simonpj@microsoft.com [Wed, 10 Sep 2008 09:58:16 +0000 (09:58 +0000)]
Check the *right* set of type variables for escape!
I did the wrong checkSigTyVars, which (happily) triggered an ASSERT
failure. This should fix it.
simonpj@microsoft.com [Wed, 10 Sep 2008 08:51:21 +0000 (08:51 +0000)]
More refactoring of instance declarations (fixes Trac #2572)
In refactoring instance declarations I'd taken a short cut over
scoped type variables, but it wasn't right as #2572 shows.
Fixing it required a significant chunk of further refactoring,
alas. But it's done! Quite tidily as it turns out.
The main issue is that when typechecking a default method, we
need two sets of type variables in scope
class C a where
op :: forall b. ...
op = e
In 'e', *both* 'a' and 'b' are in scope. But the type of the
default method has a nested flavour
op :: forall a. C a => forall b. ....
and our normal scoping mechanisms don't bring 'b' into scope.
(And probably shouldn't.)
Solution (which is done for instance methods too) is to use
a local defintion, like this:
$dmop :: forall a. C a => forall b. ....
$dmop a d = let
op :: forall b. ...
op = e
in op
and now the scoping works out. I hope I have now see the
last of this code for a bit!
simonpj@microsoft.com [Wed, 10 Sep 2008 08:30:14 +0000 (08:30 +0000)]
Fix Trac #2581: inlining of record selectors
Bryan discovered that a non-trivial record selector (non-trivial in
the sense that it has to reconstruct the result value because of
UNPACK directives) weren't being inlined. The reason was that the
unfolding generated by MkId.mRecordSelId was never being optimised
*at all*, and hence looked big, and hence wasn't inlined.
(The out-of-line version *is* put into the code of the module
and *is* optimised, which made this bug pretty puzzling. But the
unfolding inside the record-selector-Id itself, which is a GlobalId
and hence does not get its inlining updated like LocalIds, was
big and fat.)
Solution: I wrote a very simple optimiser, CoreUnfold.simplOptExpr,
which does enough optimisation to solve this particular problem.
It's short, simple, and will be useful in other contexts.
simonpj@microsoft.com [Wed, 10 Sep 2008 08:23:13 +0000 (08:23 +0000)]
Fix Trac #2573; and explanatory comment
simonpj@microsoft.com [Wed, 10 Sep 2008 08:20:03 +0000 (08:20 +0000)]
Fix the zonking of HsWrappers
HsWrappers are horribly inconsistent at the moment. I intended that
WpLam, WpApp are for evidence abstraction/application
WpTyLam, WpTyApp are for type abstraction/application
But when we zonk (WpApp co), where co is a coercion variable, we
get a *coercion* not a coercion *variable*. So for now I'm making
it into a WpTyApp, which the desugarer handles perfectly well.
(I'd forgotten to zonk it properly at all; that is the bug that
this patch fixes.)
simonpj@microsoft.com [Wed, 10 Sep 2008 08:19:13 +0000 (08:19 +0000)]
Add newDictOcc, newDictOccs
simonpj@microsoft.com [Wed, 10 Sep 2008 08:18:15 +0000 (08:18 +0000)]
Layout and type synonyms only
simonpj@microsoft.com [Wed, 10 Sep 2008 08:15:28 +0000 (08:15 +0000)]
Robustify the setting of implied flags
When setting implied flags, do so recursively. So if -Xa implies -Xb,
and -Xb implies -Xc, we do the right thing.
I thought we needed this, but we don't. But it seems like a good idea
anyway.
simonpj@microsoft.com [Wed, 10 Sep 2008 08:15:10 +0000 (08:15 +0000)]
Rename a variable
simonpj@microsoft.com [Wed, 10 Sep 2008 06:50:30 +0000 (06:50 +0000)]
Comments only
simonpj@microsoft.com [Wed, 10 Sep 2008 07:35:29 +0000 (07:35 +0000)]
Simplify the type signature for tcPolyBinds
simonpj@microsoft.com [Tue, 9 Sep 2008 16:21:07 +0000 (16:21 +0000)]
Tell GHC which module nON_EXHAUSTIVE_GUARDS_ERROR_ID comes from
This one was missed when the error-ids all moved to Control.Exception.Base
(The nofib test 'fluid' showed it up.)
simonpj@microsoft.com [Tue, 9 Sep 2008 15:57:09 +0000 (15:57 +0000)]
Improve float-in somewhat
See Note [Floating in past a lambda group]. The new thing
here is that we previously were not floating in past a group
that had (\(a:*) \(x:State# a). e), with a type var *and* a
one-shot lambda.
