[Libre-soc-dev] MP3 DCT36

[Luke Kenneth Casson Leighton]

DCT36 next! that's actually going to be really exciting, because
it's such a well-known algorithm.

    for (j = 0; j < 2; j++) {
        tmp1 = tmp + j;
        in1 = in + j;

        t2 = in1[2*4] + in1[2*8] - in1[2*2];

hmmmm... i wonder... is it worthwhile doing REMAP right now?
that would allow "jumping" (4x) over a span, in-place:

 buf[4 * ( 9 + 8 - j     )]

https://libre-soc.org/openpower/sv/remap/

this could be covered with
* invxyz setting J to invert,
* permute= YX
* Xdim = 4, Ydim=4

however... and this is complicated.... we also have to implement
context-propagation, to do it:
https://libre-soc.org/openpower/sv/propagation/

context-propagation says, "in the next NN instructions, please tell
instructions 1, 3, 5, 6, 7, and 19 ahead of THIS one that they are to
apply context X on TOP"

thus it is a bit of a pain in the ass to calculate, and unless there is
a really compelling reason i'd like to leave it for the "advanced"
version if that's ok.

l.