openssl compilation seems not to use any "-O*" flag? openssl test "speed rsa" seems to be about 5 times quicker after hand editing openssl's file "Configure" and adding "-O3" flag to "linux-ppc64" line; I used gcc-4.1.0*, but methink that their is some benefit (3x maybe?) with gcc-3.4.* too... Reproducible: Always Steps to Reproduce: 1. 2. 3.
from the ebuild: # ppc64's current toolchain sucks at optimization and will break this package [[ $(tc-arch) != "ppc64" ]] && NEW_CFLAGS="${NEW_CFLAGS} ${CFLAGS}" Sometimes it's better to be safe than sorry.
Just FYI, -O3 is totally unsafe and often produces broken code w/ gcc-4.x
> Just FYI, -O3 is totally unsafe and often produces broken code w/ gcc-4.x Yes, but I think that it is totally unacceptable that there is no optimization at all for openssl-ppc64. I think that "-O2" with gcc-3.4.* should be used at least. On the other hand, gcc-4.1.* "-O3" works here without props...
> # ppc64's current toolchain sucks at optimization and will break this package btw. I'm using openssl-0.9.7e; this "sucking" seems to be case with 0.9.7g/0.9.8 ?
Closing this bug. We won't "optimize" pretty essential packages to break them...
Is there somewhere any explanation why optimization breaks things apart?