I heard some complains about performance of ariba.
We know that ariba is not perfectly optimized. However,
due to the rather complex bitwise serialization, you may
get a HUGE performance penalty when using -O0 compiler options.
I've been analysing the serializers assembly output. When
compiling with -O2/-O3 compiler options gcc compiles the
serializers into highly efficient code (almost like hand-written).
Just wanted to let you know...