Alternative GCs for Kaffe ?

Pontus Rydin P.Rydin at astea.se
Sat Jul 18 07:54:44 PDT 1998


I did some playing around with an early version of kaffe (0.56 or something)
Boehm GC (formerely known as PARC GC). Adapting that version of Kaffe to
Boehm GC was pretty straight forward, I think it took me one or two days to
get everying (including finalzers and stuff) to work. From what I remember,
the only tricky part was to make Boehm GC to recognize the "pseudo-thread"
stacks allocated by kaffe. 

I think I still have the code somewhere, but as I said, it was written for
an almost archaic version of kaffe. If you're interested, drop me a note. 

> ----------
> From: 	dyoung at mork.uni.uiuc.edu[SMTP:dyoung at mork.uni.uiuc.edu]
> Reply To: 	kaffe at rufus.w3.org
> Sent: 	Wednesday, July 15, 1998 9:51 PM
> To: 	kaffe at rufus.w3.org
> Subject: 	Alternative GCs for Kaffe ?
> 
> 
> 	Has anyone tried Kaffe with alternative garbage collectors, like
> the one by Boehm (sp?) et al. ? I have an application that could
> benefit from Kaffe's JIT compilation, but Kaffe is so slow to allocate
> and reclaim memory that the application actually runs about twice as
> fast with Sun's interpreter. If anyone has experimented with
> alternative GCs, please let me know.
> 
> 	Also, can someone explain to me the meaning of these blurbs of
> info, emitted when kaffe is run with the -verbosegc option ? -->
> 
> <GC: heap 6144K, total 52790K, alloc 1452K, marked 914K, freeing 51875K>
> <GC: heap 6144K, total 53462K, alloc 1462K, marked 916K, freeing 52545K>
> <GC: heap 6144K, total 53563K, alloc 1453K, marked 914K, freeing 52648K>
> <GC: heap 6144K, total 53883K, alloc 1457K, marked 926K, freeing 52956K>
> 
> Dave
> 
> 
> -- 
> David Young * dcy2 cornell edu
> 


More information about the kaffe mailing list