[llvm-dev] [VPlan] Dead instructions are invariant to VFs when build vplan
Caballero, Diego via llvm-dev
llvm-dev at lists.llvm.org
Fri Jun 1 09:48:15 PDT 2018
>> Hi,
>>
>> I noticed that the dead instructions collected when build vplan are
>> invariant to different ranges of VFs since the original loop is not
>> changed. Maybe DeadInstructions should be treated as a data member of
>> LoopVectorizationPlanner and initialized by
>> collectTriviallyDeadInstructions() in plan()?
>>
>
>Yep you are right. There's a patch under review, which also addresses
>that:
>https://urldefense.proofpoint.com/v2/url?u=https-3A__reviews.llvm.org_D
>47477&d=DwIC->g&c=aUq983L2pue2FqKFoP6PGHMJQyoJ7kl3s3GZ->_haXqY&r=FZEW1c
>DCyCU3ZbRFatG9st_R0bbu9fLOaIEKyKAe7wg&m=1H53Xvb7phmsnD0jJKQxe1->wa0mbde
>wKcT8B8DbRzYk&s=GtJno4-QfCaG0-yzw_SCrQ65uTZXwWVuWUL72Y_ySGc&e=
Thanks for bringing this up! Yes, it should be better after the patch and I think we should be able to remove the DeadInstructions set once we have a full VPInstruction-based vectorization pipeline (including CG) properly modeling the evolution of the incoming IR.
Thanks,
Diego
-----Original Message-----
From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of Shixiong Xu via llvm-dev
Sent: Friday, June 1, 2018 4:09 AM
To: Florian Hahn <florian.hahn at arm.com>
Cc: llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] [VPlan] Dead instructions are invariant to VFs when build vplan
Thanks. I didn't notice the patch. 😊
Shixiong
-----Original Message-----
From: Florian Hahn <florian.hahn at arm.com>
Sent: Friday, June 01, 2018 11:21 AM
To: Shixiong Xu <shixiong at cadence.com>; llvm-dev at lists.llvm.org
Cc: Caballero, Diego <diego.caballero at intel.com>; nd <nd at arm.com>
Subject: Re: [llvm-dev] [VPlan] Dead instructions are invariant to VFs when build vplan
EXTERNAL MAIL
Hi
On 01/06/2018 10:56, Shixiong Xu via llvm-dev wrote:
> Hi,
>
> I noticed that the dead instructions collected when build vplan are
> invariant to different ranges of VFs since the original loop is not
> changed. Maybe DeadInstructions should be treated as a data member of
> LoopVectorizationPlanner and initialized by
> collectTriviallyDeadInstructions() in plan()?
>
Yep you are right. There's a patch under review, which also addresses
that: https://urldefense.proofpoint.com/v2/url?u=https-3A__reviews.llvm.org_D47477&d=DwIC-g&c=aUq983L2pue2FqKFoP6PGHMJQyoJ7kl3s3GZ-_haXqY&r=FZEW1cDCyCU3ZbRFatG9st_R0bbu9fLOaIEKyKAe7wg&m=1H53Xvb7phmsnD0jJKQxe1-wa0mbdewKcT8B8DbRzYk&s=GtJno4-QfCaG0-yzw_SCrQ65uTZXwWVuWUL72Y_ySGc&e=
Thanks,
Florian
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
More information about the llvm-dev
mailing list