------- Additional Comments From rakdver at atrey dot karlin dot mff dot cuni 
dot cz  2005-05-24 14:24 -------
Subject: Re:  poisoned ggc memory used for -ftree-vectorize

> > Is there a rule that ggc_collect should not be called during loop 
> optimizations?
> 
> I don't know. (Zdenek?)

there are several places in loop opts that are not GGC-safe (in
particular the tree nodes refered from loop structures are not
seen by garbage collector, and I think there are some other instances).

So at the moment, you cannot run ggc_collect inside loop opts.

> I think passes within the loop optimizer can assume that the scev information 
> cached in the scev htab can be safely used. During vectorization analysis 
> some 
> new entries in the scev htab are initialized, and these are valid to be used 
> by 
> subsequent passes. If no transformation takes place during vectorization then 
> there's no need to invalidate this information. So I don't think we're 
> missing 
> a call to scev_reset in the vectorizer.
> 
> The problem is more likely around the fact that ggc_collect, for some reason 
> seems to free up data that is still pointed to from scalar_evolution_info. 
> 
> 
> > If you have some time to spare, use a --enable-checking=gcac compiler, 
> and the failure will likely happen much earlier.
> 
> thanks for the tip 
> 
> 
> > You have to make sure that ... the contents of 
> the scalar evolution hash table are reached during garbage collection.
> 
> any (other) hints on how to do that? 
> 
> (by the way, I actually found a few places where don't properly free up 
> information in the vectorizer, but fixing those didn't solve this particular 
> problem).
> 
> -- 
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>                  CC|                            |rakdver at atrey dot karlin
>                    |                            |dot mff dot cuni dot cz
> 
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21639
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21639

Reply via email to