summaryrefslogtreecommitdiff
path: root/compiler/GHC/Driver
Commit message (Collapse)AuthorAgeFilesLines
* Add fused multiply-add instructionssheaf2023-05-113-1/+33
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch adds eight new primops that fuse a multiplication and an addition or subtraction: - `{fmadd,fmsub,fnmadd,fnmsub}{Float,Double}#` fmadd x y z is x * y + z, computed with a single rounding step. This patch implements code generation for these primops in the following backends: - X86, AArch64 and PowerPC NCG, - LLVM - C WASM uses the C implementation. The primops are unsupported in the JavaScript backend. The following constant folding rules are also provided: - compute a * b + c when a, b, c are all literals, - x * y + 0 ==> x * y, - ±1 * y + z ==> z ± y and x * ±1 + z ==> z ± x. NB: the constant folding rules incorrectly handle signed zero. This is a known limitation with GHC's floating-point constant folding rules (#21227), which we hope to resolve in the future.
* Move "target has RTS linker" out of settingsRodrigo Mesquita2023-05-111-2/+3
| | | | | | | | | | | | We move the "target has RTS linker" information out of configure into a predicate in GHC, and remove this option from the settings file where it is unnecessary -- it's information statically known from the platform. Note that previously we would consider `powerpc`s and `s390x`s other than `powerpc-ibm-aix*` and `s390x-ibm-linux` to have an RTS linker, but the RTS linker supports neither platform. Closes #23361
* Add structured error messages for GHC.Rename.ModuleTorsten Schmits2023-05-051-0/+1
| | | | | | | | | | | Tracking ticket: #20115 MR: !10361 This converts uses of `mkTcRnUnknownMessage` to newly added constructors of `TcRnMessage`. Only addresses the single warning missing from the previous MR.
* docs: Add Note [Timing of plugin initialization]Matthew Pickering2023-05-052-0/+2
|
* Rework plugin initialisation pointsAaron Allen2023-05-053-13/+20
| | | | | | | | | | | | | | | | | | In general this patch pushes plugin initialisation points to earlier in the pipeline. As plugins can modify the `HscEnv`, it's imperative that the plugins are initialised as soon as possible and used thereafter. For example, there are some new tests which modify hsc_logger and other hooks which failed to fire before (and now do) One consequence of this change is that the error for specifying the usage of a HPT plugin from the command line has changed, because it's now attempted to be loaded at initialisation rather than causing a cyclic module import. Closes #21279 Co-authored-by: Matthew Pickering <matthewtpickering@gmail.com>
* driver: Use hooks from plugin_hsc_envMatthew Pickering2023-05-051-1/+1
| | | | | | This fixes a bug in oneshot mode where hooks modified in a plugin wouldn't be used in oneshot mode because we neglected to use the right hsc_env. This was observed by @csabahruska.
* Hardwire a better unit-id for ghcromes2023-05-041-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Previously, the unit-id of ghc-the-library was fixed as `ghc`. This was done primarily because the compiler must know the unit-id of some packages (including ghc) a-priori to define wired-in names. However, as seen in #20742, a reinstallable `ghc` whose unit-id is fixed to `ghc` might result in subtle bugs when different ghc's interact. A good example of this is having GHC_A load a plugin compiled by GHC_B, where GHC_A and GHC_B are linked to ghc-libraries that are ABI incompatible. Without a distinction between the unit-id of the ghc library GHC_A is linked against and the ghc library the plugin it is loading was compiled against, we can't check compatibility. This patch gives a slightly better unit-id to ghc (ghc-version) by (1) Not setting -this-unit-id to ghc, but rather to the new unit-id (modulo stage0) (2) Adding a definition to `GHC.Settings.Config` whose value is the new unit-id. (2.1) `GHC.Settings.Config` is generated by Hadrian (2.2) and also by cabal through `compiler/Setup.hs` This unit-id definition is imported by `GHC.Unit.Types` and used to set the wired-in unit-id of "ghc", which was previously fixed to "ghc" The commits following this one will improve the unit-id with a cabal-style package hash and check compatibility when loading plugins. Note that we also ensure that ghc's unit key matches unit id both when hadrian or cabal builds ghc, and in this way we no longer need to add `ghc` to the WiringMap.
