.
- We added the idea of "intra" vs. "inter" procedural simplification
only as an afterthought. `genericValueTraversal` did offer an option
but `AAValueSimplify` did not. Thus, we might end up with "too much"
simplification in certain situations and then gave up on it.
- Because `genericValueTraversal` was not a real `AA` we ended up with
problems like the infinite recursion bug (#54981) as well as code
duplication.
This patch introduces `AAPotentialValues` and replaces the
`AAValueSimplify` uses with it. `genericValueTraversal` is folded into
`AAPotentialValues` as are the instruction simplifications performed in
`AAValueSimplify` before. We further distinguish "intra" and "inter"
procedural simplification now.
`AAValueSimplify` was not deleted as we haven't ported the
re-materialization of instructions yet. There are other differences over
the former handling, e.g., we may not fold trivially foldable
instructions right now, e.g., `add i32 1, 1` is not folded to `i32 2`
but if an operand would be simplified to `i32 1` we would fold it still.
We are also even more aware of function/SCC boundaries in CGSCC passes,
which is good even if some tests look like they regress.
Fixes: https://github.com/llvm/llvm-project/issues/54981
Note: A previous version was flawed and consequently reverted in
6555558a80.
157 lines
7.9 KiB
LLVM
157 lines
7.9 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes --check-globals
|
|
; RUN: opt -attributor -enable-new-pm=0 -attributor-manifest-internal -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=2 -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_CGSCC_NPM,NOT_CGSCC_OPM,NOT_TUNIT_NPM,IS__TUNIT____,IS________OPM,IS__TUNIT_OPM
|
|
; RUN: opt -aa-pipeline=basic-aa -passes=attributor -attributor-manifest-internal -attributor-max-iterations-verify -attributor-annotate-decl-cs -attributor-max-iterations=2 -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_CGSCC_OPM,NOT_CGSCC_NPM,NOT_TUNIT_OPM,IS__TUNIT____,IS________NPM,IS__TUNIT_NPM
|
|
; RUN: opt -attributor-cgscc -enable-new-pm=0 -attributor-manifest-internal -attributor-annotate-decl-cs -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_TUNIT_NPM,NOT_TUNIT_OPM,NOT_CGSCC_NPM,IS__CGSCC____,IS________OPM,IS__CGSCC_OPM
|
|
; RUN: opt -aa-pipeline=basic-aa -passes=attributor-cgscc -attributor-manifest-internal -attributor-annotate-decl-cs -S < %s | FileCheck %s --check-prefixes=CHECK,NOT_TUNIT_NPM,NOT_TUNIT_OPM,NOT_CGSCC_OPM,IS__CGSCC____,IS________NPM,IS__CGSCC_NPM
|
|
|
|
define void @f() {
|
|
; IS________OPM-LABEL: define {{[^@]+}}@f() {
|
|
; IS________OPM-NEXT: entry:
|
|
; IS________OPM-NEXT: [[A:%.*]] = alloca i32, align 1
|
|
; IS________OPM-NEXT: call void @g(i32* noalias nocapture nofree noundef nonnull readonly dereferenceable(4) [[A]])
|
|
; IS________OPM-NEXT: ret void
|
|
;
|
|
; IS__TUNIT_NPM-LABEL: define {{[^@]+}}@f() {
|
|
; IS__TUNIT_NPM-NEXT: entry:
|
|
; IS__TUNIT_NPM-NEXT: [[A:%.*]] = alloca i32, align 1
|
|
; IS__TUNIT_NPM-NEXT: [[TMP0:%.*]] = load i32, i32* [[A]], align 1
|
|
; IS__TUNIT_NPM-NEXT: call void @g(i32 [[TMP0]])
|
|
; IS__TUNIT_NPM-NEXT: ret void
|
|
;
|
|
; IS__CGSCC_NPM-LABEL: define {{[^@]+}}@f() {
|
|
; IS__CGSCC_NPM-NEXT: entry:
