summaryrefslogtreecommitdiff
path: root/compiler/nativeGen
Commit message (Collapse)AuthorAgeFilesLines
...
* Cmm: Add support for undefined unwinding statementsBen Gamari2017-02-083-20/+37
| | | | | | | | | | | | | | And use to mark `stg_stack_underflow_frame`, which we are unable to determine a caller from. To simplify parsing at the moment we steal the `return` keyword to indicate an undefined unwind value. Perhaps this should be revisited. Reviewers: scpmw, simonmar, austin, erikd Subscribers: dfeuer, thomie Differential Revision: https://phabricator.haskell.org/D2738
* Generalize CmmUnwind and pass unwind information through NCGBen Gamari2017-02-087-56/+211
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | As discussed in D1532, Trac Trac #11337, and Trac Trac #11338, the stack unwinding information produced by GHC is currently quite approximate. Essentially we assume that register values do not change at all within a basic block. While this is somewhat true in normal Haskell code, blocks containing foreign calls often break this assumption. This results in unreliable call stacks, especially in the code containing foreign calls. This is worse than it sounds as unreliable unwinding information can at times result in segmentation faults. This patch set attempts to improve this situation by tracking unwinding information with finer granularity. By dispensing with the assumption of one unwinding table per block, we allow the compiler to accurately represent the areas surrounding foreign calls. Towards this end we generalize the representation of unwind information in the backend in three ways, * Multiple CmmUnwind nodes can occur per block * CmmUnwind nodes can now carry unwind information for multiple registers (while not strictly necessary; this makes emitting unwinding information a bit more convenient in the compiler) * The NCG backend is given an opportunity to modify the unwinding records since it may need to make adjustments due to, for instance, native calling convention requirements for foreign calls (see #11353). This sets the stage for resolving #11337 and #11338. Test Plan: Validate Reviewers: scpmw, simonmar, austin, erikd Subscribers: qnikst, thomie Differential Revision: https://phabricator.haskell.org/D2741
* Nix typo and redundant where-clausesRyan Scott2017-01-251-2/+0
|
* nativeGen: Use `foldl'` instead of `foldr` in free register accumulationBen Gamari2017-01-246-9/+13
| | | | | | | | | | | | | | | | | | | Manipulations of `FreeRegs` values are all just bit-operations on a word. Turning these `foldr`s into `foldl'`s has a very small but consistent effect on compiler allocations, ``` -1 s.d. ----- -0.065% +1 s.d. ----- -0.018% Average ----- -0.042% ``` Test Plan: Validate Reviewers: austin Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2966
* Fix terminal corruption bug and clean up SDoc interface.Phil Ruffwind2017-01-101-4/+3
| | | | | | | | | | | | | | | | | | | | | | | - Fix #13076 by wrapping `printDoc_` so that the terminal color is reset even if an exception occurs. - Add `printSDoc`, `printSDocLn`, and `bufLeftRenderSDoc` to keep `SDoc` values abstract (they are wrappers of `printDoc_`, `printDoc`, and `bufLeftRender` respectively). - Remove unused function: `printForAsm` Test Plan: manual Reviewers: RyanGlScott, austin, dfeuer, bgamari Reviewed By: dfeuer, bgamari Subscribers: dfeuer, mpickering, thomie Differential Revision: https://phabricator.haskell.org/D2932 GHC Trac Issues: #13076
* Typofixes in manual and comments [ci skip]Gabor Greif2017-01-041-1/+1
|
* Fix string merging with -split-sectionsSimon Brenner2016-12-161-3/+14
| | | | | | | | | | | | | | | | | | The added flags for string literal merging ended up printed in the middle of the section name when -split-sections was enabled. Break it up to put the flags after the name. Test Plan: validate with SplitSections=YES Reviewers: austin, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2865 GHC Trac Issues: #9577
* CLabel: Kill redundant UnitId argument from labelDynamicBen Gamari2016-12-161-8/+8