* Add sized primitive literal syntaxBen Orchard2023-05-031-0/+1
| | | | | | | | | | | | | | Adds a new LANGUAGE pragma ExtendedLiterals, which enables defining unboxed numeric literals such as `0xFF#Word8 :: Word8#`. Implements GHC proposal 0451: https://github.com/ghc-proposals/ghc-proposals/blob/b384a538b34f79d18a0201455b7b3c473bc8c936/proposals/0451-sized-literals.rst Fixes #21422. Bumps haddock submodule. Co-authored-by: Krzysztof Gogolewski <krzysztof.gogolewski@tweag.io>
* Implement -jsem: parallelism controlled by semaphoressheaf2023-04-204-56/+638
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | See https://github.com/ghc-proposals/ghc-proposals/pull/540/ for a complete description for the motivation for this feature. The `-jsem` option allows a build tool to pass a semaphore to GHC which GHC can use in order to control how much parallelism it requests. GHC itself acts as a client in the GHC jobserver protocol. ``` GHC Jobserver Protocol ~~~~~~~~~~~~~~~~~~~~~~ This proposal introduces the GHC Jobserver Protocol. This protocol allows a server to dynamically invoke many instances of a client process, while restricting all of those instances to use no more than <n> capabilities. This is achieved by coordination over a system semaphore (either a POSIX semaphore [6]_ in the case of Linux and Darwin, or a Win32 semaphore [7]_ in the case of Windows platforms). There are two kinds of participants in the GHC Jobserver protocol: - The *jobserver* creates a system semaphore with a certain number of available tokens. Each time the jobserver wants to spawn a new jobclient subprocess, it **must** first acquire a single token from the semaphore, before spawning the subprocess. This token **must** be released once the subprocess terminates. Once work is finished, the jobserver **must** destroy the semaphore it created. - A *jobclient* is a subprocess spawned by the jobserver or another jobclient. Each jobclient starts with one available token (its *implicit token*, which was acquired by the parent which spawned it), and can request more tokens through the Jobserver Protocol by waiting on the semaphore. Each time a jobclient wants to spawn a new jobclient subprocess, it **must** pass on a single token to the child jobclient. This token can either be the jobclient's implicit token, or another token which the jobclient acquired from the semaphore. Each jobclient **must** release exactly as many tokens as it has acquired from the semaphore (this does not include the implicit tokens). ``` Build tools such as cabal act as jobservers in the protocol and are responsibile for correctly creating, cleaning up and managing the semaphore. Adds a new submodule (semaphore-compat) for managing and interacting with semaphores in a cross-platform way. Fixes #19349
* Convert interface file loading errors into proper diagnosticsMatthew Pickering2023-04-187-37/+43
| | | | | | | | | | | | | | This patch converts all the errors to do with loading interface files into proper structured diagnostics. * DriverMessage: Sometimes in the driver we attempt to load an interface file so we embed the IfaceMessage into the DriverMessage. * TcRnMessage: Most the time we are loading interface files during typechecking, so we embed the IfaceMessage This patch also removes the TcRnInterfaceLookupError constructor which is superceded by the IfaceMessage, which is now structured compared to just storing an SDoc before.
* Account for special GHC.Prim import in warnUnusedPackagesMatthew Pickering2023-04-171-3/+7
| | | | | | | | | | | | The GHC.Prim import is treated quite specially primarily because there isn't an interface file for GHC.Prim. Therefore we record separately in the ModSummary if it's imported or not so we don't go looking for it. This logic hasn't made it's way to `-Wunused-packages` so if you imported GHC.Prim then the warning would complain you didn't use `-package ghc-prim`. Fixes #23212
* Add import lists to few GHC.Driver.Session importsOleg Grenrus2023-04-161-1/+2
| | | | | | Related to https://gitlab.haskell.org/ghc/ghc/-/issues/23261. There are a lot of GHC.Driver.Session which only use DynFlags, but not the parsing code.
