aboutsummaryrefslogtreecommitdiffstats
path: root/vespalib
Commit message (Collapse)AuthorAgeFilesLines
* Allow for faster access of thread local variable.Henning Baldersheim2020-10-201-1/+3
| | | | | This is possible since we do not load our code explicit as dynamic library. Also replace __thread by standard c++ linkage thread_local.
* Add noexcept specifier to function that cannot throw.Tor Egge2020-10-161-1/+1
|
* monitor -> guardHenning Baldersheim2020-10-151-6/+7
|
* GC unused vespalib::MOnitor.Henning Baldersheim2020-10-155-540/+38
|
* - LockGuard -> unique_lockHenning Baldersheim2020-10-156-47/+44
| | | | | - Pass cond to avoid having protected members. - monitor -> lock
* vespalib::Monitor -> std:.mutex/std::condition_variableHenning Baldersheim2020-10-143-14/+20
|
* Merge pull request #14881 from vespa-engine/balder/modernize-rendevouzHenning Baldersheim2020-10-142-11/+13
|\ | | | | Use c++11 primitives for synchronization
| * Use c++11 primitives for synchronizationHenning Baldersheim2020-10-142-11/+13
| |
* | Use c++11 primitives for synchronizationHenning Baldersheim2020-10-142-16/+21
| |
* | Use c++11 primitives for synchronizationHenning Baldersheim2020-10-142-22/+20
|/
* Merge pull request #14877 from ↵Henning Baldersheim2020-10-142-12/+22
|\ | | | | | | | | vespa-engine/balder/gc-unused-code-and-use-condition_variable GC unuse code and use std::mutex/std:condition_variable over vespalib…
| * GC unuse code and use std::mutex/std:condition_variable over vespalib::MonitorHenning Baldersheim2020-10-142-12/+22
| |
* | vespalib::Monitor -> std:.mutex/std::condition_variableHenning Baldersheim2020-10-142-11/+22
|/
* GC unused includeHenning Baldersheim2020-10-141-3/+1
|
* Verify that atomic duration time_point is lock free.Henning Baldersheim2020-10-131-0/+6
|
* Eliminate noexcept warnings.Tor Egge2020-10-117-8/+8
|
* Merge pull request #14800 from ↵Geir Storli2020-10-091-1/+1
|\ | | | | | | | | vespa-engine/vekterli/use-xxh3-for-bucket-to-stripe-distribution Use XXH3 for bucket-to-stripe distribution
| * Use XXH3 for bucket-to-stripe distributionTor Brede Vekterli2020-10-091-1/+1
| |
* | - GC the last usages of vespalib::Lock.Henning Baldersheim2020-10-089-86/+14
|/ | | | - Now it is only vespalib::Monitor left
* Use XXH3 over XXH64 as vespalib string hash functionTor Brede Vekterli2020-10-082-2/+2
| | | | | | XXH3 is faster than XXH64 across the board, especially for short inputs. Change around some tests that implicitly depended on the old hashing.
* Make vespalib hash functors noexceptTor Brede Vekterli2020-10-082-26/+26
| | | | | | Make ducktyped functor conditionally noexcept on `hash()` member function. Follow up on `-Werror=noexcept` warnings indicating where `noexcept` can be trivially added for such functions.
* - GC unused code.Henning Baldersheim2020-10-0724-1830/+87
| | | | - vespalib::Lock -> std::mutex
* Remove broken copy and move constructor and assignment operatoos on ↵Henning Baldersheim2020-10-0714-170/+236
| | | | | | vespalib::Lock and vespalib::Monitor. Also repair broken usages of the same.
* Add noexcept as indicated by -WnoeexceptHenning Baldersheim2020-10-0722-88/+91
|
* Use noexcept as advise by gcc -WnoexceptHenning Baldersheim2020-10-069-25/+29
|
* Darwin needs std::nullptr_t workaround when using AppleClang.Tor Egge2020-10-061-2/+2
|
* Merge pull request #14696 from ↵Henning Baldersheim2020-10-031-0/+1
|\ | | | | | | | | vespa-engine/toregge/include-pthread-header-when-using-pthread-functions Include pthread.h when using pthread functions.
| * Include pthread.h when using pthread functions.Tor Egge2020-10-031-0/+1
| |
* | Add wakeup call on the executors to allow to urgent wakeup of lazy exeutors ↵Henning Baldersheim2020-10-033-2/+9
|/ | | | optimized for throughput.
* Merge pull request #14635 from ↵Henning Baldersheim2020-09-302-2/+4
|\ | | | | | | | | vespa-engine/toregge/limit-memory-stats-used-for-reference-attribute-compaction Ignore unique store dictionary memory usage when deciding if compaction is needed
| * Use memory usage for unique store values but not for unique store dictionaryTor Egge2020-09-302-2/+4
| | | | | | | | when checking if compaction should be performed.
* | Consider a buffer on hold to have zero dead elements.Tor Egge2020-09-302-4/+6
|/
* Use && qualified member functions.Henning Baldersheim2020-09-262-5/+2
|
* Add && to make it evident that the object must be moved.Henning Baldersheim2020-09-261-1/+1
|
* Replace the dangerous stealBuffer method with a static one that requires ↵Henning Baldersheim2020-09-261-1/+4
| | | | std::move to make destruction more visible.
* inline typeid(A)Håvard Pettersen2020-09-251-3/+2
|
* benchmark type detectionHåvard Pettersen2020-09-254-1/+161
|
* postpone change in DirectSparseTensorBuilderArne Juul2020-09-162-3/+3
|
* shrink stash for very small sparse tensorsArne Juul2020-09-162-4/+4
|
* Correct allocated memory accounting for vespalib::string.Tor Egge2020-09-111-1/+1
|
* account for accounting infoArne Juul2020-09-091-1/+1
|
* add get_memory_usage for StashArne Juul2020-09-092-3/+25
|
* add methods to extract memory usageArne Juul2020-09-091-0/+7
|
* Merge pull request #14295 from ↵Tor Brede Vekterli2020-09-072-1/+8
|\ | | | | | | | | vespa-engine/vekterli/add-content-bucket-db-metrics-and-reduce-memory-footprint Add content bucket DB metrics and reduce memory footprint
| * Reintroduce DataStore ctor taking explicit min_arrays argumentTor Brede Vekterli2020-09-042-1/+8
| | | | | | | | | | | | | | | | | | Lets caller specify a reasonable minimum array count without needing to create an explicit buffer. Use explicit `min_arrays=1024` for content node bucket DB `DataStore`. Reduces default memory footprint of an empty (or sparsely populated) content node bucket DB with a factor of more than 1200x.
* | adapt and convert between vespa and onnx typesHåvard Pettersen2020-09-051-2/+6
|/
* Merge pull request #14239 from ↵Geir Storli2020-09-024-40/+51
|\ | | | | | | | | vespa-engine/geirst/memory-usage-tracking-in-direct-tensor-store Memory usage tracking in direct tensor store
| * Track heap allocated memory for tensors in DirectTensorStore.Geir Storli2020-09-022-7/+10
| |
| * Extra used bytes must also be accounted in allocated bytes when creating ↵Geir Storli2020-09-022-33/+41
| | | | | | | | memory stats.
* | - Add contained support for current serialization formats.Henning Baldersheim2020-09-023-13/+14
|/ | | | - Add support for future compressed formats.