[PATCH] D54889: Fiber support for thread sanitizer

Thu Jan 10 09:47:58 PST 2019

dvyukov added a comment.

Benchmark results with clang:

write8:
HEAD:	9.74
fibers(old):	9.40
fibers(new):	8.66

read8:
HEAD:	9.92
fibers(new):	9.26

read4:
HEAD:	10.20
fibers(new):	9.36

fibers(new) is the current version of the change. fibers(old) is the one I used with gcc (no spot optimizations).

So this change indeed makes it faster for clang, but the fastest clang version is still slower then gcc, so this suggests that there is something to improve in clang codegen. If we improve clang codegen, will this change again lead to slower code or not? It's bad that we have that open clang regression...

Repository:
  rCRT Compiler Runtime

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D54889/new/

https://reviews.llvm.org/D54889