The fix makes wave4main allocate 2% less, and doesn't change any other
nofib number, so it's not a big deal.
simonpj@microsoft.com [Tue, 9 Sep 2008 15:50:11 +0000 (15:50 +0000)]
Important performance wibble to callSiteInline (the n_vals_wanted > 0 thing)
See Note [Inlining in ArgCtxt]. This very small change gives quite a
big performance win. Just showing the bigger ones:
Program Size Allocs Runtime
--------------------------------------------------------------------------------
anna -0.7% -4.3% 0.15
cichelli -0.6% -6.4% 0.15
fulsom -0.4% -18.5% -8.1%
gcd -0.6% -12.0% 0.06
integer -0.6% -16.2% -8.4%
power -0.7% -19.3% -4.8%
--------------------------------------------------------------------------------
Min -0.7% -19.3% -15.7%
Max -0.1% +0.1% +5.7%
Geometric Mean -0.6% -1.9% -4.3%
The original change was to improve a case that Roman found (see test
eyeball/inline1) but that seems to work ok now anyway.
Simon Marlow [Tue, 9 Sep 2008 14:51:22 +0000 (14:51 +0000)]
when a memory leak is detected, report which blocks are unreachable
Simon Marlow [Tue, 9 Sep 2008 13:52:51 +0000 (13:52 +0000)]
Separate pruning from marking of spark pools
Fixes crash when using compacting GC in parallel programs
Simon Marlow [Tue, 9 Sep 2008 13:43:02 +0000 (13:43 +0000)]
un-disable warnings in some modules
Simon Marlow [Tue, 9 Sep 2008 13:42:47 +0000 (13:42 +0000)]
Fix warnings
Simon Marlow [Tue, 9 Sep 2008 13:37:48 +0000 (13:37 +0000)]
More sanity checking for the TSO write barrier
Check that all threads marked as dirty are really on the mutable list.
Simon Marlow [Tue, 9 Sep 2008 13:34:53 +0000 (13:34 +0000)]
Move checkGlobalTSOList() call from GarbageCollect() to checkSanity()
Simon Marlow [Tue, 9 Sep 2008 13:33:07 +0000 (13:33 +0000)]
small bugfix in traverseBlackHoleQueue()
Simon Marlow [Tue, 9 Sep 2008 13:32:23 +0000 (13:32 +0000)]
Fix race condition in wakeupThreadOnCapability() (#2574)
wakeupThreadOnCapbility() is used to signal another capability that
there is a thread waiting to be added to its run queue. It adds the
thread to the (locked) wakeup queue on the remote capability. In
order to do this, it has to modify the TSO's link field, which has a
write barrier. The write barrier might put the TSO on the mutable
list, and the bug was that it was using the mutable list of the
*target* capability, which we do not have exclusive access to. We
should be using the current Capabilty's mutable list in this case.
Simon Marlow [Tue, 9 Sep 2008 13:22:34 +0000 (13:22 +0000)]
Fix compacting GC bug: don't forget to thread the blackhole_queue
berthold@mathematik.uni-marburg.de [Tue, 19 Aug 2008 15:18:42 +0000 (15:18 +0000)]
Capability stopping when waiting for GC
berthold@mathematik.uni-marburg.de [Tue, 2 Sep 2008 16:13:13 +0000 (16:13 +0000)]
Scheduler code cleanup
This patch removes old code from the Schedule.c file.
I removed GRAN code for GranSim, a simulator for parallel Haskell execution with GpH model.
This code is inactive since ghc-4.x.
Code for PARALLEL_HASKELL has been partially removed. The remaining code is valid, but can
refer to nonexisting functionality in other files.
Simon Marlow [Mon, 8 Sep 2008 14:56:52 +0000 (14:56 +0000)]
Fix parallel GC bug (crash in concprog001(threaded2))
Two threads were trying to move the same TSO. I like this test, it
has caught plenty of bugs in the past.
