Using swap, ext4, and the gcc-mips native compiler on the Ben--they all work!

Sergey Kvachonok ravenexp at
Fri Dec 3 01:40:11 EST 2010

> 6.  To test gcc, I found and downloaded a C-language version of the
> long-used linpack benchmark for floating-point performance.

Measuring floating point performance on a machine without a FPU is an
interesting exercise, but the results are a bit misleading. gcc fp-bit
softfloat performance is known to be miserable. This CPU is capable of
some 300+ MIPS and has SIMD extensions operating on quad 16 bit /
double 32bit words. It should be capable of decoding 320x240  H.264
video realtime, there are optimized mplayer plugins at Ingenics site.

I compiled both mpg123, which uses floating point decoder, and mpg321,
which uses integer-only libmad for Ben, and compared performance. The
results: mpg123 can't play mp3 realtime at all, while mpg321 eats some
25% CPU time. I think libvorbis/libtremor benchmarking would give
similar results, so I'd avoid libmpg123 and libvorbis in any media
software on NanoNote.

Thanks for sharing your data.

More information about the discussion mailing list