| | | | | | | | | | | | | | | | | | | | | | | | | It already has access to the current package's UnitId via the Module. Edward Yang pointed out that there is one wrinkle, however: the following invariant isn't true at all stages of compilation, if I am compiling the module (this_mod :: Module), then thisPackage dflags == moduleUnitId this_mod. Specifically, this is only true after desugaring; it may be broken when typechecking an indefinite signature. However, it's safe to assume this in the native codegen. I've updated Note to state this invariant more directly. Test Plan: Validate Reviewers: austin, ezyang, simonmar Reviewed By: ezyang, simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2863
* NCG: Implement trivColorable for PowerPC 64-bitPeter Trommler2016-12-091-4/+4
| | | | | | | | | | | | | | | Define constants for 64-bit PowerPC in graph coloring register allocator. Test Plan: ./validate Reviewers: simonmar, austin, erikd, bgamari, hvr Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2791
* BlockId: remove BlockMap and BlockSet synonymsMichal Terepeta2016-12-0814-23/+32
| | | | | | | | | | | | | | | | | | | | This continues removal of `BlockId` module in favor of Hoopl's `Label`. Most of the changes here are mechanical, apart from the orphan `Outputable` instances for `LabelMap` and `LabelSet`. For now I've moved them to `cmm/Hoopl`, since it's already trying to manage all imports from Hoopl (to avoid any collisions). Signed-off-by: Michal Terepeta <michal.terepeta@gmail.com> Test Plan: validate Reviewers: bgamari, austin, simonmar Reviewed By: simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2800
* nativeGen: Fix string merging on WindowsBen Gamari2016-12-081-3/+6
| | | | | | | | | | | | | | | | | | | D1290 places string constants in the `.rodata.str` section with `aMS` section flags so that the linker can merge them. However, it seems that ld doesn't understand these flags. It appears that `gcc -fmerge-constants` uses the `dr` flags on Windows. Make GHC do the same. Test Plan: Validate on Windows Reviewers: xnyhps, austin, Phyx Reviewed By: Phyx Subscribers: thomie, trommler Differential Revision: https://phabricator.haskell.org/D2797 GHC Trac Issues: #9577
* Reduce the size of string literals in binaries.Thijs Alkemade2016-12-064-5/+33
| | | | | | | | | | | | | | | Removed the alignment for strings and mark then as cstring sections in the generated asm so the linker can merge duplicate sections. Reviewers: rwbarton, trofi, austin, trommler, simonmar, hvr, bgamari Reviewed By: hvr, bgamari Subscribers: simonpj, hvr, thomie Differential Revision: https://phabricator.haskell.org/D1290 GHC Trac Issues: #9577
* Remove most functions from cmm/BlockIdMichal Terepeta2016-11-295-11/+12
| | | | | | | | | | | | | | | | | | | | It seems that `BlockId` module could simply go away in favor of Hoopl's `Label`. This is the first step to do that. In a few places I had to add some type signatures, but most of them seem to help with code readability. Signed-off-by: Michal Terepeta <michal.terepeta@gmail.com> Test Plan: ./validate Reviewers: austin, simonmar, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2765
* NCGMonad: Add MonadUnique NatM instanceBen Gamari2016-11-291-0/+8
| | | | | | | | | | | | Test Plan: Validate Reviewers: austin, simonmar Reviewed By: simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2737
* AsmCodeGen: Refactor worker in cmmNativeGensBen Gamari2016-11-291-10/+13
| | | | | | | | | | | | Test Plan: Validate Reviewers: austin, simonmar, michalt Reviewed By: simonmar, michalt Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2736
* Purge GHC of literate PerlDemi Obenour2016-11-291-1/+1
| | | | | | | | | | | | | | Test Plan: GHC CI Reviewers: austin, bgamari Reviewed By: bgamari Subscribers: snowleopard, thomie Maniphest Tasks: T74 Differential Revision: https://phabricator.haskell.org/D2732
* Correct spelling of command-line option in commentGabor Greif2016-11-171-1/+1
|
* Inline compiler/NOTES into X86/Ppr.hsMatthew Pickering2016-11-163-2/+20
| | | | | | | | Reviewers: austin, bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2721
* Test for unnecessary register spillsThomas Jakway2016-11-151-1/+12
| | | | | | | | | | | | Reviewers: mainland, simonmar, michalt, bgamari, austin Reviewed By: bgamari Subscribers: simonpj, mpickering, thomie Differential Revision: https://phabricator.haskell.org/D2638 GHC Trac Issues: #12744, #12745
* Typos in commentsGabor Greif2016-10-171-1/+1
|
* RegAlloc: Make some pattern matched completeJoachim Breitner2016-10-062-5/+8
| | | | | | | these actually are complete, but due to the use of pattern guards, the compiler does not see that. Refactor the code that it does. Differential Revision: https://phabricator.haskell.org/D2574
