Comments on: Evolving Faster Haskell Programs (now with LLVM!) http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/ A Journal of Haskell Programming Mon, 03 Oct 2011 02:09:43 +0000 hourly 1 http://wordpress.com/ By: Quiet Nights and Quiet Stars | The Phaneron http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-944 Mon, 03 Oct 2011 02:09:43 +0000 http://donsbot.wordpress.com/?p=476#comment-944 [...] configuration management on my 4-core i7 using genetic algorithms as described by Don Stewart in Evolving Faster Haskell Programs (now with LLVM). I also hope to get access to some cooler hardware, possibly through some academic [...]

]]> By: ghc-gc-tune: Tuning Haskell GC settings for fun and profit « Control.Monad.Writer http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-514 Tue, 06 Jul 2010 00:43:54 +0000 http://donsbot.wordpress.com/?p=476#comment-514 [...] multi-variable spaces for optimizations with GAs in Haskell, to find strictness flags and LLVM flag settings, so I was keen to see what the GC space looked like. In this initial GC search, however, I [...]

]]> By: Robert Massaioli http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-500 Sat, 05 Jun 2010 04:13:53 +0000 http://donsbot.wordpress.com/?p=476#comment-500 I just want to give credit for using acovea to try and optimise the executables that GHC produces. I’m excited to see even more speedups coming in via LLVM and couple that with some ‘Data Parallel Haskell’ videos that I have seen around, well…lets just say that it has got me excited.

]]> By: Getting GHC HEAD and LLVM working together « Alp Mestanogullari's Blog http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-394 Thu, 11 Mar 2010 00:14:22 +0000 http://donsbot.wordpress.com/?p=476#comment-394 [...] and LLVM working together March 11, 2010 alpmestan Leave a comment Go to comments Since dons’ two posts, I have seen many people on the haskell-cafe mailing-list and the haskell-related IRC [...]

]]> By: saynte http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-390 Sun, 07 Mar 2010 19:31:12 +0000 http://donsbot.wordpress.com/?p=476#comment-390 David, that is indeed great news, you’ve really made a complete job of this thesis, kudos!

]]> By: dons00 http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-386 Sat, 06 Mar 2010 19:15:20 +0000 http://donsbot.wordpress.com/?p=476#comment-386 GHC uses its own runtime and own excellent garbage collector, so that’s not an issue here.

]]> By: Marc http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-385 Sat, 06 Mar 2010 19:11:24 +0000 http://donsbot.wordpress.com/?p=476#comment-385 I read on the LHC blog that LLVM is not a good backend choice compared to C–, because garbage collection with LLVM incurs quite a performance overhead. Did you run any tests exploring this?

]]> By: David Terei http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-383 Fri, 05 Mar 2010 05:44:39 +0000 http://donsbot.wordpress.com/?p=476#comment-383 Really interesting stuff, wish there had been enough time to include things like this in my paper, oh well. I’d love to see applied to nofib as you suggested since it was never clear to me if using the provided -O sets was the best approach for the general case.

saynte, I’m currently talking to the LLVM developers about getting the changes needed by GHC merged in, it should actually make in in by the 2.7 release which is great news since 2.7 is due out in April.

]]> By: Reid http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-382 Thu, 04 Mar 2010 18:33:01 +0000 http://donsbot.wordpress.com/?p=476#comment-382 mem2reg is an enabling optimization, which simply promotes stack slots that don’t have their addresses taken into virtual registers. The IL is basically impossible to analyze without that pass, if the frontend does the obvious thing of allocating stack slots for every temporary value. So it makes sense that acovea would choose to enable it. :-D

]]> By: thu http://donsbot.wordpress.com/2010/03/01/evolving-faster-haskell-programs-now-with-llvm/#comment-381 Thu, 04 Mar 2010 07:43:24 +0000 http://donsbot.wordpress.com/?p=476#comment-381 I was wondering what about using GHC’s C output with LLVM, how does it compare ?

]]>