* driver: Unit State Data.Map -> GHC.Unique.UniqMapdoyougnu2023-04-011-2/+4
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | In pursuit of #22426. The driver and unit state are major contributors. This commit also bumps the haddock submodule to reflect the API changes in UniqMap. ------------------------- Metric Decrease: MultiComponentModules MultiComponentModulesRecomp T10421 T10547 T12150 T12234 T12425 T13035 T16875 T18140 T18304 T18698a T18698b T18923 T20049 T5837 T6048 T9198 -------------------------
* Handle records in the renamersheaf2023-03-291-14/+17
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch moves the field-based logic for disambiguating record updates to the renamer. The type-directed logic, scheduled for removal, remains in the typechecker. To do this properly (and fix the myriad of bugs surrounding the treatment of duplicate record fields), we took the following main steps: 1. Create GREInfo, a renamer-level equivalent to TyThing which stores information pertinent to the renamer. This allows us to uniformly treat imported and local Names in the renamer, as described in Note [GREInfo]. 2. Remove GreName. Instead of a GlobalRdrElt storing GreNames, which distinguished between normal names and field names, we now store simple Names in GlobalRdrElt, along with the new GREInfo information which allows us to recover the FieldLabel for record fields. 3. Add namespacing for record fields, within the OccNames themselves. This allows us to remove the mangling of duplicate field selectors. This change ensures we don't print mangled names to the user in error messages, and allows us to handle duplicate record fields in Template Haskell. 4. Move record disambiguation to the renamer, and operate on the level of data constructors instead, to handle #21443. The error message text for ambiguous record updates has also been changed to reflect that type-directed disambiguation is on the way out. (3) means that OccEnv is now a bit more complex: we first key on the textual name, which gives an inner map keyed on NameSpace: OccEnv a ~ FastStringEnv (UniqFM NameSpace a) Note that this change, along with (2), both increase the memory residency of GlobalRdrEnv = OccEnv [GlobalRdrElt], which causes a few tests to regress somewhat in compile-time allocation. Even though (3) simplified a lot of code (in particular the treatment of field selectors within Template Haskell and in error messages), it came with one important wrinkle: in the situation of -- M.hs-boot module M where { data A; foo :: A -> Int } -- M.hs module M where { data A = MkA { foo :: Int } } we have that M.hs-boot exports a variable foo, which is supposed to match with the record field foo that M exports. To solve this issue, we add a new impedance-matching binding to M foo{var} = foo{fld} This mimics the logic that existed already for impedance-binding DFunIds, but getting it right was a bit tricky. See Note [Record field impedance matching] in GHC.Tc.Module. We also needed to be careful to avoid introducing space leaks in GHCi. So we dehydrate the GlobalRdrEnv before storing it anywhere, e.g. in ModIface. This means stubbing out all the GREInfo fields, with the function forceGlobalRdrEnv. When we read it back in, we rehydrate with rehydrateGlobalRdrEnv. This robustly avoids any space leaks caused by retaining old type environments. Fixes #13352 #14848 #17381 #17551 #19664 #21443 #21444 #21720 #21898 #21946 #21959 #22125 #22160 #23010 #23062 #23063 Updates haddock submodule ------------------------- Metric Increase: MultiComponentModules MultiLayerModules MultiLayerModulesDefsGhci MultiLayerModulesNoCode T13701 T14697 hard_hole_fits -------------------------
* Allow WARNING pragmas to be controlled with custom categoriesAdam Gundry2023-03-243-17/+130
| | | | | | | | | | | | | | | | Closes #17209. This implements GHC Proposal 541, allowing a WARNING pragma to be annotated with a category like so: {-# WARNING in "x-partial" head "This function is undefined on empty lists." #-} The user can then enable, disable and set the severity of such warnings using command-line flags `-Wx-partial`, `-Werror=x-partial` and so on. There is a new warning group `-Wextended-warnings` containing all these warnings. Warnings without a category are treated as if the category was `deprecations`, and are (still) controlled by the flags `-Wdeprecations` and `-Wwarnings-deprecations`. Updates Haddock submodule.
