GitLab

Summary:
This patch implements a new code layout algorithm.
It has been tested for x86 and is disabled on other platforms.

Performance varies slightly be CPU/Machine but in general seems to be better
by around 2%.
Nofib shows only small differences of about +/- ~0.5% overall depending on
flags/machine performance in other benchmarks improved significantly.

Other benchmarks includes at least the benchmarks of: aeson, vector, megaparsec, attoparsec,
containers, text and xeno.

While the magnitude of gains differed three different CPUs where tested with
all getting faster although to differing degrees. I tested: Sandy Bridge(Xeon), Haswell,
Skylake

* Library benchmark results summarized:
  * containers: ~1.5% faster
  * aeson: ~2% faster
  * megaparsec: ~2-5% faster
  * xml library benchmarks: 0.2%-1.1% faster
  * vector-benchmarks: 1-4% faster
  * text: 5.5% faster

On average GHC compile times go down, as GHC compiled with the new layout
is faster than the overhead introduced by using the new layout algorithm,

Things this patch does:

* Move code responsilbe for block layout in it's own module.
* Move the NcgImpl Class into the NCGMonad module.
* Extract a control flow graph from the input cmm.
* Update this cfg to keep it in sync with changes during
  asm codegen. This has been tested on x64 but should work on x86.
  Other platforms still use the old codelayout.
* Assign weights to the edges in the CFG based on type and limited static
  analysis which are then used for block layout.
* Once we have the final code layout eliminate some redundant jumps.

  In particular turn a sequences of:
      jne .foo
      jmp .bar
    foo:
  into
      je bar
    foo:
      ..

Test Plan: ci

Reviewers: bgamari, jmct, jrtc27, simonmar, simonpj, RyanGlScott

Reviewed By: RyanGlScott

Subscribers: RyanGlScott, trommler, jmct, carter, thomie, rwbarton

GHC Trac Issues: #15124

Differential Revision: https://phabricator.haskell.org/D4726

Test Plan:
I can't validate this because of existing errors with the debug runtime. I'll
see if this introduces any new failures.

Reviewers: simonmar, bgamari, erikd

Reviewed By: simonmar

Subscribers: rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5337

We now allocate the key to spt on C stack rather than in Haskell heap,
avoiding allocating in `unsafeLookupStaticPtr`. This should be slightly
more efficient.

Test Plan: Validated locally

Reviewers: simonmar, hvr, bgamari, erikd

Reviewed By: simonmar

Subscribers: rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5333

Summary:
ae2c9b40 introduced some changes that
broke haddock, [1] fixed them and this patch bumps the haddock submodule
to include the fixes.

[1]: https://github.com/haskell/haddock/pull/970

Test Plan: build haddock (make or hadrian)

Reviewers: bgamari, AndreasK

Reviewed By: AndreasK

Subscribers: osa1, AndreasK, rwbarton, carter

GHC Trac Issues: #15900

Differential Revision: https://phabricator.haskell.org/D5341

Fix Trac #15898, by being smarter about when to print
a space before a promoted data constructor, in a HsType.
I had to implement a mildly tiresome function
    HsType.lhsTypeHasLeadingPromotionQuote
It has multiple cases, of course, but it's very simple.

The patch improves the error-message output in a bunch of
cases, and (to my surprise) actually fixes a bug in the
output of T14343 (Trac #14343), thus

  -  In the expression: _ :: Proxy '('( 'True,  'False),  'False)
  +  In the expression: _ :: Proxy '( '( 'True, 'False), 'False)

I discovered that there were two copies of the PromotionFlag
type (a boolean, with helpfully named data cons), one in
IfaceType and one in HsType.  So I combined into one,
PromotionFlag, and moved it to BasicTypes.  That's why
quite a few files are touched, but it's all routine.

See Trac #15704 comment:8ff

Summary:
The logic in `Note [recursive SRTs]` was correct. However, my
implementation of it wasn't: I got the associativity of
`Set.difference` wrong, which led to an extremely subtle and difficult
to find bug.

Fortunately now we have a test case. I was able to cut down the code
to something manageable, and I've added it to the test suite.

