[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
Eric Botcazou wrote:
Has it ever been enabled in the first place ?
Considering that it's completely broken, I'd say not really :)
What probably happened is this:
- asmcapa was never included in by the asm code in the i386 directory
- I wrote the SSE routines without the #ifdef ALLEGRO_SSE guards
- I benchmaked them using the test program
- I then added the #include guard, and added some checks for P3 vs
Athlon XP, and changed some little things in hopes of making it
faster/better/more solid
- Since the code compiled perfectly, and seemed to work fine on test
programs (since it wasn't compiled at all, due to #include ALLEGRO_SSE
not being defined), the code was committed to CVS.
- Yesterday, I tried to add some MMX mask blit code
- I noticed there was no performence difference in between it and the
SSE code.
- To make sure the code was being compiled/run correctly, I added a
"hlt" instruction in the middle of the SSE code, which should force the
program to crash.
- The program ran without crashing
- I noticed that even if I typed garbage in the SSE code, it would still
compile and run :/
I have a fixed copy locally, for the asmcapa problem, but the SSE code
is completely broken right now. I'll try to fix it all up tonight or
tomorrow.
--
Eric Botcazou