June 2016 - ghc-tickets

[GHC] #10352: Properly link Haskell shared libs on ELF systems
by GHC 05 Jan '17

05 Jan '17

#10352: Properly link Haskell shared libs on ELF systems -------------------------------------+------------------------------------- Reporter: duncan | Owner: Type: task | Status: new Priority: normal | Milestone: Component: Package | Version: 7.11 system | Operating System: Linux Keywords: | Type of failure: None/Unknown Architecture: | Blocked By: Unknown/Multiple | Related Tickets: Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Since the early days of building Haskell shared libs on Linux we have been using a scheme that is really a bit of a hack. We should do it properly. In my blog post on this from 2009 (http://www.well-typed.com/blog/30/) I said: > If we use ldd again to look at the libfoo.so that we've made we will notice that it is missing a dependency on the rts library. This is problem that we've yet to sort out, so for the moment we can just add the dependency ourselves: > > $ ghc --make -dynamic -shared -fPIC Foo.hs -o libfoo.so \ > -lHSrts-ghc6.11 -optl-Wl,-rpath,/opt/ghc/lib/ghc-6.11/ > > The reason it's not linked in yet is because we need to be able to switch which version of the rts we're using without having to relink every library. For example we want to be able to switch between the debug, threaded and normal rts versions. It's quite possible to do this and it just needs a bit more rearranging in the build system to sort it out. Once it's done you'll even be able to switch rts at runtime, eg: > > $ LD_PRELOAD=/opt/ghc/lib/ghc-6.11/libHSrts_debug-ghc6.11.so > $ ./Hello So in general, if a shared lib requires symbols from another shared lib then it should depend on it. In ELF terminology that means a NEEDED entry to say this lib needs that other lib. This is important to be able to link and load these shared libraries, otherwise they can have dangling dependencies. But we don't do this. For the specific case of the RTS we do not link Haskell shared libs against the RTS. So they have lots of dangling symbols. These libraries cannot be loaded on their own, e.g. with `dlopen()`. This is bad, and has other knock-on consequences. Why don't we link to the RTS? It's because historically (with static linking) GHC had had the ability to select the flavour of the RTS when final executables are linked, not when intermediate libraries are created. This works because the RTS flavours share a common ABI. This is a useful feature as it lets us select the SMP or debug or other RTS at final link time. So when we made up the first shared lib scheme on ELF we had to support this. Our initial scheme was like this: don't link Haskell library DSOs against the RTS, only like the final exe against the RTS. Each RTS flavour has a separate SONAME, e.g. `libHSrts_thr-ghc7.8.4.so` or `libHSrts_debug- ghc7.8.4.so`. This works because the runtime linker looks at the final exe first and loads the RTS, and then when other libs are loaded the symbols all resolve. Why can't we link all the libraries against the RTS? Currently each RTS flavour has a different SONAME, which is the key that the dynamic linker uses to identify each library. So if we did link all the Haskell libs against "the" RTS we would have to pick which one at the point at which we create the library, and that'd stop us from being able to choose later. So, can we use a better scheme? We want one that doesn't leave dangling undefined references in intermediate Haskell libs, and is also compatible with the ability to select the flavour of the RTS at final exe link time (or even override it at load time). Yes we can! The first thing to note is that to be interchangeable, all the RTS flavours (that share a compatible ABI) need to have the same SONAME. So for example, all the (non-profiling) RTS DSO files have to have the internal SONAME of `libHSrts-ghc7.8.4.so`. Once they all have the same SONAME, then it's ok for all the Haskell libs to specify a NEEDED dependency on that rts SONAME. But if they have the same SONAME, what do the files get called, where do they live and how are they found? The trick is to make use of the search path. Put each RTS flavour in a different directory, but otherwise with the same filename, e.g. `lib/rts-1.0/thr/libHSrts-ghc7.8.4.so`, `lib/rts-1.0/debug/libHSrts-ghc7.8.4.so` etc. Each library DSO and exe has its list of NEEDED entries, and it has an RPATH entry used to find those libraries if they're not loaded yet. The key is the "if they're not loaded yet" bit. Remember that the linker uses the SONAME as the key to decide if the lib is loaded yet or not. So the libraries could all have an RPATH entry to say to look for the RTS in the directory containing the default RTS flavour. But then the top level exe (or foreign/export shared lib) can also link to the RTS directly (ie an NEEDED entry) and can specify an RPATH which can be for any of the rts flavours. When the linker loads the top level exe, it will loads the selected RTS using the exe's RPATH, and then when the linker sees other Haskell libs that have a NEEDED entry on the RTS it will ignore them because the RTS's SONAME is already loaded. So concretely, instead of: lib/ghc-${ver}/rts-1.0/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/libHSrts_thr-ghc${ver}.so lib/ghc-${ver}/rts-1.0/libHSrts_debug-ghc${ver}.so lib/ghc-${ver}/rts-1.0/libHSrts_l-ghc${ver}.so lib/ghc-${ver}/rts-1.0/libHSrts_thr_l-ghc${ver}.so lib/ghc-${ver}/rts-1.0/libHSrts_thr_debug-ghc${ver}.so each with a different SONAME we'd have lib/ghc-${ver}/rts-1.0/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/thr/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/debug/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/l/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/thr_l/libHSrts-ghc${ver}.so lib/ghc-${ver}/rts-1.0/thr_debug/libHSrts-ghc${ver}.so each with the same SONAME When linking libs we would always use `-lHSrts -rpath lib/ghc-${ver}/rts-1.0` When linking exes (or shared libs for external consumption) we would use both `-lHSrts` and `-rpath lib/ghc-${ver}/${rtsflavour}/rts-1.0`. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10352> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 22

