Open arno-lunarg opened 2 weeks ago
Should this be a CTS test?
@ziga-lunarg no, this is an issue with our TestICD (the new Mock ICD) ... we have no plans to make a CTS conformance mock ICD and so a CTS test would not help
@arno-lunarg what does
aligned to the strictest alignment mentioned in the spec
mean, how would TestICD determine this, another issue is things like shaderGroupBaseAlignment
we set in the profile JSON and not the ICD itself
@arno-lunarg what does
aligned to the strictest alignment mentioned in the spec
mean, how would TestICD determine this, another issue is things like
shaderGroupBaseAlignment
we set in the profile JSON and not the ICD itself
No way to query that directly, so I would make sure the returned pointer is aligned to 8192, that should be enough
what did 8192
come from?
Totally made up, just go for a somewhat big power of 2!
It currently does not do that, leading to unexpected alignment issues when filling shader binding tables in the ray tracing framework