#haskell - Mon 19 Feb 2007 between 09:37 and 09:42

NY Lost Funds



actiondcoutts -> office
dcoutts-> office
ndmdons: you may need a quite recently Yhc build
donsso i grabbed todays darcs yhc, and just built it with the 'release' way. its about 20%of the way through now
i'll inspect the logs in the morning. but you should be also able to just darcs get the benchmark suite and run it on yhc
ndmdons: i think the makefile would not work on my machine
donsoh, windows. hmm
ndmdons: cool, i'll check logs later
swiert@tell dcoutts If you're coming to Fun in the Afternoon, you can borrow my copy of Graham's book.
lambdabotConsider it noted.
donsndm, anything other than building it with type=release required?
or that should be it?
CaleWell, that's a shame. I'm going to install 7.0 and see if it still goes okay.
ndmdons: not as far as i know, unless we are beating nhc though something is wrong...
Caledons: any idea about what to do when cabal runs ar with too many parameters?
donsCale, hmm. xargs ar ?
ndm, ok. and you've confirmed this on a linux box?
CaleI tried splitting the commandline into two.
With ar q libname.a ...; ar q libname.a ...
ndmdons: yes, i've not seen an example where we are slower than nhc
dons: unless you did loads of FFI, but i think thats unlikely
CaleBut then I got magic number errors, I'm not sure if they're related.
donsndm, could you maybe get the benchmarks, and compile some of them by hand?
any that look odd to you
ndmdons: i can't compile them against nhc, which is where the oddness looks most obvious
dons: i'll get Tom to take a look
donsoh, you don't have an nhc build?
ndmnhc /= windows :)
donsyeah, nhc is definitely ahead, but doing type=release did help somewhat.
Test: recursive (naive recursive functions)
http://www.cse.unsw.edu.au/~dons/code/nobench/imaginary/recursive
ghc 2.334 seconds (1.0 x)
ghc-old 2.353 seconds (1.0 x)
ghci 99.580 seconds (42.7 x)
lambdabotTitle: Index of /~dons/code/nobench/imaginary/recursive
donsnhc98 111.865 seconds (47.9 x)
yhc 143.878 seconds (61.6 x)
for example, just ran
whereas it was hugs 208.266 seconds (72.7 x)
yhc 242.875 seconds (84.8 x)
before

Page: 2 9 16 23 30 37 44 51 58 65 72 79 86 

IrcArchive

NY Lost Funds