[GHC] #9248: Document -XHaskell98 and -XHaskell2010 in flag reference
by GHC 05 Jan '17

05 Jan '17

#9248: Document -XHaskell98 and -XHaskell2010 in flag reference ------------------------------------+------------------------------------- Reporter: goldfire | Owner: goldfire Type: bug | Status: new Priority: normal | Milestone: Component: Documentation | 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: | ------------------------------------+------------------------------------- -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9248> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #11096: Builtin encoder/decoder should be used for Latin-1
by GHC 05 Jan '17

05 Jan '17

#11096: Builtin encoder/decoder should be used for Latin-1 -------------------------------------+------------------------------------- Reporter: bgamari | Owner: hvr Type: feature | Status: new request | Priority: normal | Milestone: 8.0.1 Component: Compiler | Version: 7.10.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Filed on behalf of hvr. Currently `mkTextEncoding'` delegates decoding/encoding from/to latin-1 to `iconv`. This shouldn't be necessary as we have latin-1 an encoder in `base`. Let's use it. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11096> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #10680: Make Backpack order-independent (again)
by GHC 04 Jan '17

04 Jan '17

#10680: Make Backpack order-independent (again) -------------------------------------+------------------------------------- Reporter: ezyang | Owner: ezyang Type: task | Status: new Priority: normal | Milestone: Component: Package | Version: 7.11 system | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- When we moved to the new `bkp` file format, we also went back to the a format which is order-dependent: that is to say, the order in which you put the declarations matters. So if you write: {{{ unit p where module A where import B module B where ... }}} this fails to type-check, GHC complaining that `B` is not in scope. I did this, in part because it's what the Backpack paper described, and because it was "simpler" to implement. I think we should move back to an order-independent scheme, for the following reasons: 1. Haskell users are used to not needing pay particularly close attention to the ordering of their modules, and forcing people to linearize their module descriptions would be spectacularly disruptive with large amounts of modules. So un-ordered modules are "more natural. 2. Order-independence imposes some constraints on how expressive programs are (with order-dependent Backpack, you can do some pretty tricky things by ordering things certain ways); this could simplify some aspects of compiler implementation and make Backpack easier to explain. 3. A particular case of (2): it seems a lot simpler UX-wise to let a user assume that if you import a module `M` in a unit, it doesn't matter where you import it: you always get the same set of identifiers brought into scope. Thus, the incremental results of signatures should not be visible, c.f. #10679 The main idea is that only the surface-syntax is un-ordered: the internal representation of units is a DAG which we work out in an elaboration phase, not altogether unsimilar from what `GhcMake` computes. Here are the details: **The intermediate representation.** We translate into an intermediate representation which consists of a directed graph between modules, signatures and includes. Edges in the graph indicate a "depends on" relation: 1. `include p` depends on `include q` if, for some module name `H`, `p` requires `H` and `q` provides `H`. 2. A module/signature `M` depends on `include p` if `M` imports a module provided or required by `p`. 3. A module/signature `M` depends on a module/signature `S` if `M` imports `S`. 