| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
| |
And mention more prominently that `-DDEBUG` and tests that call
`compiler_stats_num_fields` don't play well together
Reviewed By: austin
Differential Revision: https://phabricator.haskell.org/D799
|
|
|
|
| |
Signed-off-by: Austin Seipp <austin@well-typed.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This re-implements the code generation for case expressions at the Stg →
Cmm level, both for data type cases as well as for integral literal
cases. (Cases on float are still treated as before).
The goal is to allow for fancier strategies in implementing them, for a
cleaner separation of the strategy from the gritty details of Cmm, and
to run this later than the Common Block Optimization, allowing for one
way to attack #10124. The new module CmmSwitch contains a number of
notes explaining this changes. For example, it creates larger
consecutive jump tables than the previous code, if possible.
nofib shows little significant overall improvement of runtime. The
rather large wobbling comes from changes in the code block order
(see #8082, not much we can do about it). But the decrease in code size
alone makes this worthwhile.
```
Program Size Allocs Runtime Elapsed TotalMem
Min -1.8% 0.0% -6.1% -6.1% -2.9%
Max -0.7% +0.0% +5.6% +5.7% +7.8%
Geometric Mean -1.4% -0.0% -0.3% -0.3% +0.0%
```
Compilation time increases slightly:
```
-1 s.d. ----- -2.0%
+1 s.d. ----- +2.5%
Average ----- +0.3%
```
The test case T783 regresses a lot, but it is the only one exhibiting
any regression. The cause is the changed order of branches in an
if-then-else tree, which makes the hoople data flow analysis traverse
the blocks in a suboptimal order. Reverting that gets rid of this
regression, but has a consistent, if only very small (+0.2%), negative
effect on runtime. So I conclude that this test is an extreme outlier
and no reason to change the code.
Differential Revision: https://phabricator.haskell.org/D720
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This implements (roughly) the plan put forward in comment:14:ticket:7788,
fixing #7788, #8550, #9554, #10139, and addressing concerns raised in #10079.
There are some regressions w.r.t. GHC 7.8, but only with pathological type
families (like F a = F a). This also (hopefully -- don't have a test case)
fixes #10158. Unsolved problems include #10184 and #10185, which are both
known deficiencies of the approach used here.
As part of this change, the plumbing around detecting infinite loops has
changed. Instead of -fcontext-stack and -ftype-function-depth, we now have
one combined -freduction-depth parameter. Setting it to 0 disbales the
check, which is now the recommended way to get (terminating) code to
typecheck in releases. (The number of reduction steps may well change between
minor GHC releases!)
This commit also introduces a new IntWithInf type in BasicTypes
that represents an integer+infinity. This type is used in a few
places throughout the code.
Tests in
indexed-types/should_fail/T7788
indexed-types/should_fail/T8550
indexed-types/should_fail/T9554
indexed-types/should_compile/T10079
indexed-types/should_compile/T10139
typecheck/should_compile/T10184 (expected broken)
typecheck/should_compile/T10185 (expected broken)
This commit also changes performance testsuite numbers, for the better.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary: Added packageName to GHC.Generics.Datatype class definition
Reviewers: hvr, dreixel, austin
Reviewed By: dreixel, austin
Subscribers: thomie
Differential Revision: https://phabricator.haskell.org/D631
GHC Trac Issues: #10030
|
|
|
|
|
|
| |
Updates the performance numbers for 32-bit windows. I tried to do
this before with 387f1d1ec, but accidentally included some unrelated
changes.
|
|
|
|
|
|
|
| |
This reverts commit 387f1d1ec334788c3e891e9304d427bc804998f4.
Bizarrely, this commit accidentally added libraries/haskell98.
So I'll revert it entirely and start again.
|
|
|
|
|
| |
to have less annoying false negatives on
http://perf.ghc.haskell.org/#graph/testsuite/unexpected%20stats
|
|
|
|
| |
Less for GHC, more for Haddock
|
|
|
|
| |
Signed-off-by: Austin Seipp <austin@well-typed.com>
|
|
|
|
|
|
|
|
| |
As with stripTicksTop, this is because we often need the stripped
expression but not the ticks (at least not right away). This makes a big
difference for CSE, see #9961.
