July 2016 - ghc-tickets

[GHC] #10229: setThreadAffinity assumes a certain CPU virtual core layout
by GHC 30 Sep '17

30 Sep '17

#10229: setThreadAffinity assumes a certain CPU virtual core layout -------------------------------------+------------------------------------- Reporter: nh2 | Owner: simonmar Type: bug | Status: new Priority: normal | Milestone: Component: Runtime | Version: 7.10.1 System | Operating System: Unknown/Multiple Keywords: | Type of failure: Runtime Architecture: | performance bug Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- The {{{RTS -qa}}} option that can set thread affinity was implemented in https://git.haskell.org/ghc.git/commitdiff/31caec794c3978d55d79f715f21fb729… {{{ // Schedules the thread to run on CPU n of m. m may be less than the // number of physical CPUs, in which case, the thread will be allowed // to run on CPU n, n+m, n+2m etc. void setThreadAffinity (nat n, nat m) }}} Today I discovered that on some machines, this option helps parallel performance (e.g. {{{+RTS -N4}}}) a lot, while on others it doesn't. Together with thomie on #ghc, I found out the reason: Lets assume I have 4 real cores with hyperthreading, so 8 virtual cores. The mapping of hyperthreading cores to physical cores is different across machines. On my one machine (Intel i5), the layout is 11223344, meaning that the first two vCPUs (hyperthreads) that the OS announces (visible e.g. in HTOP) map to the first physical core in the system, and so on. On my other machine (Intel Xeon), the layout is 12341234; here the 1st and the 5th vCPU map to the same physical core. This layout can be (on Linux) observed by running: {{{ cat /proc/cpuinfo|egrep "processor|physical id|core id" |sed 's/^processor/\nprocessor/g' }}} I do not know whether this layout is dictated by the processor, chosen by the OS, or even changing across reboots; what is clear is that the layout can vary across machines. Now, as explained by thomie: {{{ -qa will set your 4 capabilities to cores [(1,5), (2,6), (3,7), (4,8)], and then the os randomly chooses out of those tuples }}} This strategy is optimal for the 12341234 layout; for example, when running with -N4, it ensures that two threads are not scheduled onto vCPUs that are on the same physical core. The possible {{{+RTS -aq}}} choice {{{1__4_23_}}} is a great assignment in this case, as is {{{1234____}}} ({{{_}}} means the vCPU is not chosen). But for the 11223344, the choice {{{1234____}}} isn't good, because it uses only 2 of our 4 physical cores; our program now takes twice as long to run. ---- It seems likely to me that {{{setThreadAffinity}}} was written on a machine with 12341234 layout, and with the assumption that all machines have this layout. It would be great if we could change it to take the actual layout into account. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10229> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #10941: Large heap address space breaks valgrind
by GHC 29 Sep '17

29 Sep '17

#10941: Large heap address space breaks valgrind -------------------------------------+------------------------------------- Reporter: nomeata | Owner: Type: task | Status: new Priority: normal | Milestone: Component: Runtime | Version: 7.11 System | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: #9706 Differential Rev(s): | -------------------------------------+------------------------------------- It seems that programs compiled with a GHC `--enable-large-address-space` (see #9706) no longer run under valgrind, which is a pitty, as that is useful to get more reliable numbers out of nofib. It fails with `internal error: getMBlock: mmap: Invalid argument` This bug is a general “Can we do anything about this”. In particular, is there a possibility to make this a RTS option instead of a compile time option? Or maybe it is possible to talk to the valgrind authors if they can support such large allocations? If neither, can we maybe extend the error message to say “If you are using a tool like valgrind, you might want to recompile the compiler with `--disable-large-address-space.” -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10941> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #11228: Interaction between ORF and record pattern synonyms needs to be resolved.
by GHC 29 Sep '17

29 Sep '17

#11228: Interaction between ORF and record pattern synonyms needs to be resolved. -------------------------------------+------------------------------------- Reporter: mpickering | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 Keywords: | Operating System: Unknown/Multiple PatternSynonyms, orf | Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Currently the two extensions do not work well together. When defining a record pattern synonym all field names must be distinct even with the extension turned on. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11228> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #11342: Character kind
by GHC 28 Sep '17

