- Mar 05, 2024
-
-
Fix a bug in the x86 ncg where results would be wrong when the desired output register and one of the input registers were the same global. Also adds a tiny optimization to make use of the memory addressing support when convenient. Fixes #24496
-
- Mar 04, 2024
-
-
Finley McIlwaine authored
* Refactor late cost centre insertion for extensibility * Add two more late cost centre insertion methods that add SCCs to overloaded top level bindings and call sites with dictionary arguments. * Some tests for the basic functionality of the new insertion methods Resolves: #24500
-
- Mar 02, 2024
-
-
-
cmmToRawCmmHook was added 4 years ago in d561c8f6. Its only user is the Asterius project, which has been archived and deprecated in favor of the ghc wasm backend. This patch starts deprecating cmmToRawCmmHook by placing a DEPRECATED pragma, and actual removal shall happen in a future GHC major release if no issue to oppose the deprecation has been raised in the meantime.
-
- Mar 01, 2024
-
-
This enriched the `CtOrigin` for non-linear patterns to include data of the pattern that created the constraint (which can be quite useful if it occurs nested in a pattern) as well as an explanation why the pattern is non-restricted in (at least in some cases).
-
-
This implements Proposal 0475, introducing the `ListTuplePuns` extension which is enabled by default. Disabling this extension makes it invalid to refer to list, tuple and sum type constructors by using built-in syntax like `[Int]`, `(Int, Int)`, `(# Int#, Int# #)` or `(# Int | Int #)`. Instead, this syntax exclusively denotes data constructors for use with `DataKinds`. The conventional way of referring to these data constructors by prefixing them with a single quote (`'(Int, Int)`) is now a parser error. Tuple declarations have been moved to `GHC.Tuple.Prim` and the `Solo` data constructor has been renamed to `MkSolo` (in a previous commit). Unboxed tuples and sums now have real source declarations in `GHC.Types`. Unit and solo types for tuples are now called `Unit`, `Unit#`, `Solo` and `Solo#`. Constraint tuples now have the unambiguous type constructors `CTuple<n>` as well as `CUnit` and `CSolo`, defined in `GHC.Classes` like before. A new parser construct has been added for the unboxed sum data constructor declarations. The type families `Tuple`, `Sum#` etc. that were intended to provide nicer syntax have been omitted from this change set due to inference problems, to be implemented at a later time. See the MR discussion for more info. Updates the submodule utils/haddock. Updates the cabal submodule due to new language extension. Metric Increase: haddock.base Metric Decrease: MultiLayerModulesTH_OneShot size_hello_artifact Proposal document: https://github.com/ghc-proposals/ghc-proposals/blob/master/proposals/0475-tuple-syntax.rst Merge request: !8820 Tracking ticket: #21294 -
T23540 is fragile on 32-bit platforms. The root cause is usage of `getEvidenceTreesAtPoint`, which internally relies on `Name`'s `Ord` instance, which is indeterministic. The solution is adding a deterministic `Ord` instance for `EvidenceInfo` and sorting the evidence trees before pretty printing. Fixes #24449.
-
- Feb 29, 2024
-
-
A small change that fixes #24489
-
- Feb 27, 2024
-
-
In pursuit of #15560 and #17957 and generally removing redundancy.
-
Its payload was used only for a small optimization in genAlts, avoiding a few assignments for programs of this form: case NormalDataCon arg1 arg2 of x { NormalDataCon x1 x2 -> ... ; } But when compiling with optimizations, this sort of code is generally eliminated by case-of-known-constructor in Core-to-Core. So it doesn't seem worth tracking and cleaning up again in StgToJS.
-
- Feb 25, 2024
-
-
Ben Gamari authored
-
Ben Gamari authored
Don't lookupIds unless we actually need them.
-
Ben Gamari authored
Really, these references should be via known-key names anyways. I have fixed the proximate issue here but have opened #24472 to track the additional needed refactoring.
-
Ben Gamari authored
Bumps haddock submodule due to testsuite output changes.
-
Ben Gamari authored
Here we extend the parser and AST to preserve docstrings following export items. We then extend Haddock to parse `@since` annotations in such docstrings, allowing changes in export structure to be properly documented.
-
- Feb 24, 2024
-
-
- Feb 23, 2024
-
-
Here we extend the parser and AST to preserve docstrings following export items. We then extend Haddock to parse `@since` annotations in such docstrings, allowing changes in export structure to be properly documented. Bumps haddock submodule.
