thr3ads.net - llvm dev - [LLVMdev] i1 types in MergeConsecutiveStores [May 2015]

If this information is useful, please help other people find it:
Share via:

Steve King

2015-May-12 04:40 UTC

[LLVMdev] i1 types in MergeConsecutiveStores

Hello LLVM,

In DAGCombiner.cpp, MergeConsecutiveStores uses

int64_t ElementSizeBytes = MemVT.getSizeInBits()/8;

https://github.com/llvm-mirror/llvm/blob/master/lib/CodeGen/SelectionDAG/DAGCombiner.cpp#L10669

which is broken for i1 types where getSizeInBits() == 1.  My
out-of-tree target hits this case and eventually LLVM asserts in
Type.cpp.

Is there some reason MergeConsecutiveStores should not expect to see i1 types?

Thanks,
-steve

Hal Finkel

2015-May-12 05:56 UTC

head link

[LLVMdev] i1 types in MergeConsecutiveStores

----- Original Message -----> From: "Steve King" <steve at metrokings.com>
> To: llvmdev at cs.uiuc.edu
> Sent: Monday, May 11, 2015 11:40:56 PM
> Subject: [LLVMdev] i1 types in MergeConsecutiveStores
> 
> Hello LLVM,
> 
> In DAGCombiner.cpp, MergeConsecutiveStores uses
> 
> int64_t ElementSizeBytes = MemVT.getSizeInBits()/8;
> 
>
https://github.com/llvm-mirror/llvm/blob/master/lib/CodeGen/SelectionDAG/DAGCombiner.cpp#L10669
> 
> which is broken for i1 types where getSizeInBits() == 1.  My
> out-of-tree target hits this case and eventually LLVM asserts in
> Type.cpp.
> 
> Is there some reason MergeConsecutiveStores should not expect to see
> i1 types?
My impression is that there are a lot of things that are (still) broken for i1
memory operations (and i1 vectors). Patches welcome.

 -Hal
> 
> Thanks,
> -steve
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> 
-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory

James Y Knight

2015-May-12 14:01 UTC

head link

[LLVMdev] i1 types in MergeConsecutiveStores

That seems likely, however this particular issue had already been fixed in
r236927.

On Tue, May 12, 2015 at 1:56 AM, Hal Finkel <hfinkel at anl.gov> wrote:
> ----- Original Message -----
> > From: "Steve King" <steve at metrokings.com>
> > To: llvmdev at cs.uiuc.edu
> > Sent: Monday, May 11, 2015 11:40:56 PM
> > Subject: [LLVMdev] i1 types in MergeConsecutiveStores
> >
> > Hello LLVM,
> >
> > In DAGCombiner.cpp, MergeConsecutiveStores uses
> >
> > int64_t ElementSizeBytes = MemVT.getSizeInBits()/8;
> >
> >
>
https://github.com/llvm-mirror/llvm/blob/master/lib/CodeGen/SelectionDAG/DAGCombiner.cpp#L10669
> >
> > which is broken for i1 types where getSizeInBits() == 1.  My
> > out-of-tree target hits this case and eventually LLVM asserts in
> > Type.cpp.
> >
> > Is there some reason MergeConsecutiveStores should not expect to see
> > i1 types?
>
> My impression is that there are a lot of things that are (still) broken
> for i1 memory operations (and i1 vectors). Patches welcome.
>
>  -Hal
>
> >
> > Thanks,
> > -steve
> > _______________________________________________
> > LLVM Developers mailing list
> > LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
> >
>
> --
> Hal Finkel
> Assistant Computational Scientist
> Leadership Computing Facility
> Argonne National Laboratory
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150512/cac40f43/attachment.html>

llvm dev - May 2015 - [LLVMdev] i1 types in MergeConsecutiveStores

[LLVMdev] i1 types in MergeConsecutiveStores

[LLVMdev] i1 types in MergeConsecutiveStores

[LLVMdev] i1 types in MergeConsecutiveStores