| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
They are readily derivable from other fields, so this is more
efficient, and less error prone.
Fixes #18494
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Specifically:
#13253 exponential inlining
#10421 ditto
#18140 strict constructors
#18282 another nested-function call case
This patch makes one really significant changes: change the way that
mkDupableCont handles StrictArg. The details are explained in
GHC.Core.Opt.Simplify Note [Duplicating StrictArg].
Specific changes
* In mkDupableCont, when making auxiliary bindings for the other arguments
of a call, add extra plumbing so that we don't forget the demand on them.
Otherwise we haev to wait for another round of strictness analysis. But
actually all the info is to hand. This change affects:
- Make the strictness list in ArgInfo be [Demand] instead of [Bool],
and rename it to ai_dmds.
- Add as_dmd to ValArg
- Simplify.makeTrivial takes a Demand
- mkDupableContWithDmds takes a [Demand]
There are a number of other small changes
1. For Ids that are used at most once in each branch of a case, make
the occurrence analyser record the total number of syntactic
occurrences. Previously we recorded just OneBranch or
MultipleBranches.
I thought this was going to be useful, but I ended up barely
using it; see Note [Note [Suppress exponential blowup] in
GHC.Core.Opt.Simplify.Utils
Actual changes:
* See the occ_n_br field of OneOcc.
* postInlineUnconditionally
2. I found a small perf buglet in SetLevels; see the new
function GHC.Core.Opt.SetLevels.hasFreeJoin
3. Remove the sc_cci field of StrictArg. I found I could get
its information from the sc_fun field instead. Less to get
wrong!
4. In ArgInfo, arrange that ai_dmds and ai_discs have a simpler
invariant: they line up with the value arguments beyond ai_args
This allowed a bit of nice refactoring; see isStrictArgInfo,
lazyArgcontext, strictArgContext
There is virtually no difference in nofib. (The runtime numbers
are bogus -- I tried a few manually.)
Program Size Allocs Runtime Elapsed TotalMem
--------------------------------------------------------------------------------
fft +0.0% -2.0% -48.3% -49.4% 0.0%
multiplier +0.0% -2.2% -50.3% -50.9% 0.0%
--------------------------------------------------------------------------------
Min -0.4% -2.2% -59.2% -60.4% 0.0%
Max +0.0% +0.1% +3.3% +4.9% 0.0%
Geometric Mean +0.0% -0.0% -33.2% -34.3% -0.0%
Test T18282 is an existing example of these deeply-nested strict calls.
We get a big decrease in compile time (-85%) because so much less
inlining takes place.
Metric Decrease:
T18282
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch eta-expands the Simplifier's monad, using the method
explained in GHC.Core.Unify Note [The one-shot state monad trick].
It's part of the exta-expansion programme in #18202.
It's a tiny patch, but is worth a 1-2% reduction in bytes-allocated
by the compiler. Here's the list, based on the compiler-performance
tests in perf/compiler:
Reduction in bytes allocated
T10858(normal) -0.7%
T12425(optasm) -1.3%
T13056(optasm) -1.8%
T14683(normal) -1.1%
T15164(normal) -1.3%
T15630(normal) -1.4%
T17516(normal) -2.3%
T18282(normal) -1.6%
T18304(normal) -0.8%
T1969(normal) -0.6%
T4801(normal) -0.8%
T5321FD(normal) -0.7%
T5321Fun(normal) -0.5%
T5642(normal) -0.9%
T6048(optasm) -1.1%
T9020(optasm) -2.7%
T9233(normal) -0.7%
T9675(optasm) -0.5%
T9961(normal) -2.9%
WWRec(normal) -1.2%
Metric Decrease:
T12425
T9020
T9961
|
|
|
|
|
|
| |
* Create a dedicated production for type operators
* Create a dedicated type for the UNPACK pragma
* Remove an outdated part of Note [Parsing data constructors is hard]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Before this patch, NegativeLiterals used to parse x-1 as x (-1).
This may not be what the user expects, and now it is fixed:
x-1 is parsed as (-) x 1.
We achieve this by the following requirement:
* When lexing a negative literal,
it must not be preceded by a 'closing token'.
This also applies to unboxed literals, e.g. -1#.
See GHC Proposal #229 for the definition of a closing token.
A nice consequence of this change is that -XNegativeLiterals becomes a
subset of -XLexicalNegation. In other words, enabling both of those
extensions has the same effect as enabling -XLexicalNegation alone.
|
|
|
|
|
| |
Adds commentary on the rationale for the changes made in merge request
!3689.
|
|
|
|
|
|
|
| |
Previously it was in ghc-boot so that ghc-pkg could use it. However it
wasn't necessary because ghc-pkg only uses a subset of it: reading
target arch and OS from the settings file. This is now done via
GHC.Platform.ArchOS (was called PlatformMini before).
|
|
|
|
|
|
|
|
|
|
| |
Platform constant wrappers took a DynFlags parameter, hence implicitly
used the target platform constants. We removed them to allow support
for several platforms at once (#14335) and to avoid having to pass
the full DynFlags to every function (#17957).
