Author: Tom Likens
Date: 13:19:07 07/17/05
I'm curious if anyone else has used Intel's new 9.0 compiler. I recompiled Djinn for 64-bits and I'm seeing around 1.3M nps (on average middlegame positions). The best I could get out of the 64-bit version of g++ 3.3.5 was around 900k nps (all of this is on an Athlon FX-51). For the curious my Intel compile settings were (note, this uses profiling info)... intel64-prof: CC = /opt/intel/cce/9.0/bin/icpc intel64-prof: OPT = -DX86_64 -DFAST -pipe -O3 -fomit-frame-pointer intel64-prof: OPT := $(OPT) -no-prec-div -fno-rtti -funroll-loops intel64-prof: OPT := $(OPT) -align -Zp16 -static -march=pentium4 intel64-prof: OPT := $(OPT) -prof_genx -prof_dir ./profile -fp-model fast intel64-prof: OPT := $(OPT) -rcd intel64-prof: DBG = intel64-prof: LFLAGS = -static I'm not sure if these switches are optimal, but they seem reasonable and in my limited number of experiments seem to give the best results. regards, --tom
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.