mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2025-01-07 11:51:13 +00:00
[BasicAA] Use GEP as context for computeKnownBits in aliasGEP.
In order to use assumptions, computeKnownBits needs a context instruction. We can use the GEP, if it is an instruction. We already pass the assumption cache, but it cannot be used without a context instruction. Reviewers: anemet, asbirlea, hfinkel, spatel Reviewed By: asbirlea Differential Revision: https://reviews.llvm.org/D71264
This commit is contained in:
Florian Hahn
parent
8950bb7c13
commit
1991907bad
@ -1482,7 +1482,8 @@ AliasResult BasicAAResult::aliasGEP(
|
||||
// give up if we can't determine conditions that hold for every cycle:
|
||||
const Value *V = DecompGEP1.VarIndices[i].V;
|
||||
|
||||
KnownBits Known = computeKnownBits(V, DL, 0, &AC, nullptr, DT);
|
||||
KnownBits Known =
|
||||
computeKnownBits(V, DL, 0, &AC, dyn_cast<Instruction>(GEP1), DT);
|
||||
bool SignKnownZero = Known.isNonNegative();
|
||||
bool SignKnownOne = Known.isNegative();
|
||||
|
||||
|
||||
116
test/Analysis/BasicAA/assume-index-positive.ll
Normal file
116
test/Analysis/BasicAA/assume-index-positive.ll
Normal file
@ -0,0 +1,116 @@
|
||||
; RUN: opt -basicaa -aa-eval -print-all-alias-modref-info %s 2>&1 | FileCheck %s
|
||||
|
||||
; %col.ptr.1 and %col.ptr.2 do not alias, if we know that %skip >= 0, because
|
||||
; the distance between %col.ptr.1 and %col.ptr.2 is %skip + 6 and we load 6
|
||||
; elements.
|
||||
define void @test1(double* %ptr, i32 %skip) {
|
||||
; CHECK-LABEL: Function: test1: 4 pointers, 1 call sites
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.1, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: double* %col.ptr.2, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.1, double* %col.ptr.2
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.2.cast, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.1, <6 x double>* %col.ptr.2.cast
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.2.cast, double* %col.ptr.2
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %ptr <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.1 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %col.ptr.2 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.2.cast <-> call void @llvm.assume(i1 %gt)
|
||||
;
|
||||
%gt = icmp sgt i32 %skip, -1
|
||||
call void @llvm.assume(i1 %gt)
|
||||
%stride = add nsw nuw i32 %skip, 6
|
||||
%col.ptr.1 = bitcast double* %ptr to <6 x double>*
|
||||
%lv.1 = load <6 x double>, <6 x double>* %col.ptr.1, align 8
|
||||
%col.ptr.2= getelementptr double, double* %ptr, i32 %stride
|
||||
%col.ptr.2.cast = bitcast double* %col.ptr.2 to <6 x double>*
|
||||
%lv.2 = load <6 x double>, <6 x double>* %col.ptr.2.cast, align 8
|
||||
%res.1 = fadd <6 x double> %lv.1, %lv.1
|
||||
%res.2 = fadd <6 x double> %lv.2, %lv.2
|
||||
store <6 x double> %res.1, <6 x double>* %col.ptr.1, align 8
|
||||
store <6 x double> %res.2, <6 x double>* %col.ptr.2.cast, align 8
|
||||
ret void
|
||||
}
|
||||
|
||||
; Same as @test1, but now we do not have an assume guaranteeing %skip >= 0.
