This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: FAIL: gcc.c-torture/execute/950628-1.c execution, -O1
- To: rth at redhat dot com
- Subject: Re: FAIL: gcc.c-torture/execute/950628-1.c execution, -O1
- From: kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)
- Date: Fri, 26 Jan 01 15:26:02 EST
- Cc: gcc at gcc dot gnu dot org
I think it's confusing to have "unchanging" memory change.
I agree.
We both agree that we have to reuse "const" temporaries as otherwise
stack frames explode.
Well, I wouldn't say "explode", but I do think there would be a significant
impact.
You mentioned using alias sets for .got data; I think we can use them
everwhere we currently use /u.
I think so too, but I'm not totally sure. Do we gain anything other than
aliasing by knowing that something is unchanging? In other words, what
was the effect of my change that made the memory allocated by a SAVE_EXPR
unchanging? I did it on the grounds that it's always better for the
compiler to know more about things, but what's the precise benefit?
Indeed, what, precisely, are we using the knowlege that something is
"constant" for and are we undoing all of that advantage when we reuse
the slots?