Author: Jens Kahlenberg
Date: 09:23:32 07/06/03
Go up one level in this thread
On July 06, 2003 at 05:02:50, Gerd Isenberg wrote: >Hi Dieter, > >Thanks for the nice lession in macro programming and gnu/ata-assembler. >Conclusion: branchless code on random, not predictable data is clearly >favorable. The sar instructions is not that slow on P4. >The five byte sequence with cdq is fastest: > >00401026 99 cdq >00401027 33 C2 xor eax,edx >00401029 2B C2 sub eax,edx > >With mvc using math.h abs is fastest. With gcc cdq inline assembly abs or omids >c-abs is much faster than the branching lib abs (maybe a macro from some header >file?). > >Regards, >Gerd Hi Gerd, abs is a builtin in gcc. As long as you don't compile with -fno-builtin compiler doesn't take any library into account and only does lookup abs-code for target system. Would be interesting to know what version Dieter is using and to find out why his version seems to have a __builtin_abs with branching target-code. Dieter might even (with a lot of work) optimize further by compiling a new (P4-targeted) gcc-compiler on his system with the existing one and perhaps abs will be alright then. Regards, Jens
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.