[Speex-dev] [PATCH] compute_weighted_codebook a little bit faster
Jean-Marc.Valin at USherbrooke.ca
Fri Dec 9 18:14:32 PST 2005
> here is a patch making the function compute_weighted_codebook a little
> bit faster. Not so impressive but avoid a loop and is really faster on
> small platforms like the MIPS I'm working on.
I see your patch does two things. First, it defined
sbv_size = i*subvect_size;
which shouldn't change anything with any compiler that can do CSE. The
second is to not copy shape_cb into shape. This is wrong because
MAC16_16() requires 16-bit inputs. Also, most DSPs are slower at
loading/multiplying chars than shorts.
BTW, have you tested your LSP patch on fixed-point yet?
More information about the Speex-dev