summaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Merge pull request #1085 from ↵Henning Baldersheim2016-11-1511-224/+215
|\ | | | | | | | | yahoo/balder/find-it-even-if-lidspace-compaction-has-moved-it-away-rebased-2 Balder/find it even if lidspace compaction has moved it away rebased 2
| * Sample twice and take the minimum.Henning Baldersheim2016-11-151-1/+2
| |
| * Remove redundant test.Henning Baldersheim2016-11-151-1/+0
| |
| * Move to spam.Henning Baldersheim2016-11-141-1/+1
| |
| * Add ability to find a gid that has moved.Henning Baldersheim2016-11-147-4/+41
| |
| * Log at the correct place and with correct level.Henning Baldersheim2016-11-141-1/+2
| |
| * If a possibility for lidmovement, actually verify that it has not moved.Henning Baldersheim2016-11-141-1/+18
| |
| * Retry docsum fetching if there has been lidspace compaction and your request ↵Henning Baldersheim2016-11-146-152/+116
| | | | | | | | might have been affected.
| * Refactor the summary fetching out so that it can be easily retried if ↵Henning Baldersheim2016-11-143-84/+56
| | | | | | | | documents have moved lid.
* | added rank setup verify test for unsupported AST nodeHaavard2016-11-141-0/+12
|/
* Implement disk write speed sampling.Tor Egge2016-11-0811-5/+260
| | | | Test that sampling info can be saved and overridden.
* Merge pull request #1019 from ↵Tor Egge2016-11-041-1/+1
|\ | | | | | | | | yahoo/geirst/move-datastore-code-into-separate-sub-library Geirst/move datastore code into separate sub library
| * Move datastore code from search::btree -> search::datastore.Geir Storli2016-11-041-1/+1
| |
* | Add skeleton of class that samples hardware to derive infoTor Egge2016-11-045-1/+44
| | | | | | | | (e.g. check if disk is slow for sequential writes).
* | Move HwInfo from vespalib to searchcore and simplify code.Tor Egge2016-11-0426-89/+103
| |
* | Pass on hwinfo from old attribute manager to new attribute manager.Tor Egge2016-11-041-1/+2
|/
* Merge pull request #1007 from ↵Geir Storli2016-11-0327-69/+259
|\ | | | | | | | | yahoo/toregge/use-memory-buffer-for-attribute-save-if-disk-is-slow-rebased Toregge/use memory buffer for attribute save if disk is slow rebased
| * Test that flushed attribute can be loaded again, regardless ofTor Egge2016-11-031-7/+50
| | | | | | | | flushing strategy.
| * Flush attribute to memory first if attribute saver has aTor Egge2016-11-0327-65/+206
| | | | | | | | | | | | | | read guard and we are using a spinning disk. This reduces the time the read guard is held (and corresponding extra long term memory usage due to large hold lists) at the cost of a temporary memory usage spike while flushing attribute.
| * Sync code structure with FlushableAttribute, factor out document meta storeTor Egge2016-11-032-10/+16
| | | | | | | | saving to separate method.
* | log level info for memory_flush_config_updaterBjørn Meland2016-11-031-2/+2
|/
* Removed TestApp::GetSourceDirectoryArnstein Ressem2016-11-025-13/+15
|
* Keep the readguard until docsumfetching is complete to avoid identitychange ↵Henning Baldersheim2016-10-311-5/+2
| | | | or ABA issue.
* Default bucketorder.Henning Baldersheim2016-10-281-2/+2
|
* Add validation of tensor attribute tensor type changes.Geir Storli2016-10-273-12/+50
|
* Merge pull request #903 from ↵Henning Baldersheim2016-10-183-18/+52
|\ | | | | | | | | yahoo/havardpe/support-fewer-documents-than-threads-with-adaptive-docid-range-scheduling Havardpe/support fewer documents than threads with adaptive docid range scheduling
| * fix corner case, add test and add time bomb to other testsHaavard2016-10-183-18/+52
| |
* | Merge pull request #900 from yahoo/toregge/move-tensor-attribute-sourceGeir Storli2016-10-184-4/+4
|\ \ | | | | | | Move tensor attribute source to separate directory.
| * | Move tensor attribute source to separate directory.Tor Egge2016-10-184-4/+4
| |/
* / Remove memory leak at out of threads.Henning Baldersheim2016-10-181-4/+6
|/
* Merge pull request #894 from yahoo/balder/always-provide-coverageHenning Baldersheim2016-10-1710-103/+17
|\ | | | | Balder/always provide coverage
| * GC long gone PCODE_MLD_GETDOCSUMS and PCODE_GETDOCSUMSHenning Baldersheim2016-10-141-5/+3
| |
| * remove unused PCODE_MONITORQUERYHenning Baldersheim2016-10-141-7/+2
| |
| * Only one version of QUERYXHenning Baldersheim2016-10-141-2/+1
| |
| * QFLAG_REPORT_QUEUELEN is not used, nor does it work, nor is it very likely ↵Henning Baldersheim2016-10-141-7/+0
| | | | | | | | that it will work.
| * Always produce coverage.Henning Baldersheim2016-10-143-3/+0
| |
| * Remove always true QFLAG_REPORT_COVERAGEHenning Baldersheim2016-10-142-8/+3
| |
| * Always allow error packets.Henning Baldersheim2016-10-145-23/+9
| |
| * Remove unused flag QFLAG_CACHE_MASKHenning Baldersheim2016-10-142-49/+0
| |
* | Remove the XXXFactory indirection. Just use static method on Alloc.Henning Baldersheim2016-10-161-2/+1
|/
* make sure search coverage information is updated after query limitingHaavard2016-10-142-14/+11
|
* Revert "Revert "Balder/split match loop in inner and outer loop""Henning Baldersheim2016-10-142-160/+239
|
* Revert "Balder/split match loop in inner and outer loop"Henning Baldersheim2016-10-132-239/+160
|
* Merge pull request #858 from ↵Henning Baldersheim2016-10-132-160/+239
|\ | | | | | | | | yahoo/balder/split-match-loop-in-inner-and-outer-loop Balder/split match loop in inner and outer loop
| * clean up match limiting and improve coverage estimationHaavard2016-10-132-19/+30
| |
| * isolate work sharing (detangle from iterator init/step)Haavard2016-10-132-22/+17
| |
| * normalize code orderHaavard2016-10-132-122/+124
| |
| * Use the correct one.Henning Baldersheim2016-10-121-1/+1
| |
| * Params -> ContextHenning Baldersheim2016-10-122-20/+20
| |
| * Checkpoint that build and test fine.Henning Baldersheim2016-10-122-55/+64
| |