> Hey Luca -- I have to ask how far you're thinking of going with rtasm on > x86-64... I can say from experience that while it's a nice utility for > generating very simple assembly (as in translate), as soon as you start > getting into anything more complex, llvm is by far the better choice...
The code here is very minimal, mostly to get a version of translate_sse designed for x86 working on x86-64 too (and it's also quite experimental and could have issues, even though translate_sse passes the testsuite) I don't plan to expand it further than this personally. LLVM is indeed generally a better option, but translate is probably one of the places where it is less useful, due to the very specific purpose of the code, and the fact that optimal performance requires anyway to use architecture-specific code, and possibly even processor-specific code for things like doing unaligned accesses in the best way for the specific microarchitecture (which I don't attempt to do in this patchset). It would still be interesting to have an LLVM path at least for all other architectures, but I'll leave that for future work. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
