[http://sourceforge.net/p/ufoai/patches/386 Item 386] imported from sourceforge.net tracker on 2013-01-28 20:34:58

A number of Mem_PoolAlloc replaced with Mem_PoolAllocExt where it seems no memset to 0 was necessary.
R_FilterTextures replaced with R_FilterTextures+R_FilterTextures0+R_FilterTextures1 where x0 is the old version and x1 is the new and the new is called only for !monochrome && ! inverted (=100% for me). The proper way to keep the improved execution time would be to specialize into R_FilterTexturesX for each case and specialize each X for the cases.
src/ports/windows/win_main.c: Sys_Quit now logs Sys_Milliseconds() "Time elapsed between WinMain entry and Sys_Quit entry" by Com_Printf.
Also a typo in src/tools/radiant/libs/picomodel.h was fixed that becomes apparent in other compilers and src/shared/shared.h where _stricmp wasn't defined (those that are mre versed in win32 may be able to find an appropriate #include)..

===== Comments Ported from Sourceforge =====

====== peteran (2008-12-05 23:00:18) ======


====== peteran (2008-12-05 23:20:37) ======

Forgot to add: In order to get maximum performance for a give system you may want to experiment with the compiler flags. As I am currently stuck in windows land I do the following for the windows target:
I set -g to get some debug information as I use Intel vtune to assist me
Set -O3 and -march=pentium4 for my system
under other options for compiler i have
-ffast-math
-fno-math-errno
-funsafe-math-optimizations
-ffinite-math-only
-fno-trapping-math
-fsingle-precision-constant
-mfpmath=sse
-mno-ieee-fp
-mmmx
-msse
-msse2
-fomit-frame-pointer

now some of those might be very bad (at least -funsafe-math-optimizations), but I haven't seen anything extreme yet. In any case, the -mmmx, -msse, -msse2 and -mfpmath=sse should be the most interesting to generate nice fast code I think.

Also for those of you that might consider running the vtune, you may like to add -Wl,-pie for the linker so you can run the (really slow) call graphs.
====== peteran (2008-12-06 08:36:58) ======

SSE implementation of R_FilterTexture1 with intrinsics