summaryrefslogtreecommitdiff
path: root/compiler/GHC/Cmm
Commit message (Collapse)AuthorAgeFilesLines
* Enable USE_INLINE_SRT_FIELD on ARM64Sylvain Henry2022-05-301-11/+2
| | | | | | | It was previously disabled because of: - a confusion about "SRT inlining" (see removed comment in this commit) - a linker bug (overflow) in the handling of ARM64_RELOC_SUBTRACTOR relocation: fixed by a previous commit.
* Some fixes to SRT documentationSylvain Henry2022-05-301-25/+50
| | | | | | | | | - reordered the 3 SRT implementation cases from the most general to the most specific one: USE_SRT_POINTER -> USE_SRT_OFFSET -> USE_INLINE_SRT_FIELD - added requirements for each - found and documented a confusion about "SRT inlining" not supported with MachO. (It is fixed in the following commit)
* Change `Backend` type and remove direct dependencieswip/backend-as-recordNorman Ramsey2022-05-211-8/+1
| | | | | | | | | | | | | | | | | | | With this change, `Backend` becomes an abstract type (there are no more exposed value constructors). Decisions that were formerly made by asking "is the current back end equal to (or different from) this named value constructor?" are now made by interrogating the back end about its properties, which are functions exported by `GHC.Driver.Backend`. There is a description of how to migrate code using `Backend` in the user guide. Clients using the GHC API can find a backdoor to access the Backend datatype in GHC.Driver.Backend.Internal. Bumps haddock submodule. Fixes #20927
* document fields of `DominatorSet`Andreas Klebinger2022-05-201-2/+2
|
* add HasDebugCallStack; remove unneeded extensionsNorman Ramsey2022-05-201-16/+19
|
* add dominator-tree functionNorman Ramsey2022-05-201-0/+12
|
* add dominator analysis of `CmmGraph`Norman Ramsey2022-05-201-0/+200
| | | | | | | | | | | | This commit adds module `GHC.Cmm.Dominators`, which provides a wrapper around two existing algorithms in GHC: the Lengauer-Tarjan dominator analysis from the X86 back end and the reverse postorder ordering from the Cmm Dataflow framework. Issue #20726 proposes that we evaluate some alternatives for dominator analysis, but for the time being, the best path forward is simply to use the existing analysis on `CmmGraph`s. This commit addresses a bullet in #21200.
* Don't store LlvmConfig into DynFlagsSylvain Henry2022-05-171-15/+15
| | | | | | | | | | | | | | | | | | | | | LlvmConfig contains information read from llvm-passes and llvm-targets files in GHC's top directory. Reading these files is done only when needed (i.e. when the LLVM backend is used) and cached for the whole compiler session. This patch changes the way this is done: - Split LlvmConfig into LlvmConfig and LlvmConfigCache - Store LlvmConfigCache in HscEnv instead of DynFlags: there is no good reason to store it in DynFlags. As it is fixed per session, we store it in the session state instead (HscEnv). - Initializing LlvmConfigCache required some changes to driver functions such as newHscEnv. I've used the opportunity to untangle initHscEnv from initGhcMonad (in top-level GHC module) and to move it to GHC.Driver.Main, close to newHscEnv. - I've also made `cmmPipeline` independent of HscEnv in order to remove the call to newHscEnv in regalloc_unit_tests.
* codeGen: Ensure that static datacon apps are included in SRTsBen Gamari2022-05-171-43/+86
| | | | | | | | | | | | | | | | When generating an SRT for a recursive group, GHC.Cmm.Info.Build.oneSRT filters out recursive references, as described in Note [recursive SRTs]. However, doing so for static functions would be unsound, for the reason described in Note [Invalid optimisation: shortcutting]. However, the same argument applies to static data constructor applications, as we discovered in #20959. Fix this by ensuring that static data constructor applications are included in recursive SRTs. The approach here is not entirely satisfactory, but it is a starting point. Fixes #20959.
