Author: Matt Taylor
Date: 02:00:16 01/18/03
Go up one level in this thread
On January 17, 2003 at 22:00:16, Russell Reagan wrote: >On January 17, 2003 at 20:28:03, David Rasmussen wrote: > >>I profiled my code and found (not very surprisingly) that >> >>INLINE BitBoard Mask(Square square) { return mask[square]; } >> >>was a hotspot. >> >>I tried to change it to >> >>INLINE BitBoard Mask(Square square) { return BitBoard(1) << square; } >> >>I've tried that many times before, and this time too, the latter was slower. But >>why? I mean, the latter is one instruction on x86 or something like that, and >>square is probably in a register anyway. I would think that it would be faster >>than a table lookup which accesses slow memory. >> >>How come? > >Well, neither of those are one instruction unless you are using 32-bit >bitboards, or you are writing a checkers program, or you have a 64-bit computer. >The table lookup shouldn't have any "slow memory accesses", but rather in the >cache. I would think that the table lookup in the cache would be faster than a >64-bit shift. 64-bit shift in something like 3 cycles when count is < 32. Pentium 4 L1 cache latency -- 2 clocks. Athlon L1 cache latency -- 3 clocks. Bad performance of former indicitive of poor optimization. -Matt
This page took 0.01 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.