Porting Boehm-gc to embedded m68k environment

Boehm, Hans hans_boehm@hp.com
Wed Nov 12 19:41:00 GMT 2003


I think that's a bit too restrictive.  I would use bitmap descriptors
whenever all the actual pointer offsets in the object are a
multiple of sizeof(void *), and then eventually, vaguely aim for
making that true for the most common objects in the runtime.

Hans

> -----Original Message-----
> From: java-owner@gcc.gnu.org 
> [mailto:java-owner@gcc.gnu.org]On Behalf Of
> Andrew Haley
> Sent: Wednesday, November 12, 2003 11:06 AM
> To: Boehm, Hans
> Cc: 'tromey@redhat.com'; John Neil; 'Jeff Sturm'; java@gcc.gnu.org
> Subject: RE: Porting Boehm-gc to embedded m68k environment
> 
> 
> Boehm, Hans writes:
>  > According to John, the bug is in gcj itself; the libgcj 
> code is correct.
> 
> Ah, okay.  So, let me get this right: we need to add a test so that we
> only use bitmap descriptors if sizeof (void*) == alignof 
> (void*).  Right?
> 
> Andrew.
> 
> 
>  > > -----Original Message-----
>  > > From: java-owner@gcc.gnu.org 
>  > > [mailto:java-owner@gcc.gnu.org]On Behalf Of
>  > > Tom Tromey
>  > > Sent: Wednesday, November 12, 2003 10:27 AM
>  > > To: Boehm, Hans
>  > > Cc: 'Andrew Haley'; John Neil; 'Jeff Sturm'; java@gcc.gnu.org
>  > > Subject: Re: Porting Boehm-gc to embedded m68k environment
>  > > 
>  > > 
>  > > >>>>> "Hans" == Boehm, Hans <hans_boehm@hp.com> writes:
>  > > 
>  > > Hans> 1) It uncovered a bug in the GC descriptor 
> generation code.  It
>  > > Hans> should fall back to procedure descriptors if it sees any
>  > > Hans> misaligned pointers.
>  > > 
>  > > Does this bug occur in gcj itself, or just the 
> descriptor generation
>  > > code in libgcj?  (I'd suspect the latter.)
>  > > 
>  > > Anyway, we can fix this pretty easily.
>  > > 
>  > > Tom
>  > > 
> 



More information about the Java mailing list