* CFG not taken care
@ 2003-12-10 10:36 Vimal Kumar
2003-12-10 11:48 ` Jan Hubicka
0 siblings, 1 reply; 3+ messages in thread
From: Vimal Kumar @ 2003-12-10 10:36 UTC (permalink / raw)
To: gcc
Hi,
I was just going through the GCC code when I happen
to notice something strange. It may not be bug, but
I'm confused.
Defined in rtlanal.c, the function reg_set_between_p
returns true if a register is set between two insns;
all three things passed in its arguments. To my
surprise, there was just a loop like:
for (insn = (NEXT_INSN (from_insn); insn != to_insn;
insn = NEXT_INSN (insn))
{
........
}
It is called by many passes; but the important point
is that its traversal is linear. The CFG is not taken
care of. For example, consider the scenario:
.L1
from_insn;
bra .L2 (unconditional jump)
.L3
......
to_insn;
.L2
.....
/* set the reg here */
bra .L2
Certainly, these cases cannot be caught by
reg_set_between_p. There is a whole set of such
functions like reg_referenced_between_p, etc. Is there
any other set of functions covering CFG??
I may be missing some important observation. Sorry
for my ignorance!!
Thanks and Regards,
=====
__________________________________
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
http://photos.yahoo.com/
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: CFG not taken care
2003-12-10 10:36 CFG not taken care Vimal Kumar
@ 2003-12-10 11:48 ` Jan Hubicka
2003-12-10 12:28 ` Theodore Papadopoulo
0 siblings, 1 reply; 3+ messages in thread
From: Jan Hubicka @ 2003-12-10 11:48 UTC (permalink / raw)
To: Vimal Kumar; +Cc: gcc
> Hi,
>
> I was just going through the GCC code when I happen
> to notice something strange. It may not be bug, but
> I'm confused.
>
> Defined in rtlanal.c, the function reg_set_between_p
> returns true if a register is set between two insns;
> all three things passed in its arguments. To my
> surprise, there was just a loop like:
>
> for (insn = (NEXT_INSN (from_insn); insn != to_insn;
> insn = NEXT_INSN (insn))
> {
> ........
> }
>
> It is called by many passes; but the important point
> is that its traversal is linear. The CFG is not taken
> care of. For example, consider the scenario:
>
> .L1
> from_insn;
> bra .L2 (unconditional jump)
>
> .L3
> ......
> to_insn;
>
> .L2
> .....
> /* set the reg here */
> bra .L2
>
> Certainly, these cases cannot be caught by
> reg_set_between_p. There is a whole set of such
> functions like reg_referenced_between_p, etc. Is there
> any other set of functions covering CFG??
>
> I may be missing some important observation. Sorry
> for my ignorance!!
These functions are supposed to be used only locally inside superblock
where this don't matter. We do similar local analysis at many places in
GCC still :(
Honza
>
> Thanks and Regards,
>
>
> =====
>
>
> __________________________________
> Do you Yahoo!?
> New Yahoo! Photos - easier uploading and sharing.
> http://photos.yahoo.com/
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: CFG not taken care
2003-12-10 11:48 ` Jan Hubicka
@ 2003-12-10 12:28 ` Theodore Papadopoulo
0 siblings, 0 replies; 3+ messages in thread
From: Theodore Papadopoulo @ 2003-12-10 12:28 UTC (permalink / raw)
To: Jan Hubicka; +Cc: gcc, Vimal Kumar
hubicka@ucw.cz said:
> These functions are supposed to be used only locally inside superblock
> where this don't matter. We do similar local analysis at many places
> in GCC still :(
Can we deduce of this sentence (and the question) that the name of the
function and/or the comment should be updated to explicitly give this
information ??
--------------------------------------------------------------------
Theodore Papadopoulo
Email: Theodore.Papadopoulo@sophia.inria.fr Tel: (33) 04 92 38 76 01
--------------------------------------------------------------------
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2003-12-10 11:48 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-12-10 10:36 CFG not taken care Vimal Kumar
2003-12-10 11:48 ` Jan Hubicka
2003-12-10 12:28 ` Theodore Papadopoulo
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).