* CafAnal: Improve code clarityBen Gamari2022-05-172-100/+131
| | | | | | | | | | Here we implement a few measures to improve the clarity of the CAF analysis implementation. Specifically: * Use CafInfo instead of Bool since the former is more descriptive * Rename CAFLabel to CAFfyLabel, since not all CAFfyLabels are in fact CAFs * Add numerous comments
* Move CmmParserConfig and PDConfig into GHC.Cmm.Parser.ConfigAndre Marianiello2022-05-123-11/+27
|
* Decouple dynflags in Cmm parser (related to #17957)Andre Marianiello2022-05-122-23/+29
|
* Add a Note describing lack of object merging on WindowsBen Gamari2022-04-061-6/+6
| | | | See #21068.
* Refactor handling of global initializersBen Gamari2022-04-014-2/+147
| | | | | | | | | | | | | | | | | | | | | | | | | | | GHC uses global initializers for a number of things including cost-center registration, info-table provenance registration, and setup of foreign exports. Previously, the global initializer arrays which referenced these initializers would live in the object file of the C stub, which would then be merged into the main object file of the module. Unfortunately, this approach is no longer tenable with the move to Clang/LLVM on Windows (see #21019). Specifically, lld's PE backend does not support object merging (that is, the -r flag). Instead we are now rather packaging a module's object files into a static library. However, this is problematic in the case of initializers as there are no references to the C stub object in the archive, meaning that the linker may drop the object from the final link. This patch refactors our handling of global initializers to instead place initializer arrays within the object file of the module to which they belong. We do this by introducing a Cmm data declaration containing the initializer array in the module's Cmm stream. While the initializer functions themselves remain in separate C stub objects, the reference from the module's object ensures that they are not dropped from the final link. In service of #21068.
* Fix all invalid haddock comments in the compilerZubin Duggal2022-03-293-12/+12
| | | | Fixes #20935 and #20924
* Avoid redundant imports of GHC.Driver.SessionSylvain Henry2022-03-232-2/+0
| | | | | | Remove GHC.Driver.Session imports that weren't considered as redundant because of the reexport of PlatformConstants. Also remove this reexport as modules using this datatype should import GHC.Platform instead.
* Ticky profiling improvements.Matthew Pickering2022-03-021-0/+1
| | | | | | | | | | | | | | | | | | | | This adds a number of changes to ticky-ticky profiling. When an executable is profiled with IPE profiling it's now possible to associate id-related ticky counters to their source location. This works by emitting the info table address as part of the counter which can be looked up in the IPE table. Add a `-ticky-ap-thunk` flag. This flag prevents the use of some standard thunks which are precompiled into the RTS. This means reduced cache locality and increased code size. But it allows better attribution of execution cost to specific source locations instead of simple attributing it to the standard thunk. ticky-ticky now uses the `arg` field to emit additional information about counters in json format. When ticky-ticky is used in combination with the eventlog eventlog2html can be used to generate a html table from the eventlog similar to the old text output for ticky-ticky.
* CLabel cleanup:Andreas Klebinger2022-02-282-14/+6
| | | | | | | | Remove these smart constructors for these reasons: * mkLocalClosureTableLabel : Does the same as the non-local variant. * mkLocalClosureLabel : Does the same as the non-local variant. * mkLocalInfoTableLabel : Decide if we make a local label based on the name and just use mkInfoTableLabel everywhere.
* Suggestions due to hlintMatthew Pickering2022-02-242-2/+0
| | | | | It turns out this job hasn't been running for quite a while (perhaps ever) so there are quite a few failures when running the linter locally.
