aboutsummaryrefslogtreecommitdiffstats
path: root/eval
Commit message (Collapse)AuthorAgeFilesLines
...
* added skeleton for lambda peek optimizerHåvard Pettersen2020-04-034-0/+34
|
* delay preparing tensor lambda function for executionHåvard Pettersen2020-04-033-13/+25
| | | | | | This will allow implementation-specific tensor lambda optimizations to look at the lambda function and perform appropriate optimizations before it is converted to an interpreted function.
* make tensor engine available when compiling tensor functionsHåvard Pettersen2020-04-0324-55/+67
|
* remove parameter count from interpreted functionsHåvard Pettersen2020-04-035-12/+7
|
* added support for exporting a subset of node typesHåvard Pettersen2020-04-033-5/+78
| | | | | | | This is needed to store type information about tensor lambda inner functions until it is needed; we want to delay making it into an interpreted function until after the actual tensor engine implementation gets a chance to come up with a better optimization.
* Eliminate redundant move in return statement.Tor Egge2020-03-211-1/+1
|
* Merge pull request #12651 from vespa-engine/havardpe/improve-rank-feature-errorsHenning Baldersheim2020-03-202-6/+16
|\ | | | | Havardpe/improve rank feature errors
| * better tensor lambda type errorsHåvard Pettersen2020-03-202-6/+16
| | | | | | | | | | - report actual return type when not double - import type errors from lambda function type resolving
* | ReinlineHenning Baldersheim2020-03-202-16/+9
| |
* | Stick with one way of getting an accelrator.Henning Baldersheim2020-03-201-6/+6
| |
* | Use a common accelrator instance.Henning Baldersheim2020-03-191-1/+1
|/
* Merge pull request #12619 from vespa-engine/balder/optimize-value-excutorsHenning Baldersheim2020-03-195-48/+74
|\ | | | | Balder/optimize value excutors.
| * Use vespalib::hash_set instead of std::set to reduce number of allocation ↵Henning Baldersheim2020-03-185-48/+74
| | | | | | | | and epeed it up. Also use faster 2^N AND based hash tables.
* | fix dimension list printingHåvard Pettersen2020-03-192-1/+5
| |
* | print more details about type errorsHåvard Pettersen2020-03-193-21/+89
|/
* handle tensor lambda as nested function with bindingsHåvard Pettersen2020-03-1120-96/+361
|
* - Remove unused includes.Henning Baldersheim2020-03-0515-50/+18
| | | | | | | - = default - push_back -> emplace_back - std::move on vector. No semantic changes.
* using NUM_DOCS is wrong for remove benchmark, use EFFECTIVE_DOCSArne Juul2020-02-273-2/+4
|
* Merge pull request #12321 from vespa-engine/arnej/rework-ann-filter-bmArne H Juul2020-02-2619-885/+1850
|\ | | | | Arnej/rework ann filter bm
| * add common header fileArne Juul2020-02-261-0/+203
| |
| * keep more code commonArne Juul2020-02-253-788/+427
| |
| * split out common subroutinesArne Juul2020-02-257-491/+190
| |
| * add and verify filter optionArne Juul2020-02-2413-259/+853
| | | | | | | | split out common subroutines
| * experimental HNSW with various extensionsArne Juul2020-02-241-0/+830
| |
* | - Add debug logging.Henning Baldersheim2020-02-235-42/+24
|/ | | | | - std::make_unique - Reduce code visibility.
* Fix issues detected by clang 10.Tor Egge2020-02-141-2/+2
|
* Use llvm 10 on Fedora rawhide.Tor Egge2020-02-142-0/+4
|
* cannot use std::aligned_allocArne Juul2020-02-112-8/+10
|
* * add "remove" benchmarkArne Juul2020-02-116-72/+652
| | | | | | | | * redo ops tracking * use std::aligned_alloc * more stats - measure reach This reverts commit 37fd87978ab1c3abfa840403e4e8f289d5ea4a20.
* Revert "* remove benchmark"Henning Baldersheim2020-02-116-652/+72
|
* avoid actual HNSW library hereArne Juul2020-02-071-2/+0
|
* update copyrightArne Juul2020-02-071-1/+1
|
* * remove benchmarkArne Juul2020-02-066-72/+654
| | | | | | * redo ops tracking * use std::aligned_alloc * more stats - measure reach
* verify that immutable inputs are not modifiedHåvard Pettersen2020-01-312-0/+16
|
* update cblas includesHåvard Pettersen2020-01-302-2/+2
|
* use openblas for dot productHåvard Pettersen2020-01-302-31/+38
|
* use openblas for matrix vector multiplicationHåvard Pettersen2020-01-303-114/+109
|
* Adjust include diretories for compilation on RHEL, CentOS and Fedora.Tor Egge2020-01-301-4/+0
|
* Compile with openblas on Darwin.Tor Egge2020-01-301-0/+4
|
* use openblas for matrix multiplicationHåvard Pettersen2020-01-293-33/+52
|
* Merge pull request #11995 from vespa-engine/toregge/fix-format-stringsArne H Juul2020-01-292-12/+14
|\ | | | | Fix format strings.
| * Fix format strings.Tor Egge2020-01-292-12/+14
| |
* | better coverage of cell type and parameter orderingHåvard Pettersen2020-01-281-47/+69
|/
* Merge pull request #11960 from vespa-engine/arnej/refactor-hnsw-algo-onlyGeir Storli2020-01-275-156/+385
|\ | | | | refactor and unify
| * fix copyright year etcArne Juul2020-01-271-2/+3
| |
| * refactor and unifyArne Juul2020-01-275-156/+384
| | | | | | | | | | | | | | * also: rename QueueEntry -> HnswHit * add statistics * add removeDoc implementation * add hnsw-wrap (not compiled) for comparison
* | Merge pull request #11901 from vespa-engine/havardpe/tensor-matmulHåvard Pettersen2020-01-277-0/+437
|\ \ | |/ |/| matrix multiplication
| * matrix multiplicationHåvard Pettersen2020-01-227-0/+437
| |
* | Eliminate clang warning when casting to pointer with larger alignment.Tor Egge2020-01-241-1/+1
| |
* | Merge pull request #11921 from ↵Geir Storli2020-01-241-10/+27
|\ \ | | | | | | | | | | | | vespa-engine/geirst/simplify-running-gist-data-set Make it simpler to run the ANN_GIST1M data set.