David Peixotto
David Peixotto
In your message you say > DXC doesn't know (%3, %7) and (%7, %9) are redundant How are %7 and %9 redundant here? Should it be %5 and %9? It...
What is the performance impact of this change?
> > What is the performance impact of this change? > > I haven't run the perf tests. But I believe the impact's smaller than the one I got before....
> > > > What is the performance impact of this change? > > > > > > > > > I haven't run the perf tests. But I believe...
> I'd like to see the perf data for this change. Synced offline. Looks like a nice win overall in our test suite. Reduced ALU in ~40% of shaders (increased...
> > There is a possibility of a perf regression with this change as it may prevent generating larger loads in some backends. Should we set the alignment to what...
> This is a bug we experienced since 2019. Our workaround is to have a second conditional in the loop with a second WaveReadLaneFirst around the break. @simondeschenes I believe...
> we will use constexpr to fix it, Great, thanks for the fix! > "invalid operand for instruction: is another isssue, you can follow the instruction format Sorry, I don't...
I don't think #5302 is related to this same issue. The underlying issue there is that we are only applying the dx.break fix to PS/CS/LIB targets. https://github.com/microsoft/DirectXShaderCompiler/issues/5302#issuecomment-1597605713 > In tools/clang/lib/CodeGen/CGHLSLMS.cpp,...