[PATCH] D153800: [ARM] Adjust strd/ldrd codegen alignment requirements

Thu Jul 6 01:32:14 PDT 2023

DavidSpickett added a comment.

I recommend following what the armv8-a builder does: https://lab.llvm.org/buildbot/#/builders/178/builds/5155. If you click on the "cmake stage..." steps it will show you the commands used. You can cut out some of the enabled projects and/or targets if you want to save some time.

We are running on a 64 bit Arm machine with a 32 bit docker container.

In case the hardware is a factor (I hope not) it is a Ampere Mt. Jade with 160x Neoverse-N1 (this information can also be found in the "worker" tab of the buildbot web UI).

  $ arch
  armv8l
  $ lsb_release -a
  No LSB modules are available.
  Distributor ID: Ubuntu
  Description:    Ubuntu 20.04.6 LTS
  Release:        20.04
  Codename:       focal

If you really want the exact container you can use https://git.linaro.org/ci/dockerfiles.git/tree/focal-armhf-tcwg-base/focal-armhf-tcwg-llvmbot/Dockerfile but if you do, change the entrypoint. As by default it's looking to launch a buildbot agent.

You're probably fine using the official Ubuntu Focal container and installing things as needed.

At the time of the build the host compiler was clang 15.0.0, using the release https://github.com/llvm/llvm-project/releases/download/llvmorg-15.0.0/clang+llvm-15.0.0-armv7a-linux-gnueabihf.tar.xz. For unrelated reasons, I'm about to update the bots to use 16.0.4, but given that this is a stage 2 failure it shouldn't make a difference.

It was useful for me to build 2 copies, one with this change and one without. That was how I was able to narrow it to llvm-dwarfdump because using the dwarfdump from the working copy passed the test.

To run only the failing test:

  ./bin/llvm-lit ../llvm-project/llvm/test/tools/llvm-dwarfdump/X86/simplified-template-names.s -a

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153800/new/

https://reviews.llvm.org/D153800