October 2015 - ghc-tickets

[GHC] #8457: -ffull-laziness does more harm than good
by GHC 08 Dec '17

08 Dec '17

#8457: -ffull-laziness does more harm than good ------------------------------------+------------------------------------- Reporter: errge | Owner: Type: bug | Status: new Priority: high | Milestone: 7.8.1 Component: Compiler | Version: 7.7 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- In this bug report I'd like to argue that `-ffull-laziness` shouldn't be turned on automatically with either `-O` nor `-O2`, because it's dangerous and can cause serious memory leaks which are hard to debug or prevent. I'll also try to show that its optimization benefits are negligible. Actually, my benchmarks show that it's beneficial to turn it off even in the cases where we don't hit a space leak. We've met this issue last week, but it had been reported several times before: e.g. #917 and #5262. A typical example is the following: {{{ #!haskell main :: IO () main = task () >> task () task :: () -> IO () task () = printvalues [1..1000000 :: Int] printvalues :: [Int] -> IO () printvalues (x:xs) = print x >> printvalues xs printvalues [] = return () }}} We succeed with `-O0`, but fail with `-O`: {{{ errge@curry:~/tmp $ ~/tmp/ghc/inplace/bin/ghc-stage2 -v0 -O0 -fforce- recomp lazy && ./lazy +RTS -t >/dev/null <<ghc: 1620098744 bytes, 3117 GCs, 32265/42580 avg/max bytes residency (3 samples), 2M in use, 0.00 INIT (0.00 elapsed), 1.28 MUT (1.28 elapsed), 0.02 GC (0.02 elapsed) :ghc>> errge@curry:~/tmp $ ~/tmp/ghc/inplace/bin/ghc-stage2 -v0 -O -fforce-recomp lazy && ./lazy +RTS -t >/dev/null <<ghc: 1444098612 bytes, 2761 GCs, 3812497/13044272 avg/max bytes residency (7 samples), 28M in use, 0.00 INIT (0.00 elapsed), 1.02 MUT (1.03 elapsed), 0.12 GC (0.12 elapsed) :ghc>> }}} 28M? What the leak!? Well, it's `-ffull-laziness`: {{{ errge@curry:~/tmp $ ~/tmp/ghc/inplace/bin/ghc-stage2 -v0 -O -fno-full- laziness -fforce-recomp lazy && ./lazy +RTS -t >/dev/null <<ghc: 1484098612 bytes, 2835 GCs, 34812/42580 avg/max bytes residency (2 samples), 1M in use, 0.00 INIT (0.00 elapsed), 1.04 MUT (1.04 elapsed), 0.02 GC (0.02 elapsed) :ghc>> }}} We get constant space and the fastest run-time too, since we spare some cycles on GC. Note, that in this instance we are trying to explicity disable sharing by using `()` as a fake argument for the function. Also note, that this function may easily be a utility function in a larger code base or in a library, therefore it's impractical to say that you shouldn't use it twice "too close together". Quoting from the GHC user guide: {{{ -O2: Means: “Apply every non-dangerous optimisation, even if it means significantly longer compile times.” The avoided “dangerous” optimisations are those that can make runtime or space worse if you're unlucky. They are normally turned on or off individually. At the moment, -O2 is unlikely to produce better code than -O. }}} This seems to be false at the moment. We decided to make a broader investigation into this issue and wanted to know if we can disable this optimization without too much pain. Came up with this benchmark plan: - let's benchmark GHC, - compile all stages with -O, but hack the stage1 compiler to emit `-t` statistics for every file compiled, - gather these statistics while compiling the libraries and the stage2 compiler. On the second run we compile the stage1 compiler with `-O -fno-full-laziness`, but leave everything else unchanged in the environment. When we have both results of the compilation of ~1600 files, we match them up and compute the (logarithmic) ratio of CPU and memory difference between compilations, the final results for our benchmark. The results and the raw data can be found at https://github.com/errge/notlazy. The overall compilation time dropped from 26:20 to 25:12, which is a 4% improvement. Investigating the full matching shows that this overall result is from small improvements all around the place. The results plotted: - https://github.com/errge/notlazy/blob/master/cpu.png - https://github.com/errge/notlazy/blob/master/mem.png The graphs show the logarithmic (100*log_10(new/orig)) ratio of change in cpu and memory consumption. Therefore negative results mean that the new compilation method is faster. As can be seen on the CPU graph, in most of the cases the difference is negligible (actually smaller than what can be measured on small files, this is why we have the spike at 0). In overall we see a small improvement in CPU, and there are some outliers in both directions, but there are more drastic improvement cases than drastic regressions. On the memory graph the situation is much more close to zero. There is one big positive memory outlier: `DsListComp.lhs`. It uses 69M originally and now uses 103M. But compiles in 2 seconds both ways and there are files in the source tree which requires 400M to compile, so this is not an issue. After all this, I'd like to hear other opinions about just disabling this optimization in `-O` and `-O2` and leaving it as an option that can be turned on when needed, my reasons once more: - it's unsafe, - it's hard to debug when you hit its issues, - the optimization doesn't seem to be very productive, - it's always easy to force sharing, but it's not easy to force copying. Apparently a Haskell programmer should be lazy, but never fully lazy. Research done by Gergely Risko <errge> and Mihaly Barasz <klao>, confirmed on two different machines with no other running processes. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8457> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 14

