September 2016 - ghc-tickets

[GHC] #8279: bad alignment in code gen yields substantial perf issue
by GHC 30 Aug '18

30 Aug '18

#8279: bad alignment in code gen yields substantial perf issue ------------------------------+-------------------------------------------- Reporter: carter | Owner: Type: bug | Status: new Priority: highest | Milestone: Component: Compiler | Version: 7.7 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Runtime performance bug Unknown/Multiple | Test Case: Difficulty: Unknown | Blocking: Blocked By: | Related Tickets: | ------------------------------+-------------------------------------------- independently, a number of folks have noticed that in various ways, GHC currently has quite a few different memory alignment related performance problems that can have >= 10% perf impact! Nicolas Frisby notes {{{ On my laptop, a program showed a consistent slowdown with -fdicts-strict I didn't find any obvious causes in the Core differences, so I turned to Intel's Performance Counter Monitor for measurements. After trying a few counters, I eventuall saw that there are about an order of magnitude more misaligned memory loads with -fdicts-strict than without, so I think that may be a significant part of the slowdown. I'm not sure if these are code or data reads. Can anyone suggest how to validate this hypothesis about misaligned reads? A subsequent commit has changed the behavior I was seeing, so I'm not interested in alternatives means to determine if -fdicts-strict is somehow at fault — I'm just asking specifically about data/code memory alignment in GHC and how to diagnose/experiment with it. }}} Reid Barton has independently noted {{{ so I did a nofib run with llvm libraries, ghc quickbuild so there's this really simple benchmark tak, https://github.com/ghc/nofib/blob/master/imaginary/tak/Main.hs it doesn't use any libraries at all in the main loop because the Ints all get unboxed but it's still 8% slower with quick-llvm (vs -fasm) weird right? [14:36:30] <carter> could you post the asm it generates for that function? [14:36:49] <rwbarton> well it's identical between the two versions <rwbarton> but they get linked at different offsets because some llvm sections are different sizes <rwbarton> if I add a 128-byte symbol to the .text section to move it to the same address... then the llvm libs version is just as fast <rwbarton> well, apparently 404000 is good and 403f70 is bad <rwbarton> I guess I can test other alignments easily enough <rwbarton> I imagine it wants to start on a cache line <rwbarton> but I don't know if it's just a coincidence that it worked with the ncg libraries <rwbarton> that it got a good location <rwbarton> for this program every 32-byte aligned address is 10+% faster than any merely 16-byte aligned address <rwbarton> and by alignment I mean alignment of the start of the Haskell code section <carter> haswell, sandybridge, ivy bridge, other? <rwbarton> dunno <rwbarton> I have similar results on Intel(R) Core(TM)2 Duo CPU T7300 @ 2.00GHz <rwbarton> and on Quad-Core AMD Opteron(tm) Processor 2374 HE <carter> ok <rwbarton> trying a patch now that aligns all *_entry symbols to 32 bytes }}} the key point in there is that on the tak benchmark, better alignment for the code made a 10% perf differnce on TAk on Core2 and opteron cpus! benjamin scarlet and Luite are speculating that this may be further induced by Tables next to code (TNC) accidentally creating bad alignment so theres cache line pollution / conflicts between the L1 Instruction- cache and data-caches. So one experiment would be to have the TNC transform pad after the table so the function entry point starts on the next cacheline? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8279> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 26

[GHC] #11350: Allow visible type application in patterns
by GHC 29 Aug '18

29 Aug '18

#11350: Allow visible type application in patterns -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 8.1 Keywords: | Operating System: Unknown/Multiple TypeApplications PatternSynonyms | Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Constructors (and pattern synonyms) when treated as expressions may be applied to types: {{{#!hs {-# LANGUAGE TypeApplications #-} Nothing :: Maybe a Nothing @() :: Maybe () pattern Pair :: a -> a -> (a, a) pattern Pair x y = (x, y) Pair :: a -> a -> (a, a) Pair @Int :: Int -> Int -> (Int, Int) }}} But using them in patterns won't parse: {{{#!hs -- parse error in pattern: @Int maybeToList :: Maybe Int -> [Int] maybeToList (Nothing @Int) = [] maybeToList (Just @Int x) = [x] -- parse error in pattern: @Int add :: (Int, Int) -> Int add (Pair @Int x y) = x + y }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11350> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 23

[GHC] #8733: I/O manager causes unnecessary syscalls in send/recv loops
by GHC 27 Aug '18

