August 2015 - ghc-tickets

Re: [GHC] #7206: Implement cheap build
by GHC 04 Aug '15

04 Aug '15

#7206: Implement cheap build -------------------------------------+------------------------------------- Reporter: simonpj | Owner: simonpj Type: bug | Status: new Priority: normal | Milestone: ⊥ Component: Compiler | Version: 7.4.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: -------------------------------------+------------------------------------- Changes (by thomie): * failure: None/Unknown => Runtime performance bug -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7206#comment:14> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

Re: [GHC] #7206: Implement cheap build
by GHC 04 Aug '15

04 Aug '15

#7206: Implement cheap build -------------------------------------+------------------------------------- Reporter: simonpj | Owner: simonpj Type: bug | Status: new Priority: normal | Milestone: ⊥ Component: Compiler | Version: 7.4.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: -------------------------------------+------------------------------------- Comment (by bgamari): I brought up this issue again in yesterday's GHC meeting. Simon's opinion was that while this doesn't produce a compelling performance improve with existing producers, it is a tool that may be nice to have available for use in user code. I've ported these patches to `master` in my `cheap-build` [https://github.com/ghc/ghc/compare/master...bgamari:cheap-build?expand=1 branch]. Sadly but not surprisingly the story told by `nofib` is still rather disappointing (showing only major changes), {{{ -------------------------------------------------------------------------------- Program Size Allocs Runtime Elapsed TotalMem -------------------------------------------------------------------------------- cacheprof 0.0% +0.1% 0.173 0.173 -2.6% circsim 0.0% 0.0% +3.5% +3.6% 0.0% constraints 0.0% 0.0% +1.2% +1.2% 0.0% cryptarithm1 0.0% 0.0% -0.6% -0.4% 0.0% k-nucleotide 0.0% 0.0% -1.1% -1.1% 0.0% maillist 0.0% 0.0% 0.018 0.018 +5.6% wheel-sieve1 0.0% 0.0% +0.7% +0.9% 0.0% -------------------------------------------------------------------------------- Min 0.0% 0.0% -1.1% -1.1% -2.6% Max 0.0% +0.1% +3.5% +3.6% +5.6% Geometric Mean -0.0% +0.0% +0.2% +0.3% +0.0% }}} I'll be looking in to this later. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7206#comment:13> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

[GHC] #10676: silly assembly for comparing the result of comparisons that return Int# against 0#
by GHC 04 Aug '15

