ab310bfdb4
ck_hp: Linearize registration and recycling.
12 years ago
94475b85be
ck_epoch: Remove redundant load fence.
12 years ago
a30a89817c
ck_epoch: Follow-up to previous commit
12 years ago
acde423ce8
ck_epoch: Adopt RMO ck_pr semantics.
...
These changes primarily affect platforms with RMO semantics.
- Write out to local epoch requires atomic semantics,
adopt atomic store to epoch in ck_epoch_reclaim.
- Correct an issue where we would publish node availability
before record re-initialization.
- Execute full barrier in write_begin to serialize active flag
IFF we are not recursing into an epoch block. Serialize active
publication with respect to epoch walk.
- Use CAS-based tick operation for epoch counter. Despite higher
constant factor, we are much less likely to starve writers on
bursty workloads.
12 years ago
d37eb779cb
build: Add dist target, allow user to modify version.
12 years ago
45f310893c
ck_rwlock: Fix behavior for weakly-ordered systems.
...
As ck_pr semantics were still not molded, I was designing
under the assumption I would potentially go towards
acq/req interface. Since RMO will be the semantic norm for
the ck_pr model from now on, enforce stricter ordering
requirements on rwlock.
ck_rwlock_write_unlock function will now also serialize both
loads and stores.
12 years ago
8043f52130
ck_pr: Add ck_pr_barrier for compiler barrier.
...
CK_CC_BARRIER isn't idiomatic, roll this into PR memory model
instead.
12 years ago
1e8fe57028
ck_bitmap: Remove unnecessary whitespace.
12 years ago
e3cf015723
build: Drop dist target, it really adds little value.
...
Use git-archive instead or similar if you need to build
your own distribution tarball.
12 years ago
7e9a0d8751
regressions/ck_brlock: Reduce number of iterations.
12 years ago
d4abd39dd6
build: Remove tarball.
12 years ago
f8c19a89b4
regressions/ck_hp: LLVM's optimizer keeps getting the best of us.
...
If I see more regressions in LLVM, I may end up enforcing compiler
barriers in all inlined code.
12 years ago
bc19dccb1f
ck_cc: Add CK_CC_BARRIER for compiler barrier.
12 years ago
f567513f8c
regressions/ck_hp: Add Donner's unit test.
12 years ago
06345e7917
build: Add "dist" target for people who want to build their own distribution.
...
Based on original patch submitted by Andrew Schorr.
12 years ago
b1b19c1f74
ck_ring: Memory fence is overly pedantic for parameterized enqueue.
12 years ago
753f95fb45
Merge branch 'master' of git.concurrencykit.org:ck
12 years ago
94698736c1
ck_queue: Add appropriate load fences.
12 years ago
7530ca21b7
ck_fifo: Do not forget about node recycling.
12 years ago
2d2b49a313
ck_stack: Assume weak ordering of CAS.
...
I was actually unsure of the exact memory model
I wanted for atomic RMW operations. It was
made apparent with time that I had to adopt RMO
if I didn't want to sacrifice performance. Make
sure we can assume RMO for the stack.
12 years ago
3f6a3660eb
regressions/ck_stack: Add trypush unit tests to check target.
...
Submitted by Andrew Schorr <aschorr@telemetry-inves....>.
12 years ago
a7f51fffca
ck_hp_fifo: Serialize second CAS.
12 years ago
d454f729e0
regressions/ck_fifo: Add dependency to ck_fifo.h.
12 years ago
ccbf28f0cb
ck_fifo: Remove unused variables, serialize MPMC insertion.
12 years ago
78f00eeeb6
ck_fifo_spsc: Fix implementation.
...
Do not be overly pedantic. Fix misplaced store
barrier and make sure to atomically load stub
pointer on reader side.
12 years ago
80c56d115b
ck_ring: Do not use volatile load for parametric ck_ring_size.
12 years ago
6159940337
build: Bump version for next release.
12 years ago
fc316307c4
build: Add sysctl line for OpenBSD.
12 years ago
ad5e143bd5
ck_ring: Various improvements to readability.
...
No regressions in unit tests, so I'm dropping volatile
semantics for unnecessary volatile load/store on producer/consumer
side.
12 years ago
290add2e50
build/regressions: Add CORES option, add "check" target.
...
The check target will attempt to run the unit tests with some
sane defaults.
12 years ago
fbb88a519c
ck_bytelock: read_lock must serialize all load/store before checking owner field.
...
Thanks to Andrew Schorr for reporting the unit test failure.
12 years ago
3afa6b2b39
ck_ht: Let users define their own empty and tombstone values.
12 years ago
94c26494a6
ck_ring: Size is no longer necessary.
...
Patch submitted by Andrew Schorr.
12 years ago
1a8b3db453
ck_ht: Various performance improvements.
...
Though a new implementation is in the works, roll in
some performance improvements in the mean time.
The probe routines have been broken out into separate
reader/writer variants. These variants are much less
branch-intensive (and don't involve predict to stall
in many cases).
New implementation is attempting to deal with
interface-induced overheads.
12 years ago
e1ec55819e
regressions/common. Move edx/eax to x86_64 block.
12 years ago
a56cac3a38
regressions/ck_sequence: Work-around LLVM bug.
...
For some reason, load operations were being re-ordered with respect
to read begin operation (even with explicit compiler barriers).
12 years ago
7657df25a8
ck_sequence: Serialize read_begin, add load barrier.
12 years ago
d100c8b834
Merge pull request #6 from abelmathew/master
...
ck_bag documentation
12 years ago
b9c6532b2a
Merge branch 'master' of https://github.com/sbahra/ck
12 years ago
5ca4e884bd
doc/ck_bag: Add ck_bag documentation.
12 years ago
ac8acffe6a
Merge branch 'master' of git.concurrencykit.org:ck
13 years ago
2015f0f5d4
Merge branch 'master' of git.concurrencykit.org:ck
13 years ago
e9bd744877
doc/ck_ht_entry_set_direct: Update manual page to include latest API change.
13 years ago
37d8a5e98d
build: Bump version for next release.
13 years ago
976f017450
build: Move symbolic link to -devel package, keep .spec file in distributions.
...
Patch submitted by Andrew Schorr <schorr@te....>.
13 years ago
031d950cc0
ck_stack: Add aliased attribute to ck_stack for GCC 4.4.
13 years ago
65cf506af9
ck_cc: Add CK_CC_ALIASED attribute.
...
Submitted by John Wittrock <jwittrock@...>.
13 years ago
7dd705c86f
ck_ht: Remove ck_ht_allocator_set declaration.
13 years ago
59158c824b
ck_ht: Do not re-hash on growth for non-PP case.
...
We already store the hash value in the hash table entry
if pointer packing is not enabled.
13 years ago
ccb1fd6d86
ck_fifo: Add some load fences for SPARC/PPC.
13 years ago
Samy Al Bahra
Samy Al Bahra
Abel Mathew