4. An `include p` depends on a module `M` if `p` requires a module named `M`. (This rule is included for completeness; we are going to disallow it shortly.) We impose one restriction: a signature cannot depend on a home module. See below for how to eliminate this restriction. Rule (2) is worth remarking upon: if a module imports a signature, it depends-on every `include` which requires that signature, as well as the relevant home signature. This could easily result in a cycle; see (2) for how to break these cycles. The consequence of this, however, is that we can factor the graph to introduce the node for the "merge of signatures", which depends on each signature and include which requires it; we can use this opportunity to build and write out the merged interface file for the unit. **Elaboration.** Take a Backpack file, construct this graph, and topsort it into a DAG of SCCs. SCCs with a single node are compileable as before. SCCs with multiple nodes will have to be managed with some mutual recursion mechanism; see refinements for more thoughts on this. **Refinements:** 1. **Can a signature depend on a (home) module?** Imports of this kind require a retypecheck loop. Consider this situation: {{{ unit p where signature H where data T module M where import H data S = S T unit q where include p module Q where import M signature H where import Q data T = T S }}} Here, signature H in q depends on Q. When we typecheck `Q`, we bring `M.S` into the type environment with a `TyThing` that describes the constructor as accepting an abstract type `T`. However, when we subsequently typecheck the local signature `H`, we must refine all `TyThing`s of `T` with the true description (e.g. constructor information). So you'll need to retypecheck `Q` (and `M`) in order to make sure the `TyThing` is correct. 2. **Can an include depend on a (home) module?** If the module has no (transitive) dependency on signatures, this is fine. However, it's easy to have a circular dependency. Consider: {{{ unit p where signature A signature B module M unit q where include p module B where import A ... }}} `B` depends on `p` for `p/A.hsig`; however, `p` depends on `B` because this module is filling a requirement. Fortunately, we can untangle this knot without any mutual recursion nonsense (and the attendant efficiency loss): `A` is just an export list, we can compute it from the abstractly type-checked version of `p` without instantiating `B`. 3. **Can we deal with include-include dependency cycles?** Yes! Just use the Backpack paper's strategy for creating a recursive unit key and compile the two packages `hs-boot` style. But I'm not planning on implementing this yet. 4. **Can we deal with signature-signature dependency cycles?** Ordered Backpack would have supported this: {{{ unit a-sig where signature A where data T unit ab-sig where include a-sig signature B where import A data S = S T signature A where import B data T = T S }}} In our model, `ab-sig` has a cycle. However, I believe any such cycle can be broken by creating sufficiently many units: {{{ unit a-sig where signature B where data T signature A where data S = S T unit b-sig where signature A where data S signature B where data T = T S unit ab-sig where include a-sig include b-sig }}} In principle, GHC could automatically break import cycles by replacing an import with an import of a reduced signature that simply has abstract type definitions. (I'm not sure this is possible for all language features.) This technique would also work for normal modules, assuming that every function is explicitly annotated with a type. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10680> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #12189: sigof01m broken after D2030
by GHC 04 Jan '17

