aboutsummaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Merge pull request #13673 from vespa-engine/toregge/add-global-filter-limitGeir Storli2020-06-303-4/+6
|\ | | | | If the estimated rate of matching documents is less than global filte…
| * If the estimated rate of matching documents is less than global filter limit ↵Tor Egge2020-06-233-4/+6
| | | | | | | | | | | | then don't build a global filter.
* | Use find_package to find gtest library.Tor Egge2020-06-2911-12/+23
| |
* | Let getExecutorId be a pure interface.Henning Baldersheim2020-06-261-2/+2
| |
* | Use ForegroundThreadExecutor in unit tests.Geir Storli2020-06-255-17/+23
| |
* | Run prepare step of two-phase puts in the shared executor instead of the ↵Geir Storli2020-06-253-17/+27
| | | | | | | | | | | | | | | | attribute field executor. Benchmarking has shown that we get an uneven balance between the underlying threads when using the attribute field executor. This is a sequenced task executor, where each underlying thread only is allowed to handle a subset of the prepare tasks (based on the local document id). Using the shared executor should give an even balance.
* | Expose the shared executor via attribute manager interface.Geir Storli2020-06-2514-52/+95
| |
* | Rename function to be more descriptive.Geir Storli2020-06-251-3/+3
| |
* | Let the executor create an executor id to its liking.Henning Baldersheim2020-06-258-12/+14
| | | | | | | | Taske full advantage over strands instead of being limited to threads.
* | Use std::make_xxxx to avoid newHenning Baldersheim2020-06-255-24/+19
| |
* | With adadptive mode there is no need to cap number of threads.Henning Baldersheim2020-06-241-7/+11
|/
* Add a rank feature that will use the distribution key to provide a globally ↵Henning Baldersheim2020-06-226-10/+20
| | | | | | unique number. unique = (docId << 16 | distributionKey)
* Merge pull request #13645 from ↵Geir Storli2020-06-192-5/+5
|\ | | | | | | | | vespa-engine/geirst/multi-threaded-hnsw-indexing-flag Multi-threaded hnsw indexing flag
| * Wire flag to enable multi-threaded indexing for a hnsw index.Geir Storli2020-06-192-5/+5
| |
* | Wire in nearest neighbor brute force limit.Tor Egge2020-06-194-5/+27
|/
* Test more edge cases for two-phase put.Geir Storli2020-06-171-16/+78
|
* Merge pull request #13618 from ↵Geir Storli2020-06-171-1/+1
|\ | | | | | | | | vespa-engine/toregge/searchsummary-test-is-a-normal-static-library Fix compilation on CentOS 8.2.
| * Use same kludge for CentOS 8.2 as for RHEL 8.2 to avoid unresolvedTor Egge2020-06-171-1/+1
| | | | | | | | symbols with gcc-toolset-9.
* | Avoid deadlock if field value is not set in two phase put.Geir Storli2020-06-171-4/+4
| |
* | Remove default tensor spec in doctype builder.Geir Storli2020-06-172-6/+7
| |
* | Test that attribute writer can handle put in two phases.Geir Storli2020-06-171-9/+86
| |
* | Implement initial support for two-phase puts in attribute writer.Geir Storli2020-06-175-37/+314
|/ | | | This is only turned on for tensor attributes with a hnsw index that allows multi-threaded indexing.
* Rename variable '_m' -> '_mgr'.Geir Storli2020-06-161-16/+16
|
* Simplify attribute writer tests by using mock attribute manager.Geir Storli2020-06-163-31/+46
|
* Revert "Two phase put in attribute writer"Harald Musum2020-06-159-463/+93
|
* Avoid deadlock if field value is not set in two phase put.Geir Storli2020-06-151-4/+4
|
* Remove default tensor spec in doctype builder.Geir Storli2020-06-152-6/+7
|
* Test that attribute writer can handle put in two phases.Geir Storli2020-06-151-9/+86
|
* Implement initial support for two-phase puts in attribute writer.Geir Storli2020-06-155-37/+314
| | | | This is only turned on for tensor attributes with a hnsw index that allows multi-threaded indexing.
* Rename variable '_m' -> '_mgr'.Geir Storli2020-06-151-16/+16
|
* Simplify attribute writer tests by using mock attribute manager.Geir Storli2020-06-153-31/+46
|
* avoid deadlock when using compile cache with blocking executorsHåvard Pettersen2020-06-142-3/+3
| | | | | | | | - avoid taking cache lock in compile task destructor - avoid holding cache lock while posting compile task - add unit test trying to provoke deadlock - bonus: fix blocking executor implementation (did not work with threads <= task limit)
* Merge pull request #13542 from ↵Geir Storli2020-06-103-323/+302
|\ | | | | | | | | vespa-engine/geirst/rewrite-attribute-tests-to-gtest Rewrite attribute tests to gtest
| * Simplify.Geir Storli2020-06-101-3/+1
| |
| * Rewrite attribute (writer) tests to use gtest.Geir Storli2020-06-102-316/+302
| |
| * Stop using script to run the attribute test.Geir Storli2020-06-102-7/+2
| | | | | | | | DirectoryHandler is already used to handle creation and deletion of test folder.
* | Merge pull request #13540 from ↵Geir Storli2020-06-109-63/+169
|\ \ | | | | | | | | | | | | vespa-engine/toregge/adjust-prepare-restart-flush-strategy-for-differentiated-replay-operation-costs Adjust prepare restart strategy for differentiated replay operation costs.
| * | Adjust prepare restart strategy for differentiated replay operation costs.Tor Egge2020-06-109-63/+169
| |/
* / redo fetchPostings for completenessArne Juul2020-06-101-0/+2
|/
* Merge pull request #13514 from ↵Geir Storli2020-06-096-34/+1
|\ | | | | | | | | vespa-engine/arnej/cleanup-unused-white-list-provider WhiteListProvider no longer used
| * WhiteListProvider no longer usedArne Juul2020-06-096-34/+1
| |
* | keep createFilterSearch public in subclassesArne Juul2020-06-091-10/+9
|/
* compute and apply global filter after fetchPostingsArne Juul2020-06-073-10/+21
|
* Merge pull request #13497 from vespa-engine/toregge/add-rhel-8-2-kludgeArnstein Ressem2020-06-052-0/+17
|\ | | | | Add kludge to avoid unresolved symbols with gcc-toolset-9 on RHEL 8.2
| * Add kludge to avoid unresolved symbols with gcc-toolset-9 on RHEL 8.2Tor Egge2020-06-052-0/+17
| |
* | use vector of UP as MultiSearch::ChildrenArne Juul2020-06-041-14/+17
|/ | | | | | * add helper class for constructing MultiSearch::Children (mostly for unit tests) * rewrite as needed to adapt
* Take read guard before creating match context.Henning Baldersheim2020-06-031-4/+1
|
* Use the docid limit in the base class.Henning Baldersheim2020-06-023-13/+9
|
* To avoid issues during feeding and resize stick to using the docidLimit used ↵Henning Baldersheim2020-06-021-1/+1
| | | | for the search.
* Merge pull request #13421 from vespa-engine/arnej/more-intermediate-filtersArne H Juul2020-06-021-2/+5
|\ | | | | Arnej/more intermediate filters