* [tree-ssa] -fdisable-gimple
@ 2003-10-03 13:21 Jeff Sturm
2003-10-03 14:23 ` Daniel Berlin
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Jeff Sturm @ 2003-10-03 13:21 UTC (permalink / raw)
To: rth, dnovillo; +Cc: gcc
How important is it that the branch continue to support -fdisable-gimple?
For C or all frontends?
I see that Java has been completely broken with -fdisable-gimple for some
time. The flag gets no routine testing, making it worse than useless IMO.
More urgently, I don't think we can do tree_rest_of_compilation in any
sane fashion without resorting to
void
java_expand_body (tree fndecl)
{
if (! flag_disable_gimple)
tree_rest_of_compilation (fndecl);
else
{
/* ... preserve all old cruft here ... */
}
}
Jeff
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [tree-ssa] -fdisable-gimple
2003-10-03 13:21 [tree-ssa] -fdisable-gimple Jeff Sturm
@ 2003-10-03 14:23 ` Daniel Berlin
2003-10-03 16:27 ` Jeff Sturm
2003-10-03 15:04 ` Steven Bosscher
2003-10-03 17:02 ` Diego Novillo
2 siblings, 1 reply; 6+ messages in thread
From: Daniel Berlin @ 2003-10-03 14:23 UTC (permalink / raw)
To: Jeff Sturm; +Cc: rth, dnovillo, gcc
On Fri, 3 Oct 2003, Jeff Sturm wrote:
> How important is it that the branch continue to support -fdisable-gimple?
> For C or all frontends?
I believe it's only actual use is by the SPEC tester to have something to
use as the baseline.
>
> I see that Java has been completely broken with -fdisable-gimple for some
> time. The flag gets no routine testing, making it worse than useless IMO.
Actually, the SPEC tester uses it every night, right Diego?
>
> More urgently, I don't think we can do tree_rest_of_compilation in any
> sane fashion without resorting to
>
> void
> java_expand_body (tree fndecl)
> {
> if (! flag_disable_gimple)
> tree_rest_of_compilation (fndecl);
> else
> {
> /* ... preserve all old cruft here ... */
> }
> }
>
> Jeff
>
>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [tree-ssa] -fdisable-gimple
2003-10-03 14:23 ` Daniel Berlin
@ 2003-10-03 16:27 ` Jeff Sturm
0 siblings, 0 replies; 6+ messages in thread
From: Jeff Sturm @ 2003-10-03 16:27 UTC (permalink / raw)
To: Daniel Berlin; +Cc: rth, dnovillo, gcc
On Fri, 3 Oct 2003, Daniel Berlin wrote:
> On Fri, 3 Oct 2003, Jeff Sturm wrote:
> > I see that Java has been completely broken with -fdisable-gimple for some
> > time. The flag gets no routine testing, making it worse than useless IMO.
>
> Actually, the SPEC tester uses it every night, right Diego?
It might. SPEC doesn't touch Java.
So I can try to make it work again, leave it broken or just kill
-fdisable-gimple in Java. What do tree-ssa folks prefer?
Jeff
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [tree-ssa] -fdisable-gimple
2003-10-03 13:21 [tree-ssa] -fdisable-gimple Jeff Sturm
2003-10-03 14:23 ` Daniel Berlin
@ 2003-10-03 15:04 ` Steven Bosscher
2003-10-03 17:11 ` Gerald Pfeifer
2003-10-03 17:02 ` Diego Novillo
2 siblings, 1 reply; 6+ messages in thread
From: Steven Bosscher @ 2003-10-03 15:04 UTC (permalink / raw)
To: Jeff Sturm; +Cc: rth, dnovillo, gcc
Op vr 03-10-2003, om 15:21 schreef Jeff Sturm:
> How important is it that the branch continue to support -fdisable-gimple?
> For C or all frontends?
There already are C and C++ programs that do not compile with
-fdisable-gimple (such as the test case for PR8361), and nobody has
complained so apparently it's not all that important.
Gr.
Steven
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [tree-ssa] -fdisable-gimple
2003-10-03 15:04 ` Steven Bosscher
@ 2003-10-03 17:11 ` Gerald Pfeifer
0 siblings, 0 replies; 6+ messages in thread
From: Gerald Pfeifer @ 2003-10-03 17:11 UTC (permalink / raw)
To: Steven Bosscher; +Cc: Jeff Sturm, rth, Diego Novillo, gcc
On Fri, 3 Oct 2003, Steven Bosscher wrote:
>> How important is it that the branch continue to support -fdisable-gimple?
> There already are C and C++ programs that do not compile with
> -fdisable-gimple (such as the test case for PR8361), and nobody has
> complained so apparently it's not all that important.
Well, it's interesting in that the test case for PR8361 fails both
with -fdisable-gimple and without -- and with two different failure
modes! :-(
It would be really, really good could someone finally debug this and
fix tree-ssa.
Gerald
PS: I'm even willing to take a bet that fixing this bug will help
uncover further tree-ssa bugs triggered by my code.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [tree-ssa] -fdisable-gimple
2003-10-03 13:21 [tree-ssa] -fdisable-gimple Jeff Sturm
2003-10-03 14:23 ` Daniel Berlin
2003-10-03 15:04 ` Steven Bosscher
@ 2003-10-03 17:02 ` Diego Novillo
2 siblings, 0 replies; 6+ messages in thread
From: Diego Novillo @ 2003-10-03 17:02 UTC (permalink / raw)
To: Jeff Sturm; +Cc: Richard Henderson, gcc
On Fri, 2003-10-03 at 09:21, Jeff Sturm wrote:
> How important is it that the branch continue to support -fdisable-gimple?
> For C or all frontends?
>
It's not. At first it was a way of debugging the gimplifier, then it
became a quick and dirty way of comparing tree-ssa against
something-resembling-mainline when running SPEC. I never modified the
SPEC scripts to do comparison runs between different compilers. I
should.
Unless any FE folk have a use for it, I'd say we should discard the
flag.
Diego.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2003-10-03 17:11 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-10-03 13:21 [tree-ssa] -fdisable-gimple Jeff Sturm
2003-10-03 14:23 ` Daniel Berlin
2003-10-03 16:27 ` Jeff Sturm
2003-10-03 15:04 ` Steven Bosscher
2003-10-03 17:11 ` Gerald Pfeifer
2003-10-03 17:02 ` Diego Novillo
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).