04 Jan '17

#12189: sigof01m broken after D2030 -------------------------------------+------------------------------------- Reporter: niteria | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: sigof01m | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- With [https://phabricator.haskell.org/D2030 D2030] I've changed the implementation of `Ord UnitId` to go from being `Unique` defined to being lexicographic. That somehow broke `sigof01m`. {{{ <ezyang> I think I know what the problem is (ghc --make isn't specifying a dependency from B to A) }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12189> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 2

[GHC] #10838: hsig files don't have good enough error checking for wired-in types
by GHC 04 Jan '17

04 Jan '17

#10838: hsig files don't have good enough error checking for wired-in types -------------------------------------+------------------------------------- Reporter: ezyang | Owner: ezyang Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.11 (Type checker) | Keywords: backpack | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- If I have the signature: {{{ {-# LANGUAGE NoImplicitPrelude #-} module MyPrelude where data Int a }}} And I compile it with `ghc MyPrelude.hsig -sig-of "MyPrelude is base:Prelude"` this won't actually give an error, despite `Int` having the wrong kind. The bug seems to be because `Int` is wired in and this case is not handled. Filing this so I don't forget to fix. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10838> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 1

[GHC] #11564: Possible overzealous unfolding
by GHC 04 Jan '17

04 Jan '17

#11564: Possible overzealous unfolding -------------------------------------+------------------------------------- Reporter: simonmar | Owner: simonpj Type: bug | Status: new Priority: high | Milestone: Component: Compiler | Version: 8.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- I was investigating why (>>=) in the Haxl monad is being inlined more than I would expect, and I ran into something I don't fully understand, and looks dubious. Start from this standalone example: {{{ {-# LANGUAGE ScopedTypeVariables, ExistentialQuantification #-} module Haxl where import Data.IORef import Control.Exception newtype GenHaxl u a = GenHaxl { unHaxl :: Int -> IORef () -> IO (Result u a) } data Result u a = Done a | Throw SomeException | Blocked (Cont u a) data Cont u a = forall b. Cont u b :>>= (b -> GenHaxl u a) | forall b. (b -> a) :<$> (Cont u b) instance Monad (GenHaxl u) where return a = GenHaxl $ \_env _ref -> return (Done a) GenHaxl m >>= k = GenHaxl $ \env ref -> do e <- m env ref case e of Done a -> unHaxl (k a) env ref Throw e -> return (Throw e) Blocked cont -> return (Blocked (cont :>>= k)) instance Functor (GenHaxl u) instance Applicative (GenHaxl u) }}} (it could be simplified further, but I've intentionally used the exact definition of `>>=` that is used in Haxl to be sure I'm not investigating the wrong thing) Compile like this: {{{ ghc -O -c Haxl.hs }}} and look at the .hi file: {{{ ghc --show-iface Haxl.hi }}} see this: {{{ ea159c3b107c307a4e76cd310efcc8bc $fMonadGenHaxl2 :: GenHaxl u a -> (a -> GenHaxl u b) -> Int -> IORef () -> State# RealWorld -> (# State# RealWorld, Result u b #) {- Arity: 5, HasNoCafRefs, Strictness: <C(C(C(S(SS)))),1*C1(C1(C1(U(U,1*U))))><L,U><L,U><L,U><S,U>, Unfolding: InlineRule (5, True, False) (\ @ u @ a @ b (ds :: GenHaxl u a) (k :: a -> GenHaxl u b) (env :: Int) (ref :: IORef ()) (s :: State# RealWorld)[OneShot] -> case (ds `cast` (N:GenHaxl[0] <u>_P <a>_R) env ref) `cast` (N:IO[0] <Result u a>_R) s of ds1 { (#,#) ipv ipv1 -> case ipv1 of wild { Done a1 -> ((k a1) `cast` (N:GenHaxl[0] <u>_P <b>_R) env ref) `cast` (N:IO[0] <Result u b>_R) ipv Throw e -> (# ipv, Throw @ u @ b e #) Blocked cont -> (# ipv, Blocked @ u @ b (:>>= @ u @ b @ a cont k) #) } }) -} }}} That right there is the definition of `>>=`. Note that it has an `InlineRule`, which means that it will be unconditionally unfolded pretty much everywhere. I don't think this is right - there's no benefit to be had in inlining it unconditionally. I delved in a bit more, and it seems this unfolding arises during worker- wrapper. Before WW we have {{{ a_sVM [LclId, Arity=5, Str=DmdType <C(C(C(S(SS)))),1*C1(C1(C1(U(U,1*U))))><L,U><L,U><L,U><S,U>, Unf=Unf{Src=<vanilla>, TopLvl=True, Value=True, ConLike=True, WorkFree=True, Expandable=True, Guidance=IF_ARGS [0 60 0 0 0] 94 60}] a_sVM = \ @ u_XQR @ a_aPN @ b_aPO ds_dQP [Dmd=<C(C(C(S(SS)))),1*C1(C1(C1(U(U,1*U))))>] k_aEC env_aED ref_aEE s_aVC [Dmd=<S,U>, OS=OneShot] -> case (((ds_dQP `cast` ...) env_aED ref_aEE) `cast` ...) s_aVC of _ [Occ=Dead, Dmd=<L,A>] { (# ipv_aVF [Dmd=<S,U>], ipv1_aVG [Dmd=<S,1*U>] #) -> case ipv1_aVG of _ [Occ=Dead, Dmd=<L,A>] { Done a_aEG -> ((((k_aEC a_aEG) `cast` ...) env_aED ref_aEE) `cast` ...) ipv_aVF; Throw e_aEH -> (# ipv_aVF, Haxl.Throw e_aEH #); Blocked cont_aEI -> (# ipv_aVF, Haxl.Blocked (Haxl.:>>= cont_aEI k_aEC) #) } } }}} and after WW we have {{{ a_sVM [LclId, Arity=5, Str=DmdType <C(C(C(S(SS)))),1*C1(C1(C1(U(U,1*U))))><L,U><L,U><L,U><S,U>, Unf=Unf{Src=InlineStable, TopLvl=True, Value=True, ConLike=True, WorkFree=True, Expandable=True, Guidance=ALWAYS_IF(arity=5,unsat_ok=True,boring_ok=False) Tmpl= \ @ u_XQR @ a_aPN @ b_aPO ds_dQP [Occ=Once] k_aEC [Occ=Once*] env_aED ref_aEE s_aVC [Occ=Once, OS=OneShot] -> case (((ds_dQP `cast` ...) env_aED ref_aEE) `cast` ...) s_aVC of _ [Occ=Dead] { (# ipv_aVF [Occ=Once*], ipv1_aVG [Occ=Once!] #) -> case ipv1_aVG of _ [Occ=Dead] { Done a_aEG [Occ=Once] -> ((((k_aEC a_aEG) `cast` ...) env_aED ref_aEE) `cast` ...) ipv_aVF; Throw e_aEH [Occ=Once] -> (# ipv_aVF, Haxl.Throw e_aEH #); Blocked cont_aEI [Occ=Once] -> (# ipv_aVF, Haxl.Blocked (Haxl.:>>= cont_aEI k_aEC) #) } }}] a_sVM = \ @ u_XQR @ a_aPN @ b_aPO ds_dQP [Dmd=<C(C(C(S(SS)))),1*C1(C1(C1(U(U,1*U))))>] k_aEC env_aED ref_aEE s_aVC [Dmd=<S,U>, OS=OneShot] -> case (((ds_dQP `cast` ...) env_aED ref_aEE) `cast` ...) s_aVC of _ [Occ=Dead, Dmd=<L,A>] { (# ipv_aVF [Dmd=<S,U>], ipv1_aVG [Dmd=<S,1*U>] #) -> case ipv1_aVG of _ [Occ=Dead, Dmd=<L,A>] { Done a_aEG -> ((((k_aEC a_aEG) `cast` ...) env_aED ref_aEE) `cast` ...) ipv_aVF; Throw e_aEH -> (# ipv_aVF, Haxl.Throw e_aEH #); Blocked cont_aEI -> (# ipv_aVF, Haxl.Blocked (Haxl.:>>= cont_aEI k_aEC) #) } } }}} For some unknown reason, this binding has acquired an always-on unfolding. There's no wrapper, we're just unfolding the whole thing. Simon, can you shed any light here? I would like to tune unfolding sizes to reduce code bloat in our codebase, but with this unfolding being unconditional it doesn't work to use `-funfolding-use-threshold`, I can only use `NOINLINE` but that's too blunt. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11564> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 15

