[MP3 ENCODER] Profiling LAME on Altivec Machine

[Martin Hairer]

Hi,

There was some discussion on this list about interfacing lame with Apple's
G4 optimized FFT routines. Surprisingly, it seems that lame spends basically
no time computing FFT's on my G4 machine, but mostly square roots... At the
lowest quality settings, the percentages spent in the various routines according
to Shikari are as follows:

25.9% sqrt
19.6% L3psycho_anal
7.9% exp
6.3% fht
5.3% window_subband
3.6% log
2.9% quantize_xrpow_ISO
2.0% count_bits

At the highest quality setting, the picture is quite different:

23.2% quantize_xrpow
14.9% calc_noise
10.0% count_bits
8.6% log10
6.6% log
4.7% outer_loop
3.8% sqrt
3.7% ix_max
3.0% L3psycho_anal

So interfacing with vecLib might not be such a good idea after all...
Just my 2 cents ;-)

Martin

_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder