Hello, On Sun, Jan 14, 2007 at 02:57:46PM +0100, Michael Niedermayer wrote: > > Main bottleneck seems to be SUBSTSHIFTROWS, for which I have no idea how > > to optimize it further right now in C. > > SSE shuffle instruction could probably be very useful to optimize the > > shifting, but the invsubst table lookup still seems not nice - except > > maybe by using a 64k size table and thus always looking up two at once... > > cant the substitution stuff be merged into the multbl? i think it should > be possible ... I currently don't see a way unless combining it also with the XORBLOCK which would mean a different tbl for each round, too, and would mean rebuilding the tables when changing the key which I would prefer to avoid, I think it would be nice for the key changing operation being a not too slow operation. Greetings, Reimar D?ffinger
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4