* Tag inference work.Andreas Klebinger2022-02-123-10/+41
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This does three major things: * Enforce the invariant that all strict fields must contain tagged pointers. * Try to predict the tag on bindings in order to omit tag checks. * Allows functions to pass arguments unlifted (call-by-value). The former is "simply" achieved by wrapping any constructor allocations with a case which will evaluate the respective strict bindings. The prediction is done by a new data flow analysis based on the STG representation of a program. This also helps us to avoid generating redudant cases for the above invariant. StrictWorkers are created by W/W directly and SpecConstr indirectly. See the Note [Strict Worker Ids] Other minor changes: * Add StgUtil module containing a few functions needed by, but not specific to the tag analysis. ------------------------- Metric Decrease: T12545 T18698b T18140 T18923 LargeRecord Metric Increase: LargeRecord ManyAlternatives ManyConstructors T10421 T12425 T12707 T13035 T13056 T13253 T13253-spj T13379 T15164 T18282 T18304 T18698a T1969 T20049 T3294 T4801 T5321FD T5321Fun T783 T9233 T9675 T9961 T19695 WWRec -------------------------
* Introduce alignment to CmmStoreBen Gamari2022-02-047-15/+22
|
* Introduce alignment in CmmLoadBen Gamari2022-02-0412-44/+60
|
* cmm: Introduce cmmLoadBWord and cmmLoadGCWordBen Gamari2022-02-044-5/+13
|
* Check type first then expressionPeter Trommler2022-02-021-1/+1
|
* Cmm: fix equality of expressionsPeter Trommler2022-02-021-1/+1
| | | | | Compare expressions and types when comparing `CmmLoad`s. Fixes #21016
* Fix a few Note inconsistenciesBen Gamari2022-02-0115-60/+39
|
* Consistently upper-case "Note ["Ben Gamari2022-02-011-1/+1
| | | | | | This was achieved with git ls-tree --name-only HEAD -r | xargs sed -i -e 's/note \[/Note \[/g'
* StgToCmm: decouple DynFlags, add StgToCmmConfigdoyougnu2022-01-314-47/+35
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | StgToCmm: add Config, remove CgInfoDownwards StgToCmm: runC api change to take StgToCmmConfig StgToCmm: CgInfoDownad -> StgToCmmConfig StgToCmm.Monad: update getters/setters/withers StgToCmm: remove CallOpts in StgToCmm.Closure StgToCmm: remove dynflag references StgToCmm: PtrOpts removed StgToCmm: add TMap to config, Prof - dynflags StgToCmm: add omit yields to config StgToCmm.ExtCode: remove redundant import StgToCmm.Heap: remove references to dynflags StgToCmm: codeGen api change, DynFlags -> Config StgToCmm: remove dynflags in Env and StgToCmm StgToCmm.DataCon: remove dynflags references StgToCmm: remove dynflag references in DataCon StgToCmm: add backend avx flags to config StgToCmm.Prim: remove dynflag references StgToCmm.Expr: remove dynflag references StgToCmm.Bind: remove references to dynflags StgToCmm: move DoAlignSanitisation to Cmm.Type StgToCmm: remove PtrOpts in Cmm.Parser.y DynFlags: update ipInitCode api StgToCmm: Config Module is single source of truth StgToCmm: Lazy config breaks IORef deadlock testsuite: bump countdeps threshold StgToCmm.Config: strictify fields except UpdFrame Strictifying UpdFrameOffset causes the RTS build with stage1 to deadlock. Additionally, before the deadlock performance of the RTS is noticeably slower. StgToCmm.Config: add field descriptions StgToCmm: revert strictify on Module in config testsuite: update CountDeps tests StgToCmm: update comment, fix exports Specifically update comment about loopification passed into dynflags then stored into stgToCmmConfig. And remove getDynFlags from Monad.hs exports Types.Name: add pprFullName function StgToCmm.Ticky: use pprFullname, fixup ExtCode imports Cmm.Info: revert cmmGetClosureType removal StgToCmm.Bind: use pprFullName, Config update comments StgToCmm: update closureDescription api StgToCmm: SAT altHeapCheck StgToCmm: default render for Info table, ticky Use default rendering contexts for info table and ticky ticky, which should be independent of command line input. testsuite: bump count deps pprFullName: flag for ticky vs normal style output convertInfoProvMap: remove unused parameter StgToCmm.Config: add backend flags to config StgToCmm.Config: remove Backend from Config StgToCmm.Prim: refactor Backend call sites StgToCmm.Prim: remove redundant imports StgToCmm.Config: refactor vec compatibility check StgToCmm.Config: add allowQuotRem2 flag StgToCmm.Ticky: print internal names with parens StgToCmm.Bind: dispatch ppr based on externality StgToCmm: Add pprTickyname, Fix ticky naming Accidently removed the ctx for ticky SDoc output. The only relevant flag is sdocPprDebug which was accidental set to False due to using defaultSDocContext without altering the flag. StgToCmm: remove stateful fields in config fixup: config: remove redundant imports StgToCmm: move Sequel type to its own module StgToCmm: proliferate getCallMethod updated api StgToCmm.Monad: add FCodeState to Monad Api StgToCmm: add second reader monad to FCode fixup: Prim.hs: missed a merge conflict fixup: Match countDeps tests to HEAD StgToCmm.Monad: withState -> withCgState To disambiguate it from mtl withState. This withState shouldn't be returning the new state as a value. However, fixing this means tackling the knot tying in CgState and so is very difficult since it changes when the thunk of the knot is forced which either leads to deadlock or to compiler panic.
* Fix typosKrzysztof Gogolewski2021-12-251-1/+1
|
* Cmm: DynFlags to CmmConfig refactordoyougnu2021-12-226-101/+121
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | add files GHC.Cmm.Config, GHC.Driver.Config.Cmm Cmm: DynFlag references --> CmmConfig Cmm.Pipeline: reorder imports, add handshake Cmm: DynFlag references --> CmmConfig Cmm.Pipeline: DynFlag references --> CmmConfig Cmm.LayoutStack: DynFlag references -> CmmConfig Cmm.Info.Build: DynFlag references -> CmmConfig Cmm.Config: use profile to retrieve platform Cmm.CLabel: unpack NCGConfig in labelDynamic Cmm.Config: reduce CmmConfig surface area Cmm.Config: add cmmDoCmmSwitchPlans field Cmm.Config: correct cmmDoCmmSwitchPlans flag The original implementation dispatches work in cmmImplementSwitchPlans in an `otherwise` branch, hence we must add a not to correctly dispatch Cmm.Config: add cmmSplitProcPoints simplify Config remove cmmBackend, and cmmPosInd Cmm.CmmToAsm: move ncgLabelDynamic to CmmToAsm Cmm.CLabel: remove cmmLabelDynamic function Cmm.Config: rename cmmOptDoLinting -> cmmDoLinting testsuite: update CountDepsAst CountDepsParser
* codeGen: Introduce flag to bounds-check array accessesBen Gamari2021-12-211-1/+3
| | | | | | | Here we introduce code generator support for instrument array primops with bounds checking, enabled with the `-fcheck-prim-bounds` flag. Introduced to debug #20769.
* generalize GHC.Cmm.Dataflow to work over any node typeNorman Ramsey2021-12-071-31/+46
| | | | | | See #20725. The commit includes source-code changes and a test case.
* Require all dirty_MUT_VAR callers to do explicit stg_MUT_VAR_CLEAN_info ↵nineonine2021-12-021-1/+4
| | | | comparison (#20088)
* nativeGen/aarch64: Fix handling of subword valuesBen Gamari2021-12-022-3/+39
| | | | | | | | | | | | | | | | Here we rework the handling of sub-word operations in the AArch64 backend, fixing a number of bugs and inconsistencies. In short, we now impose the invariant that all subword values are represented in registers in zero-extended form. Signed arithmetic operations are then responsible for sign-extending as necessary. Possible future work: * Use `CMP`s extended register form to avoid burning an instruction in sign-extending the second operand. * Track sign-extension state of registers to elide redundant sign extensions in blocks with frequent sub-word signed arithmetic.
* cmm/opt: Fold away shifts larger than shiftee widthBen Gamari2021-12-021-2/+12
| | | | | This is necessary for lint-correctness since we no longer allow such shifts in Cmm.
* cmm: Disallow shifts larger than shifteeBen Gamari2021-12-022-0/+19
| | | | | | | | | | | | Previously primops.txt.pp stipulated that the word-size shift primops were only defined for shift offsets in [0, word_size). However, there was no further guidance for the definition of Cmm's sub-word size shift MachOps. Here we fix this by explicitly disallowing (checked in many cases by CmmLint) shift operations where the shift offset is larger than the shiftee. This is consistent with LLVM's shift operations, avoiding the miscompilation noted in #20637.
