Hi all,
I've made a change to quantize.c/quantize_xrpow() which results in a 12%
overall speedup under Win32/MSVC. All I've done is implement the double ->
int conversions using inline assembler, following the lead of the gcc/i386
version. On the test file I'm using (*) I get a speed of 2.37 for the new
version vs. 2.12 for the original. AFAICT, output is identical. The new
version is attached.
-- Mat.
* all the test wavs from sulaco.org, plus the SQAM files, concatenated in
alphabetical order.
quantize_xrpow.c