04 Aug '15

#10676: silly assembly for comparing the result of comparisons that return Int# against 0# -------------------------------------+------------------------------------- Reporter: rwbarton | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.1 (CodeGen) | Keywords: | Operating System: Unknown/Multiple Architecture: x86_64 | Type of failure: Runtime (amd64) | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- Consider the module {{{ {-# LANGUAGE MagicHash #-} module Min where import GHC.Exts fgood :: Int# -> Int# -> Int fgood x# y# = case isTrue# (x# <# y#) of False -> I# y# True -> I# x# fbad :: Int# -> Int# -> Int fbad x# y# = case x# <# y# of 0# -> I# y# _ -> I# x# }}} The code for `fgood` looks fine: {{{ 0000000000000130 <Min_fgood_info>: 130: 49 83 c4 10 add $0x10,%r12 134: 4d 3b a5 58 03 00 00 cmp 0x358(%r13),%r12 13b: 77 1a ja 157 <Min_fgood_info+0x27> 13d: 49 39 f6 cmp %rsi,%r14 140: 7c 29 jl 16b <Min_fgood_info+0x3b> 142: 49 c7 44 24 f8 00 00 movq $0x0,-0x8(%r12) 149: 00 00 147: R_X86_64_32S ghczmprim_GHCziTypes_Izh_con_info 14b: 49 89 34 24 mov %rsi,(%r12) 14f: 49 8d 5c 24 f9 lea -0x7(%r12),%rbx 154: ff 65 00 jmpq *0x0(%rbp) 157: 49 c7 85 88 03 00 00 movq $0x10,0x388(%r13) 15e: 10 00 00 00 162: bb 00 00 00 00 mov $0x0,%ebx 163: R_X86_64_32 Min_fgood_closure 167: 41 ff 65 f8 jmpq *-0x8(%r13) 16b: 49 c7 44 24 f8 00 00 movq $0x0,-0x8(%r12) 172: 00 00 170: R_X86_64_32S ghczmprim_GHCziTypes_Izh_con_info 174: 4d 89 34 24 mov %r14,(%r12) 178: 49 8d 5c 24 f9 lea -0x7(%r12),%rbx 17d: ff 65 00 jmpq *0x0(%rbp) }}} But the code for `fbad` has several problems: {{{ 0000000000000018 <Min_fbad_info>: 18: 48 8d 45 f0 lea -0x10(%rbp),%rax 1c: 4c 39 f8 cmp %r15,%rax 1f: 72 3a jb 5b <Min_fbad_info+0x43> 21: 48 89 f0 mov %rsi,%rax 24: 4c 89 f3 mov %r14,%rbx 27: 49 39 f6 cmp %rsi,%r14 2a: 0f 9c c1 setl %cl 2d: 0f b6 c9 movzbl %cl,%ecx 30: 48 85 c9 test %rcx,%rcx 33: 75 51 jne 86 <c1Sm_info+0xe> 35: 49 83 c4 10 add $0x10,%r12 39: 4d 3b a5 58 03 00 00 cmp 0x358(%r13),%r12 40: 0f 87 aa 00 00 00 ja f0 <c1Su_info+0x10> 46: 49 c7 44 24 f8 00 00 movq $0x0,-0x8(%r12) 4d: 00 00 4b: R_X86_64_32S ghczmprim_GHCziTypes_Izh_con_info 4f: 49 89 04 24 mov %rax,(%r12) 53: 49 8d 5c 24 f9 lea -0x7(%r12),%rbx 58: ff 65 00 jmpq *0x0(%rbp) 5b: bb 00 00 00 00 mov $0x0,%ebx 5c: R_X86_64_32 Min_fbad_closure 60: 41 ff 65 f8 jmpq *-0x8(%r13) ... ; c1Sm_info is the other case with its own heap check and GC entry code ; c1Su_info is another GC entry ; in total, another 160 bytes of code }}} For some reason, the heap checks were moved into the alternatives, which was not a good decision in this case. But the silly thing here is the `cmp/setl/movzbl/test/jne` sequence in `Min_fbad_info`, which should be replaced by a `cmp/jl` as in `Min_fgood_info`. Same behavior in 7.8 and HEAD. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10676> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 10

Re: [GHC] #7679: Regression in -fregs-graph performance
by GHC 04 Aug '15

04 Aug '15

#7679: Regression in -fregs-graph performance -------------------------------------+------------------------------------- Reporter: simonmar | Owner: Type: bug | Status: new Priority: low | Milestone: 7.12.1 Component: Compiler (NCG) | Version: 7.6.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: #7192, #8657 | Differential Revisions: -------------------------------------+------------------------------------- Changes (by bgamari): * priority: high => low Comment: After discussing this with Simon Marlow we are going to drop this in severity as the gains offered by `-fregs-graph` weren't so great. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/7679#comment:8> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

[GHC] #10730: Spectral norm allocations increased 17% between 7.6 and 7.8
by GHC 04 Aug '15

04 Aug '15

#10730: Spectral norm allocations increased 17% between 7.6 and 7.8 -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: low | Milestone: Component: Compiler | Version: 7.8.3 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Runtime Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- It appears that `spectral-norm` regressed rather severely from 7.6 to 7.8 in total allocations. Meanwhile the binary size also increased by 7% or so. However, it also appears that runtime was unaffected by this change. The rule firings are very similar between 7.6 and 7.8. Three rules fire in 786 that do not fire in 7.6, {{{ 1 Class op formatArg 1 Class op parseFormat 1 tagToEnum# }}} The difference in allocation persists even if one reverts the `Text.Printf` rewrite that gives rise to the `formatArg` and `parseFormat` rules, so these can be safely ignored. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10730> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 1

[GHC] #10664: T8131 times out on master
by GHC 04 Aug '15

04 Aug '15

#10664: T8131 times out on master -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.12.1 Component: Compiler | Version: 7.11 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Other Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- The testsuite test `codeGen/should_fail/T8131` appears to time out on the `master` branch as of today. This isn't counted as a testsuite failure as it is a should-fail test, but nevertheless something is not right as it doesn't terminate despite being allowed to run for several minutes. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10664> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 10