Simon Marlow [Mon, 8 Sep 2008 14:43:48 +0000 (14:43 +0000)]
Make LOOKS_LIKE_{INFO,CLOSURE}_PTR into inline functions, instead of macros
The macros were duplicating their arguments, which was normally
harmless, but in the parallel GC was actually wrong and caused
spurious assertion failures.
Ian Lynagh [Mon, 8 Sep 2008 22:46:18 +0000 (22:46 +0000)]
Use test -f rather than test -e
Hopefully this will fix the SunOS builbot slave.
Ian Lynagh [Mon, 8 Sep 2008 18:48:29 +0000 (18:48 +0000)]
Make a pdf, rather than ps.gz, of teh ext-core docs
Simon Marlow [Mon, 8 Sep 2008 14:24:14 +0000 (14:24 +0000)]
FIX BUILD on non-Windows
Simon Marlow [Mon, 8 Sep 2008 10:28:55 +0000 (10:28 +0000)]
make this build with GHC 6.7+
Simon Marlow [Mon, 8 Sep 2008 10:28:43 +0000 (10:28 +0000)]
add (c) Lennart Augustsson (part of #740)
Simon Marlow [Wed, 3 Sep 2008 10:50:18 +0000 (10:50 +0000)]
sysErrorBelch: don't put an extra \n on Windows
Simon Marlow [Wed, 3 Sep 2008 10:49:51 +0000 (10:49 +0000)]
Windows: print an error message in addDLL
Also, look for libXXX.dll in addition to XXX.dll (see #1883, this
isn't really a proper fix, but it'll help in some cases).
And I tidied up the error message for a DLL load failure, though it's
still a bit of a mess because addDLL is supposed to return a (static)
string with the error message, but this isn't possible on Windows.
Ian Lynagh [Sun, 7 Sep 2008 18:52:32 +0000 (18:52 +0000)]
bindist fixes
Ian Lynagh [Sun, 7 Sep 2008 16:48:13 +0000 (16:48 +0000)]
Fix the ext-core doc in bindists
daniel@phasevelocity.org [Fri, 22 Aug 2008 06:21:36 +0000 (06:21 +0000)]
Add extern flag to avoid multiple symbol errors on Mac os X
Manuel M T Chakravarty [Sun, 7 Sep 2008 11:21:28 +0000 (11:21 +0000)]
Type families: new algorithm to solve equalities
- This adds the new equational solver based on the notion of normalised
equalities.
- The new algorithm is conceptually much simpler and will eventually enable us
to implement a fully integrated solver that solves equality and dictionary
constraints together.
- More details are at
<http://hackage.haskell.org/trac/ghc/wiki/TypeFunctionsSolving>
- The code is there, but it is not being used yet.
Ian Lynagh [Sat, 6 Sep 2008 18:49:12 +0000 (18:49 +0000)]
Don't build latex docs by default
"LATEX_DOCS = YES" enables them
Ian Lynagh [Sat, 6 Sep 2008 18:46:33 +0000 (18:46 +0000)]
Now that haddock is in-tree, build haddock docs by default
Ian Lynagh [Sat, 6 Sep 2008 15:55:19 +0000 (15:55 +0000)]
Add an explanation of the format of the packages file format
Ian Lynagh [Sat, 6 Sep 2008 14:25:46 +0000 (14:25 +0000)]
Install the stage 2 ghc package when installing; fixes trac #2567
Ian Lynagh [Sat, 6 Sep 2008 14:02:43 +0000 (14:02 +0000)]
Update the users guide to point at the in-tree core.ps.gz
It used to point to a file on haskell.org, which didn't necessarily
describe the right version of core.
Ian Lynagh [Sat, 6 Sep 2008 13:56:16 +0000 (13:56 +0000)]
Compress the ext-core docs
Ian Lynagh [Sat, 6 Sep 2008 13:51:52 +0000 (13:51 +0000)]
Install the ext-core docs
Ian Lynagh [Sat, 6 Sep 2008 12:14:15 +0000 (12:14 +0000)]
Make ext-core when making all in docs
Ian Lynagh [Fri, 5 Sep 2008 18:25:53 +0000 (18:25 +0000)]
Remove the haddock detection
Now that we build haddock in the GHC build, we don't need to look for it
on the path.