[GHC] #10535: Float out causes major space leak
by GHC 08 Dec '17

08 Dec '17

#10535: Float out causes major space leak -------------------------------------+------------------------------------- Reporter: AlexET | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- This issue originally was on haskell-cafe https://mail.haskell.org/pipermail/haskell-cafe/2015-June/120113.html The following code has a space leak {{{ #!hs numbers = [1..200] replicateM' :: Monad m => Int -> m a -> m [a] replicateM' 0 xs = return [] replicateM' n xs = do a <- xs b <- replicateM' (n-1) xs return (a:b) test :: [[Int]] test = replicateM' 4 numbers main = print test }}} The recursive call in replicateM' gets floated out. This then causes a very large (200^3 elements) list to be kept in memory instead of being lazily produced and consumed each time. A similar thing happens to the replicateM in Control.Monad but it is harder to spot as a lot of specialisation and fusion goes on before the float-out happens This is similar to #7367 but in this case we have a huge space-leak instead of just adding allocations. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10535> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #9118: Can't eta-reduce representational coercions
by GHC 19 Nov '17

19 Nov '17

#9118: Can't eta-reduce representational coercions ------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- When I say {{{ import Data.Type.Coercion import Data.Coerce import Data.Proxy eta2 :: Coercible (f a) (g a) => Proxy a -> Coercion f g eta2 _ = Coercion }}} I get {{{ Could not coerce from ‘f’ to ‘g’ because ‘f’ and ‘g’ are different types. arising from a use of ‘Coercion’ from the context (Coercible (f a) (g a)) bound by the type signature for eta2 :: Coercible (f a) (g a) => Proxy a -> Coercion f g at /Users/rae/temp/Roles.hs:5:9-56 In the expression: Coercion In an equation for ‘eta2’: eta2 _ = Coercion }}} Unlike the case for #9116, this one is ''not'' expressible in Core. (At least, I don't see a way to do it.) So, to do this, we would have to update Core and then update the constraint solver. I ''do'' think this would be type safe. But, I also think it's reasonable to wait for someone with a real use case to shout before doing this. The only use case I have is to be able to do this: {{{ -- says that a's parameter is representational class Rep (a :: k1 -> k2) where co :: Coercible x y => Coercible (a x) (a x) instance Rep f => Rep (Compose f) where ... }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9118> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 9

[GHC] #9052: Support a "stable heap" which doesn't get garbage collected
by GHC 13 Nov '17