* cmm: narrow when folding signed quotientsBen Gamari2021-12-021-2/+2
| | | | | | | | | | | | | | | | | | Previously the constant-folding behavior for MO_S_Quot and MO_S_Rem failed to narrow its arguments, meaning that a program like: %zx64(%quot(%lobits8(0x00e1::bits16), 3::bits8)) would be miscompiled. Specifically, this program should reduce as %lobits8(0x00e1::bits16) == -31 %quot(%lobits8(0x00e1::bits16), 3::bits8) == -10 %zx64(%quot(%lobits8(0x00e1::bits16), 3::bits8)) == 246 However, with this bug the `%lobits8(0x00e1::bits16)` would instead be treated as `+31`, resulting in the incorrect result of `75`. (cherry picked from commit 94e197e3dbb9a48991eb90a03b51ea13d39ba4cc)
* rts: Refactor SRT representation selectionBen Gamari2021-12-012-9/+31
| | | | | | The goal here is to make the SRT selection logic a bit clearer and allow configurations which we currently don't support (e.g. using a full word in the info table even when TNTC is used).
* Add specific optimization flag for Cmm control flow analysis (#20500)Gergo ERDI2021-11-251-1/+1
|
* CmmSink: Be more aggressive in removing no-op assignments.Andreas Klebinger2021-11-231-3/+57
| | | | | | | | | | No-op assignments like R1 = R1 are not only wasteful. They can also inhibit other optimizations like inlining assignments that read from R1. We now check for assignments being a no-op before and after we simplify the RHS in Cmm sink which should eliminate most of these no-ops.
* compiler: Improve Haddocks of atomic MachOpsBen Gamari2021-10-041-2/+4
|
* Use Info Table Provenances to decode cloned stack (#18163)Sven Tennie2021-09-231-0/+2
| | | | | | | | | | | | | | | | Emit an Info Table Provenance Entry (IPE) for every stack represeted info table if -finfo-table-map is turned on. To decode a cloned stack, lookupIPE() is used. It provides a mapping between info tables and their source location. Please see these notes for details: - [Stacktraces from Info Table Provenance Entries (IPE based stack unwinding)] - [Mapping Info Tables to Source Positions] Metric Increase: T12545
* Move `/includes` to `/rts/include`, sort per package betterJohn Ericson2021-08-092-6/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | In order to make the packages in this repo "reinstallable", we need to associate source code with a specific packages. Having a top level `/includes` dir that mixes concerns (which packages' includes?) gets in the way of this. To start, I have moved everything to `rts/`, which is mostly correct. There are a few things however that really don't belong in the rts (like the generated constants haskell type, `CodeGen.Platform.h`). Those needed to be manually adjusted. Things of note: - No symlinking for sake of windows, so we hard-link at configure time. - `CodeGen.Platform.h` no longer as `.hs` extension (in addition to being moved to `compiler/`) so as not to confuse anyone, since it is next to Haskell files. - Blanket `-Iincludes` is gone in both build systems, include paths now more strictly respect per-package dependencies. - `deriveConstants` has been taught to not require a `--target-os` flag when generating the platform-agnostic Haskell type. Make takes advantage of this, but Hadrian has yet to.
* Delete ToDo about incorrect optimisation [skip ci]Peter Trommler2021-07-291-4/+0
| | | | | On big-endian systems a narrow after a load cannot be replaced with a narrow load.
* Cmm.Opt: Fix type of shift amount in constant foldingBen Gamari2021-07-231-2/+2
| | | | | | | | Previously the `MO_S_Quot` constant folding rule would incorrectly pass the shift amount of the same width as the shifted value. However, the machop's type expects the shift amount to be a Word. Fixes #20142.