28 Sep '17

#11342: Character kind -------------------------------------+------------------------------------- Reporter: alexvieth | Owner: alexvieth Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 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: -------------------------------------+------------------------------------- We know and love the Symbol kind, but as far as I know it's not possible to analyse a Symbol type, as we would a String term. It would be nice to have a Character kind and some related type families: {{{#!hs data Character -- Pattern match a Symbol. 'Nothing when empty, 'Just '(head, tail) -- otherwise. type family UnconsSymbol (s :: Symbol) :: Maybe (Character, Symbol) -- Put a Character at the head of a Symbol. type family ConsSymbol (c :: Character) (s :: Symbol) :: Symbol -- Work with Characters. Ideally we'd have counterparts for every -- function on Data.Char. type family ToUpper (c :: Character) :: Character type family IsAlpha (c :: Character) :: Bool }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11342> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 12

[GHC] #12376: Allow function definitions in record syntax
by GHC 27 Sep '17

27 Sep '17

#12376: Allow function definitions in record syntax -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 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: -------------------------------------+------------------------------------- Would it be possible to write {{{#!hs propertyMap :: (Monad m, Ord k) => v -> PropertyMap m k v propertyMap v0 = go v0 Map.empty where go v m = PropertyMap { getP = \k -> return $ maybe v id (Map.lookup k m) , putP = \k v' -> return $ go v (Map.insert k v' m) } }}} as {{{#!hs propertyMap :: (Monad m, Ord k) => v -> PropertyMap m k v propertyMap v0 = go v0 Map.empty where go v m = PropertyMap { getP k = return $ maybe v id (Map.lookup k m) , putP k v' = return $ go v (Map.insert k v' m) } }}} Simpler example `MkFoo { id = \x -> x }` as {{{#!hs MkFoo { id x = x } }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12376> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 9

[GHC] #9898: Couldn't match type `(Char, ())' with `()'
by GHC 27 Sep '17

27 Sep '17

#9898: Couldn't match type `(Char, ())' with `()' -------------------------------------+------------------------------------- Reporter: erisco | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler (Type | Version: 7.8.3 checker) | Operating System: Windows Keywords: | Type of failure: GHC Architecture: x86_64 (amd64) | rejects valid program Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: -------------------------------------+------------------------------------- 'test3' is expected to type check given that 'test1' and 'test2' type check. 'test4' is not expected to type check. {{{ ghc_bug2.hs:24:9: Couldn't match type `(Char, ())' with `()' Expected type: Filter (Equal Int) (Char, Filter (Equal Int) ()) Actual type: () In the expression: () :: Filter (Equal Int) (Char, Filter (Equal Int) ()) In an equation for `test2': test2 = () :: Filter (Equal Int) (Char, Filter (Equal Int) ()) ghc_bug2.hs:28:9: Couldn't match type `(Char, ())' with `()' Expected type: Filter (Equal Int) (Char, ()) Actual type: () In the expression: () In an equation for `test3': test3 = () }}} {{{#!hs {-# LANGUAGE TypeFamilies, UndecidableInstances, DataKinds #-} module Main where type family Filter f xs where Filter f (x, xs) = ApplyFilter (f x) (x, Filter f xs) Filter f () = () -- type family ApplyFilter p xs where ApplyFilter False (x, xs) = xs ApplyFilter p xs = xs -- type family Equal x y where Equal x x = True Equal x y = False -- -- Type checks test1 :: ApplyFilter ((Equal Int) Char) (Char, Filter (Equal Int) ()) test1 = () -- Couldn't match type `(Char, ())' with `()' test2 = () :: Filter (Equal Int) (Char, Filter (Equal Int) ()) -- Couldn't match type `(Char, ())' with `()' test3 :: Filter (Equal Int) (Char, ()) test3 = () -- Type checks, should not test4 :: Filter (Equal Int) (Char, ()) test4 = ('x', ()) }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9898> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #12110: Windows exception handler change causes segfault with API Monitor
by GHC 26 Sep '17

