| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* Rename pprCLabel to pprCLabelStyle, and use the name pprCLabel
for a function using CStyle (analogous to pprAsmLabel)
* Move LabelStyle to the CLabel module, it no longer needs to be in Outputable.
* Move calls to 'text' right next to literals, to make sure the text/str
rule is triggered.
* Remove FastString/String roundtrip in Tc.Deriv.Generate
* Introduce showSDocForUser', which abstracts over a pattern in
GHCi.UI
|
|
|
|
|
|
|
|
|
|
| |
I had forgotten to add the auxiliary dict bindings to the
/unfolding/ of a specialised function. This caused #22358,
which reports failures when compiling Hackage packages
fixed-vector
indexed-traversable
Regression test T22357 is snarfed from indexed-traversable
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change aims to minimize source location information leaking
into interface files, which makes ABI hashes dependent on the
build location.
The `Binary (Located a)` instance has been removed completely.
It seems that the HIE interface still needs the ability to
serialize SrcSpans, but by wrapping the instances, it should
be a lot more difficult to inadvertently add source location
information.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When faced with VDQ in the type of a term, GHC generates the following
error message:
Illegal visible, dependent quantification in the type of a term
(GHC does not yet support this)
Prior to this patch, there were two ways this message could have been
generated and represented:
1. with the dedicated constructor TcRnVDQInTermType
(see check_type in GHC.Tc.Validity)
2. with the transitional constructor TcRnUnknownMessage
(see noNestedForallsContextsErr in GHC.Rename.Utils)
Not only this led to duplication of code generating the final SDoc,
it also made it tricky to track the origin of the error message.
This patch fixes the problem by using TcRnVDQInTermType exclusively.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Before this patch, GHC used withHsDocContext to attach an HsDocContext
to an error message:
addErr $ mkTcRnUnknownMessage $ mkPlainError noHints (withHsDocContext ctxt msg)
The problem with this approach is that it only works with
TcRnUnknownMessage. But could we attach an HsDocContext to a
structured error message in a generic way? This patch solves
the problem by introducing a new constructor to TcRnMessage:
data TcRnMessage where
...
TcRnWithHsDocContext :: !HsDocContext -> !TcRnMessage -> TcRnMessage
...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch fixes two distinct (but closely related) buglets that were uncovered
in #22235:
* `liftEnvSubst` used an empty in-scope set, which was not wide enough to cover
the variables in the range of the substitution. This patch fixes this by
populating the in-scope set from the free variables in the range of the
substitution.
* `composeTCvSubst` applied the first substitution argument to the range of the
second substitution argument, but the first substitution's in-scope set was
not wide enough to cover the range of the second substutition. We similarly
fix this issue in this patch by widening the first substitution's in-scope set
before applying it.
Fixes #22235.
|
|
|
|
|
|
|
| |
Remove unused mkPtrString and isUnderscoreFS.
We no longer use mkPtrString since 1d03d8bef96.
Remove unnecessary conversions between FastString and String and back.
|
|
|
|
| |
Move doCpp out of the driver to be able to use it in the upcoming JS backend.
|
|
|
|
|
| |
The comment applies only when host's word size < target's word size.
So we can relax the guard.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
ghc-bignum needs a way to raise numerical exceptions defined in base
package. At the time we used FFI calls into primops defined in the RTS.
These FFI calls had to be wrapped into hacky bottoming functions because
"foreign import prim" syntax doesn't support giving a bottoming demand
to the foreign call (cf #16929).
These hacky wrapper functions trip up the JavaScript backend (#21078)
because they are polymorphic in their return type. This commit
replaces them with primops very similar to raise# but raising predefined
exceptions.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ticket #13873 unexpectedly showed that a SPECIALISE pragma made a
program run (a lot) slower, because less specialisation took place
overall. It turned out that the specialiser was missing opportunities
because of quantified type variables.
It was quite easy to fix. The story is given in
Note [Specialising polymorphic dictionaries]
Two other minor fixes in the specialiser
* There is no benefit in specialising data constructor /wrappers/.
(They can appear overloaded because they are given a dictionary
to store in the constructor.) Small guard in canSpecImport.
* There was a buglet in the UnspecArg case of specHeader, in the
case where there is a dead binder. We need a LitRubbish filler
for the specUnfolding stuff. I expanded
Note [Drop dead args from specialisations] to explain.
There is a 4% increase in compile time for T15164, because we generate
more specialised code. This seems OK.
Metric Increase:
T15164
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch fixes #21229 properly, by avoiding doing a
binder-swap on dictionary Ids. This is pretty subtle, and explained
in Note [Care with binder-swap on dictionaries].
Test is already in simplCore/should_run/T21229
This allows us to restore a feature to the specialiser that we had
to revert: see Note [Specialising polymorphic dictionaries].