* PPC/CodeGen: fix lwa instruction generationPeter Trommler2016-10-011-4/+12
| | | | | | | | | | | | | | | | | | | Opcode lwa is a 64-bit opcode and allows a DS-form only. This patch generates lwa opcodes only when the offset is a multiple of 4. Fixes #12621 Test Plan: validate Reviewers: erikd, hvr, simonmar, austin, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2547 GHC Trac Issues: #12621
* CodeGen X86: fix unsafe foreign calls wrt inliningSylvain HENRY2016-10-011-36/+75
| | | | | | | | | | | | | | | | | | | | | | | | | Foreign calls (unsafe and safe) interact badly with inlining and register passing ABIs (see #11792 and #12614): the inlined code to compute a parameter of the call may overwrite a register already set to pass a preceding parameter. With this patch, we compute all parameters which are not simple expressions before assigning them to fixed registers required by the ABI. Test Plan: - Add test (test both reg and stack parameters) - Validate Reviewers: osa1, bgamari, austin, simonmar Reviewed By: simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2263 GHC Trac Issues: #11792, #12614
* Comments and manual only: spellingGabor Greif2016-09-281-1/+1
|
* Fix codegen bug in PIC version of genSwitch (#12433)Simon Marlow2016-09-151-3/+6
| | | | | | | | | | | | | | | | | | Summary: * getNonClobberedReg instead of getSomeReg, because the reg needs to survive across t_code * Use a new reg for the table offset calculation instead of clobbering the reg returned by expr (this was the bug affecting #12433) Test Plan: New unit test; validate Reviewers: rwbarton, bgamari, austin, erikd Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2529 GHC Trac Issues: #12433
* PPC NCG: Implement minimal stack frame header.Peter Trommler2016-08-312-8/+27
| | | | | | | | | | | | | | | | | | | | | | | | According to the ABI specifications a minimal stack frame consists of a header and a minimum size parameter save area. We reserve the minimal size for each ABI. On PowerPC 64-bil Linux and AIX the parameter save area can accomodate up to eight parameters. So calls with eight parameters and fewer can be done without allocating a new stack frame and deallocating that stack frame after the call. On AIX one additional spill slot is available on the stack. Code size for all nofib benchmarks is 0.3 % smaller on powerpc64. Test Plan: validate on AIX Reviewers: hvr!, erikd, austin, simonmar, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2445
* Revert "codeGen: Remove binutils<2.17 hack, fixes T11758"Simon Peyton Jones2016-08-192-3/+35
| | | | | | | This reverts commit e3e2e49a8f6952e1c8a19321c729c17b294d8c92. I'm reverting because it makes ghc-stage2 seg-fault on 64-bit Windows machines. Even ghc-stage2 --version seg-faults.
* RegAlloc: Use IntSet/IntMaps instead of generic Set/MapsÖmer Sinan Ağacan2016-08-063-28/+22
|
* RegAlloc: Remove duplicate seqList (use seqList from Util)Ömer Sinan Ağacan2016-08-061-14/+7
|
* AsmCodeGen: Give linear-scan and coloring reg. allocators different cc namesÖmer Sinan Ağacan2016-08-061-2/+2
|
* codeGen: Remove binutils<2.17 hack, fixes T11758Alex Dzyoba2016-08-052-35/+3
| | | | | | | | | | | | | | | | | | | There was a complication on the x86_64 platform, where pointers were 64 bits, but the tools didn't support 64-bit relative relocations. This was true before binutils 2.17, which nowadays is quite standart (even CentOs 5 is shipped with 2.17). Hacks were removed from x86 genSwitch and asm pretty printer. Also [x86-64-relative] note was dropped from includes/rts/storage/InfoTables.h as it's not referenced anywhere now. Reviewers: austin, simonmar, rwbarton, erikd, bgamari Reviewed By: simonmar, erikd, bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2426
* Document some codegen nondeterminismBartosz Nitka2016-07-077-15/+23
| | | | | | | | | Bit-for-bit reproducible binaries are not a goal for now, so this is just marking places that could be a problem. Doing this will allow eltsUFM to be removed and will leave only nonDetEltsUFM. GHC Trac: #4012
* Remove uniqSetToListBartosz Nitka2016-07-017-32/+61
| | | | | | | This documents nondeterminism in code generation and removes the nondeterministic ufmToList function. In the future someone will have to use nonDetEltsUFM (with proper explanation) or pprUFM.
