This is the same implementation as mulxc3
Details
Details
Diff Detail
Diff Detail
- Lint
Lint Skipped - Unit
Tests Skipped
Event Timeline
Comment Actions
Looks good to me. Thanks!
Question: this hasn't been upstreamed yet, right? If looks like compiler-rt has a version for ppc that uses a similar algorithm, but is structured differently. Does this version supersede that one?
Would it make sense to submit this upstream first? I have no strong preference. :-)
Comment Actions
That's correct. I've started a thread on llvmdev about missing compiler-rt functionality for quad precision, but want to get a concrete version of the change into review/discussion.
If looks like compiler-rt has a version for ppc that uses a similar algorithm, but is structured differently. Does this version supersede that one?
Hmm, good point. I think this could supersede that one - the algorithm is about the same, with a key difference that the ppc one uses explicit calls to __gcc_qmul rather than just a * c and relying on the compiler generating the calls itself.
Would it make sense to submit this upstream first? I have no strong preference. :-)
My immediate goal is to get libcompiler-rt building for arm64 for us, and I'm a bit leery of the latency involved in potential upstream discussion (or bikeshed). I'd like to say it's being simultaneously submitted/discussed upstream, and I'll replace our local change with the upstream implementation when it's resolved.