Signed-off-by: Austin Seipp <austin@well-typed.com>
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Suppose we have a handful H of entries in a TrieMap, each with a very large
key, size K. If you fold over such a TrieMap you'd expect time O(H). That would
certainly be true of an association list! But with TrieMap we actually have to
navigate down a long singleton structure to get to the elements, so it takes
time O(K*H). The point of a TrieMap is that you need to navigate to the point
where only one key remains, and then things should be fast.
This is a starting point: we can improve the patch by generalizing the
singleton constructor so it applies to CoercionMap and CoreMap; I'll do this
in a later commit.
Summary: Signed-off-by: Edward Z. Yang <ezyang@cs.stanford.edu>
Test Plan: validate
Reviewers: simonpj, austin
Subscribers: carter, thomie
Differential Revision: https://phabricator.haskell.org/D606
GHC Trac Issues: #9960
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The idea was promted by Trac #9939, but it was Christmas, so I did
some recreational programming that went much further.
The idea is to warn when a constraint in a user-supplied context is
redundant. Everything is described in detail in
Note [Tracking redundant constraints]
in TcSimplify.
Main changes:
* The new ic_status field in an implication, of type ImplicStatus.
It replaces ic_insol, and includes information about redundant
constraints.
* New function TcSimplify.setImplicationStatus sets the ic_status.
* TcSigInfo has sig_report_redundant field to say whenther a
redundant constraint should be reported; and similarly
the FunSigCtxt constructor of UserTypeCtxt
* EvBinds has a field eb_is_given, to record whether it is a given
or wanted binding. Some consequential chagnes to creating an evidence
binding (so that we record whether it is given or wanted).
* AbsBinds field abs_ev_binds is now a *list* of TcEvBiinds;
see Note [Typechecking plan for instance declarations] in
TcInstDcls
* Some significant changes to the type checking of instance
declarations; Note [Typechecking plan for instance declarations]
in TcInstDcls.
* I found that TcErrors.relevantBindings was failing to zonk the
origin of the constraint it was looking at, and hence failing to
find some relevant bindings. Easy to fix, and orthogonal to
everything else, but hard to disentangle.
Some minor refactorig:
* TcMType.newSimpleWanteds moves to Inst, renamed as newWanteds
* TcClassDcl and TcInstDcls now have their own code for typechecking
a method body, rather than sharing a single function. The shared
function (ws TcClassDcl.tcInstanceMethodBody) didn't have much code
and the differences were growing confusing.
* Add new function TcRnMonad.pushLevelAndCaptureConstraints, and
use it
* Add new function Bag.catBagMaybes, and use it in TcSimplify
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The purpose of silent superclass parameters was to solve the
awkward problem of superclass dictinaries being bound to bottom.
See THE PROBLEM in Note [Recursive superclasses] in TcInstDcls
Although the silent-superclass idea worked,
* It had non-local consequences, and had effects even in Haddock,
where we had to discard silent parameters before displaying
instance declarations
* It had unexpected peformance costs, shown up by Trac #3064 and its
test case. In monad-transformer code, when constructing a Monad
dictionary you had to pass an Applicative dictionary; and to
construct that you neede a Functor dictionary. Yet these extra
dictionaries were often never used. (All this got much worse when
we added Applicative as a superclass of Monad.) Test T3064
compiled *far* faster after silent superclasses were eliminated.
* It introduced new bugs. For example SilentParametersOverlapping,
T5051, and T7862, all failed to compile because of instance overlap
directly because of the silent-superclass trick.
So this patch takes a new approach, which I worked out with Dimitrios
in the closing hours before Christmas. It is described in detail
in THE PROBLEM in Note [Recursive superclasses] in TcInstDcls.
Seems to work great!
Quite a bit of knock-on effect
* The main implementation work is in tcSuperClasses in TcInstDcls
Everything else is fall-out
* IdInfo.DFunId no longer needs its n-silent argument
* Ditto IDFunId in IfaceSyn
* Hence interface file format changes
* Now that DFunIds do not have silent superclass parameters, printing
out instance declarations is simpler. There is tiny knock-on effect
in Haddock, so that submodule is updated
* I realised that when computing the "size of a dictionary type"
in TcValidity.sizePred, we should be rather conservative about
type functions, which can arbitrarily increase the size of a type.
Hence the new datatype TypeSize, which has a TSBig constructor for
"arbitrarily big".
* instDFunType moves from TcSMonad to Inst
* Interestingly, CmmNode and CmmExpr both now need a non-silent
(Ord r) in a couple of instance declarations. These were previously
silent but must now be explicit.