Ian Lynagh [Fri, 5 Sep 2008 15:50:10 +0000 (15:50 +0000)]
Add a comment to utils/Makefile
simonpj@microsoft.com [Fri, 5 Sep 2008 17:26:54 +0000 (17:26 +0000)]
Fix up the instance-declaration re-engineering story
This patch deals with a rather complicated situation involving
overlapping instances. It's all explained in the commments
Note [Subtle interaction of recursion and overlap]
The absence of this case make DoCon and regex-base fail with
an error about overlapping instances. Now they work properly
again.
simonpj@microsoft.com [Fri, 5 Sep 2008 17:25:53 +0000 (17:25 +0000)]
Comments only
simonpj@microsoft.com [Fri, 5 Sep 2008 17:25:07 +0000 (17:25 +0000)]
Minor refactoring to share InstEnv.instanceBindFun
simonpj@microsoft.com [Fri, 5 Sep 2008 17:24:24 +0000 (17:24 +0000)]
Remove unused constructors AvoidMe, WildCard
simonpj@microsoft.com [Fri, 5 Sep 2008 17:21:12 +0000 (17:21 +0000)]
More specialiser wibbles
Several things
* Only gather call details for local things, not imported ones
* When making auxiliary dictionary bindings in specDefn, remember
to give them an unfolding. Otherwise we don't gather call details
for functions applied to them.
simonpj@microsoft.com [Fri, 5 Sep 2008 17:18:14 +0000 (17:18 +0000)]
Retain unfoldings even with SimplGently
When binding x = e, we now attach an unfolding to 'x' even if
it won't be used because SimplGently is on.
Reason: the specialiser runs right after SimplGently, and it (now)
only gathers call information for calls whose dictionary arguments are
"interesting" -- i.e. have an unfolding of some kind.
simonpj@microsoft.com [Fri, 5 Sep 2008 17:17:31 +0000 (17:17 +0000)]
Comments only
simonpj@microsoft.com [Fri, 5 Sep 2008 17:16:39 +0000 (17:16 +0000)]
Better debug panic messages in applyTys
simonpj@microsoft.com [Fri, 5 Sep 2008 17:14:58 +0000 (17:14 +0000)]
Make ASSERT2 mention msg even when debug is off (avoid warnings)
Ian Lynagh [Fri, 5 Sep 2008 12:21:09 +0000 (12:21 +0000)]
Move the "This is a generated file" to the top of GHC.Prim
Ian Lynagh [Fri, 5 Sep 2008 12:14:43 +0000 (12:14 +0000)]
Make genprimopcode generate code that haddock 2 understands
Mainly this means adding a binding for all the exports, e.g. as well as
gtAddr# :: Addr# -> Addr# -> Bool
we also generate
gtAddr# = let x = x in x
Ian Lynagh [Thu, 4 Sep 2008 18:50:42 +0000 (18:50 +0000)]
Define _BSD_SOURCE in Stg.h
This means S_ISSOCK gets defined on Linux
Ian Lynagh [Thu, 4 Sep 2008 15:49:31 +0000 (15:49 +0000)]
Add a note explaining a couple of seq's
Ross Paterson [Thu, 4 Sep 2008 16:48:16 +0000 (16:48 +0000)]
fix message for missing packages
An incorrect libraries/ prefix was being added.
dias@eecs.harvard.edu [Thu, 4 Sep 2008 13:51:55 +0000 (13:51 +0000)]
add assertion to check that UniqFM is only passed "positive" uniques
The insertion code in UniqFM fails if a unique key
produces a negative FastInt. I've added an assertion to check
that each insertion uses a positive Unique.
Where do the negative uniques come from? Both Simom M and
I have run into this problem when computing hashes for data structures.
In both cases, we have avoided the problem by ensuring that
the hashes remain positive.