[GHC] #12241: Surprising constructor accumulation
by GHC 29 Dec '16

29 Dec '16

#12241: Surprising constructor accumulation -------------------------------------+------------------------------------- Reporter: dfeuer | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Runtime | Version: 7.10.3 System | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Runtime Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- `containers` version 0.5.7.1 (and a few earlier versions) uses the following implementation of `fromList` by Ross Paterson: {{{#!hs fromList :: [a] -> Seq a fromList = Seq . mkTree 1 . map_elem where {-# SPECIALIZE mkTree :: Int -> [Elem a] -> FingerTree (Elem a) #-} {-# SPECIALIZE mkTree :: Int -> [Node a] -> FingerTree (Node a) #-} mkTree :: (Sized a) => Int -> [a] -> FingerTree a mkTree !_ [] = EmptyT mkTree _ [x1] = Single x1 mkTree s [x1, x2] = Deep (2*s) (One x1) EmptyT (One x2) mkTree s [x1, x2, x3] = Deep (3*s) (One x1) EmptyT (Two x2 x3) mkTree s (x1:x2:x3:x4:xs) = case getNodes (3*s) x4 xs of (ns, sf) -> case mkTree (3*s) ns of !m -> Deep (3*size x1 + size m + size sf) (Three x1 x2 x3) m sf getNodes :: Int -> a -> [a] -> ([Node a], Digit a) getNodes !_ x1 [] = ([], One x1) getNodes _ x1 [x2] = ([], Two x1 x2) getNodes _ x1 [x2, x3] = ([], Three x1 x2 x3) getNodes s x1 (x2:x3:x4:xs) = (Node3 s x1 x2 x3:ns, d) where (ns, d) = getNodes s x4 xs map_elem :: [a] -> [Elem a] #if __GLASGOW_HASKELL__ >= 708 map_elem xs = coerce xs #else map_elem xs = Data.List.map Elem xs #endif {-# INLINE map_elem #-} }}} This uses one lazy list per "level" in the tree being constructed. I believe Paterson (and pretty much everyone else) expected that there would be `O(log n)` pair constructors and conses live at any given time. Wadler's technique in [http://homepages.inf.ed.ac.uk/wadler/papers/leak/leak.ps.gz Fixing some space leaks with a garbage collector], which the GHC commentary indicates is used in GHC, should clean up the pairs in `getNodes`'s `d` thunks as they reach WHNF. Lennart Spitzner dug into the unimpressive performance of the above code and using {{{#!hs main = evaluate $ S.fromList [(0::Int)..999999] }}} produced [http://heap.ezyang.com/view/72d4d1eb879a2085ffd49d270b03c7a037b4d5c2 this heap profile]. If I'm reading it right, this suggests that there are lots of `(,)` and also `(:)` constructors live, more `O(n)` than `O(log n)`. I had previously found that I could improve performance by building the intermediate lists strictly, but that violates the generational hypothesis and leads to a slow-down for very large arguments ([http://heap.ezyang.com/view/2e80598f73d9281f5eadfeb041f5b9aef6e448b0 Spitzner's heap profile]). Spitzner was able to come up with a very clever (but much trickier) implementation that skirted all these problems ([http://heap.ezyang.com/view/7f1994c0417931360fe7c41bd995c37ebc3fd6c5 profile]) and avoids ever allocating the troublesome pairs. So the problem is thoroughly bypassed for `containers`, but it seems like something is not quite right here, and it might bear looking into. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12241> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 2

