Disclaimer: I don't have access to any hardware capable of AVX512, so I haven't tested it.
Details
Diff Detail
- Repository
- rLLDB LLDB
Event Timeline
Included dotest test removal for completeness. I'm going to commit it separately though.
For the record, GCC refuses to build this with:
Inputs/x86-zmm-read.cpp: In function ‘int main()’: Inputs/x86-zmm-read.cpp:193:4: error: more than 30 operands in ‘asm’ ); ^
Not sure if we care about it though; clang doesn't seem to have this limitation.
The test fails on linux even with the right hardware, but that's not surprising as the original test was also @skipUnlessDarwin (and I see no trace of zmm support in the linux part of code). I also don't have mac hardware around that I could try the mac side of things on (maybe @JDevlieghere has?), but I guess you can also just watch the bots carefully after committing this.
I don't think we have to worry about that now. If that ever becomes a problem, we can always make the zmm things into an array and pass them via a single memory argument.