rmaprath added inline comments.
================ Comment at: include/__threading_support:593 + using namespace chrono; + milliseconds ms = duration_cast<milliseconds>(ns); + if (ms.count() == 0 || ns > duration_cast<nanoseconds>(ms)) ---------------- joerg wrote: > Use (ns + 999999) so that the cast rounds up. So, this code was lifted from the sources as-is. I will do this change, I think it makes sense. ================ Comment at: include/__threading_support:594 + milliseconds ms = duration_cast<milliseconds>(ns); + if (ms.count() == 0 || ns > duration_cast<nanoseconds>(ms)) + ++ms; ---------------- joerg wrote: > Why is ns == 0 supposed to sleep at all? In fact, the caller already ensures > that can't happen? IIUC, this is trying to detect round-downs and then compensate for it. With the above suggestion, this should no longer be needed. Will get rid of it as well. https://reviews.llvm.org/D29630 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