* Quite a bit of wibbling in error messages
|
| |
|
|
|
|
|
| |
This has a demonstrated 2x speed boost on the T9872{a,b,c} tests.
(#9872)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit fixes some performance regressions introduced by 0cc47eb,
adding more `Coercible` magic to the solver. See Note
[flatten_many performance] in TcFlatten for more info.
The improvements do not quite restore the old numbers. Given that
the solver is really more involved now, I am accepting this regression.
The way forward (I believe) would be to have *two* flatteners: one
that deals only with nominal equalities and thus never checks roles,
and the more general one. A nice design of keeping this performant
without duplicating code eludes me, but someone else is welcome
to take a stab.
|
|
|
|
| |
The actual gitlink update got lost in 0c9c2d899e63b810e7ab6b486f7244826b4a2e33
|
| |
|
|
|
|
|
|
|
|
|
| |
The previous attempt failed, but hopefully this one succeeds...
This also updates the perf-numbers for `haddock.base` and `haddock.Cabal`
NB: this switches from `ghc-head` to `master` branch temporarily
until GHC 7.10 has been properly branched off.
|
|
|
|
|
|
|
|
| |
This reverts commit 7f634320a2c39f4f81f631deb844acbaebaced66 again for now as
it causes validate's bindist phase to fail with
haddock: internal error: .../install dir/lib/ghc-7.9.20141214/html:
getDirectoryContents: does not exist (No such file or directory)
|
|
|
|
|
|
|
| |
This also updates the perf-numbers for `haddock.base` and `haddock.Cabal`
NB: this switches from `ghc-head` to `master` branch temporarily
until GHC 7.10 has been properly branched off.
|
|
|
|
| |
Things get faster, except T5030
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In particular this patch
- Accepts that rewriting with the inert CTyEqCans should be done recursively
(hence removing the Bool result from flattenTyVarOuter)
- Refines the kick-out criterion, in paticular to avoid kick-out of (a -f-> ty)
when f cannot rewrite f. This is true of Wanteds and hence reduces kick-outs
of Wanteds, perhaps by a lot
This stuff is not fully documented because the details are still settling, but
it's looking good.
(And it validates.)
This patch includes the testsuite wibbles. perf/compiler/T5030 and
T5837 both improve in bytes-allocated (by 11% and 13% resp), which is
good. I'm not sure which of today's short series of patches is
responsible, nor do I mind much. (One could find out if necessary.)
|
| |
|
|
|
|
|
|
|
|
| |
This should silence the
perf/haddock haddock.compiler [stat not good enough] (normal)
test-failure...
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is done as a separate `integer-gmp2` backend library because it
turned out to become a complete rewrite from scratch.
Due to the different (over)allocation scheme and potentially different
accounting (via the new `{shrink,resize}MutableByteArray#` primitives),
some of the nofib benchmarks actually results in increased allocation
numbers (but not necessarily an increase in runtime!). I believe the
allocation numbers could improve if `{resize,shrink}MutableByteArray#`
could be optimised to reallocate in-place more efficiently.
Here are the more apparent changes in the latest nofib comparision
between `integer-gmp` and `integer-gmp2`:
------------------------------------------------------------------
Program Size Allocs Runtime Elapsed TotalMem
------------------------------------------------------------------
...
bernouilli +1.6% +15.3% 0.132 0.132 0.0%
...
cryptarithm1 -2.2% 0.0% -9.7% -9.7% 0.0%
...
fasta -0.7% -0.0% +10.9% +10.9% 0.0%
...
kahan +0.6% +38.9% 0.169 0.169 0.0%
...
lcss -0.7% -0.0% -6.4% -6.4% 0.0%
...
mandel +1.6% +33.6% 0.049 0.049 0.0%
...
pidigits +0.8% +8.5% +3.9% +3.9% 0.0%
power +1.4% -23.8% -18.6% -18.6% -16.7%
...
primetest +1.3% +50.1% 0.085 0.085 0.0%
...
rsa +1.6% +53.4% 0.026 0.026 0.0%
...
scs +1.2% +6.6% +6.5% +6.6% +14.3%
...
symalg +1.0% +9.5% 0.010 0.010 0.0%
...
transform -0.6% -0.0% -5.9% -5.9% 0.0%
...