27 Aug '18

#8733: I/O manager causes unnecessary syscalls in send/recv loops ------------------------------------+------------------------------------- Reporter: tibbe | Owner: simonmar Type: bug | Status: new Priority: normal | Milestone: Component: Runtime System | 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: | ------------------------------------+------------------------------------- Network applications often call send followed by recv, to send a message and then read an answer. This causes syscall traces like this one: {{{ recvfrom(13, ) -- Haskell thread A sendto(13, ) -- Haskell thread A recvfrom(13, ) = -1 EAGAIN -- Haskell thread A epoll_ctl(3, ) -- Haskell thread A (a job for the IO manager) recvfrom(14, ) -- Haskell thread B sendto(14, ) -- Haskell thread B recvfrom(14, ) = -1 EAGAIN -- Haskell thread B epoll_ctl(3, ) -- Haskell thread B (a job for the IO manager) }}} The recvfrom call always fails, as the response from the partner we're communicating with won't be available right after we send the request. We ought to consider descheduling the thread as soon as sending is "done". The hard part is to figure out when that is. See http://www.yesodweb.com/blog/2014/02/new-warp for a real world example. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8733> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 21

[GHC] #11470: Support changing cross compiler target at runtime
by GHC 25 Aug '18

25 Aug '18

#11470: Support changing cross compiler target at runtime -------------------------------------+------------------------------------- Reporter: ezyang | Owner: 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: -------------------------------------+------------------------------------- Relevant mailing list thread: https://mail.haskell.org/pipermail/ghc- devs/2016-January/011064.html At the moment, GHC's cross-compiling support requires you to build a cross-compiler for every target platform from scratch. (https://ghc.haskell.org/trac/ghc/wiki/CrossCompilation) There are two reasons for this: 1. GHC uses a configure script to interrogate the GCC toolchain about the vagaries of cross-compilation, some parameters of which get baked into various build system scripts and header files and such. Some of this information wends its way into the final built GHC executable. 2. GHC needs to build the boot libraries, using the stage 1 compiler, targeting the cross compiler. It would be very useful to move this configuration from configure time to run time, so that a user can use GHC as a cross-compiler without having to recompile GHC. That is, we would like to (1) ensure that any information from the configure script can alternately be supplemented at runtime, (2) that users can cross-compile the boot libraries on-demand (since GHC won't ship with the ARM versions of the boot libraries), and (3) rearchitect GHC internally so that it handles knows to keep separate the interface files/package database for various cross compilation targets. Such a change will make cross-compiling more convenient (the iOS cross- compiler currently requires GHC to be built twice, once for ARM and once for the simulator!) and will also pave the way for supporting Template Haskell and compiler plugins in the cross-compiler, since a GHC that knows how to deal with both the target and host platforms can simply ensure that it only loads code built for the host platform (a user can then, if necessary, build code twice, once for the host for Template Haskell, and once for the target). Unclear points of design: * How should GHC get the information about the GCC cross-compiler toolchain (which currently is gotten by `configure`?) Preferably, not by running a `configure` script every time you invoke it. What is this information anyway? * How should GHC rebuild the boot libraries? Maybe `cabal-install` can simply handle this for you. Note: please update the top of this ticket as the design becomes clearer. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11470> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #10346: Cross-module SpecConstr
by GHC 24 Aug '18

24 Aug '18

#10346: Cross-module SpecConstr -------------------------------------+------------------------------------- Reporter: simonpj | 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: | -------------------------------------+------------------------------------- Type-class specialisation now happens flawlessly across modules. That is, if I define {{{ module DefineF where f :: Num a => a -> a {-# INLINEABLE f #-} f x = ...f x'.... }}} then modules that import `DefineF` and call `f` at some particular type (say `Int`) will generate a specialised copy of `f`'s code. But this does not happen for `SpecConstr`; we only specialise a function for calls made in the same module. All the infrastructure is in place to allow cross-module `SpecConstr`; it just hasn't been done yet. This ticket is to record the idea. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10346> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 19

