Re: [RFA] SH4A contribution for sim

[Michael Snyder <msnyder at redhat dot com>]

Joern Rennecke wrote:
> > [...]  This is why I say that when we are making big changes anyway, we
> should also introduce variable length [01]* bitstring parsing.
> [...]
>
> > through, but I'm thinking that, for instance, "01NN" could be
> > handled in two passes: take the "01" first, then recurse and
> > let the "NN" be handled in the outer switch statement.  This
> > might also apply to 01xy and XY01.
>
>
> Yes, that is how I'd like to see this pattern handled.  But the '01' can
> bhe handled by code that processes arbitrary length constant bitstrings.
> There is no need to process only specific length of these. 

OK, I withdraw this patch, and I will submit a new one that
[only] changes expand_opcode along the lines we have been
discussing.  Then I'll submit the rest of this patch separately
(perhaps after further discussion of the movx/movy patterns).

[...]
> I made the movx expansion generate all the combinations of movx with movy
> as well as with nopy. That is this inner loop:
>
>                 for (m = 0; m < 32; m++)
>                   {
>                     /* Ignore illegal nopy */
>                     if ((m & 7) == 0 && m != 0)
>                       continue;
>                     mv = m & 3 | (m & 4) << 2 | (m & 8) << 3 | (m & 16) << 4;

Yes but the h8sx has a bunch of new ones.
We can continue this conversation, I'll separate this change
from the revamp of expand_opcode.