(This is done in a separate patch.)
I also modularised things, using a new function scrutBinderSwap_maybe
in all the places where we are (effectively) doing a binder-swap,
notably
* Simplify.Iteration.addAltUnfoldings
* SpecConstr.extendCaseBndrs
In Simplify.Iteration.addAltUnfoldings I also eliminated a guard
Many <- idMult case_bndr
because we concluded, in #22123, that it was doing no good.
|
|
|
|
|
|
|
| |
It now properly lints cases where sums end up distributed
over multiple args after unarise.
Fixes #22026.
|
|
|
|
|
|
|
| |
Replaces uses of `TcRnUnknownMessage` in `GHC.Tc.Gen.Splice` with
structured diagnostics.
closes #20116
|
|
|
|
| |
Updates the haddock submodule.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently for a top-level closure in the form
hey = unpackCString# x
we generate code like this:
Main.hey_entry() // [R1]
{ info_tbls: [(c2T4,
label: Main.hey_info
rep: HeapRep static { Thunk }
srt: Nothing)]
stack_info: arg_space: 8 updfr_space: Just 8
}
{offset
c2T4: // global
_rqm::P64 = R1;
if ((Sp + 8) - 24 < SpLim) (likely: False) goto c2T5; else goto c2T6;
c2T5: // global
R1 = _rqm::P64;
call (stg_gc_enter_1)(R1) args: 8, res: 0, upd: 8;
c2T6: // global
(_c2T1::I64) = call "ccall" arg hints: [PtrHint,
PtrHint] result hints: [PtrHint] newCAF(BaseReg, _rqm::P64);
if (_c2T1::I64 == 0) goto c2T3; else goto c2T2;
c2T3: // global
call (I64[_rqm::P64])() args: 8, res: 0, upd: 8;
c2T2: // global
I64[Sp - 16] = stg_bh_upd_frame_info;
I64[Sp - 8] = _c2T1::I64;
R2 = hey1_r2Gg_bytes;
Sp = Sp - 16;
call GHC.CString.unpackCString#_info(R2) args: 24, res: 0, upd: 24;
}
}
This code is generated for every string literal. Only difference between
top-level closures like this is the argument for the bytes of the string
(hey1_r2Gg_bytes in the code above).
With this patch we introduce a standard thunk in the RTS, called
stg_MK_STRING_info, that does what `unpackCString# x` does, except it
gets the bytes address from the payload. Using this, for the closure
above, we generate this:
Main.hey_closure" {
Main.hey_closure:
const stg_MK_STRING_info;
const 0; // padding for indirectee
const 0; // static link
const 0; // saved info
const hey1_r1Gg_bytes; // the payload
}
This is much smaller in code.
Metric Decrease:
T10421
T11195
T12150
T12425
T16577
T18282
T18698a
T18698b
Co-Authored By: Ben Gamari <ben@well-typed.com>
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
This reverts the workaround introduced in
f63c8ef33ec9666688163abe4ccf2d6c0428a7e7, which taught our response file
logic to write response files with the `latin1` encoding to workaround
`gcc`'s lacking Unicode support. This is now no longer necessary (and in
fact actively unhelpful) since we rather use Clang.
|
|
|
|
|
| |
As noted in #12971, we previously used `show` which resulted in
inappropriate escaping of non-ASCII characters.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch is motivated by the desire to remove the {-# OPTIONS_GHC
-fno-warn-incomplete-patterns #-} directive at the top of
GHC.Cmm.ContFlowOpt. (Based on the text in this coding standards doc, I
understand it's a goal of the project to remove such directives.) I
chose this task because I'm a new contributor to GHC, and it seemed like
a good way to get acquainted with the patching process.
In order to address the warning that arose when I removed the no-warn
directive, I added a case to removeUnreachableBlocksProc to handle the
CmmData constructor. Clearly, since this partial function has not been
erroring out in the wild, its inputs are always in practice wrapped by
the CmmProc constructor. Therefore the CmmData case is handled by a
precise panic (which is an improvement over the partial pattern match
from before).
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The assertion that checked TyEq:N in canEqCanLHSFinish incorrectly
triggered in the case of an unsaturated newtype TyCon heading the RHS,
even though we can't unwrap such an application. Now, we only trigger
an assertion failure in case of a saturated application of a newtype
TyCon.
