September 2015 - ghc-tickets

[GHC] #9291: Don't reconstruct sum types if the type subtly changes
by GHC 28 Dec '17

28 Dec '17

#9291: Don't reconstruct sum types if the type subtly changes ----------------------------------------------+---------------------------- Reporter: schyler | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Moderate (less than a day) | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: ----------------------------------------------+---------------------------- Consider this test case: {{{ module Main (main) where fun :: Either Int String -> Either String String fun x = case x of Left int -> Left (show int) Right str -> Right str {-# NOINLINE fun #-} main :: IO () main = do l <- getLine print $ fun (Right l) }}} The core we get for fun looks like: {{{ Main.fun [InlPrag=NOINLINE] :: Data.Either.Either GHC.Types.Int GHC.Base.String -> Data.Either.Either GHC.Base.String GHC.Base.String [GblId, Arity=1, Caf=NoCafRefs, Str=DmdType <S,1*U>] Main.fun = \ (x_aur :: Data.Either.Either GHC.Types.Int GHC.Base.String) -> case x_aur of _ [Occ=Dead] { Data.Either.Left int_aus -> Data.Either.Left @ GHC.Base.String @ GHC.Base.String (GHC.Show.$fShowInt_$cshow int_aus); Data.Either.Right str_aCG -> Data.Either.Right @ GHC.Base.String @ GHC.Base.String str_aCG } }}} There would be less allocations and probably perform better if we just coerced `x` into the new type. Because the coercion is common code, this also means that if hypothetically some other sum type which had 15 constructors, and only 3 had subtle type changes, 12 of the case branches could be CSE'd into the single coerce greatly reducing code generated and also hinting the inliner better. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9291> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 50

[GHC] #10012: Cheap-to-compute values aren't pushed into case branches inducing unnecessary register pressure
by GHC 22 Dec '17

22 Dec '17

#10012: Cheap-to-compute values aren't pushed into case branches inducing unnecessary register pressure -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.4 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Runtime Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- When characterizing bytestring's `Builder` interface[1] I noticed that some benchmarks involving repeated appends (e.g. `Host endian/1MB of Word8 in chunks of 16`) perform inexplicably much worse than others. A glance at the assembly revealed that a large number of `Word8`s are being evaluated and saved to registers, only to be later stored. E.g., ```#!asm # First we evaluate a number of Word8s, saving them in registers movzbl %bl,%ebx leaq 1(%r14),%rcx movzbl %cl,%ecx leaq 2(%r14),%rdx movzbl %dl,%edx leaq 3(%r14),%rsi movzbl %sil,%esi leaq 4(%r14),%r9 movzbl %r9b,%r9d leaq 5(%r14),%r10 movzbl %r10b,%r10d ... # Eventually we run out of registers and start spilling to the stack movq %rax,64(%rsp) leaq 7(%r14),%rax movzbl %al,%eax movq %rbx,72(%rsp) ... # Only after evaluating all of the needed words do we actually consume # them movq %rax,-152(%r12) movq 72(%rsp),%rax movq %rax,-144(%r12) movq 80(%rsp),%rax movq %rax,-136(%r12) movq 88(%rsp),%rax movq %rax,-128(%r12) ... movq %rsi,-56(%r12) movq %r9,-48(%r12) movq %r10,-40(%r12) movq %r11,-32(%r12) movq %r14,-24(%r12) ``` This is due to the fact that the `Word`s are bound outside of a case analysis and GHC is reluctant to push them inside of the branches. The float-in pass will only float a binding into a case if the value is "small" and at least one branch doesn't use the binding. Unfortunately the case expression in question has only two branches. This is demonstrated in the attached testcase. [1] https://github.com/kolmodin/binary/pull/65 [2] https://www.haskell.org/pipermail/ghc-devs/2015-January/007997.html -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10012> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 20

[GHC] #9244: Compiler could warn about type variable shadowing, and hint about ScopedTypeVariables
by GHC 18 Dec '17

18 Dec '17

#9244: Compiler could warn about type variable shadowing, and hint about ScopedTypeVariables -------------------------+------------------------------------------------- Reporter: | Owner: stusmith | Status: new Type: | Milestone: feature request | Version: 7.6.3 Priority: | Operating System: Unknown/Multiple normal | Type of failure: Incorrect warning at Component: | compile-time Compiler | Test Case: Keywords: | Blocking: Architecture: | Unknown/Multiple | Difficulty: | Unknown | Blocked By: | Related Tickets: | -------------------------+------------------------------------------------- GHC already warns about variable shadowing: {{{ timesTwoPlusOne x = timesTwo x + 1 where timesTwo x = x * 2 Warning: This binding for `x' shadows the existing binding bound at <location> }}} However the similar warning doesn't happen for type variables. {{{ tryMaybe :: IO a -> IO (Maybe a) tryMaybe action = do result <- (try action) :: IO (Either SomeException a) return $ case result of Left _ -> Nothing Right v -> Just v Couldn't match type `a' with `a1' `a' is a rigid type variable bound by the type signature for tryMaybe :: IO a -> IO (Maybe a) at types.hs:<line>:13 `a1' is a rigid type variable bound by an expression type signature: IO (Either SomeException a1) at types.hs:<line>:15 Expected type: IO a1 Actual type: IO a ... }}} Here, I thought that the 'a' in the function's type declaration was the same 'a' in the expression type declaration. However in Haskell 98, they are completely different variables. Suggestion: if a type variable is renamed by the compiler due to a clash with another type variable, issue a warning that the second shadows the first, and give a hint about using -XScopedTypeVariables and forall. Alternative suggestion: if an error is displayed, where the error contains a renamed type variable, issue a hint that the second shadows the first, and give a hint about using -XScopedTypeVariables and forall. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9244> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #8272: testing if SpLim=$rbp and Sp=$rsp changed performance at all
by GHC 16 Dec '17