* Compute LambdaFormInfo when using JavaScript backend.Luite Stegeman2023-03-211-3/+16
| | | | | | | | | | | CmmCgInfos is needed to write interface files, but the JavaScript backend does not generate it, causing "Name without LFInfo" warnings. This patch adds a conservative but always correct CmmCgInfos when the JavaScript backend is used. Fixes #23053
* Fix BCO creation setting caps when -j > -NTeo Camarasu2023-03-161-15/+0
| | | | | | | | | | | | | | * Remove calls to 'setNumCapabilities' in 'createBCOs' These calls exist to ensure that 'createBCOs' can benefit from parallelism. But this is not the right place to call `setNumCapabilities`. Furthermore the logic differs from that in the driver causing the capability count to be raised and lowered at each TH call if -j > -N. * Remove 'BCOOpts' No longer needed as it was only used to thread the job count down to `createBCOs` Resolves #23049
* Constraint simplification loop now depends on `ExpansionFuel`Apoorv Ingle2023-03-061-1/+18
| | | | | | | | | | | | | | instead of a boolean flag for `CDictCan.cc_pend_sc`. Pending givens get a fuel of 3 while Wanted and quantified constraints get a fuel of 1. This helps pending given constraints to keep up with pending wanted constraints in case of `UndecidableSuperClasses` and superclass expansions while simplifying the infered type. Adds 3 dynamic flags for controlling the fuels for each type of constraints `-fgivens-expansion-fuel` for givens `-fwanteds-expansion-fuel` for wanteds and `-fqcs-expansion-fuel` for quantified constraints Fixes #21909 Added Tests T21909, T21909b Added Note [Expanding Recursive Superclasses and ExpansionFuel]
* Add -single-threaded flag to force single threaded rtsOleg Grenrus2023-02-151-3/+4
| | | | | This is the small part of implementing https://github.com/ghc-proposals/ghc-proposals/pull/240
* compiler: fix generateCgIPEStub for no-tables-next-to-code buildsCheng Shao2023-02-141-14/+14
| | | | | | | | | generateCgIPEStub already correctly implements the CmmTick finding logic for when tables-next-to-code is on/off, but it used the wrong predicate to decide when to switch between the two. Previously it switches based on whether the codegen is unregisterised, but there do exist registerised builds that disable tables-next-to-code! This patch corrects that problem. Fixes #22896.
* Introduce warning for loopy superclass solvesheaf2023-02-141-0/+3
| | | | | | | | | | | | | | | | | Commit aed1974e completely re-engineered the treatment of loopy superclass dictionaries in instance declarations. Unfortunately, it has the potential to break (albeit in a rather minor way) user code. To alleviate migration concerns, this commit re-introduces the old behaviour. Any reliance on this old behaviour triggers a warning, controlled by `-Wloopy-superclass-solve`. The warning text explains that GHC might produce bottoming evidence, and provides a migration strategy. This allows us to provide a graceful migration period, alerting users when they are relying on this unsound behaviour. Fixes #22912 #22891 #20666 #22894 #22905
* Fix some correctness issues around tag inference when targeting the bytecode ↵Andreas Klebinger2023-02-141-0/+1
| | | | | | | | | generator. * Let binders are now always assumed untagged for bytecode. * Imported referenced are now always assumed to be untagged for bytecode. Fixes #22840
* Detect the `mold` linkerRebecca Turner2023-02-101-0/+1
| | | | Enables support for the `mold` linker by rui314.
* Revert "Don't keep exit join points so much"Matthew Pickering2023-02-081-24/+30
| | | | | | | | | | | | | | | | | | | | | | | | | This reverts commit caced75765472a1a94453f2e5a439dba0d04a265. It seems the patch "Don't keep exit join points so much" is causing wide-spread regressions in the bytestring library benchmarks. If I revert it then the 9.6 numbers are better on average than 9.4. See https://gitlab.haskell.org/ghc/ghc/-/issues/22893#note_479525 ------------------------- Metric Decrease: MultiComponentModules MultiComponentModulesRecomp MultiLayerModules MultiLayerModulesRecomp MultiLayerModulesTH_Make T12150 T13386 T13719 T21839c T3294 parsing001 -------------------------
* Accept an orphan declaration (sadly)Simon Peyton Jones2023-01-271-0/+3
| | | | | | | | | | This accepts the orphan type family instance type instance DsForeignHook = ... in GHC.HsToCore.Types. See Note [The Decoupling Abstract Data Hack] in GHC.Driver.Hooks
* Factorize hptModulesBelowSylvain Henry2023-01-261-35/+3
| | | | | Create and use moduleGraphModulesBelow in GHC.Unit.Module.Graph that doesn't need anything from the driver to be used.