Metric Decrease:
T4801
|
| |
|
| |
|
| |
|
|
|
|
|
| |
We don't need to pass the whole DynFlags, just pass the logging
function, if any.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Change `Located X` usage to `XRec pass X`
This increases the scope of the LPat experiment to almost all of GHC.
Introduce UnXRec and MapXRec classes
Fixes #17587 and #18408
Updates haddock submodule
Co-authored-by: Philipp Krüger <philipp.krueger1@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Issue #18451 showed that we could get an infinite type, through
over-use of occCheckExpand in the kind of an /occurrence/ of a
type variable.
See Note [Occurrence checking: look inside kinds] in GHC.Core.Type
This patch fixes the problem by making occCheckExpand less eager
to expand synonyms in kinds.
It also improves pretty printing of kinds, by *not* suppressing
the kind on a tyvar-binder like
(a :: Const Type b)
where type Const p q = p. Even though the kind of 'a' is Type,
we don't want to suppress the kind ascription. Example: the
error message for polykinds/T18451{a,b}. See GHC.Core.TyCo.Ppr
Note [Suppressing * kinds].
|
|
|
|
|
|
|
|
| |
* for consistency, try to always use UnitPprInfo to display units to
users
* remove some uses of `unitPackageIdString` as it doesn't show the
component name and it uses String
|
|
|
|
|
| |
There's one backwards compatibility issue: GHC.Prim no longer exports
Void#, we now manually re-export it from GHC.Exts.
|
|
|
|
|
|
|
|
|
|
|
| |
It's behaviour is now unconditionally enabled as
it's slightly beneficial.
There are almost no benchmarks which benefit from
disabling it, so it's not worth the keep this
configurable.
This fixes #18429.
|
|
|
|
|
|
|
|
|
| |
They both have the same role and Backend name is more explicit.
Metric Decrease:
T3064
Update Haddock submodule
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Represent backends with a `Backend` datatype in GHC.Driver.Backend
* Don't detect the default backend to use for the target platform at
compile time in Hadrian/make but at runtime. It makes "Settings"
simpler and it is a step toward making GHC multi-target.
* The latter change also fixes hadrian which has not been updated to
take into account that the NCG now supports AIX and PPC64 (cf
df26b95559fd467abc0a3a4151127c95cb5011b9 and
d3c1dda60d0ec07fc7f593bfd83ec9457dfa7984)
* Also we don't treat iOS specifically anymore (cf
cb4878ffd18a3c70f98bdbb413cd3c4d1f054e1f)
|
|
|
|
|
|
| |
Previously we used a platform settings to detect if the native code
generator was used. This was wrong. We need to use the
`DynFlags.hscTarget` field instead.
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Metric Increase:
ManyConstructors
Metric Decrease:
T4029
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
| |
Currently we're passing a indexWord8OffAddr# type function to
utf8DecodeLazy# which then passes it on to utf8DecodeChar#. By passing one
of utf8DecodeCharAddr# or utf8DecodeCharByteArray# instead we benefit from
the inlining and specialization already done for those.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There are multiple reasons we want this:
- Fewer allocations: ByteString has 3 fields, ShortByteString just has one.
- ByteString memory is pinned:
- This can cause fragmentation issues (see for example #13110) but also
- makes using FastStrings in compact regions impossible.
Metric Decrease:
T5837
T12150
T12234
T12425
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Haddock comments are, first and foremost, comments. It's very annoying
to incorporate them into the grammar. We can take advantage of an
important property: adding a Haddock comment does not change the parse
tree in any way other than wrapping some nodes in HsDocTy and the like
(and if it does, that's a bug).
This patch implements the following:
* Accumulate Haddock comments with their locations in the P monad.
This is handled in the lexer.
* After parsing, do a pass over the AST to associate Haddock comments
with AST nodes using location info.
* Report the leftover comments to the user as a warning (-Winvalid-haddock).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There are three problems with the current API:
1. It is hard to properly write instances for ``Quote m => m (TExp a)`` as the type is the composition
of two type constructors. Doing so in your program involves making your own newtype and
doing a lot of wrapping/unwrapping.
For example, if I want to create a language which I can either run immediately or
generate code from I could write the following with the new API. ::
class Lang r where
_int :: Int -> r Int
_if :: r Bool -> r a -> r a -> r a
instance Lang Identity where
_int = Identity
_if (Identity b) (Identity t) (Identity f) = Identity (if b then t else f)
instance Quote m => Lang (Code m) where
_int = liftTyped
_if cb ct cf = [|| if $$cb then $$ct else $$cf ||]
2. When doing code generation it is common to want to store code fragments in
a map. When doing typed code generation, these code fragments contain a
type index so it is desirable to store them in one of the parameterised
map data types such as ``DMap`` from ``dependent-map`` or ``MapF`` from
``parameterized-utils``.
