[Git][ghc/ghc][wip/gc/compact-nfdata] 47 commits: rts: Non-concurrent mark and sweep

Ben Gamari gitlab at gitlab.haskell.org
Wed Jun 19 01:44:14 UTC 2019 


Ben Gamari pushed to branch wip/gc/compact-nfdata at Glasgow Haskell Compiler / GHC


Commits:
a2a9b3ad by Ömer Sinan Ağacan at 2019-06-19T01:40:07Z
rts: Non-concurrent mark and sweep

This implements the core heap structure and a serial mark/sweep
collector which can be used to manage the oldest-generation heap.
This is the first step towards a concurrent mark-and-sweep collector
aimed at low-latency applications.

The full design of the collector implemented here is described in detail
in a technical note

    B. Gamari. "A Concurrent Garbage Collector For the Glasgow Haskell
    Compiler" (2018)

The basic heap structure used in this design is heavily inspired by

    K. Ueno & A. Ohori. "A fully concurrent garbage collector for
    functional programs on multicore processors." /ACM SIGPLAN Notices/
    Vol. 51. No. 9 (presented by ICFP 2016)

This design is intended to allow both marking and sweeping
concurrent to execution of a multi-core mutator. Unlike the Ueno design,
which requires no global synchronization pauses, the collector
introduced here requires a stop-the-world pause at the beginning and end
of the mark phase.

To avoid heap fragmentation, the allocator consists of a number of
fixed-size /sub-allocators/. Each of these sub-allocators allocators into
its own set of /segments/, themselves allocated from the block
allocator. Each segment is broken into a set of fixed-size allocation
blocks (which back allocations) in addition to a bitmap (used to track
the liveness of blocks) and some additional metadata (used also used
to track liveness).

This heap structure enables collection via mark-and-sweep, which can be
performed concurrently via a snapshot-at-the-beginning scheme (although
concurrent collection is not implemented in this patch).

The mark queue is a fairly straightforward chunked-array structure.
The representation is a bit more verbose than a typical mark queue to
accomodate a combination of two features:

 * a mark FIFO, which improves the locality of marking, reducing one of
   the major overheads seen in mark/sweep allocators (see [1] for
   details)

 * the selector optimization and indirection shortcutting, which
   requires that we track where we found each reference to an object
   in case we need to update the reference at a later point (e.g. when
   we find that it is an indirection). See Note [Origin references in
   the nonmoving collector] (in `NonMovingMark.h`) for details.

Beyond this the mark/sweep is fairly run-of-the-mill.

[1] R. Garner, S.M. Blackburn, D. Frampton. "Effective Prefetch for
    Mark-Sweep Garbage Collection." ISMM 2007.

Co-Authored-By: Ben Gamari <ben at well-typed.com>

- - - - -
fbaf283c by Ben Gamari at 2019-06-19T01:40:07Z
testsuite: Add nonmoving WAY

This simply runs the compile_and_run tests with `-xn`, enabling the
nonmoving oldest generation.

- - - - -
0a07f0e7 by Ben Gamari at 2019-06-19T01:40:29Z
rts: Implement concurrent collection in the nonmoving collector

This extends the non-moving collector to allow concurrent collection.

The full design of the collector implemented here is described in detail
in a technical note

    B. Gamari. "A Concurrent Garbage Collector For the Glasgow Haskell
    Compiler" (2018)

This extension involves the introduction of a capability-local
remembered set, known as the /update remembered set/, which tracks
objects which may no longer be visible to the collector due to mutation.
To maintain this remembered set we introduce a write barrier on
mutations which is enabled while a concurrent mark is underway.

The update remembered set representation is similar to that of the
nonmoving mark queue, being a chunked array of `MarkEntry`s. Each
`Capability` maintains a single accumulator chunk, which it flushed
when it (a) is filled, or (b) when the nonmoving collector enters its
post-mark synchronization phase.

While the write barrier touches a significant amount of code it is
conceptually straightforward: the mutator must ensure that the referee
of any pointer it overwrites is added to the update remembered set.
However, there are a few details:

 * In the case of objects with a dirty flag (e.g. `MVar`s) we can
   exploit the fact that only the *first* mutation requires a write
   barrier.

 * Weak references, as usual, complicate things. In particular, we must
   ensure that the referee of a weak object is marked if dereferenced by
   the mutator. For this we (unfortunately) must introduce a read
   barrier, as described in Note [Concurrent read barrier on deRefWeak#]
   (in `NonMovingMark.c`).