-
- Feb 21, 2024
-
-
Adam Gundry authored
See https://github.com/ghc-proposals/ghc-proposals/pull/613 . Also fixes #24343 and improves the documentation of language editions. Co-authored-by:
Joachim Breitner <mail@joachim-breitner.de>
-
This commit removes the unused HAVE_C11_ATOMICS macro. We used to have a few places that have fallback paths when HAVE_C11_ATOMICS is not defined, but that is completely redundant, since the FP_CC_SUPPORTS__ATOMICS configure check will fail when the C compiler doesn't support C11 style atomics. There are also many places (e.g. in unreg backend, SMP.h, library cbits, etc) where we unconditionally use C11 style atomics anyway which work in even CentOS 7 (gcc 4.8), the oldest distro we test in our CI, so there's no value in keeping HAVE_C11_ATOMICS.
-
Namespace specifiers were added to syntax of fixity signatures: - sigdecl ::= infix prec ops | ... + sigdecl ::= infix prec namespace_spec ops | ... To preserve namespace during renaming MiniFixityEnv type now has separate FastStringEnv fields for names that should be on the term level and for name that should be on the type level. makeMiniFixityEnv function was changed to fill MiniFixityEnv in the right way: - signatures without namespace specifiers fill both fields - signatures with 'data' specifier fill data field only - signatures with 'type' specifier fill type field only Was added helper function lookupMiniFixityEnv that takes care about looking for a name in an appropriate namespace. Updates haddock submodule. Metric Decrease: MultiLayerModulesTH_OneShot
-
- Feb 19, 2024
-
-
GHC Proposal 448 introduces binders for invisible type arguments (@a-binders) in various contexts. This patch implements @-binders in lambda patterns and function equations: {-# LANGUAGE TypeAbstractions #-} id1 :: a -> a id1 @t x = x :: t -- @t-binder on the LHS of a function equation higherRank :: (forall a. (Num a, Bounded a) => a -> a) -> (Int8, Int16) higherRank f = (f 42, f 42) ex :: (Int8, Int16) ex = higherRank (\ @a x -> maxBound @a - x ) -- @a-binder in a lambda pattern in an argument -- to a higher-order function Syntax ------ To represent those @-binders in the AST, the list of patterns in Match now uses ArgPat instead of Pat: data Match p body = Match { ... - m_pats :: [LPat p], + m_pats :: [LArgPat p], ... } + data ArgPat pass + = VisPat (XVisPat pass) (LPat pass) + | InvisPat (XInvisPat pass) (HsTyPat (NoGhcTc pass)) + | XArgPat !(XXArgPat pass) The VisPat constructor represents patterns for visible arguments, which include ordinary value-level arguments and required type arguments (neither is prefixed with a @), while InvisPat represents invisible type arguments (prefixed with a @). Parser ------ In the grammar (Parser.y), the lambda and lambda-cases productions of aexp non-terminal were updated to accept argpats instead of apats: aexp : ... - | '\\' apats '->' exp + | '\\' argpats '->' exp ... - | '\\' 'lcases' altslist(apats) + | '\\' 'lcases' altslist(argpats) ... + argpat : apat + | PREFIX_AT atype Function left-hand sides did not require any changes to the grammar, as they were already parsed with productions capable of parsing @-binders. Those binders were being rejected in post-processing (isFunLhs), and now we accept them. In Parser.PostProcess, patterns are constructed with the help of PatBuilder, which is used as an intermediate data structure when disambiguating between FunBind and PatBind. In this patch we define ArgPatBuilder to accompany PatBuilder. ArgPatBuilder is a short-lived data structure produced in isFunLhs and consumed in checkFunBind. Renamer ------- Renaming of @-binders builds upon prior work on type patterns, implemented in 2afbddb0, which guarantees proper scoping and shadowing behavior of bound type variables. This patch merely defines rnLArgPatsAndThen to process a mix of visible and invisible patterns: + rnLArgPatsAndThen :: NameMaker -> [LArgPat GhcPs] -> CpsRn [LArgPat GhcRn] + rnLArgPatsAndThen mk = mapM (wrapSrcSpanCps rnArgPatAndThen) where + rnArgPatAndThen (VisPat x p) = ... rnLPatAndThen ... + rnArgPatAndThen (InvisPat _ tp) = ... rnHsTyPat ... Common logic between rnArgPats and rnPats is factored out into the rn_pats_general helper. Type checker ------------ Type-checking of @-binders builds upon prior work on lazy skolemisation, implemented in f5d3e03c. This patch extends tcMatchPats to handle @-binders. Now it takes and