Il giorno lun, 15/11/2010 alle 00.41 +0100, Thomas Kahle ha scritto: > Using ‘--enable-fat’ selects a “fat binary” build on x86 or x86 64 > systems, where optimized low level subroutines are chosen at runtime > according to the CPU de- tected. This means more code, but gives > reasonable performance from a single bi- nary for all x86 chips, or > similarly for all x86 64 chips. (This option might become available > for > more architectures in the future.)
Okay this then has nothing to do with FatELF, and the name "fat binary" here is misused by most canons. What you have is instead a "runtime cpu detection", which is something glibc (via ifunc), ffmpeg, mplayer, xine and so on already supports. My personal suggestion having worked with that kind of code is not to enable it unless it has more fine-grained logic than the simple arch-based one (such as FFmpeg's way to detect broken SSE3 implementations). And if you do enable it, the USE flag you need is "cpudetection" (already used by ffmpeg, mplayer and jack-audio-connection-kit). -- Diego Elio Pettenò — “Flameeyes” http://blog.flameeyes.eu/ If you found a .asc file in this mail and know not what it is, it's a GnuPG digital signature: http://www.gnupg.org/
