#haskell - Mon 19 Feb 2007 between 12:05 and 12:28

NY Lost Funds



donswow. hbc -O beats ghc -O by 25% or so..
Eeliswhen i compile main = usleep 1000000 (where usleep is imported from System.Posix.Unistd) using ghc without -threaded, the program nicely sleeps for one second, as expected. however, when i compile the same code with -threaded, the program seemingly sleeps forever. any clue as to what might be going on? (i'm using linux x86_64, ghc 6.6)
pejoosfameron, there can't be that many users of it, unfortunately.
donsfor this one little benchmark
hpastenominolo pasted "Monadic Stream Functions, Take 2, still not working ..." at http://hpaste.org/559
Igloodons: Doesn't that equal a bug according to the GHC docs?
donsIgloo: I believe so. lennart's quite a good compiler writer, I think.
I'll try to get a full hbc run tomorrow
very exciting
now all we need is the hbc source!
ndmdons: what GCC are you using? we can't get nhc compiling with gcc 4....
dons;)
ndmdons: and i've got a patch for getArgs, just putting it in
donsits an nhc from 2005-03, so probably 3.35 at the time?
great!
ndm++
nominolocan someone give me a hint why i get List-monad like bahaviour, in my pasted code?
oh .. hm. i guess my mapML is wrong
ndmtypical, nhc won't even compile generated code with gcc 4, so we're having real difficulties here
actionmux hears GCC4 compilation problems and gets interested
muxhears GCC4 compilation problems and gets interested
got bitten by strict aliasing maybe?
pejodons, http://www.cs.chalmers.se/pub/haskell/chalmers/0.9999.3/hbc-0.9999.3.src.tar.gz?
lambdabothttp://tinyurl.com/2f4j4s
pejodons, hm, looks like .4 exists: http://www.cs.chalmers.se/pub/haskell/chalmers/hbc/hbc-0.9999.4.src.tar.gz
lambdabothttp://tinyurl.com/2dn5fe
donsi wonder if that's the one i have
or is that a newer version?
pejodons, I have no idea. There exists a .5 obviosuly, since there are binaries for it - but no src.
ndmmux: nhc writes out an array in a source code file, and then prods it and does evil things to it - its basically a broken nhc, not gcc's fault
dons: patch committed, give it 20 minutes to buildbot test itself then you can have getArg's in System
muxndm: ah okay.
actionmux wonders if we got the full FreeBSD source tree GCC 4 clean by now or not
muxwonders if we got the full FreeBSD source tree GCC 4 clean by now or not
donsndm, the full benchmark run has gone through,
that's the version using type=release
ndmdons: cheers, i don't suppose you could get instruction counts out of nhc on teh queens benchmark?
dons: we really can't get it to compile anywhere, so are having issues debugging it
donsno linux/x86 box?
i'm most of the way though an nhc darcs build atm.
from darcs get --partial http://darcs.haskell.org/nhc98
lambdabotTitle: Index of /nhc98

Page: 4 11 18 25 32 39 46 53 60 67 74 81 88 

IrcArchive

NY Lost Funds