I'm looking at the crash today. I suspect that this is during the RTL-to-instrcution conversion. I've added a new expression to the mix, so maybe that code can't handle it properly.

Yep, that's the case. GCC dies in final_scan_insn() on the subreg expression I added.

Maybe not, I traced the crash down to a debug statement I put in earlier, apparently there is no debug file for the final pass, and so the debug fprintf's cause a crash. I just keep shooting myself in the foot with this debug output.

There's still a few minor bugs, and the MOVQI pattern needs more work, but I think this has some promise. I may need to rewrite all instructions and redo all testing once MOVQI is working. Oh boy...

I've been busy, and haven't really had time for TI stuff lately. I did have bits of time here and there to fiddle with GCC, and in that time I got the 159r.combine step working. So now I'm looking into 172r.ira.

The subreg instructions are removed somewhere in reload(), but since that's represents a huge amount of work, I've got some spelunking ahead of me.

I had a lot more trouble finding reload() than I thought. It's called everywhere, but is located in reload1.c. I'm noting that fact here to possibly save some time later.

The next step is final.c::cleanup_subreg_operands(), called from reload(). That seems to delete the subreg expressions.

simplify_rtx.c::simplify_subreg() is the ultimate cause, which is called by final.c::alter_subreg(), which also makes some changes. But it looks like those changes don't matter for me.

I modified simplify_subreg() to explicitly emit a subreg expression if a word-to-byte conversion is used. That RTL expression seems to be preserved until 201r.shorten. At that point, there's a segmentation fault in GCC. Getting closer...