Dxil 1.4 update + packed dot intrinsics and SV_ShadingRate - #1620
Conversation
- Move/fix flag collection based on signature properties to be computed per entry function (including for libraries)
|
This change seems to break the Linux build. Please consider patching it if the fix is trivial. Thanks! |
I'm not sure what the issue is. It says The Travis CI build passed to me. When I look into that, it looks like each individual build failed, but I can't figure out why, since it looks like it passed everything, then exits with return code 1 at the end. But the overall result says passed. I see a bunch of build warnings, but these appear to be unrelated to this change, and don't prevent the build. |
|
It's a sanity check we added at the early stage of supporting non-Windows systems to make sure DXIL output from the command-line is good. The check is against binaries directly, so it's gonna break every time DXIL version updates. Not really needed right now, so I removed it. |
|
The Travis CI is set up to allow all build jobs to fail: https://github.com/Microsoft/DirectXShaderCompiler/blob/master/.travis.yml#L36-L38. So Travis CI will always return success no matter what. We initially set up it this way because non-Windows support is experimental and we don't want to disturb Microsoft workflow too much. It was mainly for us to see how it goes and pay attention to non-Windows platform status so that we can fix it if something went wrong. From the experience of past few months, it seems it rarely breaks; maybe only on large changes. I'd happy to flip it as correctly showing the build status if you guys think that's fine. |
|
Travis CI is happy now. :) |
Update to DXIL 1.4 and Shader Model 6.4
Add dot2add and dot4add_*8packed intrinsics
Add SV_ShadingRate plus optional feature flag …
per entry function (including for libraries)