16 Dec '17

#8272: testing if SpLim=$rbp and Sp=$rsp changed performance at all ----------------------------------------------+---------------------------- Reporter: carter | Owner: carter Type: task | Status: new Priority: normal | Milestone: 7.10.1 Component: Compiler | Version: 7.7 Keywords: callingConvention | Operating System: Architecture: x86_64 (amd64) | Unknown/Multiple Difficulty: Project (more than a week) | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: ----------------------------------------------+---------------------------- testing if SpLim=$rbp and Sp=$rsp changed performance at all would need a stack check but then push could be used to spill to the stack Idea via Nathan Howell. At the very least, the x86 PUSH instruction has a more succinct encoding than MOV. worth hacking out to see if this can measurable shift ghc perf on nofib or not. this would be part of a larger effort to explore ways to improve GHC's calling convention for modern hardware -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8272> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #9364: GHCi (or haskeline?) confused by non-single-width characters
by GHC 14 Dec '17

14 Dec '17

#9364: GHCi (or haskeline?) confused by non-single-width characters -------------------------------------+------------------------------------- Reporter: cheater | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: GHCi | Version: 7.8.3 Keywords: linux, terminal, | Operating System: colour, color, prompt | Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: Other Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: -------------------------------------+------------------------------------- Hi, there's an issue in the GHCi prompt. To reproduce: 1. Do everything in Linux (I don't have other operating systems); I'm using a version of the gnome 2 gnome-terminal. 2. Use the following to colourise the prompt in GHCi: echo -e :set prompt '"\033[32;1m%s\033[34;1m>\033[0m "' >> ~/.ghc/ghci.conf 3. Import quite a few modules (say, 5 to 10 should be enough) 4. Give the prompt enough text that it wraps around. It's best if you start the text with -- so that it's a comment and doesn't produce any output. For example, say I entered "-- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb", I now see the following: *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> -- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb (do note that the first set of b's is located on the same line as the prompt, and the second is located under the prompt) 5. Press Enter. You should now see: *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> -- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> 6. Press Ctrl-P or the up arrow key to select the previous history item. You will now see the following: *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> -- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> -- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb 7. Press Ctrl-A or the Home cursor key to go to the beginning of the text entry, and then type X. You will now see the last prompt to be: *Types Control.Monad.Writer Control.Monad Data.Functor Control.Applicative Data.Monoid> -- bbbbbbbbbbbbX-- bbbbbbbbbbbbbbbbb bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb 8. If you press Enter, you will find out that prompt was actually (without line breaks): X-- bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb I tried resizing the terminal window to 45 chars to make the lines above shorter; I imported Data.Monoid and Data.Functor in a new GHCi. The prompt contained garbage, most likely badly wrapped escape codes. See attachment. As both likely are a single bug in disguise, I include both reports in this ticket. Creating a test case for whether things actually work in a terminal might be difficult, but creating test for whether things get wrapped correctly (and not testing how things work out in a terminal) should be simple if annoying. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9364> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #9669: Long compile time/high memory usage for modules with many deriving clauses
by GHC 12 Dec '17

