Skip to content

A
abseil-cpp
Switch branch/tag
  1. 16 Nov, 2023 2 commits
  3. 15 Nov, 2023 1 commit
    • Speed-up absl::Symbolize by ~6x via faster file reads. · aa146013 
      absl::Symbolize does tons of tiny reads. Speed this up by switching
from lseek+read to a pread, and by reading more data than requested
into a buffer.

A faster absl::Symbolize will be helpful in tests and when printing
stack traces on /threadz etc.

Results for absl::Symbolize benchmark that exercises uncached behavior
of absl::Symbolize:

```
name          old time/op  new time/op  delta
BM_Symbolize  16.4ms ±12%   2.6ms ± 0%  -84.06%  (p=0.001 n=5+9)
```

PiperOrigin-RevId: 582687566
Change-Id: I44caf189d81867f3fd8c050a3100a4b9a8e744d7
      Abseil Team committed
  5. 14 Nov, 2023 4 commits
  7. 13 Nov, 2023 2 commits
  9. 09 Nov, 2023 1 commit
  11. 08 Nov, 2023 5 commits
  13. 07 Nov, 2023 2 commits
    • Properly handle signal stacks and frame-size calculations · bb7bbb12 
      We can determine the signal stack, so use that information to
make better decisions about when to calculate the frame size
and when not to. This fixes a several tests where the memory
layout had the signal stack and main stack in position that
confused some of the greater-than/less-than comparisons.

Also cleanup certain types to avoid more casting than necessary.

PiperOrigin-RevId: 580221819
Change-Id: I0365b03e7893741603dc66e6d36a069d0b7f5404
      Abseil Team committed
    • Add control()/slot() functions to iterator/const_iterator. · 116ee0fe 
      This allows for avoiding e.g. it.inner_.slot_ on const iterators.

Also, refactor HashtableDebugAccess::AllocatedByteSize to use regular iteration instead of looking directly at control/slot_array of the table in order to support small object optimization.

PiperOrigin-RevId: 580194253
Change-Id: I64cd69287834ee5c7a8daf867c532258806bfb7b
      Evan Brown committed
  15. 06 Nov, 2023 1 commit
  17. 03 Nov, 2023 1 commit
  19. 02 Nov, 2023 2 commits
  21. 01 Nov, 2023 1 commit
  23. 31 Oct, 2023 1 commit
    • Mutex: Remove destructor in release build · f3760b4d 
      The Mutex destructor is needed only to clean up debug logging
and invariant checking synch events. These are not supposed
to be used in production, but the non-empty destructor has
costs for production builds.

Instead of removing synch events in destructor,
drop all of them if we accumulated too many.
For tests is should not matter (we maybe only consume
a bit more memory). Production builds should be either unaffected
(if don't use debug logging), or use periodic reset of all synch events.

PiperOrigin-RevId: 578123259
Change-Id: I0ec59183a5f63ea0a6b7fc50f0a77974e7f677be
      Dmitry Vyukov committed
  25. 30 Oct, 2023 4 commits
    • Optimize memcasecmp. Benchmarks shows slight improvement. · 6c8338c2 
      We are also avoiding potential cache-misses, by avoiding load.

name                         old speed            new speed            delta
BM_Searchcase                3.09GB/s ±13%        3.20GB/s ±16%   +3.69%  (p=0.039 n=20+17)
BM_SearchcaseMedium          1.08GB/s ± 7%        1.04GB/s ±14%     ~     (p=0.814 n=16+20)
BM_SearchcasePathological     618kB/s ±13%         652kB/s ± 6%   +5.55%  (p=0.043 n=20+16)
BM_Memcasematch              2.43GB/s ± 3%        2.45GB/s ± 3%     ~     (p=0.488 n=17+16)
BM_MemcasematchMedium         230MB/s ± 8%         261MB/s ±14%  +13.77%  (p=0.000 n=16+20)
BM_MemcasematchPathological   624kB/s ±14%         619kB/s ±14%     ~     (p=0.836 n=20+20)

PiperOrigin-RevId: 577919033
Change-Id: I31324e04b6a577c582ad630d171d3b41d826f1e4
      Ilya Tokar committed
    • Roll forward: Add sanitizer mode checks that element constructors/destructors… · dc3ef295 
      Roll forward: Add sanitizer mode checks that element constructors/destructors don't make reentrant calls to raw_hash_set member functions.

