Assembler optimized Vector lib

[elias]

Hi,

Now that we want SSE/3Dnow! support in lwjgl vector operations, what can be done to make the coding of the routines as painless as possible? I'd like to have one common assembler code base on linux and win32, since coding to different CPUs is hard enough. But coding the same routines twice in incline VC++ MASM and inline gcc GAS simply seems wrong to me. I'm considering NASM which seems to be a nice assember available on both win32 and linux.

But this problem must have been solved before, so what _is_ the best way to code "software platform independent" assembler for the i386 hardware platform?

- elias

[vrm]

nasm (not Masm) is a good choice for me it work well with GCC (my life).

Apparently CrystalSpace use it and they build on both VC++ and GCC/MINGW (never tried it with VC++).