c - Intel based hardware speed ups for DCT? -
c - Intel based hardware speed ups for DCT? -
we writing image processing algorithm targeting intel hardware. prefer generic c implementations, have identified algorithm @ core ton of discrete cosine transforms (dct's) works extremely well. unfortunately, our throughput requirements such generic c implementation 2 orders of magnitude slow. can 1 order of magnitude through other tricks, if can improve dct's order of magnitude have path towards success.
is intel mmx way @ hardware acceleration these dct's? there other intel specific libraries and/or hardware can exploit speed these bad boys up?
where start look? new job me, , first time digging hard intel hardware, pointers appreciated.
take @ intel's integrated performance primitives library. contains wealth of routines optimized heavily take utilize of intel architecture, mmx , sse. among many other things, ipp contains routines dct (documentation here).
c image-processing intel dct
Comments
Post a Comment