- jhc - Haskell.org

darcs patch: fix environment cleanups in regress test (and 2 more)
by John Meacham 10 Jul '10

10 Jul '10

Thu Jul 8 03:34:50 PDT 2010 John Meacham <john(a)repetae.net> * fix environment cleanups in regress test Thu Jul 8 14:26:50 PDT 2010 John Meacham <john(a)repetae.net> * remove unused and obsolete flags Thu Jul 8 18:40:12 PDT 2010 John Meacham <john(a)repetae.net> * remove Query dependency for now

1 0

darcs patch: allow plain 'maxBound/minBound/umaxBound... (and 2 more)
by John Meacham 08 Jul '10

08 Jul '10

Mon Jul 5 00:22:43 PDT 2010 John Meacham <john(a)repetae.net> * allow plain 'maxBound/minBound/umaxBound' primitives when the type is clear from context Mon Jul 5 00:23:55 PDT 2010 John Meacham <john(a)repetae.net> * remove special handling and magic instantiation of bounded, eq, and ord for primitive integral types Mon Jul 5 00:45:05 PDT 2010 John Meacham <john(a)repetae.net> * allow case scrutinization of primitives when converting grin to c

1 0

the new GC, anyone got a bitset?
by John Meacham 08 Jun '10

08 Jun '10

So I am exceedingly pleased with what has been happening with the new GC in jhc, enough that I am going to push for a major release in not long. With some careful profiling and tweaking I have been able to ensure enabling GC is a fairly dramatic win in terms of performance and memory for the majority of haskell programs. There is just one thing that is bothering me, I still rely on libJudy in the runtime for its extremely fast bitset operation. Basically, in order to determine if an address is allocated on the heap vs on the stack or in the bss or data segments, I use a bitmap, adding an entry for each 4M chunk of memory dedicated to heap pages. on a 64 bit system, using a plain bit array is just not feasable, libJudy has a very fast bitset that I utilize, but it is not exactly lightweight and is LGPL which may cause issues for creating statically linked executables from jhc. So, I am looking for a very fast (in querying, updates need not be that fast) bitset that maps a 'uintptr_t' to either true or false, chances are it will be very sparse. I'd like something I can embed in the runtime, so it should not be very big. jhc executables are on the order of 8k maybe compiled, so adding 4k of bitmap code for just this isn't really ideal. So, anyone have something in the public domain sitting around? John -- John Meacham - ⑆repetae.net⑆john⑈ - http://notanumber.net/

2 2

darcs patch: add explicit registers to grin
by John Meacham 10 Apr '10

10 Apr '10

Thu Apr 8 04:02:41 PDT 2010 John Meacham <john(a)repetae.net> * add explicit registers to grin

1 0

good news everyone
by John Meacham 09 Apr '10

09 Apr '10

I have finally gotten around to what I have been promising to do for years and implemented a garbage collector for jhc. It is still fairly rough, but is very promising. In truth, I was worried it might kill jhc's performance advantage, but quite the opposite, it actually improves the performance of several benchmarks, and doesn't cut too deep into jhc's speed advantage in others. In particular, it seems to help ones where jhc did particularly worse than ghc, I had always assumed those cases were due to some pathological case in an inner loop that ghc was able to catch but jhc wasn't, but it turns out the presence of the GC played some role in their speed. I am not sure whether it is the short circuiting redirects or memory locality that is helping. In any case, JGC seems to be a couple times faster than the boehm collector so all in all, it is a good option to have. Right now, using jgc means you have to link against the judy library, this restriction may or may not be lifted in the future. libJudy is an amazingly useful and versatile library, keeping the GC information stored independently of the heap allows me to garbage collect without trouncing all over the memory space and I think this is key to getting acceptable performance from the GC. I am actually quite pleased with the results, the implementation is a very naive mark-and-don't-quite-sweep collector and achieves good performance, in particular, it isn't even generational. So there is still a fair amount of room for speed improvement. The GC is still experimental, it still doesn't support some features like IORefs and the method for collecting gc roots is fairly brute force and sloppy, but the essential work is done. John -- John Meacham - ⑆repetae.net⑆john⑈ - http://notanumber.net/

3 3

darcs patch: switch terminology from pages and arenas to blocks and...
by John Meacham 08 Apr '10

08 Apr '10

Wed Apr 7 19:30:38 PDT 2010 John Meacham <john(a)repetae.net> * switch terminology from pages and arenas to blocks and megablocks to agree with ghc and reduce confusion. introduce megablocks as an allocation unit.

1 0

darcs patch: clean up GC code, make distinction of po... (and 3 more)
by John Meacham 08 Apr '10

08 Apr '10

Wed Apr 7 12:09:43 PDT 2010 John Meacham <john(a)repetae.net> * clean up GC code, make distinction of pointers more type safe to catch errors Wed Apr 7 12:52:09 PDT 2010 John Meacham <john(a)repetae.net> * add page cache for recently freed pages Wed Apr 7 17:15:45 PDT 2010 John Meacham <john(a)repetae.net> * optimize and clean up code based on profiling Wed Apr 7 17:50:39 PDT 2010 John Meacham <john(a)repetae.net> * get rid of special handling of the what_t in data and the needed bookeeping.

1 0

darcs patch: switch tail queue to singly linked list ... (and 7 more)
by John Meacham 07 Apr '10

07 Apr '10

Tue Apr 6 21:04:07 PDT 2010 John Meacham <john(a)repetae.net> * switch tail queue to singly linked list for keeping track of pages Tue Apr 6 21:18:24 PDT 2010 John Meacham <john(a)repetae.net> * move page info to page_info struct Tue Apr 6 21:41:24 PDT 2010 John Meacham <john(a)repetae.net> * switch size to be in units of words rather than bytes Tue Apr 6 21:59:22 PDT 2010 John Meacham <john(a)repetae.net> * clean up jhc_jgc.c Tue Apr 6 22:27:55 PDT 2010 John Meacham <john(a)repetae.net> * get rid of entry header on each memory location, use header at top of page. Tue Apr 6 23:02:01 PDT 2010 John Meacham <john(a)repetae.net> * add 'best' caching to page lists, helping with fragmentation issues Wed Apr 7 00:30:31 PDT 2010 John Meacham <john(a)repetae.net> * add needed files to generated c code for GC Wed Apr 7 00:44:59 PDT 2010 John Meacham <john(a)repetae.net> tagged tokfekyuvi

1 0

darcs patch: start adding slub allocator (and 4 more)
by John Meacham 07 Apr '10

07 Apr '10

Mon Apr 5 01:42:27 PDT 2010 John Meacham <john(a)repetae.net> * start adding slub allocator Mon Apr 5 20:14:20 PDT 2010 John Meacham <john(a)repetae.net> * add per-object caches Tue Apr 6 15:32:10 PDT 2010 John Meacham <john(a)repetae.net> * start using bits in pages directly as grey set Tue Apr 6 19:20:03 PDT 2010 John Meacham <john(a)repetae.net> * keep seperate page lists for partial and fully used pages Tue Apr 6 20:56:53 PDT 2010 John Meacham <john(a)repetae.net> * performance tweaks and fix ups of jgc code

1 0

darcs patch: TAG ishbymnoabi (and 2 more)
by John Meacham 07 Apr '10

07 Apr '10

Mon Mar 29 16:15:06 PDT 2010 John Meacham <john(a)repetae.net> tagged ishbymnoabi Mon Mar 29 19:19:46 PDT 2010 John Meacham <john(a)repetae.net> * bug updates Tue Mar 30 01:48:38 PDT 2010 John Meacham <john(a)repetae.net> * c generation clean ups

1 0