> What version of gcc do you have? $ gcc --version gcc (Ubuntu 4.4.3-4ubuntu5) 4.4.3
> Also, on OS X I'm using the -fast option to gcc, which is often 2-3x faster > than -O3. --fast is a Mac OX X only option (http://developer.apple.com/mac/library/documentation/DeveloperTools/gcc-4.0.1/gcc/Optimize-Options.html) I've dig and read a little about compilation flags: - gcc man pages - gentoo Compilation Optimization Guide (a linux distro where they recompile every bit), and they recommend "-march=k8 -O2 -pipe" - AMD GCC Quick Reference Guide (my CPU is AMD Athlon 64 x2), and they recommend "-O3 -mtune=k8 -funroll-all-loops -fprefetch-loop-arrays -ffast-math" Here are the results: With defaults from "./acprep opt" : CFLAGS = -O3 -fomit-frame-pointer CXXFLAGS = -fpch-deps -Wconversion -pthread -O3 -fomit-frame-pointer -pipe time ~/ledger/ledger reg -X € -J > /dev/null real 2m5.204s With CXXFLAGS = CFLAGS = -march=k8 -O2 -pipe time ~/ledger-O2/ledger reg -X € -J > /dev/null real 2m12.285s With CXXFLAGS = CFLAGS = -O3 -mtune=k8 -funroll-all-loops -fprefetch-loop-arrays -ffast-math time ~/ledger-O3/ledger reg -X € -J > /dev/null real 2m13.937s My conclusion is that I will stick with defaults compiler flags from "acprep opt"... Thierry