13 Nov '17

#9052: Support a "stable heap" which doesn't get garbage collected ----------------------------------------------+---------------------------- Reporter: ezyang | Owner: Type: feature request | simonmar Priority: normal | Status: new Component: Runtime System | Milestone: Keywords: | Version: 7.9 Architecture: Unknown/Multiple | Operating System: Difficulty: Moderate (less than a day) | Unknown/Multiple Blocked By: | Type of failure: Related Tickets: | None/Unknown | Test Case: | Blocking: ----------------------------------------------+---------------------------- Simon Marlow was explaining the concept on IRC. The problem we're trying to solve is you might have some very long-lived pointer data-structures, and you don't want to waste time traversing this structure on a major GC when you know that it is still going to be alive. Here's the basic idea: 1. We have a new generation on the heap called the 'stable heap' (along- side the nursery, gen0, gen1, etc). By default, no objects are automatically promoted from the old gen to the stable heap. 2. A user can request an object be moved to the stable heap, creating a stable pointer which points to the object. 3. During garbage collection, all blocks in the stable heap are always considered reachable. We don't include the stable pointers in the root set (so we don't trace the stable heap), however, write barriers are processed as usual, and we promote any objects pointed to by the stable heap to the stable heap. 4. Users manually request the stable heap to be garbage-collected, usually after deallocating stable pointers. In short, the stable heap is just another generation, but one which (1) is only collected by user request, and (2) never receives objects from automatic promotion. This project fits in very nicely with the existing GC infrastructure, so if you understand how everything is laid out this should be reasonably easy to implement. Could be a good ticket for someone who wants to get their hands dirty with the GC. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9052> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #9420: Impredicative type instantiation without -XImpredicativeTypes
by GHC 11 Nov '17

11 Nov '17

#9420: Impredicative type instantiation without -XImpredicativeTypes -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.9 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- Consider this module: {{{ {-# LANGUAGE RankNTypes #-} module Bug where rank2 :: ((forall a. a -> a) -> b) -> () rank2 _ = () foo :: () -> () foo x = x quux :: (forall a. a -> a) -> Int quux _ = 5 bar = foo . rank2 $ quux }}} The `(.)` in the definition of `bar` requires an impredicative instantiation -- that is, one of its type variables is instantiated to a forall-type. Yet, the module compiles without `-XImpredicativeTypes`. To confirm this behavior, the following is an excerpt from `-ddump-simpl`: {{{ Bug.bar = break<7>() (break<6>() GHC.Base.. @ () @ () @ ((forall a_a1Tj. a_a1Tj -> a_a1Tj) -> GHC.Types.Int) Bug.foo (Bug.rank2 @ GHC.Types.Int)) Bug.quux }}} Note the third type argument to `(.)`. Fixing this would be a breaking change that could affect users. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9420> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #9642: LANGUAGE pragma synonyms
by GHC 10 Nov '17