 * Stable names are also a bit tricky as described in Note [Sweeping
   stable names in the concurrent collector] (`NonMovingSweep.c`).

We take quite some pains to ensure that the high thread count often seen
in parallel Haskell applications doesn't affect pause times. To this end
we allow thread stacks to be marked either by the thread itself (when it
is executed or stack-underflows) or the concurrent mark thread (if the
thread owning the stack is never scheduled). There is a non-trivial
handshake to ensure that this happens without racing which is described
in Note [StgStack dirtiness flags and concurrent marking].

Co-Authored-by: Ömer Sinan Ağacan <omer at well-typed.com>

- - - - -
f98864fa by Ben Gamari at 2019-06-19T01:40:29Z
Nonmoving: Disable memory inventory with concurrent collection

- - - - -
b77fed86 by Ben Gamari at 2019-06-19T01:41:21Z
rts: Tracing support for nonmoving collection events

This introduces a few events to mark key points in the nonmoving
garbage collection cycle. These include:

 * `EVENT_CONC_MARK_BEGIN`, denoting the beginning of a round of
   marking. This may happen more than once in a single major collection
   since we the major collector iterates until it hits a fixed point.

 * `EVENT_CONC_MARK_END`, denoting the end of a round of marking.

 * `EVENT_CONC_SYNC_BEGIN`, denoting the beginning of the post-mark
   synchronization phase

 * `EVENT_CONC_UPD_REM_SET_FLUSH`, indicating that a capability has
   flushed its update remembered set.

 * `EVENT_CONC_SYNC_END`, denoting that all mutators have flushed their
   update remembered sets.

 * `EVENT_CONC_SWEEP_BEGIN`, denoting the beginning of the sweep portion
   of the major collection.

 * `EVENT_CONC_SWEEP_END`, denoting the end of the sweep portion of the
   major collection.

- - - - -
9a16493d by Ben Gamari at 2019-06-19T01:41:21Z
rts: Introduce non-moving heap census

This introduces a simple census of the non-moving heap (not to be
confused with the heap census used by the heap profiler). This
collects basic heap usage information (number of allocated and free
blocks) which is useful when characterising fragmentation of the
nonmoving heap.

- - - - -
3efdc8f2 by Ben Gamari at 2019-06-19T01:41:21Z
rts/Eventlog: More descriptive error message

- - - - -
5b1b7789 by Ben Gamari at 2019-06-19T01:41:21Z
Allow census without live word count

Otherwise the census is unsafe when mutators are running due to
concurrent mutation.

- - - - -
7b06486e by Ben Gamari at 2019-06-19T01:41:21Z
NonmovingCensus: Emit samples to eventlog

- - - - -
7384acc9 by Ben Gamari at 2019-06-19T01:41:21Z
rts: Add GetMyThreadCPUTime helper

- - - - -
86376581 by Ben Gamari at 2019-06-19T01:41:21Z
rts/Stats: Track time usage of nonmoving collector

- - - - -
fa4e6e79 by Ben Gamari at 2019-06-19T01:41:34Z
Nonmoving: Allow aging and refactor static objects logic

This commit does two things:

 * Allow aging of objects during the preparatory minor GC
 * Refactor handling of static objects to avoid the use of a hashtable

- - - - -
c3df5d61 by Ben Gamari at 2019-06-19T01:41:34Z
Disable aging when doing deadlock detection GC

- - - - -
1a125cc6 by Ben Gamari at 2019-06-19T01:41:34Z
More comments for aging

- - - - -
b44c267f by Ben Gamari at 2019-06-19T01:41:49Z
testsuite: Add nonmoving_thr way

- - - - -
37cf9173 by Ben Gamari at 2019-06-19T01:41:49Z
testsuite: Add nonmoving_thr_ghc way

This uses the nonmoving collector when compiling the testcases.

- - - - -
1ac956e0 by Ben Gamari at 2019-06-19T01:41:50Z
testsuite: Don't run T15892 in nonmoving ways

The nonmoving GC doesn't support `+RTS -G1`, which this test insists on.