[GHC] #11474: incorrect redundant-constraints warning
by GHC 24 Aug '18

24 Aug '18

#11474: incorrect redundant-constraints warning -------------------------------------+------------------------------------- Reporter: hvr | Owner: Type: bug | Status: new Priority: highest | Milestone: 8.0.1 Component: Compiler | Version: 8.0.1-rc1 (Type checker) | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Incorrect Unknown/Multiple | warning at compile-time Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- The following code when compiled with GHC 8 {{{#!hs {-# LANGUAGE Haskell2010, FunctionalDependencies, KindSignatures, ScopedTypeVariables, TypeOperators, UndecidableInstances #-} import GHC.Generics data Options data Value newtype Tagged s b = Tagged {unTagged :: b} class GToJSON f where gToJSON :: Options -> f a -> Value class SumToJSON f allNullary where sumToJSON :: Options -> f a -> Tagged allNullary Value class AllNullary (f :: * -> *) allNullary | f -> allNullary instance ( AllNullary (a :+: b) allNullary, -- <- removing this line causes a compile error SumToJSON (a :+: b) allNullary ) => GToJSON (a :+: b) where gToJSON opts = (unTagged :: Tagged allNullary Value -> Value) . sumToJSON opts }}} emits a warning {{{ bug.hs:19:10: warning: • Redundant constraint: AllNullary (a :+: b) allNullary • In the instance declaration for ‘GToJSON (a :+: b)’ }}} when commenting out the `AllNullary` constraint, this however results the compile error {{{ bug.hs:19:10: error: • Could not deduce (SumToJSON (a :+: b) allNullary0) from the context: SumToJSON (a :+: b) allNullary bound by an instance declaration: SumToJSON (a :+: b) allNullary => GToJSON (a :+: b) at redconstr.hs:(19,10)-(22,24) The type variable ‘allNullary0’ is ambiguous • In the ambiguity check for an instance declaration To defer the ambiguity check to use sites, enable AllowAmbiguousTypes In the instance declaration for ‘GToJSON (a :+: b)’ }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11474> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 12

[GHC] #9688: Improve the interaction between CSE and the join point transformation
by GHC 24 Aug '18

24 Aug '18

#9688: Improve the interaction between CSE and the join point transformation -------------------------------------+------------------------------------- Reporter: dfeuer | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.9 Keywords: CSE | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: Runtime Blocked By: | performance bug Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- It appears that the join point transformation sometimes interferes with CSE when CSE would be much better. Two examples: === digitToIntMaybe === Suppose we define {{{#!hs isHexDigit :: Char -> Bool isHexDigit c = (fromIntegral (ord c - ord '0')::Word) <= 9 || (fromIntegral (ord c - ord 'a')::Word) <= 5 || (fromIntegral (ord c - ord 'A')::Word) <= 5 digitToInt c | (fromIntegral dec::Word) <= 9 = dec | (fromIntegral hexl::Word) <= 5 = hexl + 10 | (fromIntegral hexu::Word) <= 5 = hexu + 10 | otherwise = error ("Char.digitToInt: not a digit " ++ show c) -- sigh where dec = ord c - ord '0' hexl = ord c - ord 'a' hexu = ord c - ord 'A' -- We could also expand this out in cases manually, but it makes no -- difference as far as I can tell. }}} Suppose we then write a naive `digitToIntMaybe` function: {{{#!hs digitToIntMaybe c | isHexDigit c = Just (digitToInt c) | otherwise = Nothing }}} What I would want this to do is "zip" the nested cases and give Core like this: {{{ $wdigitToIntMaybe $wdigitToIntMaybe = \ ww_s2Ag -> let { x#_a2yy x#_a2yy = -# (ord# ww_s2Ag) 48 } in case tagToEnum# (leWord# (int2Word# x#_a2yy) (__word 9)) of _ { False -> let { x#1_X2z7 x#1_X2z7 = -# (ord# ww_s2Ag) 97 } in case tagToEnum# (leWord# (int2Word# x#1_X2z7) (__word 5)) of _ { False -> let { x#2_X2zh x#2_X2zh = -# (ord# ww_s2Ag) 65 } in case tagToEnum# (leWord# (int2Word# x#2_X2zh) (__word 5)) of _ { False -> Nothing; True -> Just (I# (+# x#2_X2zh 10)) }; True -> Just (I# (+# x#1_X2z7 10)) }; True -> Just (I# x#_a2yy) } digitToIntMaybe digitToIntMaybe = \ w_s2Ad -> case w_s2Ad of _ { C# ww1_s2Ag -> $wdigitToIntMaybe ww1_s2Ag } }}} But instead, the join point transformation triggers, and we get this: {{{ digitToIntMaybe1 digitToIntMaybe1 = \ ww_s2Cp -> error (unpackAppendCString# "Char.digitToInt: not a digit "# ($w$cshowsPrec15 ww_s2Cp ([]))) $wdigitToIntMaybe $wdigitToIntMaybe = \ ww_s2Cp -> let { $j_s2Bc $j_s2Bc = \ _ -> Just (let { a_s2B5 a_s2B5 = int2Word# (-# (ord# ww_s2Cp) 48) } in case tagToEnum# (leWord# a_s2B5 (__word 9)) of _ { False -> let { a1_s2B7 a1_s2B7 = int2Word# (-# (ord# ww_s2Cp) 97) } in case tagToEnum# (leWord# a1_s2B7 (__word 5)) of _ { False -> let { a2_s2B9 a2_s2B9 = int2Word# (-# (ord# ww_s2Cp) 65) } in case tagToEnum# (leWord# a2_s2B9 (__word 5)) of _ { False -> digitToIntMaybe1 ww_s2Cp; True -> I# (+# (word2Int# a2_s2B9) 10) }; True -> I# (+# (word2Int# a1_s2B7) 10) }; True -> I# (word2Int# a_s2B5) }) } in case tagToEnum# (leWord# (int2Word# (-# (ord# ww_s2Cp) 48)) (__word 9)) of _ { False -> case tagToEnum# (leWord# (int2Word# (-# (ord# ww_s2Cp) 97)) (__word 5)) of _ { False -> case tagToEnum# (leWord# (int2Word# (-# (ord# ww_s2Cp) 65)) (__word 5)) of _ { False -> Nothing; True -> $j_s2Bc void# }; True -> $j_s2Bc void# }; True -> $j_s2Bc void# } digitToIntMaybe digitToIntMaybe = \ w_s2Cm -> case w_s2Cm of _ { C# ww1_s2Cp -> $wdigitToIntMaybe ww1_s2Cp } }}} We perform the same three tests twice each, and test for an error condition that obviously can't happen. === `quotRem` and `divMod` === If we define {{{#!hs x `quot` y = fst (x `quotRem` y) x `rem` y = snd (x `quotRem` y) }}} and then write something like {{{#!hs f x y | x `rem` y == 0 = x `quot` y | otherwise = 17 }}} then CSE works some magic and we only calculate `quotRem x y` once. Unfortunately, if we do this: {{{#!hs whatever x y = if x `myRem` y == 0 then (x `myQuot` y) + 14 else x `myQuot` y }}} then the join point transformation fires, collecting the `myQuot x y` expressions in the case branches and preventing CSE from recognizing the much better opportunity to eliminate those calculations altogether. The situation with `divMod` is much worse. The join point transformation applied to the cases defining `divMod` prevents CSE from working magic on it in even simple situations, unless one of the arguments is known, making this definition unusable (the resulting Core is too horrifyingly long to paste here). It would probably be possible to improve the `divMod` situation to something close to the `quotRem` one by making `divMod` `NOINLINE` and adding special `divModLit` rules, but I'd much rather see a general solution. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9688> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #10100: Bogus "redundant constraint" warning with functional dependencies
by GHC 23 Aug '18

