Slower Compilation on GHC 7.8.4 (vs. 7.6.3)

Richard Eisenberg eir at cis.upenn.edu
Mon Apr 6 10:39:46 UTC 2015


>  
> ·         Everyone.  It’s really hard to keep focus on keeping GHC’s compilation time and space down.  Everyone is usually focused on bugs and features.  It would be incredibly helpful if someone, or a small group, could build a profiled version of GHC 7.8, 7.10, and HEAD, and track down what is happening.  Usually there are multiple causes, but a factor of 7.5 ought not to be hard to nail down.
>  
> Please!

This is a great newcomer-to-ghc-hacking task. It's both relatively easy to do and gives a newcomer a chance to tour the codebase. The person doing the profiling doesn't also have to fix the bug. Just knowing what module or component of GHC is causing the trouble helps immensely.

As a further plug: when a change I made introduced a performance regression, I went into the debugging process very warily, expecting profiling GHC to be painful. It was just the opposite -- remarkably straightforward and able to produce results without terribly much effort. On the other hand, it does generally require having multiple (profiled) builds of GHC to get meaningful comparisons.

Thanks to whoever takes this on!

Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-devs/attachments/20150406/11425aac/attachment.html>


More information about the ghc-devs mailing list