::
compiler :: Env -> AST a -> Code Q a
data AST a where ...
data Ident a = ...
type Env = MapF Ident (Code Q)
newtype Code m a = Code (m (TExp a))
In this example, the ``MapF`` maps an ``Ident String`` directly to a ``Code Q String``.
Using one of these map types currently requires creating your own newtype and constantly
wrapping every quotation and unwrapping it when using a splice. Achievable, but
it creates even more syntactic noise than normal metaprogramming.
3. ``m (TExp a)`` is ugly to read and write, understanding ``Code m a`` is
easier. This is a weak reason but one everyone
can surely agree with.
Updates text submodule.
|
|
|
|
|
|
|
|
|
| |
* add StgPprOpts datatype
* remove Outputable instances for types that need `StgPprOpts` to be
pretty-printed and explicitly call type specific ppr functions
* add default `panicStgPprOpts` for panic messages (when it's not
convenient to thread StgPprOpts or DynFlags down to the ppr function
call)
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This MR (for #18449) refactors the Simplifier's treatment
of join-point binders.
Specifically, it puts together, into
GHC.Core.Opt.Simplify.Env.adjustJoinPointType
two currently-separate ways in which we adjust the type of
a join point. As the comment says:
-- (adjustJoinPointType mult new_res_ty join_id) does two things:
--
-- 1. Set the return type of the join_id to new_res_ty
-- See Note [Return type for join points]
--
-- 2. Adjust the multiplicity of arrows in join_id's type, as
-- directed by 'mult'. See Note [Scaling join point arguments]
I think this actually fixes a latent bug, by ensuring that the
seIdSubst and seInScope have the right multiplicity on the type
of join points.
I did some tidying up while I was at it. No more
setJoinResTy, or modifyJoinResTy: instead it's done locally in
Simplify.Env.adjustJoinPointType
|
|
|
|
|
|
|
| |
This pragma has no effect since 2011.
It was introduced for External Core, which no longer exists.
Updates haddock submodule.
|
|
|
|
|
|
|
|
|
|
| |
This patch (due to Richard Eisenberg) improves
documentation of the wrapper returned by tcSubMult
(see Note [Wrapper returned from tcSubMult] in
GHC.Tc.Utils.Unify).
And, more substantially, it cleans up the multiplicity
handling in the typechecking of NPlusKPat
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As #18412 points out, it should be OK for multiple case alternatives
to have a higher rank type, provided they are all the same.
This patch implements that change. It sweeps away
GHC.Tc.Gen.Match.tauifyMultipleBranches, and friends, replacing it
with an enhanced version of fillInferResult.
The basic change to fillInferResult is to permit the case in which
another case alternative has already filled in the result; and in
that case simply unify. It's very simple actually.
See the new Note [fillInferResult] in TcMType
Other refactoring:
- Move all the InferResult code to one place, in GHC.Tc.Utils.TcMType
(previously some of it was in Unify)
- Move tcInstType and friends from TcMType to Instantiate, where it
more properly belongs. (TCMType was getting very long.)
|
| |
|
| |
|
|
|
|
|
|
|
| |
Previously we would generate expressions of the form
`CmmRegOff BaseReg 0`. This should do no harm (and really should be
handled by the NCG anyways) but it's better to just generate a plain
`CmmReg`.
|
|
|
|
|
|
|
|
|
| |
With the recent change away from first finding and then loading a
framework, loadFramework had no way of communicating the real reason
why loadDLL failed if it was any reason other than the framework
missing from the file system. It now collects all loading attempt
errors into a list and concatenates them into a string to return to
the caller.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
macOS Big Sur makes the following change to how frameworks are shipped
with the OS:
> New in macOS Big Sur 11 beta, the system ships with a built-in
> dynamic linker cache of all system-provided libraries. As part of
> this change, copies of dynamic libraries are no longer present on
> the filesystem. Code that attempts to check for dynamic library
> presence by looking for a file at a path or enumerating a directory
> will fail. Instead, check for library presence by attempting to
> dlopen() the path, which will correctly check for the library in the
> cache. (62986286)
https://developer.apple.com/documentation/macos-release-notes/macos-big-sur-11-beta-release-notes/
Therefore, the previous method of checking whether a library exists
before attempting to load it makes GHC.Runtime.Linker.loadFramework
fail to find frameworks installed at /System/Library/Frameworks.
