Recent performance regressions

[Joachim Breitner]

Dear Developers,

yesterday, there were three commits showing up at
https://perf.haskell.org/ghc/:



https://perf.haskell.org/ghc/#revision/60297486fddd5c9695e2263c2ae46fa9
0f0feb9e
60297486fddd5c9695e2263c2ae46fa90f0feb9e
Author: Ben Gamari <bgamari.foss at gmail.com>
Drop custom mapM impl for []

nofib/allocs/cryptarithm2   + 64.55%
nofib/allocs/k-nucleotide    - 5.03%
nofib/time/fannkuch-redux    - 3.62%

Did someone forget to run nofib here, or is this serious drop in
performance in one case expected and something we condone? In that
case, it should have been clearly noted in the commit message!



https://perf.haskell.org/ghc/#revision/b12dba7829742de98a483645142c7962
b9dd9f3f
b12dba7829742de98a483645142c7962b9dd9f3f
Author: RyanGlScott <ryan.gl.scott at gmail.com>

Make Exception datatypes into newtypes

nofib/time/cryptarithm1  + 7.22%

If you look at the graph, this is quite reproducible:
https://perf.haskell.org/ghc/#graph/nofib/time/cryptarithm1;hl=b12dba78
29742de98a483645142c7962b9dd9f3f
but I have no idea how that change could have that effect. Maybe some
weird cache effect due to binary sizes or different code layout? It
would be great to understand that!



https://perf.haskell.org/ghc/#revision/22bbc1cf209d44b8bb8897ae7a35f9eb
af411b10
commit 22bbc1cf209d44b8bb8897ae7a35f9ebaf411b10
Author: Takano Akio <aljee at hyper.cx>

Make sure that `all`, `any`, `and`, and `or` fuse (#9848)

nofib/allocs/circsim 	 - 4.02%
nofib/allocs/multiplier  - 8.70%

At least I do not only have bad news...



Greetings,
Joachim




-- 
Joachim “nomeata” Breitner
  mail at joachim-breitner.de • http://www.joachim-breitner.de/
  Jabber: nomeata at joachim-breitner.de  • GPG-Key: 0xF0FBF51F
  Debian Developer: nomeata at debian.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part
URL: <http://mail.haskell.org/pipermail/ghc-devs/attachments/20150806/fc12510c/attachment.sig>