> Date: Mon, 11 Feb 2019 10:22:12 +0100 > From: Jakub Jelinek <ja...@redhat.com>
> Is PREFERRED_STACK_BOUNDARY what we want here? Shouldn't that be > STACK_BOUNDARY, or PARM_BOUNDARY? Though, PARM_BOUNDARY on cris is 32... Hm. I wish there was a better distinction both in the code and in peoples minds between boundary (or whatever you'd call the multiple to which the size of something must be rounded) and (memory/stack-)alignment. Alas, there are lots of mixups in usage and the documentation even calls this alignment. I'm not even sure it was *size* boundary when I set PARM_BOUNDARY to 32 in the first place! I may have to change that to reflect the current sources. :( Still, this isn't strictly a parameter (to pass or to be expected incoming according to an ABI), but a local copy of what was incoming in registers, so picking the preferred stack boundary as per PREFERRED_STACK_BOUNDARY seems natural. brgds, H-P