Author: Slater Wold
Date: 09:10:22 03/30/04
Go up one level in this thread
On March 30, 2004 at 10:49:27, Fabien Letouzey wrote: >On March 30, 2004 at 10:30:39, Slater Wold wrote: > >>On March 30, 2004 at 04:40:06, Fabien Letouzey wrote: >> >>>On March 30, 2004 at 04:25:25, Peter Fendrich wrote: >>> >>>>On March 30, 2004 at 01:35:13, Slater Wold wrote: >>>> >>>>>64-bit Gerbil compiled with -O2 -Og -Ot -Oy -Oi -Ob2 -GL: >>>>> >>>>>1,098,625 N/sec on opening position for 30s. >>>>> >>>>>32-bit Gerbil compiled with -O2 -Og -Ot -Oy -Oi -Ob2 -GL: >>>>> >>>>>1,094,747 N/sec on opening position for 30s. >>>>> >>>>> >>>>>32-bit Gerbil that comes with source: >>>>> >>>>>1,115,953 N/sec on opening position for 30s. >>>> >>>> >>>>What is 64-bit Gerbil - a re-write to 64-bit code or the same Gerbil on a 64-bit >>>>cpu? >>>>/Peter >>> >>>I think what Slater is doing is to compile source code with the same compiler in >>>32-bit and 64-bit modes and compare the speed. I don't think he modifies the >>>source code in any way, so it is not a re-write. >>> >>>The tests are probably all made on the same 64-bit hardware. >>> >>>Fabien. >> >>Correct. >> >>I had planned to take the program with the slowest speedup, and try to clean it >>up for 64-bit. Looking like Fruit, so far! :) > >Maybe it's just that Fruit is perfectly optimised for 32-bit hardware ;) That much is obvious! >Any clue so far, apart from the index signed-extension thing? I am still trying to learn fruit. It's no small program! But yes, there are quite a few datatypes that don't run well in 64-bit. Such as sint8, and others. >Fabien.
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.