returns a list of LArgPat rather than LPat: tcMatchPats :: ... - -> [LPat GhcRn] + -> [LArgPat GhcRn] ... - -> TcM ([LPat GhcTc], a) + -> TcM ([LArgPat GhcTc], a) Invisible binders in the Match are matched up with invisible (Specified) foralls in the type. This is done with a new clause in the `loop` worker of tcMatchPats: loop :: [LArgPat GhcRn] -> [ExpPatType] -> TcM ([LArgPat GhcTc], a) loop (L l apat : pats) (ExpForAllPatTy (Bndr tv vis) : pat_tys) ... -- NEW CLAUSE: | InvisPat _ tp <- apat, isSpecifiedForAllTyFlag vis = ... In addition to that, tcMatchPats no longer discards type patterns. This is done by filterOutErasedPats in the desugarer instead. x86_64-linux-deb10-validate+debug_info Metric Increase: MultiLayerModulesTH_OneShot -
This drops GHC's dependence on the `touch` program, instead implementing it within GHC. This eliminates an external dependency and means that we have one fewer program to keep track of in the `configure` script
-
- Feb 17, 2024
-
-
In addition to existing Acquire/Release fences, this commit adds SEQ_CST fence support to GHC, allowing Cmm code to explicitly emit a fence that enforces total memory ordering. The following logic is added: - The MO_SeqCstFence callish MachOp - The %prim fence_seq_cst() Cmm syntax and the SEQ_CST_FENCE macro in Cmm.h - MO_SeqCstFence lowering logic in every single GHC codegen backend
-
-
First step towards fixing #24331. Replace foreign prim imports with real primops.
-
- Feb 16, 2024
-
-
GCC 14 on aarch64 rejects the C code written by GHC with this kind of error: error: assignment to ‘ffi_arg’ {aka ‘long unsigned int’} from ‘HsPtr’ {aka ‘void *’} makes integer from pointer without a cast [-Wint-conversion] 68 | *(ffi_arg*)resp = cret; | ^ Add the correct cast. For more information on this see: https://fedoraproject.org/wiki/Changes/PortingToModernC Tested-by:Richard W.M. Jones <rjones@redhat.com>
-
- Feb 15, 2024
-
-
The simple optimizer reduces the size of the code generated by the JavaScript backend without the complexity and performance penalty of the optimizer in GHCJS. Also see #22736 Metric Decrease: libdir size_hello_artifact -
Looking up by OccName only does not take into account when functions are only imported in a qualified way. Fixes issue #24294 Bump haddock submodule to include regression test
-
- Feb 14, 2024
-
-
This commit fixes two wasm unreg regressions caught by a nightly pipeline: - Unknown stg_scheduler_loopzh symbol when compiling scheduler.cmm - Invalid _hs_constructor(101) function name when handling ctor
-
Check for ptr/int arguments rather than 64-bit width arguments when counting integer register arguments. The old approach broke when we stopped using exclusively W64-sized types to represent sub-word sized integers. Fixes #24314
-
Leaving a few that are too tricky, maybe some other time. Also - remove some unneeded helpers from Parser.y - reduce allocations with strictness annotations Updates haddock submodule Metric Decrease: parsing001
-
- Feb 13, 2024
-
-
This commit adds JavaScriptFFI as a supported extension when the target platform is wasm32.
-
This commit adds JSFFI desugar logic for the wasm backend.
-
This commit fixes the priorities of ctors generated by GHC codegen on wasm32, see the referred note for details.
-
This commit allows the boxed JSVal type to be used as a foreign argument/result type.
-
This commit allows the javascript calling convention to be used when the target platform is wasm32.
-
The pure Haskell implementation causes i386 regression in unrelated work that can be fixed by using C-based atomic increment, see added comment for details.
-
- Feb 12, 2024
-
-
Just remove an out of date block of commented-out code, and tidy up the relevant Notes. See #8317.
-
Support linking C sources with JS output of the JavaScript backend. See the added documentation in the users guide. The implementation simply extends the JS linker to use the objects (.o) that were already produced by the emcc compiler and which were filtered out previously. I've also added some options to control the link with C functions (see the documentation about pragmas). With this change I've successfully compiled the direct-sqlite package which embeds the sqlite.c database code. Some wrappers are still required (see the documentation about wrappers) but everything generic enough to be reused for other libraries have been integrated into rts/js/mem.js.
-