|
||||
define void @test2(double* %ptr, i32 %skip) {
|
||||
; CHECK-LABEL: Function: test2: 4 pointers, 0 call sites
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.1, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: double* %col.ptr.2, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.1, double* %col.ptr.2
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.2.cast, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.1, <6 x double>* %col.ptr.2.cast
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.2.cast, double* %col.ptr.2
|
||||
;
|
||||
%stride = add nsw nuw i32 %skip, 6
|
||||
%col.ptr.1 = bitcast double* %ptr to <6 x double>*
|
||||
%lv.1 = load <6 x double>, <6 x double>* %col.ptr.1, align 8
|
||||
%col.ptr.2= getelementptr double, double* %ptr, i32 %stride
|
||||
%col.ptr.2.cast = bitcast double* %col.ptr.2 to <6 x double>*
|
||||
%lv.2 = load <6 x double>, <6 x double>* %col.ptr.2.cast, align 8
|
||||
%res.1 = fadd <6 x double> %lv.1, %lv.1
|
||||
%res.2 = fadd <6 x double> %lv.2, %lv.2
|
||||
store <6 x double> %res.1, <6 x double>* %col.ptr.1, align 8
|
||||
store <6 x double> %res.2, <6 x double>* %col.ptr.2.cast, align 8
|
||||
ret void
|
||||
}
|
||||
|
||||
; Same as @test1, but the assume just guarantees %skip > -3, which is not
|
||||
; enough to derive NoAlias
|
||||
define void @test3(double* %ptr, i32 %skip) {
|
||||
; CHECK-LABEL: Function: test3: 4 pointers, 1 call sites
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.1, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: double* %col.ptr.2, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.1, double* %col.ptr.2
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.2.cast, double* %ptr
|
||||
; CHECK-NEXT: MayAlias: <6 x double>* %col.ptr.1, <6 x double>* %col.ptr.2.cast
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.2.cast, double* %col.ptr.2
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %ptr <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.1 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %col.ptr.2 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.2.cast <-> call void @llvm.assume(i1 %gt)
|
||||
;
|
||||
%gt = icmp sgt i32 %skip, -3
|
||||
call void @llvm.assume(i1 %gt)
|
||||
%stride = add nsw nuw i32 %skip, 6
|
||||
%col.ptr.1 = bitcast double* %ptr to <6 x double>*
|
||||
%lv.1 = load <6 x double>, <6 x double>* %col.ptr.1, align 8
|
||||
%col.ptr.2= getelementptr double, double* %ptr, i32 %stride
|
||||
%col.ptr.2.cast = bitcast double* %col.ptr.2 to <6 x double>*
|
||||
%lv.2 = load <6 x double>, <6 x double>* %col.ptr.2.cast, align 8
|
||||
%res.1 = fadd <6 x double> %lv.1, %lv.1
|
||||
%res.2 = fadd <6 x double> %lv.2, %lv.2
|
||||
store <6 x double> %res.1, <6 x double>* %col.ptr.1, align 8
|
||||
store <6 x double> %res.2, <6 x double>* %col.ptr.2.cast, align 8
|
||||
ret void
|
||||
}
|
||||
|
||||
; Same as @test1, but the assume uses the sge predicate for %skip >= 0.
|
||||
define void @test4(double* %ptr, i32 %skip) {
|
||||
; CHECK-LABEL: Function: test4: 4 pointers, 1 call sites
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.1, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: double* %col.ptr.2, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.1, double* %col.ptr.2
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.2.cast, double* %ptr
|
||||
; CHECK-NEXT: NoAlias: <6 x double>* %col.ptr.1, <6 x double>* %col.ptr.2.cast
|
||||
; CHECK-NEXT: MustAlias: <6 x double>* %col.ptr.2.cast, double* %col.ptr.2
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %ptr <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.1 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: double* %col.ptr.2 <-> call void @llvm.assume(i1 %gt)
|
||||
; CHECK-NEXT: NoModRef: Ptr: <6 x double>* %col.ptr.2.cast <-> call void @llvm.assume(i1 %gt)
|
||||
;
|
||||
%gt = icmp sge i32 %skip, 0
|
||||
call void @llvm.assume(i1 %gt)
|
||||
%stride = add nsw nuw i32 %skip, 6
|
||||
%col.ptr.1 = bitcast double* %ptr to <6 x double>*
|
||||
%lv.1 = load <6 x double>, <6 x double>* %col.ptr.1, align 8
|
||||
%col.ptr.2= getelementptr double, double* %ptr, i32 %stride
|
||||
%col.ptr.2.cast = bitcast double* %col.ptr.2 to <6 x double>*
|
||||
%lv.2 = load <6 x double>, <6 x double>* %col.ptr.2.cast, align 8
|
||||
%res.1 = fadd <6 x double> %lv.1, %lv.1
|
||||
%res.2 = fadd <6 x double> %lv.2, %lv.2
|
||||
store <6 x double> %res.1, <6 x double>* %col.ptr.1, align 8
|
||||
store <6 x double> %res.2, <6 x double>* %col.ptr.2.cast, align 8
|
||||
ret void
|
||||
}
|
||||
|
||||
declare void @llvm.assume(i1 %cond)
|
||||
Loading…
Reference in New Issue
Block a user