* Add Word64#/Int64# primopsSylvain Henry2021-07-151-0/+35
| | | | | | | | | | | | | | | | | | | | | | | Word64#/Int64# are only used on 32-bit architectures. Before this patch, operations on these types were directly using the FFI. Now we use real primops that are then lowered into ccalls. The advantage of doing this is that we can now perform constant folding on Word64#/Int64# (#19024). Most of this work was done by John Ericson in !3658. However this patch doesn't go as far as e.g. changing Word64 to always be using Word64#. Noticeable performance improvements T9203(normal) run/alloc 89870808.0 66662456.0 -25.8% GOOD haddock.Cabal(normal) run/alloc 14215777340.8 12780374172.0 -10.1% GOOD haddock.base(normal) run/alloc 15420020877.6 13643834480.0 -11.5% GOOD Metric Decrease: T9203 haddock.Cabal haddock.base
* Rename getErrorMessages and getMessages function in parser codewip/adinapoli-issue-19920Alfredo Di Napoli2021-07-081-2/+2
| | | | | | | | | | | This commit renames the `getErrorMessages` and `getMessages` function in the parser code to `getPsErrorMessages` and `getPsMessages`, to avoid import conflicts, as we have already `getErrorMessages` and `getMessages` defined in `GHC.Types.Error`. Fixes #19920. Update haddock submodule
* Put tracing functions into their own moduleSylvain Henry2021-06-221-2/+1
| | | | | | | | Now that Outputable is independent of DynFlags, we can put tracing functions using SDocs into their own module that doesn't transitively depend on any GHC.Driver.* module. A few modules needed to be moved to avoid loops in DEBUG mode.
* Make Logger independent of DynFlagsSylvain Henry2021-06-074-57/+53
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Introduce LogFlags as a independent subset of DynFlags used for logging. As a consequence in many places we don't have to pass both Logger and DynFlags anymore. The main reason for this refactoring is that I want to refactor the systools interfaces: for now many systools functions use DynFlags both to use the Logger and to fetch their parameters (e.g. ldInputs for the linker). I'm interested in refactoring the way they fetch their parameters (i.e. use dedicated XxxOpts data types instead of DynFlags) for #19877. But if I did this refactoring before refactoring the Logger, we would have duplicate parameters (e.g. ldInputs from DynFlags and linkerInputs from LinkerOpts). Hence this patch first. Some flags don't really belong to LogFlags because they are subsystem specific (e.g. most DumpFlags). For example -ddump-asm should better be passed in NCGConfig somehow. This patch doesn't fix this tight coupling: the dump flags are part of the UI but they are passed all the way down for example to infer the file name for the dumps. Because LogFlags are a subset of the DynFlags, we must update the former when the latter changes (not so often). As a consequence we now use accessors to read/write DynFlags in HscEnv instead of using `hsc_dflags` directly. In the process I've also made some subsystems less dependent on DynFlags: - CmmToAsm: by passing some missing flags via NCGConfig (see new fields in GHC.CmmToAsm.Config) - Core.Opt.*: - by passing -dinline-check value into UnfoldingOpts - by fixing some Core passes interfaces (e.g. CallArity, FloatIn) that took DynFlags argument for no good reason. - as a side-effect GHC.Core.Opt.Pipeline.doCorePass is much less convoluted.
* Adds AArch64 Native Code GeneratorMoritz Angermann2021-06-055-18/+58
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In which we add a new code generator to the Glasgow Haskell Compiler. This codegen supports ELF and Mach-O targets, thus covering Linux, macOS, and BSDs in principle. It was tested only on macOS and Linux. The NCG follows a similar structure as the other native code generators we already have, and should therfore be realtively easy to follow. It supports most of the features required for a proper native code generator, but does not claim to be perfect or fully optimised. There are still opportunities for optimisations. Metric Decrease: ManyAlternatives ManyConstructors MultiLayerModules PmSeriesG PmSeriesS PmSeriesT PmSeriesV T10421 T10421a T10858 T11195 T11276 T11303b T11374 T11822 T12227 T12545 T12707 T13035 T13253 T13253-spj T13379 T13701 T13719 T14683 T14697 T15164 T15630 T16577 T17096 T17516 T17836 T17836b T17977 T17977b T18140 T18282 T18304 T18478 T18698a T18698b T18923 T1969 T3064 T5030 T5321FD T5321Fun T5631 T5642 T5837 T783 T9198 T9233 T9630 T9872d T9961 WWRec Metric Increase: T4801
View Lorry Controller Status