I am looking at the optimize options in TeXmacs: tm_optimize.m4 and tm_platform.m4.
- A minor point:
tm_platform.m4 is not needed. For
gcc, it is written in the official document that
-O2 and above already include
clang, on the other hand,
-fexpensive-optimizations is not supported, and
clang will throw a warning with that flag. It seems better to remove.
- Seemingly there is no differentiation between
clang in the optimization flags.
- I don’t know whether we have profiled the performance with
-O3 systematically? Including with
clang, I don’t know whether
-O3 is in general really faster.