Test Plan:
Before (using my stage 1 compiler without the fix):

```
====> T15892(normal) 1 of 1 [0, 0, 0]
cd "T15892.run" &&  "/home/smarlow/ghc/inplace/bin/ghc-stage1" -o T15892
T15892.hs -dcore-lint -dcmm-lint -no-user-package-db -rtsopts
-fno-warn-missed-specialisations -fshow-warning-groups
-fdiagnostics-color=never -fno-diagnostics-show-caret -Werror=compat
-dno-debug-output  -O
cd "T15892.run" && ./T15892  +RTS -G1 -A32k -RTS
Wrong exit code for T15892(normal)(expected 0 , actual 134 )
Stderr ( T15892 ):
T15892: internal error: evacuate: strange closure type 0
    (GHC version 8.7.20181113 for x86_64_unknown_linux)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
Aborted (core dumped)
*** unexpected failure for T15892(normal)
=====> T15892(g1) 1 of 1 [0, 1, 0]
cd "T15892.run" &&  "/home/smarlow/ghc/inplace/bin/ghc-stage1" -o T15892
T15892.hs -dcore-lint -dcmm-lint -no-user-package-db -rtsopts
-fno-warn-missed-specialisations -fshow-warning-groups
-fdiagnostics-color=never -fno-diagnostics-show-caret -Werror=compat
-dno-debug-output  -O
cd "T15892.run" && ./T15892 +RTS -G1 -RTS +RTS -G1 -A32k -RTS
Wrong exit code for T15892(g1)(expected 0 , actual 134 )
Stderr ( T15892 ):
T15892: internal error: evacuate: strange closure type 0
    (GHC version 8.7.20181113 for x86_64_unknown_linux)
    Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug
Aborted (core dumped)
```

After (using my stage 2 compiler with the fix):

```
=====> T15892(normal) 1 of 1 [0, 0, 0]
cd "T15892.run" &&  "/home/smarlow/ghc/inplace/test   spaces/ghc-stage2"
-o T15892 T15892.hs -dcore-lint -dcmm-lint -no-user-package-db -rtsopts
-fno-warn-missed-specialisations -fshow-warning-groups
-fdiagnostics-color=never -fno-diagnostics-show-caret -Werror=compat
-dno-debug-output
cd "T15892.run" && ./T15892  +RTS -G1 -A32k -RTS
=====> T15892(g1) 1 of 1 [0, 0, 0]
cd "T15892.run" &&  "/home/smarlow/ghc/inplace/test   spaces/ghc-stage2"
-o T15892 T15892.hs -dcore-lint -dcmm-lint -no-user-package-db -rtsopts
-fno-warn-missed-specialisations -fshow-warning-groups
-fdiagnostics-color=never -fno-diagnostics-show-caret -Werror=compat
-dno-debug-output
cd "T15892.run" && ./T15892 +RTS -G1 -RTS +RTS -G1 -A32k -RTS
```

Reviewers: bgamari, osa1, erikd

Reviewed By: osa1

Subscribers: rwbarton, carter

GHC Trac Issues: #15892

Differential Revision: https://phabricator.haskell.org/D5334

... in addition to ghc-iserv and ghc-iserv-prof, as it is required
to get 10+ tests to pass

Reviewers: bgamari, alpmestan

Reviewed By: alpmestan

Subscribers: alpmestan, rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5327

Summary:
This brings the situation of `UniqDSet` in line with `UniqSet`.

@dfeuer said in D3146#92820 that he would do this, but probably
never got around to it.

Validated locally.

Reviewers: AndreasK, mpickering, bgamari, dfeuer, simonpj

Reviewed By: simonpj

Subscribers: simonpj, rwbarton, carter, dfeuer

GHC Trac Issues: #15879, #13114

Differential Revision: https://phabricator.haskell.org/D5313

Currently, it assumes the package names are identical and this
breaks in the case where integer-gmp is in one package db and
integer-simple in another. This became a problem with
the commit: fc2ff6dd.

Instead of following the precedence information, leading to
the right choice, the current code would compare the
integer-gmp and integer-simple versions and pick integer-gmp
because it happened to have a greater version, despite having
a lower precedence. See
https://github.com/snowleopard/hadrian/issues/702 for
a comprehensive report about the problem.

This effectively un-breaks integer-simple builds with hadrian.

Test Plan: hadrian/build.sh --integer-simple

Reviewers: snowleopard, bgamari

Reviewed By: bgamari

Subscribers: snowleopard, rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5266

- The StgBinderInfo type was never used in the code gen, so the type, related
  computation in CoreToStg, and some comments about it are removed. See #15770
  for more details.

- Simplified CoreToStg after removing the StgBinderInfo computation: removed
  StgBinderInfo arguments and mfix stuff.

The StgBinderInfo values were not used in the code gen, but I still run nofib
just to make sure: 0.0% change in allocations and binary sizes.

Test Plan: Validated locally

Reviewers: simonpj, simonmar, bgamari, sgraf

Reviewed By: sgraf

Subscribers: AndreasK, sgraf, rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5232

Summary:
For holes, its necessary to "see through" the instantiation
of the hole to get accurate family instance dependencies.
For example, if B imports <A>, and <A> is instantiated with
F, we must grab and include all of the dep_finsts from
F to have an accurate transitive dep_finsts list.