10 Nov '17

#9642: LANGUAGE pragma synonyms -------------------------------------+------------------------------------- Reporter: dreixel | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.9 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- Here are the first 20 lines of a typical Haskell file of mine: {{{ {-# LANGUAGE EmptyDataDecls #-} {-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE LiberalTypeSynonyms #-} {-# LANGUAGE ScopedTypeVariables #-} {-# LANGUAGE RankNTypes #-} {-# LANGUAGE StandaloneDeriving #-} {-# LANGUAGE AutoDeriveTypeable #-} {-# LANGUAGE DeriveDataTypeable #-} {-# LANGUAGE DeriveGeneric #-} {-# LANGUAGE DefaultSignatures #-} {-# LANGUAGE OverlappingInstances #-} {-# LANGUAGE UndecidableInstances #-} {-# LANGUAGE MultiParamTypeClasses #-} {-# LANGUAGE TypeOperators #-} {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE GADTs #-} {-# LANGUAGE DataKinds #-} {-# LANGUAGE PolyKinds #-} {-# LANGUAGE ConstraintKinds #-} }}} Over a couple of modules in a small package, I find myself copy-pasting these 20 lines, as either I will use these extensions, or they won't really cause any harm to the code I'm writing. As a way to simplify this, and reduce code duplication, I propose introducing LANGUAGE pragma synonyms: {{{ {-# LANGUAGE PedrosHaskell = EmptyDataDecls, FlexibleContexts, FlexibleInstances , LiberalTypeSynonyms, ScopedTypeVariables , RankNTypes, StandaloneDeriving, AutoDeriveTypeable , DeriveDataTypeable, DeriveGeneric, DefaultSignatures , OverlappingInstances, UndecidableInstances , MultiParamTypeClasses, TypeOperators, TypeFamilies , GADTs, DataKinds, PolyKinds, ConstraintKinds #-} }}} After writing this, in that same module or any module that imports it, I could just write: {{{ {-# LANGUAGE PedrosHaskell #-} }}} And all the pragmas above would be enabled. Besides making my life easier, I suspect this will also help increase modularity of certain packages. For example, packages could define a package-specific language pragma that should be enabled in order to use that package. Then, if the package maintainer upgrades one datatype to become a GADT, the pragma would be updated to include `GADTs`, and user code (that could now be requiring that pragma) would automatically enable `GADTs` too. Furthermore, it can also make the language standardisation process simpler. After all, {{{ {-# LANGUAGE Haskell2010 = PatternGuards, NoNPlusKPatterns, RelaxedPolyRec , EmptyDataDecls, ForeignFunctionInterface #-} }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9642> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #8238: Implement unloading of shared libraries
by GHC 09 Nov '17

09 Nov '17

#8238: Implement unloading of shared libraries ------------------------------------+------------------------------------- Reporter: simonmar | Owner: Type: task | Status: new Priority: normal | Milestone: 7.10.1 Component: Runtime System | Version: 7.7 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: 8039 | ------------------------------------+------------------------------------- In #8039 we added support for unloading static objects from the runtime linker, with the GC detecting when there are no references left. We could add this functionality for shared libraries too, using `dl_iterate_phdr`. @heisenbug's comment from #8039 with the relevant pointers: [http://eli.thegreenplace.net/2011/08/25/load-time-relocation-of-shared- libraries Eli Bendersky's article] suggests to use the [http://linux.die.net/man/3/dl_iterate_phdr dl_iterate_phdr] function for finding information about loaded libraries. Seems to be linux only. There is a [http://stackoverflow.com/questions/10009043/dl-iterate-phdr- equivalent-on-mac workaround on OSX], though, on !StackOverflow. And here is how Böhm-Demers-Weiser's GC implement a [https://github.com/ivmai/bdwgc/blob/master/dyn_load.c#L451 callback function for dl_iterate_phdr]. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8238> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #10514: Generic for existential types
by GHC 09 Nov '17

09 Nov '17

#10514: Generic for existential types -------------------------------------+------------------------------------- Reporter: | Owner: ekmett andreas.abel | Status: new Type: feature | Milestone: request | Version: 7.10.1 Priority: normal | Operating System: Unknown/Multiple Component: Core | Type of failure: None/Unknown Libraries | Blocked By: Keywords: | Related Tickets: Architecture: | Unknown/Multiple | Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- I have some use for Generic for an existential type which is constraint to be Generic. {{{!#hs {-# LANGUAGE DeriveGeneric #-} {-# LANGUAGE ExistentialQuantification #-} {-# LANGUAGE StandaloneDeriving #-} import GHC.Generics data U = forall a. (Generic a) => U a -- deriving (Generic) -- Can't make a derived instance of ‘Generic U’: -- Constructor ‘U’ has existentials or constraints in its type -- Possible fix: use a standalone deriving declaration instead -- deriving instance Generic U -- Can't make a derived instance of ‘Generic U’: -- U must be a vanilla data constructor -- In the stand-alone deriving instance for ‘Generic U’ data D1Ser data C1_0Ser instance Generic U where type Rep U = D D1Ser (C1 C1_0Ser (S1 NoSelector (Rep a))) -- Not in scope: type variable ‘a’ -- How to bring the existential type `a' into scope? }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10514> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #10577: Use empty cases where appropriate when deriving instances for empty types
by GHC 02 Nov '17