* Key ModSummary cache by UnitId as well as FilePathMatthew Pickering2023-01-241-9/+10
| | | | | | | | | | | | | | | Multiple units can refer to the same files without any problem. Just another assumption which needs to be updated when we may have multiple home units. However, there is the invariant that within each unit each file only maps to one module, so as long as we also key the cache by UnitId then we are all good. This led to some confusing behaviour in GHCi when reloading, multipleHomeUnits_shared distils the essence of what can go wrong. Fixes #22679
* Improve driver diagnostic messages by including UnitId in messageMatthew Pickering2023-01-243-15/+16
| | | | | | | | | | Currently the driver diagnostics don't give any indication about which unit they correspond to. For example `-Wmissing-home-modules` can fire multiple times for each different home unit and gives no indication about which unit it's actually reporting about. Perhaps a longer term fix is to generalise the providence information away from a SrcSpan so that these kind of whole project errors can be reported with an accurate provenance. For now we can just include the `UnitId` in the error message. Fixes #22678
* Don't write o-boot files in Interactive modeMatthew Pickering2023-01-241-15/+27
| | | | | | | | We should not be producing object files when in interactive mode but we still produced the dummy o-boot files. These never made it into a `Linkable` but then confused the recompilation checker. Fixes #22669
* Recompilation checking: Don't try to find artefacts for Interactive & ↵Matthew Pickering2023-01-243-62/+74
| | | | | | | | | | hs-boot combo In interactive mode we don't produce any linkables for hs-boot files. So we also need to not going looking for them when we check to see if we have all the right objects needed for recompilation. Ticket #22669
* Use NodeKey rather than ModuleName in pruneCacheMatthew Pickering2023-01-241-3/+10
| | | | | | | | | | | The `pruneCache` function assumes that the list of `CachedInfo` all have unique `ModuleName`, this is not true: * In normal compilation, the same module name can appear for a file and it's boot file. * In multiple home unit compilation the same ModuleName can appear in different units The fix is to use a `NodeKey` as the actual key for the interfaces which includes `ModuleName`, `IsBoot` and `UnitId`. Fixes #22677
* Augment target filepath by working directory when checking if module ↵Matthew Pickering2023-01-241-1/+1
| | | | | | | | | | | | satisfies target This fixes a spurious warning in -Wmissing-home-modules. This is a simple oversight where when looking for the target in the first place we augment the search by the -working-directory flag but then fail to do so when checking this warning. Fixes #22676
* Set "since: 9.8" for TypeAbstractions and -Wterm-variable-captureVladislav Zavialov2023-01-231-1/+1
| | | | | These flags did not make it into the 9.6 release series, so the "since" annotations must be corrected.
* Minor corrections to commentsAdam Gundry2023-01-181-4/+4
|
* Refactor warning flag parsing to add missing flagsAdam Gundry2023-01-182-91/+111
| | | | | | | | This adds `-Werror=<group>` and `-fwarn-<group>` flags for warning groups as well as individual warnings. Previously these were defined on an ad hoc basis so for example we had `-Werror=compat` but not `-Werror=unused-binds`, whereas we had `-fwarn-unused-binds` but not `-fwarn-compat`. Fixes #22182.
* Enable -Wstar-is-type by default (#22759)Vladislav Zavialov2023-01-181-2/+1
| | | | | | | | | | | | | Following the plan in GHC Proposal #143 "Remove the * kind syntax", which states: In the next release (or 3 years in), enable -fwarn-star-is-type by default. The "next release" happens to be 9.6.1 I also moved the T21583 test case from should_fail to should_compile, because the only reason it was failing was -Werror=compat in our test suite configuration.