12 Dec '17

#9669: Long compile time/high memory usage for modules with many deriving clauses -------------------------------------+------------------------------------- Reporter: snoyberg | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.3 Keywords: | Operating System: Linux Architecture: x86_64 (amd64) | Type of failure: Compile- Difficulty: Unknown | time performance bug Blocked By: | Test Case: Related Tickets: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- I've seen many different people complain about this in various different contexts. One of the most common I personally see is when using Persistent with a large number of entities, which results in multi-minute build times for that single module. Usually, these reports have come from private repositories, which makes for a difficult repro. Also, the usage of Template Haskell in those modules tends to confuse the issue. So I'd like to report this issue from a completely separate project, with the guess that this represents an issue that is affecting many other users. Steps to reproduce: * `cabal unpack fpco-api-1.2.0.1` * `cd fpco-api-1.2.0.1` * `cabal install` * `ghc --make -isrc/library src/library/FP/API/Types.hs` * `touch src/library/FP/API/Types.hs` Then, to see the time and memory usage of compiling just the one module, run: * `time ghc -O2 --make -isrc/library src/library/FP/API/Types.hs +RTS -s` On my system (16GB RAM, quadcore i7), the results are: {{{ $ time ghc -O2 --make -isrc/library src/library/FP/API/Types.hs +RTS -s [3 of 3] Compiling FP.API.Types ( src/library/FP/API/Types.hs, src/library/FP/API/Types.o ) 51,846,533,568 bytes allocated in the heap 6,799,246,288 bytes copied during GC 389,297,584 bytes maximum residency (22 sample(s)) 11,330,472 bytes maximum slop 1041 MB total memory in use (0 MB lost due to fragmentation) Tot time (elapsed) Avg pause Max pause Gen 0 2041 colls, 0 par 9.19s 9.19s 0.0045s 0.0998s Gen 1 22 colls, 0 par 4.33s 4.33s 0.1969s 0.5436s TASKS: 4 (1 bound, 3 peak workers (3 total), using -N1) SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled) INIT time 0.00s ( 0.00s elapsed) MUT time 34.21s ( 37.74s elapsed) GC time 13.52s ( 13.53s elapsed) EXIT time 0.07s ( 0.07s elapsed) Total time 47.81s ( 51.34s elapsed) Alloc rate 1,515,662,937 bytes per MUT second Productivity 71.7% of total user, 66.8% of total elapsed gc_alloc_block_sync: 0 whitehole_spin: 0 gen[0].sync: 0 gen[1].sync: 0 real 0m51.375s user 0m50.864s sys 0m0.456s }}} By contrast, with optimizations turned off: {{{ $ time ghc -O0 --make -isrc/library src/library/FP/API/Types.hs +RTS -s [3 of 3] Compiling FP.API.Types ( src/library/FP/API/Types.hs, src/library/FP/API/Types.o ) 12,767,593,936 bytes allocated in the heap 1,078,202,664 bytes copied during GC 179,551,768 bytes maximum residency (13 sample(s)) 6,684,544 bytes maximum slop 439 MB total memory in use (0 MB lost due to fragmentation) Tot time (elapsed) Avg pause Max pause Gen 0 623 colls, 0 par 1.24s 1.24s 0.0020s 0.0651s Gen 1 13 colls, 0 par 0.76s 0.76s 0.0586s 0.2397s TASKS: 4 (1 bound, 3 peak workers (3 total), using -N1) SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled) INIT time 0.00s ( 0.00s elapsed) MUT time 8.28s ( 9.23s elapsed) GC time 2.00s ( 2.00s elapsed) EXIT time 0.05s ( 0.05s elapsed) Total time 10.34s ( 11.28s elapsed) Alloc rate 1,542,443,897 bytes per MUT second Productivity 80.6% of total user, 73.9% of total elapsed gc_alloc_block_sync: 0 whitehole_spin: 0 gen[0].sync: 0 gen[1].sync: 0 real 0m11.304s user 0m11.108s sys 0m0.172s }}} and -O1: {{{ $ time ghc -O1 --make -isrc/library src/library/FP/API/Types.hs +RTS -s [3 of 3] Compiling FP.API.Types ( src/library/FP/API/Types.hs, src/library/FP/API/Types.o ) 45,550,443,664 bytes allocated in the heap 5,721,700,512 bytes copied during GC 358,036,456 bytes maximum residency (21 sample(s)) 9,167,176 bytes maximum slop 906 MB total memory in use (0 MB lost due to fragmentation) Tot time (elapsed) Avg pause Max pause Gen 0 1642 colls, 0 par 7.70s 7.70s 0.0047s 0.1031s Gen 1 21 colls, 0 par 3.68s 3.69s 0.1756s 0.4968s TASKS: 4 (1 bound, 3 peak workers (3 total), using -N1) SPARKS: 0 (0 converted, 0 overflowed, 0 dud, 0 GC'd, 0 fizzled) INIT time 0.00s ( 0.00s elapsed) MUT time 30.21s ( 33.09s elapsed) GC time 11.38s ( 11.39s elapsed) EXIT time 0.07s ( 0.07s elapsed) Total time 41.68s ( 44.55s elapsed) Alloc rate 1,507,571,106 bytes per MUT second Productivity 72.7% of total user, 68.0% of total elapsed gc_alloc_block_sync: 0 whitehole_spin: 0 gen[0].sync: 0 gen[1].sync: 0 real 0m44.577s user 0m44.164s sys 0m0.368s }}} Note that this module is essentially just 900 lines of data declarations, with a large number of derived classes, and a few manually written helper functions and instances. This high memory usage hasn't just resulted in user complaints: automated Jenkins and Travis jobs will often times fail without optimizations disabled, which can be problematic for proper testing and production code. In the case of fpco-api, I've worked around this by adding `-O0` to the cabal file, but it would be much nicer to not have to do that. For some other examples of complaints along these lines: * https://groups.google.com/d/msg/yesodweb/MX1bnOFu8Hc/VLQXhnvpIkMJ * https://groups.google.com/d/msg/yesodweb/XPWixNjuOnM/FN26bmkudgwJ I believe there are a few other threads discussing this, if it would be helpful. I did my testing on GHC 7.8.3 64-bit, Ubuntu 12.04. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9669> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 19

[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