Martin Storsjo 3916055ddf Remove inline assembly optimizations that give no speedup
These inline assembly versions are about as fast as what the
compiler itself can do. However, when the compiler emits it
by itself instead of via inline assembly, it can interleave the
instructions with other instructions, improving scheduling.

Silence the warnings in fixmul.h though, since the compiler
is able to produce good enough code for the C version of these
functions.

This gives a speedup on both GCC and clang.
2016-09-07 11:54:06 +03:00
2012-09-13 23:27:27 +03:00
2013-09-07 18:32:50 +03:00
2016-08-25 22:20:09 +03:00
2014-06-23 10:45:50 +03:00
2012-09-13 23:27:27 +03:00
Description
No description provided
4.6 MiB
Languages
C++ 74.1%
C 25.7%
Makefile 0.2%