23 Aug '18

#10100: Bogus "redundant constraint" warning with functional dependencies -------------------------------------+------------------------------------- Reporter: jstolarek | 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: | -------------------------------------+------------------------------------- Compiling this code with GHC HEAD: {{{#!hs {-# LANGUAGE MultiParamTypeClasses #-} {-# LANGUAGE FunctionalDependencies #-} {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE UndecidableInstances #-} module T10100 where data Zero data Succ a class Add a b ab | a b -> ab, a ab -> b instance Add Zero b b instance (Add a b ab) => Add (Succ a) b (Succ ab) }}} Gives a bogus warning about redundant constraint: {{{ Redundant constraint: Add a b ab In the instance declaration for ‘Add (Succ a) b (Succ ab)’ }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10100> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #11409: Cannot instantiate literals using TypeApplications
by GHC 22 Aug '18

22 Aug '18

#11409: Cannot instantiate literals using TypeApplications -------------------------------------+------------------------------------- Reporter: Feuerbach | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1-rc1 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: -------------------------------------+------------------------------------- {{{ GHCi, version 8.0.1.20160110: http://www.haskell.org/ghc/ :? for help Prelude> :set -XTypeApplications -fprint-explicit-foralls Prelude> :t 42 42 :: forall {a}. Num a => a Prelude> 42 @Int <interactive>:3:1: error: • Cannot not apply expression of type ‘a0’ to a visible type argument ‘Int’ • In the expression: 42 @Int In an equation for ‘it’: it = 42 @Int }}} (Also, "cannot not" looks like a typo.) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11409> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 29

[GHC] #10869: Option to dump preprocessed source
by GHC 22 Aug '18

22 Aug '18

#10869: Option to dump preprocessed source -------------------------------------+------------------------------------- Reporter: phischu | Owner: Type: feature | Status: new request | Priority: low | Milestone: Component: Driver | 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 Revisions: | -------------------------------------+------------------------------------- It would be awesome if GHC had an option `-ddump-preprocessed` that dumps the source code for each module after preprocessing. I am not sure what the current definition of "preprocessing" is but I mean the output of at least the following tools: happy, alex, c2hs, hsc2hs and cpp. Additionally even if a module was not subject to any preprocessing it should be dumped anyway. Use case: I want to parse module files from packages from hackage with `haskell-src-exts` but find it prohibitively difficult to get the preprocessing right. The idea is that after `cabal install` with ghc options `-ddump-preprocessed -ddump-to-file -dumpdir real_modules` you get a complete working set of haskell modules that can be parsed directly without any preprocessing in folder `real_modules`. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10869> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16