However, we MUST NOT do this for regular modules.
First, for efficiency reasons, doing this
bloats the the dep_finsts list, because we *already* had
those modules in the list (it wasn't a hole module, after
all). But there's a second, more important correctness
consideration: we perform module renaming when running
--abi-hash.  In this case, GHC's contract to the user is that
it will NOT go and read out interfaces of any dependencies
(https://github.com/haskell/cabal/issues/3633); the point of
--abi-hash is just to get a hash of the on-disk interfaces
for this *specific* package.  If we go off and tug on the
interface for /everything/ in dep_finsts, we're gonna have a
bad time.  (It's safe to do do this for hole modules, though,
because the hmap for --abi-hash is always trivial, so the
interface we request is local.  Though, maybe we ought
not to do it in this case either...)
Signed-off-by: Edward Z. Yang <ezyang@fb.com>

Test Plan: validate

Reviewers: alexbiehl, goldfire, bgamari

Subscribers: ppk, shlevy, rwbarton, carter

GHC Trac Issues: #15594

Differential Revision: https://phabricator.haskell.org/D5123

Summary:
The patch https://phabricator.haskell.org/D5284
didn't respect the local naming conventions in module
compiler/rename/RnUnbound.hs:

  - Top level functions names are written in camelCase.
  -  Local function names in where clauses are written as names_with_underscores.

This patch restores these conventions.

Test Plan: make test TESTS="T15611a T15611b"

Reviewers: DavidEichmann, monoidal, hvr, mpickering, bgamari

Reviewed By: mpickering

Subscribers: rwbarton, carter

GHC Trac Issues: #15611

Differential Revision: https://phabricator.haskell.org/D5308

Summary: I tend to accumulate these and they are often quite useful to keep around.

Reviewers: monoidal

Reviewed By: monoidal

Subscribers: rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5320

Reviewers: bgamari, RyanGlScott

Reviewed By: RyanGlScott

Subscribers: RyanGlScott, rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5311

Summary: In GHCi we don't check anymore, whether a main function is exported.

Test Plan: make test TEST=T11647

Reviewers: hvr, osa1, monoidal, mpickering, bgamari

Reviewed By: osa1, mpickering

Subscribers: rwbarton, carter

GHC Trac Issues: #11647

Differential Revision: https://phabricator.haskell.org/D5162

Summary:
Both #9692 and #14179 were caused by GHC being careless
about using eta-reduced data family instance axioms. Each of those
tickets were fixed by manually whipping up some code to eta-expand
the axioms. The same sort of issue has now caused #15845, so I
figured it was high time to factor out the code that each of these
fixes have in common.

This patch introduces the `etaExpandFamInstLHS` function, which takes
a family instance's type variables, LHS types, and RHS type, and
returns type variables and LHS types that have been eta-expanded if
necessary, in the case of a data family instance. (If it's a type
family instance, `etaExpandFamInstLHS` just returns the supplied type
variables and LHS types unchanged).

Along the way, I noticed that many references to
`Note [Eta reduction for data families]` (in `FamInstEnv`) had
slightly bitrotted (they either referred to a somewhat different
name, or claimed that the Note lived in a different module), so
I took the liberty of cleaning those up.

Test Plan: make test TEST="T9692 T15845"

Reviewers: goldfire, bgamari

Reviewed By: goldfire

Subscribers: rwbarton, carter

GHC Trac Issues: #15845

Differential Revision: https://phabricator.haskell.org/D5294

This patch makes the following improvement:
  - Automatically records test metrics (per test environment) so that
    the programmer need not supply nor update expected values in *.T
    files.
    - On expected metric changes, the programmer need only indicate the
      direction of change in the git commit message.
  - Provides a simple python tool "perf_notes.py" to compare metrics
    over time.

Issues:
  - Using just the previous commit allows performance to drift with each
    commit.
    - Currently we allow drift as we have a preference for minimizing
      false positives.
    - Some possible alternatives include:
      - Use metrics from a fixed commit per test: the last commit that
        allowed a change in performance (else the oldest metric)
      - Or use some sort of aggregate since the last commit that allowed
        a change in performance (else all available metrics)
      - These alternatives may result in a performance issue (with the
        test driver) having to heavily search git commits/notes.
  - Run locally, performance tests will trivially pass unless the tests
    were run locally on the previous commit. This is often not the case
    e.g.  after pulling recent changes.