PiperOrigin-RevId: 577877803
Change-Id: I254c589b00cadec6ff95dfd60a8a38ab303c1af5
      Evan Brown committed
    • Fix an unreachable code warning on some platforms · e7d9317b 
      #1558

PiperOrigin-RevId: 577874842
Change-Id: I1d56f3f4b445f6c4a9df2fe06fec542cb00e0e92
      Derek Mauro committed
    • absl: speed up Mutex::Lock · f4106724 
      Currently Mutex::Lock contains not inlined non-tail call:
TryAcquireWithSpinning -> GetMutexGlobals -> LowLevelCallOnce -> init closure
This turns the function into non-leaf with stack frame allocation
and additional register use. Remove this non-tail call to make the function leaf.
Move spin iterations initialization to LockSlow.

Current Lock happy path:

00000000001edc20 <absl::Mutex::Lock()>:
  1edc20:	55                   	push   %rbp
  1edc21:	48 89 e5             	mov    %rsp,%rbp
  1edc24:	53                   	push   %rbx
  1edc25:	50                   	push   %rax
  1edc26:	48 89 fb             	mov    %rdi,%rbx
  1edc29:	48 8b 07             	mov    (%rdi),%rax
  1edc2c:	a8 19                	test   $0x19,%al
  1edc2e:	75 0e                	jne    1edc3e <absl::Mutex::Lock()+0x1e>
  1edc30:	48 89 c1             	mov    %rax,%rcx
  1edc33:	48 83 c9 08          	or     $0x8,%rcx
  1edc37:	f0 48 0f b1 0b       	lock cmpxchg %rcx,(%rbx)
  1edc3c:	74 42                	je     1edc80 <absl::Mutex::Lock()+0x60>
  ... unhappy path ...
  1edc80:	48 83 c4 08          	add    $0x8,%rsp
  1edc84:	5b                   	pop    %rbx
  1edc85:	5d                   	pop    %rbp
  1edc86:	c3                   	ret

New Lock happy path:

00000000001eea80 <absl::Mutex::Lock()>:
  1eea80:	48 8b 07             	mov    (%rdi),%rax
  1eea83:	a8 19                	test   $0x19,%al
  1eea85:	75 0f                	jne    1eea96 <absl::Mutex::Lock()+0x16>
  1eea87:	48 89 c1             	mov    %rax,%rcx
  1eea8a:	48 83 c9 08          	or     $0x8,%rcx
  1eea8e:	f0 48 0f b1 0f       	lock cmpxchg %rcx,(%rdi)
  1eea93:	75 01                	jne    1eea96 <absl::Mutex::Lock()+0x16>
  1eea95:	c3                   	ret
  ... unhappy path ...

PiperOrigin-RevId: 577790105
Change-Id: I20793534050302ff9f7a20aed93791c088d98562
      Dmitry Vyukov committed
  27. 27 Oct, 2023 2 commits
    • Rollback "Mutex: Remove destructor in release build" · 89d2caa1 
      PiperOrigin-RevId: 577180526
Change-Id: Iec53709456805ca8dc5327669cc0f6c95825d0e9
      Dmitry Vyukov committed
    • Mutex: Remove destructor in release build · cdb1e7f4 
      The Mutex destructor is needed only to clean up debug logging
and invariant checking synch events. These are not supposed
to be used in production, but the non-empty destructor has
costs for production builds.

Instead of removing synch events in destructor,
drop all of them if we accumulated too many.
For tests is should not matter (we maybe only consume
a bit more memory). Production builds should be either unaffected
(if don't use debug logging), or use periodic reset of all synch events.

PiperOrigin-RevId: 577106805
Change-Id: Icaaf7166b99afcd5dce92b4acd1be661fb72f10b
      Dmitry Vyukov committed
  29. 26 Oct, 2023 2 commits
  31. 25 Oct, 2023 1 commit
  33. 24 Oct, 2023 4 commits
  35. 20 Oct, 2023 4 commits