Ian Lynagh [Thu, 4 Sep 2008 12:54:12 +0000 (12:54 +0000)]
Add darcs-all usage info
Ian Lynagh [Wed, 3 Sep 2008 22:35:08 +0000 (22:35 +0000)]
Re-merge concurrent,timeout,unique,st,getopt into base
Ian Lynagh [Wed, 3 Sep 2008 20:41:01 +0000 (20:41 +0000)]
Fix ifBuildable
Required libraries now have 3 fields in the packages file, not 2
simonpj@microsoft.com [Wed, 3 Sep 2008 12:02:28 +0000 (12:02 +0000)]
Major change in compilation of instance declarations (fix Trac #955, #2328)
This patch makes an important change to the way that dictionary
functions are handled. Before, they were unconditionally marked
INLIINE, but all the code written by the user in the instance
was inside that unconditionally-inlined function. Result: massive
code bloat in programs that use complicated instances.
This patch make instances behave rather as if all the methods
were written in separate definitions. That dramatically reduces
bloat. The new plan is described in TcInstDcls
Note [How instance declarations are translated]
Everything validates. The major code-bloat bug is squashed: in particular
DoCon is fine now (Trac #2328) and I believe that #955 is also better.
Nofib results:
Binary sizes
-1 s.d. +2.5%
+1 s.d. +3.1%
Average +2.8%
Allocations
-1 s.d. -6.4%
+1 s.d. +2.5%
Average -2.0%
Note that 2% improvement. Some programs improve by 20% (rewrite)!
Two get slightly worse: pic (2.1%), and gameteb (3.2%), but all others
improve or stay the same.
I am not absolutely 100% certain that all the corners are correct; for
example, when default methods are marked INLINE, are they inlined? But
overall it's better.
It's nice that the patch also removes a lot of code. I deleted some
out of date comments, but there's something like 100 fewer lines of
code in the new version! (In the line counts below, there are a lot
of new comments.)
simonpj@microsoft.com [Wed, 3 Sep 2008 11:56:29 +0000 (11:56 +0000)]
Improved specialisation of recursive groups
This patch significantly improves the way in which recursive groups
are specialised. This turns out ot be very important when specilising
the bindings that (now) emerge from instance declarations.
Consider
let rec { f x = ...g x'...
; g y = ...f y'.... }
in f 'a'
Here we specialise 'f' at Char; but that is very likely to lead to
a specialisation of 'g' at Char. We must do the latter, else the
whole point of specialisation is lost. This was not happening before.
The whole thing is desribed in
Note [Specialising a recursive group]
Simon
simonpj@microsoft.com [Tue, 2 Sep 2008 10:30:43 +0000 (10:30 +0000)]
Cope with recent versions of Perl that lack the $* feature
Simon Marlow [Tue, 2 Sep 2008 12:45:04 +0000 (12:45 +0000)]
FIX #2554: GHC.TopHandler.{runIO,runNonIO} should be GC roots
Simon Marlow [Mon, 1 Sep 2008 13:26:14 +0000 (13:26 +0000)]
Check the modification times of libraries in --make link step
When linking in --make we check the modification time of the
executable against the modification time of the object files, and only
re-link if any object file is newer. However, we should also check
the modification times of packages, since the recompilation checker
also tracks dependencies in packages.
In a GHC build this means that if you recompile stage2 and don't
manage to change any fingerpints, we won't recompile Main but we'll
still re-link it.
Ian Lynagh [Sun, 31 Aug 2008 15:24:28 +0000 (15:24 +0000)]
Make the dynlibs, ghci and ncg flags manual too
Ian Lynagh [Sat, 30 Aug 2008 23:54:30 +0000 (23:54 +0000)]
Squash a space leak
Ian Lynagh [Sat, 30 Aug 2008 19:26:19 +0000 (19:26 +0000)]
Fix a space leak
Ian Lynagh [Sat, 30 Aug 2008 19:25:26 +0000 (19:25 +0000)]
fix a space leak
Ian Lynagh [Sun, 31 Aug 2008 15:21:03 +0000 (15:21 +0000)]
Make ghc.cabal's editline flag manual, to simplify the Makefile
Judah Jacobson [Sat, 30 Aug 2008 21:14:29 +0000 (21:14 +0000)]
Don't build ghc in stage1 with editline (since we already don't build ghci). Fixes #2557.