02 Nov '17

#10577: Use empty cases where appropriate when deriving instances for empty types -------------------------------------+------------------------------------- Reporter: rwbarton | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.11 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- In some sense the correct instance for {{{ data X deriving instance Eq X }}} is not (what GHC currently produces) {{{ instance Eq X where (==) = error "Void ==" }}} but rather {{{ instance Eq X where a == b = case a of {} -- using the EmptyCase extension }}} See comments starting at ticket:7401#comment:28 for justification. The list of classes that GHC can currently derive is * Eq, Ord, Enum, Bounded, Show, Read (Haskell 2010 Report) * Functor, Foldable, Traversable, Typeable, Generic, Data, "any class" (`-XDerive*`) Deriving Enum and Bounded is not currently allowed for empty data types. The `showList` method of Show and the whole Read instance are easy and already implemented correctly. All the remaining methods of Haskell 2010 derivable type classes take at least one argument of the instance head type `a`. I propose that all these methods be defined as an empty case on the first such argument. Similarly in Functor, Foldable and Traversable, each method has a single argument of the form `t _` where `t :: * -> *` is the instance head type, and the method should be defined as an empty case on that argument. In all these cases so far, the derived methods for a non-empty type would, at the outermost level, be a (non-empty) case on that first argument, or the equivalent (series of pattern matches, one for each constructor), so the use of an empty case is justified for an empty type. Typeable does not care about the values or even the kind of the type at all, so it's not relevant here. Generic and especially Data are above my pay grade, but generally I expect that methods which are normally defined by case analysis on an argument of the instance head type should be defined by an empty case, and methods that (sometimes) produce a value of the instance head type should do whatever they normally do when unable to produce such a value (like `readsPrec` returning an empty list, or `read` calling `error` (if `read` were actually a method of `Read`)). `DeriveAnyClass` doesn't generate any new code, so like Typeable it's not relevant. None of this behavior is specified by the Haskell 2010 Report, which disallows deriving any class for a type with no constructors; but see #7401. So, we are entitled to do what we think is best here. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10577> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #10871: Implement "buildable" interface files which can be directly compiled without source
by GHC 02 Nov '17

02 Nov '17

#10871: Implement "buildable" interface files which can be directly compiled without source -------------------------------------+------------------------------------- Reporter: ezyang | Owner: ezyang Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 7.11 Keywords: backpack | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- A fat interface file is like a normal hi file, but with extra information so that GHC can *recompile* the associated module without having to parse, rename and typecheck the source code again. The proposed command-line interface is something like (subject to bikeshedding): {{{ ghc -c A.hs -fno-code -fwrite-buildable-interface # produces a .hi-build file, then ghc -c A.hi-build # completes building, as if ghc -c A.hs had been called }}} The primary motivation for this is to support partially compiling indefinite packages, which cannot be compiled to object code due to the fact that some of the dependencies haven't been provided yet. However, there are some other cases where this might be useful: 1. Supercompilation requires having the source of all bindings, buildable interfaces can make this information available, even beyond what storing inlinings might provide 2. A buildable interface file can be used to build variants of a source, e.g. profiling and optimization, even "on the fly" by GHC if necessary. And maybe more we haven't thought of yet. The primary function you need to implement is a function from `ModIface`/`ModDetails` to `TcGblEnv`, which can then be fed to the desugarer. I'm looking at what things we need to add to `ModIface` to support this. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10871> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 28