- - - - -
d51f62ab by Ben Gamari at 2019-06-19T01:41:50Z
testsuite: Nonmoving collector doesn't support -G1

- - - - -
b14784df by Ben Gamari at 2019-06-19T01:41:50Z
testsuite: Ensure that threaded tests are run in nonmoving_thr

- - - - -
1a286242 by Ben Gamari at 2019-06-19T01:41:50Z
testsuite: bug1010 requires -c, which isn't supported by nonmoving

- - - - -
eedd46dd by Ben Gamari at 2019-06-19T01:41:50Z
testsuite: Skip T15892 in nonmoving_thr_ghc

- - - - -
8b65a9cd by Ben Gamari at 2019-06-19T01:41:50Z
ghc-heap: Skip heap_all test with debugged RTS

The debugged RTS initializes the heap with 0xaa, which breaks the
(admittedly rather fragile) assumption that uninitialized fields are set
to 0x00:
```
Wrong exit code for heap_all(nonmoving)(expected 0 , actual 1 )
Stderr ( heap_all ):
heap_all: user error (assertClosuresEq: Closures do not match
Expected: FunClosure {info = StgInfoTable {entry = Nothing, ptrs = 0, nptrs = 1, tipe = FUN_0_1, srtlen = 0, code = Nothing}, ptrArgs = [], dataArgs = [0]}
Actual:   FunClosure {info = StgInfoTable {entry = Nothing, ptrs = 0, nptrs = 1, tipe = FUN_0_1, srtlen = 1032832, code = Nothing}, ptrArgs = [], dataArgs = [12297829382473034410]}

CallStack (from HasCallStack):
  assertClosuresEq, called at heap_all.hs:230:9 in main:Main
)
```

- - - - -
da9e3ef2 by Ben Gamari at 2019-06-19T01:41:50Z
Skip ghc_heap_all test in nonmoving ways

- - - - -
934c48ee by Ben Gamari at 2019-06-19T01:42:04Z
NonMoving: Eliminate integer division in nonmovingBlockCount

Perf showed that the this single div was capturing up to 10% of samples
in nonmovingMark. However, the overwhelming majority of cases is looking
at small block sizes. These cases we can easily compute explicitly,
allowing the compiler to turn the division into a significantly more
efficient division-by-constant.

While the increase in source code looks scary, this all optimises down
to very nice looking assembler. At this point the only remaining
hotspots in nonmovingBlockCount are due to memory access.

- - - - -
3245cd10 by Ben Gamari at 2019-06-19T01:42:04Z
Allocate mark queues in larger block groups

- - - - -
3d9b5d9b by Ben Gamari at 2019-06-19T01:42:04Z
NonMovingMark: Optimize representation of mark queue

This shortens MarkQueueEntry by 30% (one word)

- - - - -
074000f7 by Ben Gamari at 2019-06-19T01:42:04Z
NonMoving: Optimize bitmap search during allocation

Use memchr instead of a open-coded loop. This is nearly twice as fast in
a synthetic benchmark.

- - - - -
93ea36a1 by Ben Gamari at 2019-06-19T01:42:04Z
rts: Add prefetch macros

- - - - -
ca0b2f0b by Ben Gamari at 2019-06-19T01:42:04Z
NonMoving: Prefetch when clearing bitmaps

Ensure that the bitmap of the segmentt that we will clear next is in
cache by the time we reach it.

- - - - -
ee2ddc01 by Ben Gamari at 2019-06-19T01:42:04Z
NonMoving: Inline nonmovingClearAllBitmaps

- - - - -
213529aa by Ben Gamari at 2019-06-19T01:42:05Z
NonMoving: Fuse sweep preparation into mark prep

- - - - -
2367a592 by Ben Gamari at 2019-06-19T01:42:05Z
NonMoving: Pre-fetch during mark

This improved overall runtime on nofib's constraints test by nearly 10%.

- - - - -
5c31a903 by Ben Gamari at 2019-06-19T01:42:05Z
NonMoving: Prefetch segment header

- - - - -
c65cd50b by Ben Gamari at 2019-06-19T01:42:05Z
NonMoving: Optimise allocator cache behavior

Previously we would look at the segment header to determine the block
size despite the fact that we already had the block size at hand.

- - - - -
02fd2c3b by Ben Gamari at 2019-06-19T01:42:05Z
NonMovingMark: Eliminate redundant check_in_nonmoving_heaps

- - - - -
9c7e79cf by Ben Gamari at 2019-06-19T01:42:05Z
NonMoving: Don't do major GC if one is already running

Previously we would perform a preparatory moving collection, resulting
in many things being added to the mark queue. When we finished with this
we would realize in nonmovingCollect that there was already a collection
running, in which case we would simply not run the nonmoving collector.

