[release_40 branch] Merge Request of r293658 - Don't combine stores to a swifterror pointer operand to a different type
Hans Wennborg via llvm-commits
llvm-commits at lists.llvm.org
Wed Feb 1 09:08:17 PST 2017
r293781.
Thanks,
Hans
On Tue, Jan 31, 2017 at 3:37 PM, David Majnemer
<david.majnemer at gmail.com> wrote:
> SGTM
>
> On Tue, Jan 31, 2017 at 2:50 PM, Hans Wennborg <hans at chromium.org> wrote:
>>
>> David: wdyt?
>>
>> On Tue, Jan 31, 2017 at 12:40 PM, Arnold Schwaighofer via llvm-commits
>> <llvm-commits at lists.llvm.org> wrote:
>> > I would like to propose merging the patch below into the release branch.
>> >
>> > The patch affects only code that uses the ‘swifterror’ attribute — code
>> > generated by the swift compiler.
>> >
>> >
>> > Begin forwarded message:
>> >
>> > From: Arnold Schwaighofer via llvm-commits <llvm-commits at lists.llvm.org>
>> > Subject: [llvm] r293658 - Don't combine stores to a swifterror pointer
>> > operand to a different type
>> > Date: January 31, 2017 at 9:53:50 AM PST
>> > To: llvm-commits at lists.llvm.org
>> > Reply-To: Arnold Schwaighofer <aschwaighofer at apple.com>
>> >
>> > Author: arnolds
>> > Date: Tue Jan 31 11:53:49 2017
>> > New Revision: 293658
>> >
>> > URL: http://llvm.org/viewvc/llvm-project?rev=293658&view=rev
>> > Log:
>> > Don't combine stores to a swifterror pointer operand to a different type
>> >
>> > Modified:
>> > llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
>> > llvm/trunk/test/Transforms/InstCombine/load.ll
>> >
>> > Modified:
>> > llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
>> > URL:
>> >
>> > http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp?rev=293658&r1=293657&r2=293658&view=diff
>> >
>> > ==============================================================================
>> > --- llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
>> > (original)
>> > +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
>> > Tue
>> > Jan 31 11:53:49 2017
>> > @@ -502,7 +502,8 @@ static Instruction *combineLoadToOperati
>> > !DL.isNonIntegralPointerType(Ty)) {
>> > if (all_of(LI.users(), [&LI](User *U) {
>> > auto *SI = dyn_cast<StoreInst>(U);
>> > - return SI && SI->getPointerOperand() != &LI;
>> > + return SI && SI->getPointerOperand() != &LI &&
>> > + !SI->getPointerOperand()->isSwiftError();
>> > })) {
>> > LoadInst *NewLoad = combineLoadToNewType(
>> > IC, LI,
>> >
>> > Modified: llvm/trunk/test/Transforms/InstCombine/load.ll
>> > URL:
>> >
>> > http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/load.ll?rev=293658&r1=293657&r2=293658&view=diff
>> >
>> > ==============================================================================
>> > --- llvm/trunk/test/Transforms/InstCombine/load.ll (original)
>> > +++ llvm/trunk/test/Transforms/InstCombine/load.ll Tue Jan 31 11:53:49
>> > 2017
>> > @@ -219,3 +219,22 @@ entry:
>> > store %swift.error* %err.res, %swift.error** %err, align 8
>> > ret void
>> > }
>> > +
>> > +; Make sure we preseve the type of the store to a swifterror pointer.
>> > +; CHECK-LABEL: @test19(
>> > +; CHECK: [[A:%.*]] = alloca
>> > +; CHECK: call
>> > +; CHECK: [[BC:%.*]] = bitcast i8** [[A]] to
>> > +; CHECK: [[ERRVAL:%.*]] = load {{.*}}[[BC]]
>> > +; CHECK: store {{.*}}[[ERRVAL]]
>> > +; CHECK: ret
>> > +declare void @initi8(i8**)
>> > +define void @test19(%swift.error** swifterror %err) {
>> > +entry:
>> > + %tmp = alloca i8*, align 8
>> > + call void @initi8(i8** %tmp)
>> > + %swifterror = bitcast i8** %tmp to %swift.error**
>> > + %err.res = load %swift.error*, %swift.error** %swifterror, align 8
>> > + store %swift.error* %err.res, %swift.error** %err, align 8
>> > + ret void
>> > +}
>> >
>> >
>> > _______________________________________________
>> > llvm-commits mailing list
>> > llvm-commits at lists.llvm.org
>> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>> >
>> >
>> >
>> > _______________________________________________
>> > llvm-commits mailing list
>> > llvm-commits at lists.llvm.org
>> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>> >
>
>
More information about the llvm-commits
mailing list