July 2014 - ghc-devs - Haskell.org

Core-to-core transformations: current state-of-the-art?
by Jan Stolarek 15 Jul '14

15 Jul '14

Hi devs, I'm curious about the current state of core-to-core transformations inside GHC. I know about following papers that cover this subject: 1. "Compilation by Transformation in Non-Strict Functional Languages", a PhD dissertation by Santos (I assume this summarizes earlier papers by Santos and SPJ). 2. "A transformation-based optimiser for Haskell", paper by SPJ and Santos. 3. "Let-floating: moving bindings to give faster programs", paper by SPJ, Partain and Santos. 4. "Modular, Higher-Order Cardinality Analysis in Theory and Practice", paper by Sergey, Vytiniotis and SPJ. Is there any other work that I should be aware of? Are there transformations implemented in GHC that were not described in any paper? Janek PS. I'm aware of the wiki pages that list the GHC papers.

2 1

Phab vs. Trac
by Richard Eisenberg 15 Jul '14

15 Jul '14

Hi all, Sometimes, the progress on a particular issue is tracked both on Trac and on Phab. What posts should go where? I know Austin is trying to get Trac to be notified when a relevant post happens on Phab -- great. But, if I have a comment, where should I put it? Here is my proposed answer: The Phab reviews are a good place for code-specific commentary/feedback but Trac is better for design issues. A rule of thumb might be to pretend that Phab comments are all forgotten in a month or two, whereas Trac comments are expected to be around in 5 years. Another feature request around Phab/Trac: Is it possible to have a field *at the top of a Trac ticket* that links to the Differential page? Thanks! Richard

3 6

RE: how to get ghci to load compiled modules in 7.8?
by Simon Peyton Jones 14 Jul '14

14 Jul '14

| One thing, how about an addition to the release notes in the | -dynamic-too section saying "By the way, this doesn't actually work, | see https://ghc.haskell.org/trac/ghc/ticket/8736." Not a bad idea. Austin can we conveniently do that retrospectively?

1 0

T9203
by Simon Peyton Jones 14 Jul '14

14 Jul '14

Joachim In commit 194107ea9333c1d9d61abf307db2da6a699847af you reduced the allocation of T9203 by half. Butbut I get allocation of 95747304 on my 64 bit machine, so the test fails. Any idea why? Simon diff --git a/testsuite/tests/perf/should_run/all.T b/testsuite/tests/perf/should_run/all.T index 4fa77a5..e9e7ef3 100644 --- a/testsuite/tests/perf/should_run/all.T +++ b/testsuite/tests/perf/should_run/all.T @@ -371,7 +371,9 @@ test('InlineCloneArrayAlloc', test('T9203', [stats_num_field('bytes allocated', [ (wordsize(32), 50000000, 5) - , (wordsize(64), 95747304, 5) ]), + , (wordsize(64), 42946176, 5) ]), + # previously: 95747304 + # 2014-07-14: 42946176 (amd64/Linux)

2 1

Status updates
by Austin Seipp 14 Jul '14

14 Jul '14

Hello all, I've been lax on the status updates, but better late than never. :) Here are some things that I've been up to. - 7.8.3 is released! Hooray! - Phabricator can now build code reviews if you submit them. I sent an email about this earlier in the week[1] in case you missed it. Unfortunately it has a rather stupid bug that it doesn't catch stderr properly, but I'll fix that soon. :) - I also spent some time on trying to get Phabricator to post to Trac (when a code review for a ticket is posted), but to no avail yet. Hopefully I'll have this done in the next few days. - I spent a few hours yesterday and on Saturday on the AMP work still, because I missed a few things actually (broken deriving mechanism due to forgetting some changes). I also need to send one more patch to Cabal upstream I believe, but otherwise things look OK. I'll post a new code review today when I've finished. - This morning I've combed some tickets and remilestoning stuff, since there was a *lot* left around for 7.6.2 still! This should all be fixed. I'll keep doing this today and this morning probably. - I want to reorganize the wiki pages around the Git repository, and I have been meaning to do this, but not gotten around to it yet. :( Some other things on my mind: - When I released 7.8.3, I noticed we have a whole bunch of new upstream distribution packages now, but the download page for those is pretty out of date I think.[2] That should probably be updated. - There are patches in the Trac queue I need to get done with. If I've ignored your patch I'm sorry! I'll try to get this done quickly. - Johan reminded me about D4 this morning[3] which I still need to merge upstream after addressing the details in the review. [1]: https://www.haskell.org/pipermail/ghc-devs/2014-July/005509.html [2]: https://www.haskell.org/ghc/download_ghc_7_8_3#distros [3]: https://phabricator.haskell.org/D4 -- Regards, Austin Seipp, Haskell Consultant Well-Typed LLP, http://www.well-typed.com/

2 1

RE: [commit: ghc] master: Update various performance benchmarks (194107e)
by Simon Peyton Jones 14 Jul '14

14 Jul '14

| I started monitoring perfomance on a per-commit base. These seem to | be | off for a while now. Adjusting them, and from now I hope I can keep | closer tabs on them. Thank you. That's most helpful. One of the reasons that perf tests get worse is that each time the limit is reached, the author thinks "oh my bit probably only made it a tiny bit worse", so we miss big jumps. It's the big jumps we thereby miss, and nailing them to a particular commit would be great. I'ts important to monitor the actual bump, rather than the first time the limit is broken. (They aren't the same, of course.) Simon

