https://bugs.freedesktop.org/show_bug.cgi?id=29407
--- Comment #14 from steckde...@yahoo.fr 2010-08-08 07:44:55 PDT --- Created an attachment (id=37702) --> (https://bugs.freedesktop.org/attachment.cgi?id=37702) Use fneg instead of neg when inverting the sign of floats Hello, I have very good news for you ! I recompiled my LLVM with all of its debug capabilities and asserts, and I ran Neverball on it. I used GDB to catch the precise moment when a wrong instruction is emitted, and I fixed the problem. The fact is that more instructions than add, sub, mul and div need their float counterpart. My patch fixes two "neg" instructions in lp_bld_tgsi_soa.c. It was the fix I needed to be able to run Neverball. I will now see if there are more places where an instruction needs to be replaced with its float counterpart, but I need to find GL programs that use them. Maybe piglit can help me. The second good news is that Neverball runs really fine with LLVM. I have an AMD Athlon 64 X2 processor at 1,2Ghz, and Neverball ran at nearly 5 fps (800x600), with its graphics settings set to the maximum. My processor was also busy rendering a really big 3D scene, so only the half of the processing power was available to llvmpipe. They weren't any glitches on the screen, even the beautiful reflections of Neverball worked perfectly. Thanks for all :) . -- Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug. _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