* nativeGen: Allow -fregs-graph to be usedBen Gamari2016-06-301-4/+2
| | | | | | | | | | | | | | | | Previously the flag was silently ignored due the #7679 and #8657. This, however, seems unnecessarily brutal and makes experimentation unduly difficult for users. Test Plan: Validate Reviewers: austin, simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2335 GHC Trac Issues: #7679, #8657
* Remove ufmToListBartosz Nitka2016-06-305-20/+44
| | | | | | | This documents nondeterminism in code generation and removes the nondeterministic ufmToList function. In the future someone will have to use nonDetUFMToList (with proper explanation) or pprUFMWithKeys.
* Delete Ord UniqueBartosz Nitka2016-06-301-1/+22
| | | | | | | | | | | | | | | | | | | | | | Ord Unique can be a source of invisible, accidental nondeterminism as explained in Note [No Ord for Unique]. This removes it, leaving a note with rationale. It's unfortunate that I had to write Ord instances for codegen data structures by hand, but I believe that it's a right trade-off here. Test Plan: ./validate Reviewers: simonmar, austin, bgamari Reviewed By: simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2370 GHC Trac Issues: #4012
* Remove some `undefined`sÖmer Sinan Ağacan2016-06-271-5/+4
| | | | | | These get annoying when `undefined` is actually used as placeholder in WIP code. Some of these were also completely redundant (just call `deAnnotate'` instead of `deAnnotate` etc.).
* Provide Uniquable version of SCCBartosz Nitka2016-06-233-6/+6
| | | | | | | | | | | | | | | | | | | | | | | | | | | We want to remove the `Ord Unique` instance because there's no way to implement it in deterministic way and it's too easy to use by accident. We sometimes compute SCC for datatypes whose Ord instance is implemented in terms of Unique. The Ord constraint on SCC is just an artifact of some internal data structures. We can have an alternative implementation with a data structure that uses Uniquable instead. This does exactly that and I'm pleased that I didn't have to introduce any duplication to do that. Test Plan: ./validate I looked at performance tests and it's a tiny bit better. Reviewers: bgamari, simonmar, ezyang, austin, goldfire Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2359 GHC Trac Issues: #4012
* PPC NCG: Fix and refactor TOC handling.Peter Trommler2016-06-191-28/+28
| | | | | | | | | | | | | | | | | | | In a call to a fixed function the TOC does not need to be saved. The linker handles TOC saving. Refactor TOC handling by folding the two functions toc_before and toc_after into the code generating the call sequence. This saves repeating the case distinction in those two functions. Test Plan: validate on PowerPC 32-bit Linux and AIX Reviewers: hvr, simonmar, austin, erikd, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2328
* PPC NCG: Fix float parameter passing on 64-bit.Peter Trommler2016-06-192-7/+22
| | | | | | | | | | | | | | | | | | | | | | | | On Linux 64-bit PowerPC the first 13 floating point parameters are passed in registers. We only passed the first 8 floating point params. The alignment of a floating point single precision value in ELF v1.9 is the second word of a doubleword. For ELF v2 we support only little endian and the least significant word of a doubleword is the first word, so no special handling is required. Add a regression test. Test Plan: validate on powerpc Linux and AIX Reviewers: erikd, hvr, austin, simonmar, bgamari Reviewed By: simonmar Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2327 GHC Trac Issues: #12134
* Fix trac #10647: Notice about lack of SIMD supportSeraphime Kirkovski2016-06-191-1/+3
| | | | | | | | | | | | | | | | | | | | | | | | | Fixes #10647. Changes the error message when a SIMD size variable is required in the native code generation backend. Test Plan: Try compiling the test case given in the ticket : {-# LANGUAGE MagicHash #-} module Foo where import GHC.Prim data V = V Int8X16# GHC should give a clearer error message Reviewers: austin, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2325 GHC Trac Issues: #10647
* Kill foldUniqSetBartosz Nitka2016-06-062-3/+4
| | | | | | | | | | | | | | | I planned to just say that we don't care about this part. Turns out I was able to document away the uses in the codegenerator. Test Plan: ./validate Reviewers: simonmar, austin, bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2307 GHC Trac Issues: #4012
* PPC NCG: Fix pretty printing of st[wd]ux instr.Peter Trommler2016-05-161-1/+2
| | | | | | | | | | | | | | | | | Printing STU was mixed up. The tab character must appear after the 'x'. Test Plan: validate on powerpc Reviewers: bgamari, austin, erikd Reviewed By: austin, erikd Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2214 GHC Trac Issues: #12054
