GNU bug report logs - #9542
peval inlines recursive procedures

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 9542 in the body.
You can then email your comments to 9542 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Andy Wingo <wingo <at> pobox.com>
Cc: bug-guile <at> gnu.org
Subject: Re: peval error
Date: Sun, 18 Sep 2011 12:00:30 +0200

Hi Andy,

Andy Wingo <wingo <at> pobox.com> skribis:

> Still, though, the results are not great:

[...]

> Here we see that `fold' was inlined once, but to no use.  More code for
> no more speed.

Yes.  OTOH, it was inlined *and* specialized; peval only “inlines”
where’s there’s at least one static argument (known at compile-time), in
which case there’s room for specialization (in this example, *, 1,
zero?, and the anonymous lambdas were propagated/resolved as
primitives.)

In the above example, I agree that we can argue that inlining wasn’t
useful.

> Waddell and Dybvig report on similar problems in their inlining paper
> (Fast and Effective Procedure Inlining, IU CS Dept. TR 484).  See
> section 4.4 where they discuss this and similar problems.  We should
> figure out what we can do in these cases; and in the case where we can't
> fully inline a call site, perhaps we should abort the attempt to inline
> it.

Right.  I think we should avoid inlining when recursive calls appear in
the residual code, as mentioned in the referred section.

WDYT?

Now, the current heuristic in peval is to “inline” procedures only at
call sites with static arguments.  Conversely, a general procedure
inlining algorithm like Waddell’s may inline even when there are only
dynamic arguments.  This puts more pressure on the inlining criteria
because there are many more potential inlining opportunities.

I don’t think an optimization heuristic can be optimal in all cases.

So the question is not whether we can find such pathological cases, but
how frequent they are in practice.  So I think we must keep an eye on
the optimized code, like you did, to find out more.

Thanks for your feedback!

Ludo’.

Message #12 received at 9542-close <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Andy Wingo <wingo <at> pobox.com>
Cc: bug-guile <at> gnu.org, 9542-close <at> debbugs.gnu.org
Subject: Re: peval error
Date: Sun, 18 Sep 2011 23:05:55 +0200

Hi,

Andy Wingo <wingo <at> pobox.com> skribis:

> Here we see that `fold' was inlined once, but to no use.  More code for
> no more speed.
>
> Waddell and Dybvig report on similar problems in their inlining paper
> (Fast and Effective Procedure Inlining, IU CS Dept. TR 484).  See
> section 4.4 where they discuss this and similar problems.  We should
> figure out what we can do in these cases; and in the case where we can't
> fully inline a call site, perhaps we should abort the attempt to inline
> it.

Fixed in 72b2ca55f6e115927aa4e76401c992f21198681f.

Thanks!

Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.