summaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Make it compatible with c++2aHenning Baldersheim2020-12-021-2/+2
|
* Make it compatible with c++2aHenning Baldersheim2020-12-027-19/+19
|
* use SimpleValue instead of EngineOrFactoryArne Juul2020-12-028-42/+47
|
* Move to .cpp fileHenning Baldersheim2020-12-012-10/+13
|
* Avoid polluting global namespaceHenning Baldersheim2020-12-011-1/+1
|
* Also test distributr directly on chainHenning Baldersheim2020-12-011-2/+5
|
* Just take a nap when Q is full.Henning Baldersheim2020-12-012-20/+8
|
* Misc cleanup of includes and code healthHenning Baldersheim2020-12-011-5/+8
|
* Avoid creating buckets via SPI when using distributor.Tor Egge2020-11-3012-37/+1
|
* Revert cleanup that was a bit too optimistic.Henning Baldersheim2020-11-271-0/+5
|
* Merge pull request #15503 from vespa-engine/balder/gc-loadtype-setHenning Baldersheim2020-11-275-27/+7
|\ | | | | Balder/gc loadtype set
| * GC LoadType and PriorityMapper.Henning Baldersheim2020-11-271-5/+0
| |
| * GC LoadTypeSetHenning Baldersheim2020-11-275-22/+7
| |
* | use factory.from_spec() instead of tensor::makeTensor<T>Arne Juul2020-11-272-9/+11
| |
* | remove superfluous includeArne Juul2020-11-272-3/+1
| |
* | Merge pull request #15498 from vespa-engine/arnej/use-default-tensor-engine-lessArne H Juul2020-11-274-4/+0
|\ \ | |/ |/| Arnej/use default tensor engine less
| * cleanup unused includeArne Juul2020-11-274-4/+0
| |
* | Remove loadmetrics indirection.Henning Baldersheim2020-11-267-8/+6
|/
* Use correct storage cluster name.Henning Baldersheim2020-11-261-1/+1
|
* As we have have now removed the expensive Route member we can further ↵Henning Baldersheim2020-11-264-4/+16
| | | | | | | | | | compact the message objects. - Compact StorageMessageAddress to 16 bytes by - using reference to cluster name. - Use small enums for protocol and node type. - Avoid having StorageMessage as separate allocation. - Avoid default values
* Create the mbus::Route on demand instead of storing it inside ↵Geir Storli2020-11-254-4/+10
| | | | | | StorageMessageAddress. Creating and deleting the route is expensive and not used with RPC for Storage API communication.
* Merge pull request #15442 from vespa-engine/arnej/flip-engine-defaultArne H Juul2020-11-241-1/+1
|\ | | | | Flip default to FastValueBuilderFactory
| * flip default in proton.def alsoArne Juul2020-11-241-1/+1
| |
* | Remove load type from ContextHenning Baldersheim2020-11-243-43/+21
|/
* Place _serial_num member variable earlier in the class.Tor Egge2020-11-201-2/+2
|
* Simplify return statements.Tor Egge2020-11-203-3/+3
|
* Update comments.Tor Egge2020-11-201-1/+5
|
* Delay gid to lid change notifications for put operations.Tor Egge2020-11-1928-299/+260
|
* Add comment for _prunedSerialNum.Tor Egge2020-11-171-0/+1
|
* Use snake case: incSerialNum() ==> inc_serial_num()Tor Egge2020-11-178-9/+9
|
* Adjust serial number in feed handler during replay to ensureTor Egge2020-11-1711-27/+99
| | | | proper semantics when using IGetSerialNum interface.
* Enforce minimum bucked used bits at document metastore load timeTor Brede Vekterli2020-11-123-1/+29
|
* Inline default GidToLidMapKey constructor.Tor Egge2020-11-122-14/+10
|
* suammry -> summaryHenning Baldersheim2020-11-121-2/+2
|
* Control the number of rpc threads and set it to #cores/4 instead of the ↵Henning Baldersheim2020-11-123-5/+8
| | | | fixed number of 48.
* Name the threads so it is easier to see who is doing what.Henning Baldersheim2020-11-1211-22/+50
|
* Merge pull request #15291 from ↵Henning Baldersheim2020-11-1112-60/+137
|\ | | | | | | | | vespa-engine/toregge/add-gid-to-lid-map-key-class-for-document-meta-store Add 32 most significant bits of gid key to key in btree mapping from
| * Add 32 most significant bits of gid key to key in btree mapping fromTor Egge2020-11-1112-60/+137
| | | | | | | | | | gid to lid in document meta store to reduce amount of CPU cache misses during lookup.
* | Avoid stalling the piepline by just daisy chaining the feedtokens.Henning Baldersheim2020-11-111-3/+16
|/
* Remove IGidCompare interface.Tor Egge2020-11-1114-144/+22
|
* Tune parameters to show differences in performance.Henning Baldersheim2020-11-091-8/+8
|
* Merge pull request #15229 from vespa-engine/balder/add-test-scriptHenning Baldersheim2020-11-091-0/+32
|\ | | | | Add a small test script for running a set of benchmarks.
| * Use 1 client thread for spi only and 2 for rest.Henning Baldersheim2020-11-091-3/+3
| |
| * Add dashes to client-threadsHenning Baldersheim2020-11-091-1/+1
| |
| * Add a small test script for running a set of benchmarks.Henning Baldersheim2020-11-091-0/+32
| |
* | Merge pull request #15230 from vespa-engine/balder/gc-unused-legacy-flagHenning Baldersheim2020-11-091-11/+0
|\ \ | | | | | | GC unused legacy bucket db flag
| * | GC unused legacy bucket db flagHenning Baldersheim2020-11-091-11/+0
| |/
* | Merge pull request #15228 from ↵Henning Baldersheim2020-11-099-178/+140
|\ \ | | | | | | | | | | | | vespa-engine/balder/avoid-cachemiss-on-checksum-aggregation Avoid cache miss and dependency stall when have virtual small objects…
| * | Only assert on the bytes that are actually initialized.Henning Baldersheim2020-11-091-2/+1
| | |
| * | Avoid cache miss and dependency stall when have virtual small objects for ↵Henning Baldersheim2020-11-099-178/+141
| |/ | | | | | | | | | | checksum aggregation. Instead compute checksum inline with a simple switch.