[GHC] #11216: RebindableSyntax with pattern guards needs `fail`
by GHC 26 Dec '16

26 Dec '16

#11216: RebindableSyntax with pattern guards needs `fail` -------------------------------------+------------------------------------- Reporter: goldfire | 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 | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- If I say {{{ {-# LANGUAGE RebindableSyntax #-} module Bug where foo :: (a, b) -> () foo x | (_,_) <- x = () }}} I get {{{ Bug.hs:6:9: error: Not in scope: ‘fail’ }}} Yet even if I bring `fail` into scope, it's not mentioned in the Core produced by my program. And it really shouldn't: there's not a monad in sight! -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11216> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #8107: need types to express constant argument for primop correctness
by GHC 21 Dec '16

21 Dec '16

#8107: need types to express constant argument for primop correctness ------------------------------------+------------------------------------- Reporter: carter | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.6.3 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- (not sure if this is a bug or a feature request, or a task, or a dialogue, filing as a bug for now) a number of compiler intrinsics that GHC has started to support (prefetch) and will come to support (concurrent memory operations a la http://ghc.haskell.org/trac/ghc/ticket/7883) require certain arguments to be statically fixed at compile time {{{ primop PrefetchAddrOp "prefetchAddr#" GenPrimOp Addr# -> Int# -> Addr# with llvm_only = True }}} consider the specification on the llvm side for prefetch http://llvm.org/docs/LangRef.html#llvm-prefetch-intrinsic {{{ Syntax:¶ declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>) Overview: The ‘llvm.prefetch‘ intrinsic is a hint to the code generator to insert a prefetch instruction if supported; otherwise, it is a noop. Prefetches have no effect on the behavior of the program but can change its performance characteristics. Arguments: address is the address to be prefetched, rw is the specifier determining if the fetch should be for a read (0) or write (1), and locality is a temporal locality specifier ranging from (0) - no locality, to (3) - extremely local keep in cache. The cache type specifies whether the prefetch is performed on the data (1) or instruction (0) cache. The rw, locality and cache type arguments must be constant integers. }}} the punch line is that we're starting add primops that REALLY require certain arguments to be compile time constants to work correctly, otherwise there will by definition be compiler errors after the c-- phase when operations are passed along to the llvm backend and friends. (for the atomicity ops, i'll be looking into adding them to the native code gen, but the point being it will likely trigger an error at code gen time). a strawman approach would look like adding a type {{{ ReallyActuallyConstant :: * or unliftedtype -> * unliftedtype}}} roughly, then prefetch could have the type {{{ primop PrefetchAddrOp "prefetchAddr#" GenPrimOp Addr# ->ReallyActuallyConstant Int# -> Addr# with llvm_only = True }}} a way of emulating something like this functionality now, would be some template haskell that lets you lift a compile time constant of type t into the type ReallyActuallyConstant t, and is the only way to build a value of that type "ReallyActuallyConstant t". That said, that approach seems like a bad way of providing type safetype for making sure primops actually compile! Point being, some sort of way of nicely enforcing that a value needs to be available at compile time, that doesn't require doing abusing template haskell, is the idea i'm pointing at. Might also be useful for other things, but for now, i'm merely remarking on an invariant these primops have -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8107> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11