
Hi, due to #9938 ghcspeed did not measure every commit but somewhere in these commits, cryptarithm2 regressed by 8%: $ git log 625dd..8d62f9 --oneline 8d62f92 Update nofib submodule, unbreak cryptarithm2 9521a58 Refine test case for #9938 65e3e0b Test case for #9938 4e1e776 Skip T2276_ghci on Darwin, since stdcall is not supported. b32c227 Fix system linker on Mac OS X 40561cd Fix `heapSizeSuggesionAuto` typo (#9934) 58ac9c8 LlvmCodeGen cross-compiling fixes (#9895) 0cc0cc8 Support pattern synonyms in GHCi (fixes #9900) 6c86635 Update validate-settings.mk 1fefb59 Update parallel submodule to 3.2.0.6 release d6e7f5d Add export lists to some modules. 6b9e958 Update hoopl and hpc submodules c55fefc Avoid redundant-import warning (w/o CPP) bd01af9 Update hsc2hs submodule for de-tabbing 0899caa Use directory-style database for bootstrapping database c0ab767 We do emit a warning for stdcall now. 1dcef98 Run T9762 only if dynamic libraries are available 9ae78b0 Copy GHC's config.guess/sub over libffi's versions add6a30 2nd attempt to fix T9032 test-case 3e3aa92 Fix linker interaction between Template Haskell and HPC (#9762) cc510b4 Make ghc -e fail on invalid declarations 878910e Make ghc -e not exit on valid import commands (#9905) 7a2c9dd Fixup edd233acc19d269385 (T9032 test) edd233a Test earlier for self-import (Trac #9032) c3394e0 Attempt to improve cleaning 679a661 A bit of refactoring to TcErrors c407b5a Comments only 3e96d89 Add a couple of missing cases to isTcReflCo and isTcReflCo_maybe a6f0f5a Eliminate so-called "silent superclass parameters" My guess: Related to this: * It had unexpected peformance costs, shown up by Trac #3064 and its test case. In monad-transformer code, when constructing a Monad dictionary you had to pass an Applicative dictionary; and to construct that you neede a Functor dictionary. Yet these extra dictionaries were often never used. (All this got much worse when we added Applicative as a superclass of Monad.) Test T3064 compiled *far* faster after silent superclasses were eliminated. Happy new year! Joachim -- Joachim “nomeata” Breitner mail@joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata@joachim-breitner.de • GPG-Key: 0xF0FBF51F Debian Developer: nomeata@debian.org

Hi, Am Donnerstag, den 01.01.2015, 13:30 +0100 schrieb Joachim Breitner:
due to #9938 ghcspeed did not measure every commit but somewhere in these commits, cryptarithm2 regressed by 8%:
it seems that d8d003185a4bca1a1ebbadb5111118ef37bbc83a (When solving one Given from another, use the depth to control which way round) has solved this. Greetings, Joachim -- Joachim “nomeata” Breitner mail@joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata@joachim-breitner.de • GPG-Key: 0xF0FBF51F Debian Developer: nomeata@debian.org

That's interesting. You mean the 8% regression has gone away? S | -----Original Message----- | From: ghc-devs [mailto:ghc-devs-bounces@haskell.org] On Behalf Of Joachim | Breitner | Sent: 03 January 2015 20:53 | To: ghc-devs@haskell.org | Subject: Re: cryptarithm2 +8.16% | | Hi, | | Am Donnerstag, den 01.01.2015, 13:30 +0100 schrieb Joachim Breitner: | > due to #9938 ghcspeed did not measure every commit but somewhere in | > these commits, cryptarithm2 regressed by 8%: | | it seems that d8d003185a4bca1a1ebbadb5111118ef37bbc83a (When solving one | Given from another, use the depth to control which way round) has solved | this. | | | Greetings, | Joachim | | -- | Joachim “nomeata” Breitner | mail@joachim-breitner.de • http://www.joachim-breitner.de/ | Jabber: nomeata@joachim-breitner.de • GPG-Key: 0xF0FBF51F | Debian Developer: nomeata@debian.org

Hi, Am Samstag, den 03.01.2015, 22:03 +0000 schrieb Simon Peyton Jones:
That's interesting. You mean the 8% regression has gone away?
exactly. It is quite obvious when you look at the graph: http://ghcspeed-nomeata.rhcloud.com/timeline/?ben=nofib/allocs/cryptarithm2&env=1&equid=on Greetings, Joachim -- Joachim “nomeata” Breitner mail@joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata@joachim-breitner.de • GPG-Key: 0xF0FBF51F Debian Developer: nomeata@debian.org
participants (2)
-
Joachim Breitner
-
Simon Peyton Jones