GHC.Runtime.Linker.loadFramework now opportunistically loads the
framework libraries without checking for their existence first,
failing only if all attempts to load a given framework from any of the
various possible locations fail.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Executing on the plan described in #17582, this patch changes the way if expressions
are handled in the compiler in the presence of rebindable syntax. We get rid of the
SyntaxExpr field of HsIf and instead, when rebindable syntax is on, we rewrite the HsIf
node to the appropriate sequence of applications of the local `ifThenElse` function.
In order to be able to report good error messages, with expressions as they were
written by the user (and not as desugared by the renamer), we make use of TTG
extensions to extend GhcRn expression ASTs with an `HsExpansion` construct, which
keeps track of a source (GhcPs) expression and the desugared (GhcRn) expression that
it gives rise to. This way, we can typecheck the latter while reporting the former in
error messages.
In order to discard the error context lines that arise from typechecking the desugared
expressions (because they talk about expressions that the user has not written), we
carefully give a special treatment to the nodes fabricated by this new renaming-time
transformation when typechecking them. See Note [Rebindable syntax and HsExpansion]
for more details. The note also includes a recipe to apply the same treatment to
other rebindable constructs.
Tests 'rebindable11' and 'rebindable12' have been added to make sure we report
identical error messages as before this patch under various circumstances.
We also now disable rebindable syntax when processing untyped TH quotes, as per
the discussion in #18102 and document the interaction of rebindable syntax and
Template Haskell, both in Note [Template Haskell quotes and Rebindable Syntax]
and in the user guide, adding a test to make sure that we do not regress in
that regard.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ticket #18282 showed that the result discount given by conSize
was massively too large. This patch reduces that discount to
a constant 10, which just balances the cost of the constructor
application itself.
Note [Constructor size and result discount] elaborates, as
does the ticket #18282.
Reducing result discount reduces inlining, which affects perf. I
found that I could increase the unfoldingUseThrehold from 80 to 90 in
compensation; in combination with the result discount change I get
these overall nofib numbers:
Program Size Allocs Runtime Elapsed TotalMem
--------------------------------------------------------------------------------
boyer -0.2% +5.4% -3.2% -3.4% 0.0%
cichelli -0.1% +5.9% -11.2% -11.7% 0.0%
compress2 -0.2% +9.6% -6.0% -6.8% 0.0%
cryptarithm2 -0.1% -3.9% -6.0% -5.7% 0.0%
gamteb -0.2% +2.6% -13.8% -14.4% 0.0%
genfft -0.1% -1.6% -29.5% -29.9% 0.0%
gg -0.0% -2.2% -17.2% -17.8% -20.0%
life -0.1% -2.2% -62.3% -63.4% 0.0%
mate +0.0% +1.4% -5.1% -5.1% -14.3%
parser -0.2% -2.1% +7.4% +6.7% 0.0%
primetest -0.2% -12.8% -14.3% -14.2% 0.0%
puzzle -0.2% +2.1% -10.0% -10.4% 0.0%
rsa -0.2% -11.7% -3.7% -3.8% 0.0%
simple -0.2% +2.8% -36.7% -38.3% -2.2%
wheel-sieve2 -0.1% -19.2% -48.8% -49.2% -42.9%
--------------------------------------------------------------------------------
Min -0.4% -19.2% -62.3% -63.4% -42.9%
Max +0.3% +9.6% +7.4% +11.0% +16.7%
Geometric Mean -0.1% -0.3% -17.6% -18.0% -0.7%
I'm ok with these numbers, remembering that this change removes
an *exponential* increase in code size in some in-the-wild cases.
I investigated compress2. The difference is entirely caused by this
function no longer inlining
WriteRoutines.$woutputCodes
= \ (w :: [CodeEvent]) ->
let result_s1Sr
= case WriteRoutines.outputCodes_$s$woutput w 0# 0# 8# 9# of
(# ww1, ww2 #) -> (ww1, ww2)
in (# case result_s1Sr of (x, _) ->
map @Int @Char WriteRoutines.outputCodes1 x
, case result_s1Sr of { (_, y) -> y } #)
It was right on the cusp before, driven by the excessive result
discount. Too bad!
Happily, the compiler/perf tests show a number of improvements:
T12227 compiler bytes-alloc -6.6%
T12545 compiler bytes-alloc -4.7%
T13056 compiler bytes-alloc -3.3%
T15263 runtime bytes-alloc -13.1%
T17499 runtime bytes-alloc -14.3%
T3294 compiler bytes-alloc -1.1%
T5030 compiler bytes-alloc -11.7%
T9872a compiler bytes-alloc -2.0%
T9872b compiler bytes-alloc -1.2%
T9872c compiler bytes-alloc -1.5%
Metric Decrease:
T12227
T12545
T13056
T15263
T17499
T3294
T5030
T9872a
T9872b
T9872c
|
| |
|