March 10th, 2003, 18:50
Mupen64 0.1 source code
Finally i took some time to pack the mupen64 source code and i'm able to post source code of two linux plugins
Please be warned that the windows binary should be compiled with mingw, (a Dev-C++ project is included).
Last edited by Hacktarux; July 11th, 2003 at 11:12.
March 10th, 2003, 18:51
My basic input plugin
You can also take a look at blight's plugin for a more advanced one :
Last edited by Hacktarux; July 11th, 2003 at 11:15.
March 10th, 2003, 18:53
The tr64 plugin written by Icepir8 ported to linux
Last edited by Hacktarux; July 11th, 2003 at 11:16.
March 10th, 2003, 19:01
March 10th, 2003, 19:29
Gonna try compiling with some more aggressive flags to see if it influences the results
March 10th, 2003, 21:47
March 10th, 2003, 22:34
March 11th, 2003, 10:26
I compiled mupen1.0 and tr64 with this options:
-fstrict-aliasing -fexpensive-optimizations -DINLINE=static__inline__ -foptimize-register-move -foptimize-sibling-calls -fprefetch-loop-arrays -freduce-all-givs -frename-registers -frerun-cse-after-loop -frerun-loop-opt -fschedule-insns -fno-sched-interblock -fsched-spec-load -fsingle-precision-constant -fssa -fstrength-reduce -fstrict-aliasing
and there is a lttle speed up
March 11th, 2003, 11:18
Yes I saw some improvements as well. I need to test some more to see how the individual compiler flags affect the speed and stability. But initial recompiles raised speed by ~10-15%.
March 11th, 2003, 13:38
I have changed a bit the source code compared to 0.1 binary release and it should be a bit faster, nothing major but still a bit faster...
I don't think changing compiler flags in the core will really change speed but maybe you can expect some interesting improvements in the gfx plugin.
One last thing, be carefull with fpu optimizations.. the core needs that float is a 32 bits type and double is a 64 bits type and you should use the real fpu not sse...