Previously, *.T files contain statements such as:
```
stats_num_field('peak_megabytes_allocated', (2, 1))
compiler_stats_num_field('bytes allocated',
                         [(wordsize(64), 165890392, 10)])
```
This required the programmer to give the expected values and a tolerance
deviation (percentage). With this patch, the above statements are
replaced with:
```
collect_stats('peak_megabytes_allocated', 5)
collect_compiler_stats('bytes allocated', 10)
```
So that programmer must only enter which metrics to test and a tolerance
deviation. No expected value is required. CircleCI will then run the
tests per test environment and record the metrics to a git note for that
commit and push them to the git.haskell.org ghc repo. Metrics will be
compared to the previous commit. If they are different by the tolerance
deviation from the *.T file, then the corresponding test will fail. By
adding to the git commit message e.g.
```
 # Metric (In|De)crease <metric(s)> <options>: <tests>
Metric Increase ['bytes allocated', 'peak_megabytes_allocated'] \
         (test_env='linux_x86', way='default'):
    Test012, Test345
Metric Decrease 'bytes allocated':
    Test678
Metric Increase:
    Test711
```
This will allow the noted changes (letting the test pass). Note that by
omitting metrics or options, the change will apply to all possible
metrics/options (i.e. in the above, an increase for all metrics in all
test environments is allowed for Test711)

phabricator will use the message in the description

Reviewers: bgamari, hvr

Reviewed By: bgamari

Subscribers: rwbarton, carter

GHC Trac Issues: #12758

Differential Revision: https://phabricator.haskell.org/D5059

This reverts commit 406978c4.

This reverts commit adcb5fb4.

This reverts commit d8495549.

This fixes two isssues:

- Using bitcast for MO_XX_Conv
  Arguments to a bitcast must be of the same size. We should be using
  `trunc` and `zext` instead.

- Using unsupported MO_*_QuotRem for LLVM
  The two primops `MO_*_QuotRem` are not supported by the LLVM backend,
so
  we shouldn't use them for `Int8#`/`Word8#` (just as we do not use
them for
  `Int#`/`Word#`).
Signed-off-by: Michal Terepeta <michal.terepeta@gmail.com>

Test Plan: manually run tests with WAY=llvm

Reviewers: bgamari, simonmar

Reviewed By: bgamari

Subscribers: rwbarton, carter

GHC Trac Issues: #15864

Differential Revision: https://phabricator.haskell.org/D5304

Oops. Forgot to `git add`.

We thought that visible dependent quantification was impossible
in terms, but Iceland Jack discovered otherwise in #15859. This fixes an
ASSERT failure that arose.

test case: dependent/should_fail/T15859

Summary:
For the error message:
    Not in scope X.Y
    Module X does not export Y
    No module named ‘X’ is imported:
there are 2 cases, where we don't show the last "no module named is imported" line:
1. If the module X has been imported.
2. If the module X is the current module. There are 2 subcases:

   2.1 If the unknown module name is in a input source file,
       then we can use the getModule function to get the current module name.

   2.2 If the unknown module name has been entered by the user in GHCi,
       then the getModule function returns something like "interactive:Ghci1",
       and we have to check the current module in the last added entry of
       the HomePackageTable.

Test Plan: make test TESTS="T15611a T15611b"

Reviewers: monoidal, hvr, thomie, dfeuer, bgamari, DavidEichmann

Reviewed By: monoidal, DavidEichmann

Subscribers: rwbarton, carter

GHC Trac Issues: #15611

Differential Revision: https://phabricator.haskell.org/D5284

This is the first step of implementing:
https://github.com/ghc-proposals/ghc-proposals/pull/74

The main highlights/changes:

    primops.txt.pp gets two new sections for two new primitive types for
    signed and unsigned 8-bit integers (Int8# and Word8 respectively) along
    with basic arithmetic and comparison operations. PrimRep/RuntimeRep get
    two new constructors for them. All of the primops translate into the
    existing MachOPs.

    For CmmCalls the codegen will now zero-extend the values at call
    site (so that they can be moved to the right register) and then truncate
    them back their original width.

    x86 native codegen needed some updates, since it wasn't able to deal
    with the new widths, but all the changes are quite localized. LLVM
    backend seems to just work.

This is the second attempt at merging this, after the first attempt in
D4475 had to be backed out due to regressions on i386.

Bumps binary submodule.
Signed-off-by: Michal Terepeta <michal.terepeta@gmail.com>

Test Plan: ./validate (on both x86-{32,64})

Reviewers: bgamari, hvr, goldfire, simonmar

Subscribers: rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5258

Instead of using the GCC `/* fallthrough */` syntax we now use the
`__attribute__((fallthrough))`, which Phyx says should be more portable
than the former.

Also adds a missing fallthrough annotation in the MachO linker,
fixing #14613.

Reviewers: erikd, simonmar

Reviewed By: simonmar

Subscribers: rwbarton, carter

GHC Trac Issues: #14613

Differential Revision: https://phabricator.haskell.org/D5292

This introduces the `+RTS -ol` flag, which allows user to specify the
destination file for eventlog output.

Test Plan: Validate with included test

Reviewers: simonmar, erikd

Reviewed By: simonmar

Subscribers: rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5293

(cherry picked from commit ee545ff44e0ba9a165de40807548c75bf181dda3)