Re: [GHC] #3306: Improve syntax for GADT + records
by GHC 04 Aug '15

04 Aug '15

#3306: Improve syntax for GADT + records -------------------------------------+------------------------------------- Reporter: simonpj | Owner: simonpj Type: feature request | Status: closed Priority: normal | Milestone: 6.12.1 Component: Compiler | Version: 6.10.2 Resolution: fixed | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: -------------------------------------+------------------------------------- Comment (by Ben Gamari <ben@…>): In [changeset:"30c981e135033840fe1b4bcab697b412369739d7/ghc" 30c981e1/ghc]: {{{ #!CommitTicketReference repository="ghc" revision="30c981e135033840fe1b4bcab697b412369739d7" Removed deprecated syntax for GADT constuctors. Old syntax was deprecated 6 years ago in this commit 432b9c9322181a3644083e3c19b7e240d90659e7 by simonpj:"New syntax for GADT-style record declarations, and associated refactoring" discussed in Trac #3306. This patch removes 2 reduce/reduce conflicts in parser. Conflicting productions were: ``` gadt_constr -> con_list '::' sigtype gadt_constr -> oqtycon '{' fielddecls '}' '::' sigtype ``` Recursive inlining of `con_list` and `oqtycon` helped reveal the conflict: ``` gadt_constr -> '(' CONSYM ')' '::' sigtype gadt_constr -> '(' CONSYM ')' '{' fielddecls '}' '::' sigtype ``` between two types of GADT constructors (second form stands for deprecated syntax). Test Plan: `make fasttest`, one breakage TEST="records-fail" (parse error instead of typecheck error due to removal of deprecated syntax). Updated test. Reviewers: simonmar, bgamari, austin, simonpj Reviewed By: simonpj Subscribers: thomie, mpickering, trofi Differential Revision: https://phabricator.haskell.org/D1118 GHC Trac Issues: #3306 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/3306#comment:9> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

[GHC] #10711: Defining mapM_ in terms of traverse_ causes substantial blow-up in ByteCodeAsm
by GHC 04 Aug '15

04 Aug '15

#10711: Defining mapM_ in terms of traverse_ causes substantial blow-up in ByteCodeAsm -------------------------------------+------------------------------------- Reporter: bgamari | Owner: bgamari Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Compile-time Unknown/Multiple | performance bug Test Case: ghcirun004 | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- Phab:D924 has proposed that we redefine `mapM_`, currently, {{{ mapM_ = foldr ((>>) . f) (return ()) }}} as, {{{ mapM_ = foldr ((*>) . f) (return ()) = traverse_ }}} as part of the AMP proposal. However, this appears to have severe effects on the performance characteristics of the `Assembler` monad defined in `ByteCodeAsm`. In particular, the `mapM_` use in `ByteCodeAsm.assembleBCO` blows up severely, increasing the runtime of the `ghcirun004` testcase from 4 seconds to over 5 minutes. Intriguingly, defining `(*>) = (>>)` in `Assembler`'s `Applicative` instance (as done in Phab:D1097) restores reasonable runtime. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10711> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 13

[GHC] #10708: Rejection of constant functions defined using conditional pattern matching
by GHC 03 Aug '15

03 Aug '15

#10708: Rejection of constant functions defined using conditional pattern matching -------------------------------------+------------------------------------- Reporter: | Owner: HubertGaravel | Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.1 Keywords: | Operating System: Linux Architecture: x86 | Type of failure: GHC rejects | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- Hi, The Haskell program below is rejected, although it is correct. GHC gives the following error message: {{{ HASKELL/tricky.hs:12:1: Multiple declarations of `Main.d3' Declared at: HASKELL/tricky.hs:10:1 HASKELL/tricky.hs:11:1 HASKELL/tricky.hs:12:1 }}} I agree that the definition of d3 is tricky and useless, but it seems to be valid. This example was exctracted from Haskell code automatically produced by an in-house code generator. {{{#!hs data Nat = D0 | Succ Nat deriving (Show, Eq, Ord) d1 :: Nat d2 :: Nat d3 :: Nat d1 = (Succ D0) d2 | D0 == D0 = D0 d3 | D0 /= D0 = D0 d3 | (Succ D0) == D0 = D0 d3 | (Succ D0) /= D0 = (Succ D0) main = do print d1 print d2 print d3 }}} Best regards Hubert -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10708> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

Re: [GHC] #3699: Wildcards in type functions
by GHC 03 Aug '15

03 Aug '15

#3699: Wildcards in type functions -------------------------------------+------------------------------------- Reporter: | Owner: msosn MartijnVanSteenbergen | Type: feature request | Status: closed Priority: low | Milestone: 7.12.1 Component: Compiler (Type | Version: 6.10.4 checker) | Resolution: fixed | Keywords: newcomer Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Revisions: Phab:D1092 -------------------------------------+------------------------------------- Changes (by bgamari): * status: new => closed * resolution: => fixed -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/3699#comment:25> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0