[FFmpeg-devel] a64 encoder
Sat Jan 17 16:25:53 CET 2009
> ldx $de00
> lda highnib,x
> sta dest
> lda lownib,x
> sta dest+1
> 4+5+4+5+4 = 22 cycles
infact it could even work in 17 cycles (vs. 16 cycles when transferring
normal with unrolling) :
as the highnibble in colorram is ignored anyway, the STX doesn't hurt.
But loopunrolling is not yet necessary, as it would mean more code to
generate an apropriate loop, or lot of copy & paste work (I'm doing all
that in Assembler on the native machine) Yet i need 22 cycles without
unrolling/additional tables, so still no big gain.
More information about the ffmpeg-devel