2 1

Put GHC on travis for real
by Joachim Breitner 14 Jul '14

14 Jul '14

Hi, with all packages as submodules, ghc-complete (which is basically a git repository tracking the „fingerprint“ of the main repository) is obsolete. So we could move the travis-checking of the main line to run on the ghc repository directly. This would require * adding a .travis.yaml based on the contents of https://github.com/nomeata/ghc-complete/blob/master/.travis.yml and https://github.com/nomeata/ghc-complete/blob/master/validate.sh to the repository. * enabling Travis for this repository. I can do the former, but the latter needs to be done by a member of the “ghc” on GitHub. I can do the latter (and keep managing the travis instance) if someone adds me to the organization... Greetings, Joachim -- Joachim “nomeata” Breitner mail(a)joachim-breitner.de • http://www.joachim-breitner.de/ Jabber: nomeata(a)joachim-breitner.de • GPG-Key: 0xF0FBF51F Debian Developer: nomeata(a)debian.org

4 6

Huge space leak of GHC API 7.8.x?
by Kazu Yamamoto 14 Jul '14

14 Jul '14

Hi, Some guys reported to me that ghc-mod uses about 1G bytes on Mac and I can reproduce this. I tried to understand why ghc-mod uses such huge memory. I found that GHC API 7.8.x uses much more memory than GHC API 7.6.x. Attached two files demonstrate this: - A.hs -- Simple program using GHC API (copied from Wiki) - B.hs -- A target file, just hello world You can compile A.hs as follows: % ghc A.hs -package ghc -package ghc-paths The following is the result: Mac (64bit) Linux (64bit) GHC 7.6.3: 20MB 4MB GHC 7.8.3: 106MB 13MB So, I think GHC API 7.8.x has huge space leak. (And I'm wondering why Mac uses much more memory than Linux). I would like to hear opinions from you guys. --Kazu <A.hs> import Control.Concurrent import CoreMonad (liftIO) import DynFlags import GHC import GHC.Paths (libdir) main = defaultErrorHandler defaultFatalMessager defaultFlushOut $ do runGhc (Just libdir) $ do dflags <- getSessionDynFlags let dflags' = dflags {hscTarget = HscInterpreted ,ghcLink = LinkInMemory ,ghcMode = CompManager } setSessionDynFlags dflags' target <- guessTarget "B.hs" Nothing setTargets [target] load LoadAllTargets liftIO $ threadDelay 10000000 </A.hs> <B.hs> module B where main = print "Hello, World!" </B.hs>

3 3

Trac spam
by Simon Peyton Jones 14 Jul '14

14 Jul '14

Dear GHC devs Is it possible to stop spam on GHC's Trac? Simon -----Original Message----- From: ghc-tickets [mailto:ghc-tickets-bounces@haskell.org] On Behalf Of GHC Sent: 14 July 2014 04:24 Cc: ghc-tickets(a)haskell.org Subject: [GHC] #9310: This Cleanse Ultima is very beneficial to regulate a good blood circulation #9310: This Cleanse Ultima is very beneficial to regulate a good blood circulation ------------------------------------+----------------------------------- ------------------------------------+-- Reporter: joerjpool | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: Cleanse Ultima | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+----------------------------------- ------------------------------------+-- He something that would punish if you feel something in the question is how are you actually going to do I liked it physically working mostly stress myself out our food Afghani book more than half my life when hapless seafood gumbo I have Lucy golf when Wane have you know some sushi or thought not downtown today know what more you deprive yourself the more you want be changed when you get sold week we're not moving why outtakes house days fish they still hadn't fully fused week you some two dozen times but hey it happens that I wanted to take a little bit it is not a little bit better which will find few lost five straight days %uh week yeah I want to our ERP user roughly everything you worked so hard for all anyway so if not before keep good food you know people who you are happy light snack crackers and stuff like that my skate since I people I illegal you know I was clapping her [http://cleanseultima-uk.com/ Cleanse Ultima] hands they know they're not going to be. and what any death of Natalie wasting money first off secondly it’s not Billy is not going to help you get through your weight loss process some things that hike to do was keep so I may even take don't use to the fact that I sweet potato chips actually sweet potato fries around Kasha pizza its peak but I'll be healthy so what's that me he likes me happy nasty old who was involved you don't mind any something around the time I sleep he on a healthy track remind mean my healthy lifestyle ha-ha sweet shocking video must if you see how check on the food eat some other schools that joy hey stop talking hospital all major install look back up debt but soon up today work 350 pounds on top it's not worth it if you have some see accident I just have work 18th dub step make use of your own you created movie didn't happen ask. http://cleanseultima-uk.com/ -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9310> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler _______________________________________________ ghc-tickets mailing list ghc-tickets(a)haskell.org http://www.haskell.org/mailman/listinfo/ghc-tickets

3 2

HaRe and incremental type checking / type inference
by AlanKim Zimmerman 13 Jul '14

13 Jul '14

I have hit a problem in HaRe when lifting a declaration from e.g. a where clause of a function to the top level, where there is a type signature of any complexity. e.g lifting 'baz' from function 'foo' below -------------------------- foo a = baz where baz :: Int baz = xx 1 a xx :: (Num t) => t -> t -> t xx p1 p2 = p1 + p2 -------------------------------------- becomes --------------------------------------------- foo a = (baz xx a) where xx :: (Num t) => t -> t -> t xx p1 p2 = p1 + p2 -- baz:: (forall t. Num t => t -> t -> t) -> Int ->Int baz :: Num a => (a -> t1 -> t) -> t1 -> t baz xx a= xx 1 a ----------------------------------------------- For a very small subset this can be calculated easily, but for full generality it would be great to access the full power of the GHC type system. So before diving in too deeply, I thought I would test the waters as to the feasibility of doing something like this. I was hoping that perhaps the effort at an external constraint solver might be making the interfacing slightly simpler. Regards Alan

2 4