| commit | 51455b29e3c1a29eecf21ebb10317b739a30a53e | [log] [tgz] |
|---|---|---|
| author | Allan MacKinnon <allanmac@fuchsia.infra.roller.google.com> | Fri Mar 11 22:29:11 2022 +0000 |
| committer | Copybara-Service <copybara-worker@google.com> | Fri Mar 11 14:30:41 2022 -0800 |
| tree | 83a3b95dc44002be409d281f644e8d3fd2e94845 | |
| parent | fa8c81bfd2cd0f832d18b3d527211a2ff06d2126 [diff] |
[roll] Roll fuchsia [graphics][compute][spinel2] Enable AMD GCN3 Tested on host on an AMD RX560 (Polaris11) using the unittest framework. It wasn't possible to test the `spinel2-vk-bench` benchmarking application because the AMD GPU is not recognized by X despite updating xconf.org. Changes: * Raised global buffer alignment limit to 512 bytes. This is bigger than I would like but relatively harmless. This might be a good reason to make this alignment calculation a runtime decision. * A bit calculation for 64-wide subgroups was incorrect. GCN3+ is the only architecture with a 64-wide subgroup (for now). * Updated checksums for AMD. * Removed use of the GL_AMD_gpu_shader_half_float extension since this has been dropped by Mesa and even though GCN3 supports full-rate fp16 (1:1), support is not advertised in Mesa. For this reason, the GCN3 device target renders with fp32 components. Testability: ``` $ export VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/radeon_icd.x86_64.json $ out/.../host_x64/exe.unstripped/spinel2-vk-test --gtest_filter=* --gtest_random_seed=0 --gtest_throw_on_failure --gtest_repeat=25 ``` Original-Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/650665 Original-Revision: 192d591b45630ba59f7200b05b0028d5cfb174dd GitOrigin-RevId: 5b49d4f9816d196a0058c5a228b2b5c15982ff87 Change-Id: Ib7bb6afafcd5a4e2098f2ff2a07bdc1ff65b7292
This repository contains Fuchsia's Global Integration manifest files.
All changes should be made to the internal version of this repository. Our infrastructure automatically updates this version when the internal one changes.
Currently all changes must be made by a Google employee. Non-Google employees wishing to make a change can ask for assistance via the IRC channel #fuchsia on Freenode.
First install Jiri.
Next run:
$ jiri init $ jiri import minimal https://fuchsia.googlesource.com/integration $ jiri update
Third party projects should have their own subdirectory in ./third_party.