[PATCH][X86] Unsafe copysign xform in DAGCombiner

Cameron McInally via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 9 13:14:00 PDT 2016


Apologies for this one. When I went to read the Phabricator
documentation, I unwittingly left review.llvm.org and went to some
other page.

New differential on reviews.llvm.org is...

https://reviews.llvm.org/D24422

On Fri, Sep 9, 2016 at 3:30 PM, Cameron McInally
<cameron.mcinally at nyu.edu> wrote:
> Created D16532 for anyone interested...
>
> On Fri, Sep 9, 2016 at 3:07 PM, Demikhovsky, Elena
> <elena.demikhovsky at intel.com> wrote:
>> Hi Cameron,
>>
>> Are you familiar with the review board https://reviews.llvm.org ?
>> I suggest to upload the patch there.
>>
>> -  Elena
>>
>>
>>   |-----Original Message-----
>>   |From: Cameron McInally [mailto:cameron.mcinally at nyu.edu]
>>   |Sent: Friday, September 09, 2016 19:25
>>   |To: llvm-commits at lists.llvm.org
>>   |Cc: Demikhovsky, Elena <elena.demikhovsky at intel.com>; Breger,
>>   |Igor <igor.breger at intel.com>; Craig Topper
>>   |<craig.topper at gmail.com>
>>   |Subject: [PATCH][X86] Unsafe copysign xform in DAGCombiner
>>   |
>>   |Hey guys,
>>   |
>>   |There's a copysign xform in the DAGCombiner that replaces safe FP
>>   |code with unsafe code. Attached is a patch to guard this xform.
>>   |
>>   |The issue here (see test case) is the first copysign is removed before
>>   |the cvt is performed. If the input to the cvt is unordered, it will raise
>>   |an exception. The purpose of the first copysign is to create a safe
>>   |value (i.e. 1.0 or -1.0) to feed the cvt.
>>   |
>>   |Also, I'm not the most versed in FileCheck. I'd appreciate any critiques
>>   |on the test case I've added.
>>   |
>>   |Thanks,
>>   |Cam
>> ---------------------------------------------------------------------
>> Intel Israel (74) Limited
>>
>> This e-mail and any attachments may contain confidential material for
>> the sole use of the intended recipient(s). Any review or distribution
>> by others is strictly prohibited. If you are not the intended
>> recipient, please contact the sender and delete all copies.


More information about the llvm-commits mailing list