I thought I'd share my fresh compile of oggenc (optimised for Pentium 4) using the Intel C++ Compiler 7.1 on Linux. It runs exclusively on Pentium 4s
The switches I used were: -O3 -unroll -ip -xW -march=pentium4 -static-libcxa
Just did a quick benchmark comparing the gcc compiled version with my ICC version:
GCC 3.2 oggenc: 26 seconds
ICC 7.1 oggenc: 20 seconds
So that shaved off 6 seconds.
EDIT: It supports flac's now and replaced the 'buggy' GT3b2 with the original 1.0.1
Here is an ICC compile of GT3b1 using the same compiler optimisations above. It runs faster than the gcc compiled executable.
I've attached a tarball of the following linux binaries, which were compiled using the Intel C++ Compiler, Version 7.1.
These should work and be optimised for all Pentium processors that use MMX, SSE, and SSE2 which includes Pentium II, Pentium III, and Pentium 4. Depending on what processor you have, the binary will dynamically select the appropriate optimised code during runtime, thanks to the -ax flags.
The flags used to compile: -O3 -unroll -ip -axMKW -static-libcxa
Binaries in this package include:
P.S. I've tried using the latest version 8 of the Intel C++ compiler but the resulting oggenc turned out to be slower than the 7.1 compile (ICC 8: 9.3 x realtime, ICC 7.1: 10.8 x realtime)