FYI here's image load store results before a segfault occurred

Deleted: client.0.log 699 KB

client.0.log 699 KB View Download

Working on dEQP-GLES31.functional.image_load_store.*

dEQP-GLES31.functional.ssbo.* works now.  image load store tests probably have the similar write-mask/indexing issues, so here's the branch with the extra commits ontop of airlied@'s changes: 

https://github.com/gurchetansingh/virglrenderer-gles31

We should keep all of our commits in one place -- please send pull requests to the above branch!

Next, I'll focus on fixing the segfaults that occur when running the entire dEQP GLES31 suite (hopefully there's not too many).

To be more specific about comment #3,  dEQP-GLES31.functional.ssbo.layout* works now, but dEQP-GLES31.functional.ssbo.atomic* does not.  Let me look at that now, since dEQP-GLES31.functional.ssbo.atomic.compswap.lowp_uint causes a segfault in the nvidia driver.

Hello, we at Collabora are now also starting to look at the GLES 3.1 support. My first intent do try the branch with extra commits mostly failed, that is, I don't even get the tests to pass that supposedly work to work. 

Is there anything else missing?

Do you have a correct guest side Mesa?

https://cgit.freedesktop.org/~airlied/mesa/log/?h=virgl-gles31-sketch

We haven't been really working on new features.  I'm mostly trying to abstract away the annoying format conversions (floatBitsToUint, uintBitsToFloat) etc.  As we try to add more features, let's try not to make translate_image_store/translate_load/(new features) as cluttered as translate_tex. 

Thanks I've got it running now, it crashed qemu when trying to run all tests, but I guess that's to be expected at this point. 

This is probably not the best place to track this, since freedesktop now uses Gitlab.  Anyone interested in working on this should get a freedesktop account and use the upstream tracker.

https://gitlab.freedesktop.org/