soundcheck wrote: > Some words about gcc compiler options: > > I use ' Linpack' (http://elinux.org/RPi_Performance) for the PI2. And > gcc 4.9.2. > > > Case 1: > > Default gcc option > > > Code: -------------------- > > > cc -O2 -march=armv7-a -mfloat-abi=hard -mfpu=vfpv3-d16 -o linpack linpack.c -lm > > linpack = 123279.473 kflops > -------------------- > > > > Tuned gcc options > > Case 2: > > > Code: -------------------- > > > cc -O3 -march=native mtune=native -mfloat-abi=hard -o linpack linpack.c -lm > > linpack = 128987.632 > -------------------- > > > > Roughly 5% difference to Case 1! > > Case 3: > > > Code: -------------------- > > > cc -O3 -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard -funsafe-math-optimizations -o linpack linpack.c -lm > > linpack = 159337.319 kflops > -------------------- > > > > Roughly 30% difference to Case 1! > > > As you can see gcc options can make a huge difference. > Not to forget. Also the gcc compiler versions are making a huge > difference. From around revision 4.8.x ARM architecture is much better > supported.
Actually 5% is only a small difference, and the 30% big difference is gained by explicitly enabling optimisations which call themselves unsafe. Why would anyone do that? Furthermore, this is an improvement when running test code, not LMS. How much improvement do these changes make to LMS? And finally, why does this matter? If the cpu is running at less than 100% already then all this will do is increase the free cpu time, which does basically nothing. In particular I cannot see how any of this can improve sound quality. ------------------------------------------------------------------------ Owen Smith's Profile: http://forums.slimdevices.com/member.php?userid=42371 View this thread: http://forums.slimdevices.com/showthread.php?t=101624 _______________________________________________ unix mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/unix