* PPC NCG: Improve pointer de-tagging codePeter Trommler2016-04-293-5/+36
| | | | | | | | | | | | | | | | Generate a clrr[wd]i instruction to clear the tag bits in a pointer. This saves one instruction and one temporary register. Optimize signed comparison with zero after andi. operation This saves one instruction when comparing a pointer tag with zero. This reduces code size by 0.6 % in all nofib benchmarks. Test Plan: validate on AIX and 32-bit Linux Reviewed By: erikd, hvr Differential Revision: https://phabricator.haskell.org/D2093
* Kill mapUniqSetBartosz Nitka2016-04-281-6/+2
| | | | | | | | | | | | | | Note [Unsound mapUniqSet] explains why it got removed. Test Plan: build ghc Reviewers: simonpj, austin, bgamari Reviewed By: bgamari Subscribers: thomie, simonmar Differential Revision: https://phabricator.haskell.org/D2152
* Reduce default for -fmax-pmcheck-iterations from 1e7 to 2e6Herbert Valerio Riedel2016-04-101-0/+6
| | | | | | | | | | | | | | | | | | | | | | | | The commit 28f951edfe50ea5182065144340061ec326781f5 introduced the `-fmax-pmcheck-iterations` flag and set the default limit to 1e7 iterations. However, this value is still high enough that it can result GHC to exhibit memory spikes beyond 1 GiB of RAM usage (heap profile showed several `(:)`s, as well as `THUNK_2_0`, and `PmCon` during the memory spikes) A value of 2e6 seems to be a safer upper bound which still manages to let the checker not run into the limit in most cases. Test Plan: Validate, try building a few Hackage packages Reviewers: austin, gkaracha, bgamari Reviewed By: bgamari Subscribers: thomie Differential Revision: https://phabricator.haskell.org/D2095
* Fix AIX/ppc codegen in `-prof` compilation modeHerbert Valerio Riedel2016-03-271-2/+8
| | | | | | | The implementation in df26b95559fd467abc0a3a4151127c95cb5011b9 wrongly assumed that all C-ABI subroutine calls would use a 'ForeignLabel' but it turns out that calls inserted via 'emitRtsCall' use 'CmmLabel's instead.
* Remove code-duplication in the PPC NCGHerbert Valerio Riedel2016-03-241-26/+19
| | | | | | Reviewed By: bgamari, trommler Differential Revision: https://phabricator.haskell.org/D2020
* Add NCG support for AIX/ppc32Herbert Valerio Riedel2016-03-245-10/+205
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This extends the previous work to revive the unregisterised GHC build for AIX/ppc32. Strictly speaking, AIX runs on POWER4 (and later) hardware, but the PPC32 instructions implemented in the PPC NCG represent a compatible subset of the POWER4 ISA. IBM AIX follows the PowerOpen ABI (and shares many similiarites with the Linux PPC64 ELF V1 NCG backend) but uses the rather limited XCOFF format (compared to ELF). This doesn't support dynamic libraries yet. A major limiting factor is that the AIX assembler does not support the `@ha`/`@l` relocation types nor the ha16()/lo16() functions Darwin's assembler supports. Therefore we need to avoid emitting those. In case of numeric literals we simply compute the functions ourselves, while for labels we have to use local TOCs and hope everything fits into a 16bit offset (for ppc32 this gives us at most 16384 entries per TOC section, which is enough to compile GHC). Another issue is that XCOFF doesn't seem to have a relocation type for label-differences, and therefore the label-differences placed into tables-next-to-code can't be relocated, but the linker may rearrange different sections, so we need to place all read-only sections into the same `.text[PR]` section to workaround this. Finally, the PowerOpen ABI distinguishes between function-descriptors and actualy entry-point addresses. For AIX we need to be specific when emitting assembler code whether we want the address of the function descriptor `printf`) or for the entry-point (`.printf`). So we let the asm pretty-printer prefix a dot to all emitted subroutine calls (i.e. `BL`) on AIX only. For now, STG routines' entry-point labels are not prefixed by a label and don't have any associated function-descriptor. Reviewers: austin, trommler, erikd, bgamari Reviewed By: trommler, erikd, bgamari Differential Revision: https://phabricator.haskell.org/D2019
* PPC NCG: Emit more portable `fcmpu 0, ...` instead of `fcmpu cr0, ...`Herbert Valerio Riedel2016-03-211-2/+6
| | | | | | | Use `fcmpu 0, ...` rather than `fcmpu cr0, ...` for better portability since some non-GNU assembler (such as IBM's `as`) tend to not support the symbolic register name `cr0`. This matches the syntax that GCC emits for PPC targets.
View Lorry Controller Status