[PATCH] D121137: [AArch64] Lower 3 and 4 sources buildvectors to TBL

[Martin Storsjö via Phabricator via llvm-commits]

mstorsjo added a comment.

In D121137#3407452 <https://reviews.llvm.org/D121137#3407452>, @mstorsjo wrote:

> FYI, headsup, I'm seeing a misoptimization introduced by this commit.

The misoptimization can be triggered within this standalone C file: https://martin.st/temp/dctref-preproc.c
Compiled with `clang -target aarch64-linux-gnu -c -O3 dctref-preproc.c`

For a full repro, you can follow these steps:

  git clone git://source.ffmpeg.org/ffmpeg
  cd ffmpeg
  ./configure --cc=clang
  make -j$(nproc) fate-idct8x8-0

(The misoptimized object file is libavcodec/dctref.o.)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D121137/new/

https://reviews.llvm.org/D121137