Fixes #22310
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
TcSolverReportMsg
It's clear from asserts in `GHC.Tc.Errors` that `overlappingInstances_matches`
and `unsafeOverlapped` are supposed to be non-empty, and `unsafeOverlap_matches`
contains a single instance, but these invariants are immediately lost afterwards
and not encoded in types. This patch enforces the invariants by pattern matching
and makes types more precise, avoiding asserts and partial functions such as `head`.
|
| |
|
|
|
|
| |
Lets us avoid some use of `head` and `tail`, and some panics.
|
|
|
|
|
|
|
|
|
|
| |
As noted in #22297, SIMD vector registers can be used
to store different kinds of values, e.g. xmm1 can be used
both to store integer and floating point values.
The Cmm type system doesn't properly account for this, so
we weaken the Cmm register assignment lint check to only
compare widths when comparing a vector type with its
allocated vector register.
|
|
|
|
|
|
|
|
|
|
|
| |
This patch makes it so that packing/unpacking SIMD
vectors always uses the right sized types, e.g.
unpacking a Word16X4# will give a tuple of Word16#s.
As a result, we can get rid of the conversion instructions
that were previously required.
Fixes #22296
|
|
|
|
|
|
|
|
|
| |
This patch adds the missing `VecRep` case to `primRepSlot` function and
all the necessary machinery to carry this new `VecSlot` through code
generation. This allows programs involving unboxed sums of SIMD vectors
to be written and compiled.
Fixes #22187
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In many development environments, the source span is the primary means
of seeing what an error message relates to, and the In the expression:
and In an equation for: clauses are not particularly relevant. However,
they can grow to be quite long, which can make the message itself both
feel overwhelming and interact badly with limited-space areas.
It's simple to implement this flag so we might as well do it and give
the user control about how they see their messages.
Fixes #21722
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This MR implements the idea of #21731 that the printing of a diagnostic
method should be configurable at the printing time.
The interface of the `Diagnostic` class is modified from:
```
class Diagnostic a where
diagnosticMessage :: a -> DecoratedSDoc
diagnosticReason :: a -> DiagnosticReason
diagnosticHints :: a -> [GhcHint]
```
to
```
class Diagnostic a where
type DiagnosticOpts a
defaultDiagnosticOpts :: DiagnosticOpts a
diagnosticMessage :: DiagnosticOpts a -> a -> DecoratedSDoc
diagnosticReason :: a -> DiagnosticReason
diagnosticHints :: a -> [GhcHint]
```
and so each `Diagnostic` can implement their own configuration record
which can then be supplied by a client in order to dictate how to print
out the error message.
At the moment this only allows us to implement #21722 nicely but in
future it is more natural to separate the configuration of how much
information we put into an error message and how much we decide to print
out of it.
Updates Haddock submodule
|
|
|
|
| |
functions.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I had assumed that wrappers were not inlined in interactive mode.
Meaning we would always execute the compiled wrapper which properly
takes care of upholding the strict field invariant.
This turned out to be wrong. So instead we now run tag inference even
when we generate bytecode. In that case only for correctness not
performance reasons although it will be still beneficial for runtime
in some cases.
I further fixed a bug where GHCi didn't tag nullary constructors
properly when used as arguments. Which caused segfaults when calling
into compiled functions which expect the strict field invariant to
be upheld.
Fixes #22042 and #21083
-------------------------
Metric Increase:
T4801
Metric Decrease:
T13035
-------------------------
|
|
|
|
|
|
|
| |
We do so by having an explicit folding function that doesn't need to
allocate intermediate lists first.
Fixes #22196
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, the demand signature we computed upfront for a DataCon wrapper
lacked boxity information and was much less precise than the demand transformer
for the DataCon worker.
In this patch we adopt the solution to look through unfoldings of DataCon
wrappers during Demand Analysis, but still attach a demand signature for other
passes such as the Simplifier.
See `Note [DmdAnal for DataCon wrappers]` for more details.
Fixes #22241.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In adc7f108141a973b6dcb02a7836eed65d61230e8 we fixed a number of issues
to do with sign extension in the AArch64 NCG found by ghc/test-primops>.
However, this patch made a critical error, assuming that getSomeReg
would allocate a fresh register for the result of its evaluation.
However, this is not the case as `getSomeReg (CmmReg r) == r`.
Consequently, any mutation of the register returned by `getSomeReg` may
have unwanted side-effects on other expressions also mentioning `r`. In
the fix listed above, this manifested as the registers containing the
operands of binary arithmetic operations being incorrectly
sign-extended. This resulted in #22282.
Sadly, the rather simple structure of the tests generated
by `test-primops` meant that this particular case was not exercised.
Even more surprisingly, none of our testsuite caught this case.
Here we fix this by ensuring that intermediate sign extension is
performed in a fresh register.
Fixes #22282.
|
|
|
|
|
|
|
|
|
|
|
| |
When a newtype introduces GADT eq_specs due to a defaulted
RuntimeRep, we detect this and print the error message with
explicit kinds.
This also refactors newtype type checking to use the new
diagnostic infra.
