Conversation
8136e8a to
75c8356
Compare
|
Looks like this series of pull requests is responsible:
The change itself is interesting to me (I wonder if they figured out a similar trick to what I used in SPIR-T, taking advantage of the pattern in which Khronos assigns "standard" vs "vendor" enums etc.). But in terms of the build script changes, I'll try to make sense of EDIT: that worked out, it's mostly removals with one monolithic invocation being added, and it's literally copy-pasteable out of |
There was a problem hiding this comment.
I wish GitHub could render at least tiny hex diffs, because I used a hex editor to change the byte at position 5, from 0x00 to 0x03, and confirmed the SPIR-V 1.0 -> SPIR-V 1.3 effect that had, by running spirv-dis.
EDIT: diffing the output of xxd before/after this change:
@@ -1,4 +1,4 @@
-00000000: 0302 2307 0000 0100 0000 0f00 3900 0000 ..#.........9...
+00000000: 0302 2307 0003 0100 0000 0f00 3900 0000 ..#.........9...
this is expected to fail(EDIT(@eddyb): took a stab at the
ggt.pymigration, and pushed my own commits to this PR)The
generate_grammar_tables.pyscript was replaced with a newggt.pyscript that takes vastly different arguments, thus ourcargo run -p generatebinary to generate cpp headers from xml spec and such needs to be adjusted to this new script.