|
|
; IS__CGSCC_NPM-NEXT: call void @g(i32 undef)
|
|
; IS__CGSCC_NPM-NEXT: ret void
|
|
;
|
|
entry:
|
|
%a = alloca i32, align 1
|
|
call void @g(i32* %a)
|
|
ret void
|
|
}
|
|
|
|
define internal void @g(i32* %a) {
|
|
; IS________OPM-LABEL: define {{[^@]+}}@g
|
|
; IS________OPM-SAME: (i32* noalias nocapture nofree noundef nonnull readonly dereferenceable(4) [[A:%.*]]) {
|
|
; IS________OPM-NEXT: [[AA:%.*]] = load i32, i32* [[A]], align 1
|
|
; IS________OPM-NEXT: call void @z(i32 [[AA]])
|
|
; IS________OPM-NEXT: ret void
|
|
;
|
|
; IS________NPM-LABEL: define {{[^@]+}}@g
|
|
; IS________NPM-SAME: (i32 [[TMP0:%.*]]) {
|
|
; IS________NPM-NEXT: [[A_PRIV:%.*]] = alloca i32, align 4
|
|
; IS________NPM-NEXT: store i32 [[TMP0]], i32* [[A_PRIV]], align 4
|
|
; IS________NPM-NEXT: [[AA:%.*]] = load i32, i32* [[A_PRIV]], align 1
|
|
; IS________NPM-NEXT: call void @z(i32 [[AA]])
|
|
; IS________NPM-NEXT: ret void
|
|
;
|
|
%aa = load i32, i32* %a, align 1
|
|
call void @z(i32 %aa)
|
|
ret void
|
|
}
|
|
|
|
declare void @z(i32)
|
|
|
|
; Test2
|
|
; Different alignemnt privatizable arguments
|
|
define internal i32 @test(i32* %X, i64* %Y) {
|
|
; IS__CGSCC_OPM: Function Attrs: argmemonly nofree norecurse nosync nounwind readonly willreturn
|
|
; IS__CGSCC_OPM-LABEL: define {{[^@]+}}@test
|
|
; IS__CGSCC_OPM-SAME: (i32* nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[X:%.*]], i64* noalias nocapture nofree noundef nonnull readonly align 8 dereferenceable(8) [[Y:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
; IS__CGSCC_OPM-NEXT: [[A:%.*]] = load i32, i32* [[X]], align 4
|
|
; IS__CGSCC_OPM-NEXT: [[B:%.*]] = load i64, i64* [[Y]], align 8
|
|
; IS__CGSCC_OPM-NEXT: [[C:%.*]] = add i32 [[A]], 1
|
|
; IS__CGSCC_OPM-NEXT: [[D:%.*]] = add i64 [[B]], 1
|
|
; IS__CGSCC_OPM-NEXT: [[COND:%.*]] = icmp sgt i64 [[D]], -1
|
|
; IS__CGSCC_OPM-NEXT: br i1 [[COND]], label [[RETURN1:%.*]], label [[RETURN2:%.*]]
|
|
; IS__CGSCC_OPM: Return1:
|
|
; IS__CGSCC_OPM-NEXT: ret i32 [[C]]
|
|
; IS__CGSCC_OPM: Return2:
|
|
; IS__CGSCC_OPM-NEXT: ret i32 [[A]]
|
|
;
|
|
; IS__CGSCC_NPM: Function Attrs: argmemonly nofree norecurse nosync nounwind readonly willreturn
|
|
; IS__CGSCC_NPM-LABEL: define {{[^@]+}}@test
|
|
; IS__CGSCC_NPM-SAME: (i32* nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[X:%.*]], i64 [[TMP0:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
; IS__CGSCC_NPM-NEXT: [[Y_PRIV:%.*]] = alloca i64, align 8
|
|
; IS__CGSCC_NPM-NEXT: store i64 [[TMP0]], i64* [[Y_PRIV]], align 4
|
|
; IS__CGSCC_NPM-NEXT: [[A:%.*]] = load i32, i32* [[X]], align 4
|
|
; IS__CGSCC_NPM-NEXT: [[B:%.*]] = load i64, i64* [[Y_PRIV]], align 8
|
|
; IS__CGSCC_NPM-NEXT: [[C:%.*]] = add i32 [[A]], 1
|
|
; IS__CGSCC_NPM-NEXT: [[D:%.*]] = add i64 [[B]], 1
|
|
; IS__CGSCC_NPM-NEXT: [[COND:%.*]] = icmp sgt i64 [[D]], -1
|
|
; IS__CGSCC_NPM-NEXT: br i1 [[COND]], label [[RETURN1:%.*]], label [[RETURN2:%.*]]
|
|
; IS__CGSCC_NPM: Return1:
|
|
; IS__CGSCC_NPM-NEXT: ret i32 [[C]]
|
|
; IS__CGSCC_NPM: Return2:
|
|
; IS__CGSCC_NPM-NEXT: ret i32 [[A]]
|
|
;
|
|
%A = load i32, i32* %X
|
|
%B = load i64, i64* %Y
|
|
%C = add i32 %A, 1
|
|
%D = add i64 %B, 1
|
|
%cond = icmp sgt i64 %D, -1
|
|
