Possible optimization opportunities about gcc-ia16 HOT 3 CLOSED

Hello @bartoldeman,

Thanks for the report. My understanding is that -fcaller-saves (enabled with -Os) should, by right, not result in worse code than -fno-caller-saves. I will need to find out what is going on here.

Thank you!

from gcc-ia16.

Hello @bartoldeman,

I have added an optimization subpass to do something along the lines of what you proposed.

(As always, of course, I will need to do further tests on it to make sure it does not break anything.)

Thank you!

from gcc-ia16.

Hello @bartoldeman,

Further down the compiled code, there is this bit which makes it a bit hard to get rid of the frame pointer %bp:

	movw	-2(%bp),	%dx
	xorb	%cl,	%cl
	cmpw	%ax,	%dx
	jle	.L2
	movb	$1,	%cl
.L2:
	...
	movw	%bp,	%sp
	popw	%bp
	popw	%si
	ret

I think it will be nice if GCC can somehow change the move into %dx into a popw %dx, but my optimization subpass does not do that yet.

Thank you!

from gcc-ia16.