------------------------------------------------------------------
Min -2.3% -23.8% -18.6% -18.6% -16.7%
Max +1.6% +53.4% +10.9% +10.9% +14.3%
Geometric Mean -0.3% +1.9% -0.8% -0.8% +0.0%
(see P35 / https://phabricator.haskell.org/P35 for full report)
By default, `INTEGER_LIBRARY=integer-gmp2` is active now, which results
in the package `integer-gmp-1.0.0.0` being registered in the package db.
The previous `integer-gmp-0.5.1.0` can be restored by setting
`INTEGER_LIBRARY=integer-gmp` (but will probably be removed altogether
for GHC 7.12). In-tree GMP support has been stolen from the old
`integer-gmp` (while unpatching the custom memory-allocators, as well as
forcing `-fPIC`)
A minor hack to `ghc-cabal` was necessary in order to support two different
`integer-gmp` packages (in different folders) with the same package key.
There will be a couple of follow-up commits re-implementing some features
that were dropped to keep D82 minimal, as well as further
clean-ups/improvements.
More information can be found via #9281 and
https://ghc.haskell.org/trac/ghc/wiki/Design/IntegerGmp2
Reviewed By: austin, rwbarton, simonmar
Differential Revision: https://phabricator.haskell.org/D82
|
|
|
|
|
|
| |
This test seems too close to a tipping point (and thus too sensitive to the
build-env used) and T4801's `max_bytes_used` was disabled as well some time
ago for a similiar reason.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Generally clean up things relating to Applicative and Monad in `GHC.Base`
and `Control.Applicative` to make `Applicative` feel like a bit more of a
first-class citizen rather than just playing second fiddle to `Monad`. Use
`coerce` and GND to improve performance and clarity.
Change the default definition of `(*>)` to use `(<$)`, in case the
`Functor` instance optimizes that.
Moreover, some manually written instances are made into compiler-derived
instances.
Finally, this also adds a few AMP-related laws to the `Applicative` docstring.
NOTE: These changes result in a 13% decrease in allocation for T9020
Reviewed By: ekmett, hvr
Differential Revision: https://phabricator.haskell.org/D432
|
| |
|
|
|
|
|
|
|
|
| |
See Note [Lazy flattening] in TcFlatten.
Lazy flattening was an apparently good idea which actually made
the type inference engine go a LOTS slower in T3064. So I switched
it off again.
|
| |
|
|
|
|
|
| |
This disables T3064 temporarily as it puts a strain on buildbots
during validation exhausting all available memory.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This is a deliberate choice, to simplify code, invariants, and I think
performance in typical cases. The "loopy givens" case is situations like
[G] a ~ TF (a, Int)
where TF is a type function with TF (a,b) = (TF a, TF b).
See Note [An alternative story for the inert substitution] in TcFlatten.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This gets rid of all hand-unboxing in `GHC.List` and moves `Foldable`
requirements from `Data.OldList` into `GHC.List` (preparatory work for
addressing #9716). Specifically, this moves the definition of
`maximum`, `minimum`, `foldl'`, `foldl1`, `foldl1'`, `sum`, and
`product` into `GHC.List` (which now needs to import `GHC.Num`)
Make `take`, `drop`, `length`, and `!!` generally saner (see also #9510)
Performance overall seems minimally affected. Some things go up; some
things go down; nothing moves horribly much. The code is much easier to
read.
Differential Revision: https://phabricator.haskell.org/D380
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
T3064 is deactivated for now because it's currently too volatile
and causes too much noise in Phabricator's CI
C.f. 4805abf413c02a2ed1af4fbeca2476590e984e37
Reviewers: austin
Subscribers: thomie, carter, ezyang, simonmar
Differential Revision: https://phabricator.haskell.org/D381
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
This is a fixup of https://phabricator.haskell.org/D233
The only difference is in findTFiles (first commit), which
previously broke Windows runner; now I translated literally
instead attempting to improve it, and checked it works.
Test Plan:
I used validate under 2,3 on Linux and under 2 on msys2.
On Windows I've seen a large number of failures, but they don't
seem to be connected with the patch.
Reviewers: hvr, simonmar, thomie, austin
Reviewed By: austin
Subscribers: thomie, carter, ezyang, simonmar
Differential Revision: https://phabricator.haskell.org/D310
GHC Trac Issues: #9184
|
|
|
|
|
| |
which I just put it in comments. Thanks Yuras for noticing, and sorry
for the noise.
|