br i1 %cond, label %Return1, label %Return2
|
|
Return1:
|
|
ret i32 %C
|
|
Return2:
|
|
ret i32 %A
|
|
}
|
|
|
|
define internal i32 @caller(i32* %A) {
|
|
; IS__CGSCC_OPM: Function Attrs: argmemonly nofree nosync nounwind willreturn
|
|
; IS__CGSCC_OPM-LABEL: define {{[^@]+}}@caller
|
|
; IS__CGSCC_OPM-SAME: (i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
|
|
; IS__CGSCC_OPM-NEXT: [[B:%.*]] = alloca i64, align 8
|
|
; IS__CGSCC_OPM-NEXT: store i64 1, i64* [[B]], align 8
|
|
; IS__CGSCC_OPM-NEXT: [[C:%.*]] = call i32 @test(i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[A]], i64* noalias nocapture nofree noundef nonnull readonly align 8 dereferenceable(8) [[B]]) #[[ATTR3:[0-9]+]]
|
|
; IS__CGSCC_OPM-NEXT: ret i32 [[C]]
|
|
;
|
|
; IS__CGSCC_NPM: Function Attrs: argmemonly nofree nosync nounwind willreturn
|
|
; IS__CGSCC_NPM-LABEL: define {{[^@]+}}@caller
|
|
; IS__CGSCC_NPM-SAME: (i32 [[TMP0:%.*]]) #[[ATTR1:[0-9]+]] {
|
|
; IS__CGSCC_NPM-NEXT: [[A_PRIV:%.*]] = alloca i32, align 4
|
|
; IS__CGSCC_NPM-NEXT: store i32 [[TMP0]], i32* [[A_PRIV]], align 4
|
|
; IS__CGSCC_NPM-NEXT: [[C:%.*]] = call i32 @test(i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[A_PRIV]], i64 1) #[[ATTR3:[0-9]+]]
|
|
; IS__CGSCC_NPM-NEXT: ret i32 [[C]]
|
|
;
|
|
%B = alloca i64
|
|
store i64 1, i64* %B
|
|
%C = call i32 @test(i32* %A, i64* %B)
|
|
ret i32 %C
|
|
}
|
|
|
|
define i32 @callercaller() {
|
|
; IS__TUNIT____: Function Attrs: nofree norecurse nosync nounwind readnone willreturn
|
|
; IS__TUNIT____-LABEL: define {{[^@]+}}@callercaller
|
|
; IS__TUNIT____-SAME: () #[[ATTR0:[0-9]+]] {
|
|
; IS__TUNIT____-NEXT: [[B:%.*]] = alloca i32, align 4
|
|
; IS__TUNIT____-NEXT: ret i32 3
|
|
;
|
|
; IS__CGSCC_OPM: Function Attrs: nofree nosync nounwind readnone willreturn
|
|
; IS__CGSCC_OPM-LABEL: define {{[^@]+}}@callercaller
|
|
; IS__CGSCC_OPM-SAME: () #[[ATTR2:[0-9]+]] {
|
|
; IS__CGSCC_OPM-NEXT: [[B:%.*]] = alloca i32, align 4
|
|
; IS__CGSCC_OPM-NEXT: store i32 2, i32* [[B]], align 4
|
|
; IS__CGSCC_OPM-NEXT: [[X:%.*]] = call i32 @caller(i32* noalias nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) [[B]]) #[[ATTR4:[0-9]+]]
|
|
; IS__CGSCC_OPM-NEXT: ret i32 [[X]]
|
|
;
|
|
; IS__CGSCC_NPM: Function Attrs: nofree nosync nounwind readnone willreturn
|
|
; IS__CGSCC_NPM-LABEL: define {{[^@]+}}@callercaller
|
|
; IS__CGSCC_NPM-SAME: () #[[ATTR2:[0-9]+]] {
|
|
; IS__CGSCC_NPM-NEXT: [[X:%.*]] = call i32 @caller(i32 2) #[[ATTR4:[0-9]+]]
|
|
; IS__CGSCC_NPM-NEXT: ret i32 [[X]]
|
|
;
|
|
%B = alloca i32
|
|
store i32 2, i32* %B
|
|
%X = call i32 @caller(i32* %B)
|
|
ret i32 %X
|
|
}
|
|
;.
|
|
; IS__TUNIT____: attributes #[[ATTR0]] = { nofree norecurse nosync nounwind readnone willreturn }
|
|
;.
|
|
; IS__CGSCC____: attributes #[[ATTR0:[0-9]+]] = { argmemonly nofree norecurse nosync nounwind readonly willreturn }
|
|
; IS__CGSCC____: attributes #[[ATTR1:[0-9]+]] = { argmemonly nofree nosync nounwind willreturn }
|
|
; IS__CGSCC____: attributes #[[ATTR2:[0-9]+]] = { nofree nosync nounwind readnone willreturn }
|
|
; IS__CGSCC____: attributes #[[ATTR3:[0-9]+]] = { readonly willreturn }
|
|
; IS__CGSCC____: attributes #[[ATTR4:[0-9]+]] = { nounwind willreturn }
|
|
;.
|