However, it was very easy to end up in a "treadmilling" situation: all
subsequent GC following the first failed major GC would be scheduled as
major GCs. Consequently we would continuously feed the concurrent
collector with more mark queue entries and it would never finish.

This patch aborts the major collection far earlier, meaning that we
avoid adding nonmoving objects to the mark queue and allowing the
concurrent collector to finish.

- - - - -
7f6da43c by Ben Gamari at 2019-06-19T01:42:05Z
Nonmoving: Ensure write barrier vanishes in non-threaded RTS

- - - - -
6cf1164d by Ben Gamari at 2019-06-19T01:42:18Z
Merge branches 'wip/gc/optimize' and 'wip/gc/test' into wip/gc/everything

- - - - -
209a441d by Ömer Sinan Ağacan at 2019-06-19T01:42:42Z
NonMoving: Implement indirection shortcutting

This allows indirection chains residing in the non-moving heap to be
shorted-out.

- - - - -
59c627d3 by Ömer Sinan Ağacan at 2019-06-19T01:42:43Z
NonMoving: Implement selector optimisation

- - - - -
417e7bf2 by Ben Gamari at 2019-06-19T01:42:57Z
NonMoving: Introduce nonmovingSegmentLogBlockSize acccessor

This will allow us to easily move the block size elsewhere.

- - - - -
10fb49f9 by Ben Gamari at 2019-06-19T01:42:57Z
NonMoving: Move block size to block descriptor

- - - - -
a6fe38e2 by Ben Gamari at 2019-06-19T01:42:57Z
NonMoving: Move next_free_snap to block descriptor

- - - - -
c093c222 by Ben Gamari at 2019-06-19T01:43:06Z
NonMoving: Add summarizing Note

- - - - -
158d0374 by Ben Gamari at 2019-06-19T01:43:06Z
NonMoving: More comments

- - - - -
8c13a6ec by Ben Gamari at 2019-06-19T01:43:19Z
Merge branches 'wip/gc/segment-header-to-bdescr' and 'wip/gc/docs' into wip/gc/everything2

- - - - -
99dca8f2 by Ömer Sinan Ağacan at 2019-06-19T01:43:39Z
rts: COMPACT_NFDATA support for the nonmoving collector

This largely follows the model used for large objects, with appropriate
adjustments made to account for references in the sharing deduplication
hashtable.

- - - - -


30 changed files:

- compiler/cmm/CLabel.hs
- compiler/codeGen/StgCmmBind.hs
- compiler/codeGen/StgCmmPrim.hs
- compiler/codeGen/StgCmmUtils.hs
- includes/Cmm.h
- includes/Rts.h
- includes/RtsAPI.h
- includes/rts/EventLogFormat.h
- includes/rts/Flags.h
- + includes/rts/NonMoving.h
- includes/rts/storage/Block.h
- includes/rts/storage/ClosureMacros.h
- includes/rts/storage/GC.h
- includes/rts/storage/TSO.h
- includes/stg/MiscClosures.h
- libraries/base/GHC/RTS/Flags.hsc
- libraries/base/GHC/Stats.hsc
- libraries/ghc-heap/tests/all.T
- rts/Apply.cmm
- rts/Capability.c
- rts/Capability.h
- rts/Exception.cmm
- rts/GetTime.h
- rts/Messages.c
- rts/PrimOps.cmm
- rts/RaiseAsync.c
- rts/RtsFlags.c
- rts/RtsStartup.c
- rts/RtsSymbols.c
- rts/STM.c


The diff was not included because it is too large.


View it on GitLab: https://gitlab.haskell.org/ghc/ghc/compare/7246251a1a60d58b98f91b1549d81389a3c0a35b...99dca8f272907ad67bbb4387e72002873612b518

-- 
View it on GitLab: https://gitlab.haskell.org/ghc/ghc/compare/7246251a1a60d58b98f91b1549d81389a3c0a35b...99dca8f272907ad67bbb4387e72002873612b518
You're receiving this email because of your account on gitlab.haskell.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-commits/attachments/20190618/c494ef3c/attachment-0001.html>

More information about the ghc-commits mailing list