Fixes #21447
|
|
|
|
|
| |
Implements GHC proposal:
https://github.com/ghc-proposals/ghc-proposals/blob/master/proposals/0170-unrestricted-overloadedlabels.rst
|
|
|
|
|
|
|
|
|
|
|
| |
Lint errors indicate an internal error in GHC, so it makes sense to use
it instead of the user style. This is consistent with Core Lint and STG Lint:
https://gitlab.haskell.org/ghc/ghc/-/blob/22096652/compiler/GHC/Core/Lint.hs#L429
https://gitlab.haskell.org/ghc/ghc/-/blob/22096652/compiler/GHC/Stg/Lint.hs#L144
Fixes #22218.
|
|
|
|
|
|
| |
The source file name can very often be shared across many IPE entries
whereas the source coordinates are generally unique. Separate the two to
exploit sharing of the former.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Here we refactor the representation of info table provenance information
in object code to significantly reduce its size and link-time impact.
Specifically, we deduplicate strings and represent them as 32-bit
offsets into a common string table.
In addition, we rework the registration logic to eliminate allocation
from the registration path, which is run from a static initializer where
things like allocation are technically undefined behavior (although it
did previously seem to work). For similar reasons we eliminate lock
usage from registration path, instead relying on atomic CAS.
Closes #22077.
|
|
|
|
| |
isInfoTableLabel does not take Cmm info table into account. This patch is required for data section layout of wasm32 NCG to work.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ticket #22162 pointed out that the build directory was leaking into the
ABI hash of a module because the BufPos depended on the location of the
build tree.
BufPos is only used in GHC.Parser.PostProcess.Haddock, and the
information doesn't need to be propagated outside the context of a
module.
Fixes #22162
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When doing performance debugging on #22084 / !8901, I found that the
algorithm in SpecConstr.decreaseSpecCount was so aggressive that if
there were /more/ specialisations available for an outer function,
that could more or less kill off specialisation for an /inner/
function. (An example was in nofib/spectral/fibheaps.)
This patch makes it a bit more aggressive, by dividing by 2, rather
than by the number of outer specialisations.
This makes the program bigger, temporarily:
T19695(normal) ghc/alloc +11.3% BAD
because we get more specialisation. But lots of other programs
compile a bit faster and the geometric mean in perf/compiler
is 0.0%.
Metric Increase:
T19695
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We were religiously keeping exit join points throughout, which
had some bad effects (#21148, #22084).
This MR does two things:
* Arranges that exit join points are inhibited from inlining
only in /one/ Simplifier pass (right after Exitification).
See Note [Be selective about not-inlining exit join points]
in GHC.Core.Opt.Exitify
It's not a big deal, but it shaves 0.1% off compile times.
* Inline used-once non-recursive join points very aggressively
Given join j x = rhs in
joinrec k y = ....j x....
where this is the only occurrence of `j`, we want to inline `j`.
(Unless sm_keep_exits is on.)
See Note [Inline used-once non-recursive join points] in
GHC.Core.Opt.Simplify.Utils
This is just a tidy-up really. It doesn't change allocation, but
getting rid of a binding is always good.
Very effect on nofib -- some up and down.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch teachs the code generation logic of -fno-code about
-fprefer-byte-code, so that if we need to generate code for a module
which prefers byte code, then we generate byte code rather than object
code.
We keep track separately which modules need object code and which byte
code and then enable the relevant code generation for each. Typically
the option will be enabled globally so one of these sets should be empty
and we will just turn on byte code or object code generation.
We also fix the bug where we would generate code for a module which
enables Template Haskell despite the fact it was unecessary.
Fixes #22016
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit adds three new flags
* -fwrite-if-simplified-core: Writes the whole core program into an interface
file
* -fbyte-code-and-object-code: Generate both byte code and object code
when compiling a file
* -fprefer-byte-code: Prefer to use byte-code if it's available when
running TH splices.
The goal for including the core bindings in an interface file is to be able to restart the compiler pipeline
at the point just after simplification and before code generation. Once compilation is
restarted then code can be created for the byte code backend.
This can significantly speed up
start-times for projects in GHCi. HLS already implements its own version of these extended interface
files for this reason.
Preferring to use byte-code means that we can avoid some potentially
expensive code generation steps (see #21700)
* Producing object code is much slower than producing bytecode, and normally you
need to compile with `-dynamic-too` to produce code in the static and dynamic way, the
dynamic way just for Template Haskell execution when using a dynamically linked compiler.
* Linking many large object files, which happens once per splice, can be quite
expensive compared to linking bytecode.
And you can get GHC to compile the necessary byte code so
`-fprefer-byte-code` has access to it by using
`-fbyte-code-and-object-code`.
Fixes #21067
|