[llvm] r342870 - [ARM] bottom-top mul support ARMParallelDSP

[Friedman, Eli via llvm-commits]

On 9/24/2018 2:34 AM, Sam Parker via llvm-commits wrote:
> Author: sam_parker
> Date: Mon Sep 24 02:34:06 2018
> New Revision: 342870
>
> URL: http://llvm.org/viewvc/llvm-project?rev=342870&view=rev
> Log:
> [ARM] bottom-top mul support ARMParallelDSP
>
> Originally committed in rL342210 but was reverted in rL342260 because
> it was causing issues in vectorized code, because I had forgotten to
> ensure that we're operating on scalar values.
>
> Original commit message:
>
> On failing to find sequences that can be converted into dual macs,
> try to find sequential 16-bit loads that are used by muls which we
> can then use smultb, smulbt, smultt with a wide load.
>
> Differential Revision: https://reviews.llvm.org/D51983

It looks like this is causing the polly-aosp buildbot 
(http://lab.llvm.org:8011/builders/aosp-O3-polly-before-vectorizer-unprofitable) 
to fail.  Attached reduced testcase; reproduce with "llc 
aosp-convolve-crash.bc".

-Eli

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project

-------------- next part --------------
A non-text attachment was scrubbed...
Name: aosp-convolve-crash.bc
Type: application/octet-stream
Size: 2388 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180925/09fec501/attachment.obj>