public inbox for gcc-help@gcc.gnu.org
 help / color / mirror / Atom feed
* gimplify.c - missing tree codes
@ 2010-03-05 19:38 Steve Teale
  2010-03-06  0:26 ` Ian Lance Taylor
  0 siblings, 1 reply; 4+ messages in thread
From: Steve Teale @ 2010-03-05 19:38 UTC (permalink / raw)
  To: gcc-help

In 4.3.4,the switch in function gimplify_expr(...) included:


	case STATIC_CHAIN_EXPR:
	  /* The argument is used as information only.  No need to
	     gimplify */
	case STATIC_CHAIN_DECL:  
	  ret = GS_ALL_DONE;
	  break;

4.4.3 does not mention these codes, so any front end that generates them
fails at a later catch-all with an Internal compiler error.

What was the rationale for removing them? Is it a big deal to put them
back?

Thanks
Steve



^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: gimplify.c - missing tree codes
  2010-03-05 19:38 gimplify.c - missing tree codes Steve Teale
@ 2010-03-06  0:26 ` Ian Lance Taylor
  2010-03-06  7:59   ` Steve Teale
  0 siblings, 1 reply; 4+ messages in thread
From: Ian Lance Taylor @ 2010-03-06  0:26 UTC (permalink / raw)
  To: Steve Teale; +Cc: gcc-help

Steve Teale <steve.teale@britseyeview.com> writes:

> In 4.3.4,the switch in function gimplify_expr(...) included:
>
>
> 	case STATIC_CHAIN_EXPR:
> 	  /* The argument is used as information only.  No need to
> 	     gimplify */
> 	case STATIC_CHAIN_DECL:  
> 	  ret = GS_ALL_DONE;
> 	  break;
>
> 4.4.3 does not mention these codes, so any front end that generates them
> fails at a later catch-all with an Internal compiler error.
>
> What was the rationale for removing them? Is it a big deal to put them
> back?

I can't see any reference to STATIC_CHAIN_EXPR or STATIC_CHAIN_DECL in
any gcc versions from 4.2 on.  Nor can I find that specific code.
Could this be a local patch in your gcc 4.3.4?

Ian

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: gimplify.c - missing tree codes
  2010-03-06  0:26 ` Ian Lance Taylor
@ 2010-03-06  7:59   ` Steve Teale
  2010-03-08 14:33     ` Ian Lance Taylor
  0 siblings, 1 reply; 4+ messages in thread
From: Steve Teale @ 2010-03-06  7:59 UTC (permalink / raw)
  To: Ian Lance Taylor; +Cc: gcc-help


> I can't see any reference to STATIC_CHAIN_EXPR or STATIC_CHAIN_DECL in
> any gcc versions from 4.2 on.  Nor can I find that specific code.
> Could this be a local patch in your gcc 4.3.4?
> 
> Ian

Thanks Ian - yes, I had completely ignored that piece of patch, or more
likely was looking at the wrong version code and concluded I'd already
applied it. Working with differences between gdc-4.3.4/dev/gcc-4.3.4/gcc
and gdc-4.4.3/dev/gcc4.4.3/gcc can get a tad confusing if you're tired.

I have two outstanding buts of patch where I'm not making much progress.
Between 4.3.4 and 4.4.3, tree-gimple disappeared, and was replaced by
gimple, and tree-nested was radically changed. Do you know much about
those changes?

Thanks
Steve



^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: gimplify.c - missing tree codes
  2010-03-06  7:59   ` Steve Teale
@ 2010-03-08 14:33     ` Ian Lance Taylor
  0 siblings, 0 replies; 4+ messages in thread
From: Ian Lance Taylor @ 2010-03-08 14:33 UTC (permalink / raw)
  To: Steve Teale; +Cc: gcc-help

Steve Teale <steve.teale@britseyeview.com> writes:

> I have two outstanding buts of patch where I'm not making much progress.
> Between 4.3.4 and 4.4.3, tree-gimple disappeared, and was replaced by
> gimple, and tree-nested was radically changed. Do you know much about
> those changes?

In gcc 4.4 we introduced a new intermediate representation which uses
less memory and is more efficient to work with.  It is called GIMPLE
and is documented in doc/gimple.texi.  The C/C++ frontends used to
produce general trees, and the gimplification pass trimmed that down
to a language-independent set of trees known as GENERIC.  The
gimplification pass now produces a new IR, GIMPLE, rather than
GENERIC.  I don't know what your patches do or where they are supposed
to take effect.  If you are changing something which affects the
middle-end, then you are going to need to write it for GIMPLE rather
than for GENERIC.  If your new tree codes are front-end specific, then
you need to figure out how to translate them into GIMPLE.

Ian

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2010-03-08 14:33 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-03-05 19:38 gimplify.c - missing tree codes Steve Teale
2010-03-06  0:26 ` Ian Lance Taylor
2010-03-06  7:59   ` Steve Teale
2010-03-08 14:33     ` Ian Lance Taylor

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).