26 Sep '17

#12110: Windows exception handler change causes segfault with API Monitor ----------------------------------------+---------------------------------- Reporter: enolan | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Runtime System | Version: 8.0.1 Keywords: | Operating System: Windows Architecture: Unknown/Multiple | Type of failure: Runtime crash Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: ----------------------------------------+---------------------------------- [http://www.rohitab.com/apimonitor API Monitor] is a debugging tool that intercepts and logs calls to DLLs. It was extremely helpful with debugging #12010 and is broken in recent GHCs. Attaching it to a GHC compiled executable causes the message `Segmentation fault/access violation in generated code` to be printed. Launching one from inside API Monitor doesn't work either, though I can't read the console output before the window closes. Bisect shows the offending commit is 5200bdeb26 - "Replaced SEH handles with VEH handlers which should work uniformly across x86 and x64". I know this is somewhat outside of "standard usage", but there are very few useful debugging tools for stuff like this and I'd really like to not lose this one. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12110> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #11198: TypeInType error message regressions
by GHC 26 Sep '17

26 Sep '17

#11198: TypeInType error message regressions -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: high | Milestone: 8.0.1 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: -------------------------------------+------------------------------------- `TypeInType` causes error messages around certain kind errors to degrade. Here is the case: {{{ [W] (alpha :: k) ~ (Int :: *) }}} where `alpha` is a unification variable, `k` is a skolem, and there is no witness that `k ~ *`. The solver will simplify to {{{ [W] kco :: k ~ * [W] (alpha :: k) ~ (Int |> sym kco) }}} and then the latter, now homogeneous equality is solved by unification {{{ alpha := Int |> sym kco }}} This is all sound. But when reporting a kind error around `k` not equalling `*`, GHC likes to report the type error that the kind error came from. But, after zonking, the type error looks like `Int ~ Int`, when we drop coercions (as we tend to do in error messages). That's unfortunate. It affects test cases * T9196 * T8262 * T8603 * tcfail122 Simon has proposed a rule that should avoid this: never, ever use a wanted as a kind cast. That might work, but an attempt at implementing caused a loop when kind families were involved. I will keep pushing on this. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11198> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 20

[GHC] #8919: Why is xhtml library installed but not exported to users?
by GHC 20 Sep '17

20 Sep '17

#8919: Why is xhtml library installed but not exported to users? ------------------------------------+------------------------------------- Reporter: simons | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.8.1 Component: Build System | Version: 7.6.3 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: Other Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | ------------------------------------+------------------------------------- GHC 7.8.1-rc2 installs the xhtml library, but doesn't export it, i.e. it's completely invisible to users. Is there a particular reason why this setup is necessary? If GHC needs that library internally anywhere, would it be possible to link it statically and *not* install it? Or, if that's not possible, if the library needs to be installed, please make it visible to users. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8919> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 27

[GHC] #9281: Rewrite `integer-gmp` to use only non-allocating GMP functions
by GHC 19 Sep '17

19 Sep '17

#9281: Rewrite `integer-gmp` to use only non-allocating GMP functions -------------------------------------+------------------------------------- Reporter: hvr | Owner: hvr Type: task | Status: new Priority: normal | Milestone: 7.10.1 Component: libraries (other) | Version: Keywords: integer-gmp | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | -------------------------------------+------------------------------------- After some promising results with a proof-of-concept implementation, I'm optimistic we can rewrite `integer-gmp` to use only non-allocating GMP lib functions without suffering from serious regressions. If successful, this would - allow to avoid the custom GMP allocator hack, and thus - avoid issues when linking against other C libraries using GMP, - simplify code, as we would perform all heap allocations in Haskell code (and never inside Cmm/C code as its done now), - and finally maybe even remove a few more superfluous temporary heap allocations. This rewrite can be done incrementally. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9281> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 32