[FFmpeg-devel] [PATCH 0/5] Integer scalar product DSP functions
christophe.gisquet at gmail.com
Thu May 9 11:23:06 CEST 2013
I'm not going to reply all comments. This set was more as a RFC, and
indeed it seems better to try to not modify the asm.
2013/5/8 Michael Niedermayer <michaelni at gmx.at>:
>> + AV_ZERO128(buffer_a+BLOCKSIZE);
> if buffer_a is moved to the context then this could be avoided
> also it would avoid the need for LOCAL_ALIGNED()
> also trying to build the code with the patch results in:
Yes, I took a shortcut here (local modification then diff against
master of too few files).
Regarding TAK and its need for 4 elements in a batch, this is probably
too costly compared to (schematically):
scalprod(a, b, length&-4)
for(i=length&-4; i<length; i++) sum += a[i]*b[i];
ie, the issue that applies for wma lossless also applies for tak.
More information about the ffmpeg-devel