Ian Lynagh [Sat, 30 Aug 2008 19:13:05 +0000 (19:13 +0000)]
If we want a profiled GHC then use -auto-all when building the GHC lib
Ian Lynagh [Sat, 30 Aug 2008 16:16:37 +0000 (16:16 +0000)]
Fix building the GHC API with profiling
Ian Lynagh [Sat, 30 Aug 2008 12:30:32 +0000 (12:30 +0000)]
Actually, that's use a separate haddock repo for now
Everything will be changing in a couple of weeks anyway, so let's
just keep it simple for the time being.
Ian Lynagh [Fri, 29 Aug 2008 15:21:41 +0000 (15:21 +0000)]
Always build docs/{ext-core,storage-mgt}, even when making a bindist
kili@outback.escape.de [Thu, 28 Aug 2008 15:03:31 +0000 (15:03 +0000)]
Fix linkage on OpenBSD.
Ian Lynagh [Fri, 29 Aug 2008 13:00:19 +0000 (13:00 +0000)]
push-all needs to skip repos, as we don't know how to push to them
Alternatives would be to have a GHC haddock repo (which I believe David
is keen to avoid), or to put the SSH URL in the packages file too (which
would probably mean moving to a more structured file format).
Ian Lynagh [Thu, 28 Aug 2008 22:17:01 +0000 (22:17 +0000)]
Remove install-inplace directories when cleaning
Ian Lynagh [Thu, 28 Aug 2008 19:15:12 +0000 (19:15 +0000)]
Use the right ghc-pkg when installing the rts package
Ian Lynagh [Thu, 28 Aug 2008 18:24:47 +0000 (18:24 +0000)]
When installing, we can't rely on the ghc-pkg we just installed to work
If DESTDIR is defined, then on *nix machines ghc-pkg is a script that
doesn't point to the right path. Therefore we use the ghc-pkg binary
directly.
Ian Lynagh [Thu, 28 Aug 2008 18:21:48 +0000 (18:21 +0000)]
Tweaks for installs into a DESTDIR
Ian Lynagh [Thu, 28 Aug 2008 15:14:29 +0000 (15:14 +0000)]
Handle the html target in compiler/ and ghc/
Ian Lynagh [Thu, 28 Aug 2008 14:23:20 +0000 (14:23 +0000)]
Turn on optimisation for stage2 when validating
This brings my validate down from 22mins to 16 mins.
Compiling stage2 takes longer, but we gain a faster haddock, faster
running of the tests, and faster building of the utils to be installed.
Ian Lynagh [Thu, 28 Aug 2008 14:22:10 +0000 (14:22 +0000)]
haddock 2 doc fixes
Ian Lynagh [Thu, 28 Aug 2008 13:23:16 +0000 (13:23 +0000)]
Remove an erroneous comma from the Makefile
Ian Lynagh [Thu, 28 Aug 2008 12:50:18 +0000 (12:50 +0000)]
Disambiguate some names in haddock comments
Ian Lynagh [Thu, 28 Aug 2008 12:48:14 +0000 (12:48 +0000)]
Update the build system to handle building and using haddock2
One side-effect of this is that we need to build the install-utils with
stage2 rather than stage1 as we need the ghc package.
Ian Lynagh [Thu, 28 Aug 2008 12:47:21 +0000 (12:47 +0000)]
Fix a haddock2 failure
Ian Lynagh [Thu, 28 Aug 2008 12:46:08 +0000 (12:46 +0000)]
Don't check that we have found haddock
We now use an in-tree haddock instead
Ian Lynagh [Wed, 27 Aug 2008 18:50:24 +0000 (18:50 +0000)]
Don't fail if libraries/bootstrapping already exists
Ian Lynagh [Wed, 27 Aug 2008 18:27:31 +0000 (18:27 +0000)]
When haddocking the libraries, tell Cabal where haddock is
We can't do it when configuring, because we haven't built haddock yet
Ian Lynagh [Thu, 14 Aug 2008 11:22:21 +0000 (11:22 +0000)]
Build haddock with the stage1 compiler
Ian Lynagh [Wed, 13 Aug 2008 22:01:58 +0000 (22:01 +0000)]
Teach darcs-all about absolute repos, and add haddock to packages
simonpj@microsoft.com [Thu, 28 Aug 2008 12:52:11 +0000 (12:52 +0000)]
Improve documentation of stolen syntax
This patch adds a section that summarises what syntax is stolen by
which flags. The section is at the end of the "syntactic extensions"
section of the manual.