David, I am doing the XPCOMGC rewrites. There are three rewrites in the series:
* garburator implemented using oink/elsa
* GCObject inheritance implemented using dehydra and text processing
* remove-addrefs implemented using text processing
A single script runs these three all in a row without human intervention. If you want to see an example of the results, see the three “automatic-*” files from a previous revision. Caution: the patches are almost 200k lines, 20+MB a piece.
I wonder how this load would run on a Sun UltraSPARC T2 box. Say a dual socket 1.4GHz rig. That gives you 128 threads at 8 cores/socket and 8 threads/core. Seeing as how the load is probably memory and integer intensive only, the fit would probably be pretty good.
April 24th, 2008 at 1:55 pm
Hi Benjamin,
Is there anywhere I could view/browse those rewritten pieces of source code?
Thanks.
April 24th, 2008 at 2:09 pm
An infinite number of monkeys require processor time.
April 24th, 2008 at 2:55 pm
Here’s a stupid question forya: What are you rewriting?
April 24th, 2008 at 4:15 pm
And once again, CPU6 is being a slacker. *shakes head*
April 24th, 2008 at 5:39 pm
David, I am doing the XPCOMGC rewrites. There are three rewrites in the series:
* garburator implemented using oink/elsa
* GCObject inheritance implemented using dehydra and text processing
* remove-addrefs implemented using text processing
A single script runs these three all in a row without human intervention. If you want to see an example of the results, see the three “automatic-*” files from a previous revision. Caution: the patches are almost 200k lines, 20+MB a piece.
For more information about Mozilla’s automatic rewriting projects in general, see Taras and David Mandelin‘s blogs, and the mozilla.dev.static-analysis newsgroup.
April 24th, 2008 at 7:15 pm
Just curious, but can you apply the patches and compile and run the build or is the MMgc stuff not set up yet?
April 25th, 2008 at 4:29 am
I wonder how this load would run on a Sun UltraSPARC T2 box. Say a dual socket 1.4GHz rig. That gives you 128 threads at 8 cores/socket and 8 threads/core. Seeing as how the load is probably memory and integer intensive only, the fit would probably be pretty good.
April 25th, 2008 at 5:14 am
An hour for a full rewriting pass… and how long to fry an egg on that box?