David Neto
David Neto
To me, polyfilling the const case implies also polyfiling the runtime case. It's bad to be inconsistent between const and runtime like that. But if implementations polyfill the runtime case,...
the tsan build failed with: /usr/bin/ld: cannot find libtsan_preinit.o: No such file or directory
PR #270 already updated the kokoro CI to use GCC 13. Closing. The sanitizer failures are tracked with #279
> Hm, I've been looking but can't find the documentation for license-check. I also don't have admin, and am unable to ignore the check. @dneto0 do you have the ability...
Landed as #1059
At the recent F2F we decided to drop the `subgroups-f16` feature. @alan-baker
Hi, former SPIR WG chair here, and long time maintainer of spirv-as and spirv-dis. @s-perron is correct: only the numbers in the binary are normative. > I think implementing a...
The hard sticking point is that in a storage buffer, invocations writing to adjacent f16 values in the same 4byte word will race. Clients that need this could use texel...
I still agree with previous discussion: - don't split the feature - polyfill the functionality for `uniformAndStorageBuffer16BitAccess`. This is a prioritization issue for implementors.
The notes from March 11 say we agreed to not change the spec. Supporting shader-f16 on the problematic Qualcomm devices will require a polyfill of the loads from uniform buffers....