* Fix #22728: Not all diagnostics in safe check are fatalOleg Grenrus2023-01-121-13/+15
| | | | Also add tests for the issue and -Winferred-safe-imports in general
* Introduce the TypeAbstractions language flagVladislav Zavialov2023-01-111-0/+5
| | | | | | | | | | | | | | | | | | | | | | | | | GHC Proposals #448 "Modern scoped type variables" and #425 "Invisible binders in type declarations" introduce a new language extension flag: TypeAbstractions. Part of the functionality guarded by this flag has already been implemented, namely type abstractions in constructor patterns, but it was guarded by a combination of TypeApplications and ScopedTypeVariables instead of a dedicated language extension flag. This patch does the following: * introduces a new language extension flag TypeAbstractions * requires TypeAbstractions for @a-syntax in constructor patterns instead of TypeApplications and ScopedTypeVariables * creates a User's Guide page for TypeAbstractions and moves the "Type Applications in Patterns" section there To avoid a breaking change, the new flag is implied by ScopedTypeVariables and is retroactively added to GHC2021. Metric Decrease: MultiLayerModulesTH_OneShot
* Added a new warning about compatibility with RequiredTypeArgumentsHaskellMouse2023-01-112-1/+4
| | | | | | | | | This commit introduces a new warning that indicates code incompatible with future extension: RequiredTypeArguments. Enabling this extension may break some code and the warning will help to make it compatible in advance.
* packaging: Build perf builds with -split-sectionswip/various-hadrian-fixesMatthew Pickering2023-01-041-1/+1
| | | | | | | | | | | In 8f71d958 the make build system was made to use split-sections on linux systems but it appears this logic never made it to hadrian. There is the split_sections flavour transformer but this doesn't appear to be used for perf builds on linux. This is disbled on deb9 and windows due to #21670 Closes #21135
* compiler: Add -f[no-]split-sections flagsMatthew Pickering2023-01-041-8/+12
| | | | | | | | Here we add a `-fsplit-sections` flag which may some day replace `-split-sections`. This has the advantage of automatically providing a `-fno-split-sections` flag, which is useful for our packaging because we enable `-split-sections` by default but want to disable it in certain configurations.
* Force the Docs structure to prevent leaks in GHCi with -haddock without ↵wip/force-docsZubin Duggal2023-01-031-0/+4
| | | | | | | | | | -fwrite-interface Involves adding many new NFData instances. Without forcing Docs, references to the TcGblEnv for each module are retained by the Docs structure. Usually these are forced when the ModIface is serialised but not when we aren't writing the interface.
* JS: fix support for -outputdir (#22641)Sylvain Henry2022-12-223-23/+27
| | | | | | | The `-outputdir` option wasn't correctly handled with the JS backend because the same code path was used to handle both objects produced by the JS backend and foreign .js files. Now we clearly distinguish the two in the pipeline, fixing the bug.
* Make GHC.Driver.Main.hscTcRnLookupRdrName to return NonEmptyBodigrim2022-12-201-2/+7
|
* compiler: make .wasm the default executable extension on wasm32Cheng Shao2022-12-171-3/+4
| | | | Following convention as in other wasm toolchains. Fixes #22594.
* compiler: add optional tail-call support in wasm NCGCheng Shao2022-12-161-2/+1
| | | | | | | | | When the `-mtail-call` clang flag is passed at configure time, wasm tail-call extension is enabled, and the wasm NCG will emit `return_call`/`return_call_indirect` instructions to take advantage of it and avoid the `StgRun` trampoline overhead. Closes #22461.
* codeGen: Introduce ThreadSanitizer instrumentationBen Gamari2022-12-153-2/+7
| | | | | | This introduces a new Cmm pass which instruments the program with ThreadSanitizer annotations, allowing full tracking of mutator memory accesses via TSAN.
* JS: fix object file name comparison (#22578)Sylvain Henry2022-12-131-1/+8
|
* driver: Set correct UnitId when rehydrating modulesMatthew Pickering2022-12-091-2/+3
| | | | | | | | We were not setting the UnitId before rehydrating modules which just led to us attempting to find things in the wrong HPT. The test for this is the hadrian-multi command (which is now added as a CI job). Fixes #22222
* Add initial support for LoongArch Architecture.lrzlin2022-12-081-0/+1
|
View Lorry Controller Status