summaryrefslogtreecommitdiffstats
path: root/eval
Commit message (Collapse)AuthorAgeFilesLines
* use const references and stash instead of UPHåvard Pettersen2017-11-0915-222/+201
|
* clean up tensor engine APIHåvard Pettersen2017-11-0744-434/+352
| | | | make Tensor a subclass of Value
* add tests for set membership with tensorsHåvard Pettersen2017-11-064-1/+37
|
* remove 'equal' concept for tensorsHåvard Pettersen2017-11-0315-161/+27
|
* handle 'in' operator as custom (tensor) map operationHåvard Pettersen2017-11-0317-305/+224
| | | | | | free arrays no longer allowed restrict set members to be numbers or strings auto-unbox negative numbers in AST
* unify conformance test format, refactor code and update specHåvard Pettersen2017-11-017-468/+264
|
* remove letHåvard Pettersen2017-10-3117-302/+33
|
* simple optimizations for double map/joinHåvard Pettersen2017-10-277-4/+54
|
* test both typed and untyped evaluationHåvard Pettersen2017-10-271-10/+18
| | | | the reference is untyped SimpleTensorEngine
* test interpret speed for untyped simple and typed defaultHåvard Pettersen2017-10-271-10/+11
| | | | also fix printing of simple interpret speed for big function
* also test with resolved typesHåvard Pettersen2017-10-271-3/+8
|
* remove no longer needed operation abstractionsHåvard Pettersen2017-10-2716-476/+178
|
* normalize to fewer operation primitivesHåvard Pettersen2017-10-262-108/+124
|
* add whitespaceHåvard Pettersen2017-10-261-15/+15
|
* remove TensorSum (use more generic TensorReduce instead)Håvard Pettersen2017-10-2616-155/+35
|
* test both reference and production tensor enginesHåvard Pettersen2017-10-261-16/+31
|
* it is not allowed to reduce over a non-existing dimensionHåvard Pettersen2017-10-261-3/+7
|
* remove old reduce/map/apply tensor APIHåvard Pettersen2017-10-2513-409/+44
| | | | | | | | | | | | | | | remove Operation::perform and Value::apply; invoke TensorEngine map/join directly instead remove no-longer-needed operation proxy classes using thread-local storage add fast-path evaluation of double-only map/join in SimpleTensorEngine to avoid creating tensors when doing interpreted constant folding during LLVM compilation enable inlining of multiply join operation in DefaultTensorEngine::join
* use new api in tensor function irHåvard Pettersen2017-10-2411-110/+110
|
* implement new 'reduce' API in DefaultTensorEngineHåvard Pettersen2017-10-243-28/+59
| | | | | | | | | add a singleton instance of ErrorValue, and use that instead of creating new instances many places move null pointer checks inside to_value() simplify using statements
* implement new join APIHåvard Pettersen2017-10-238-1/+96
| | | | | only 'match' is inlined added new join function to tensor::Tensor
* implement new map APIHåvard Pettersen2017-10-231-4/+23
|
* use static functions for low-level operation evalHåvard Pettersen2017-10-202-81/+134
|
* use raw function pointers for map/joinHåvard Pettersen2017-10-1910-24/+94
|
* update spec with new cases (~1Mb+ json)Håvard Pettersen2017-10-131-5/+1209
|
* generate more testsHåvard Pettersen2017-10-132-7/+213
|
* split code to generate tensor specs into separate fileHåvard Pettersen2017-10-133-234/+249
|
* use new top-level convenience accessorHåvard Pettersen2017-10-131-6/+6
|
* The first argument to fastMatch() should be the tensor with fewest cells.Tor Egge2017-10-121-1/+6
|
* avoid using source tree location to locate binariesHåvard Pettersen2017-10-041-10/+10
|
* wire in spec-based tensor conformance testingHåvard Pettersen2017-10-035-109/+135
| | | | | | verify conformance spec generation verify C++ default expression evaluation bonus: verify binary format test spec generation
* Revert "Revert "Havardpe/avoid reading past json value in slime json parser""Arnstein Ressem2017-10-022-3/+3
|
* Revert "Havardpe/avoid reading past json value in slime json parser"Arnstein Ressem2017-10-022-3/+3
|
* avoid implicit cast to boolHåvard Pettersen2017-09-292-3/+3
|
* avoid requiring that json parsing consumes all inputHåvard Pettersen2017-09-292-3/+3
| | | | (the input could end with \n if human readable)
* application to facilitate cross-language tensor conformance testing - WIPHåvard Pettersen2017-09-288-3/+413
| | | | with a few smoke tests to show the format
* clean up stuffHåvard Pettersen2017-08-221-88/+10
|
* Include what you need.Henning Baldersheim2017-08-188-51/+116
|
* Use VESPA_ prefix for new cmake variables.Tor Egge2017-08-101-1/+1
|
* Prepare for using other versions of boost and llvm.Tor Egge2017-08-101-1/+1
|
* added modulo operator (%)Haavard2017-06-2713-43/+57
|
* support tensor resultsHaavard2017-06-231-2/+11
|
* Use locale insensitive strtod and strtof.Henning Baldersheim2017-06-191-5/+4
|
* Update copyright headersJon Bratseth2017-06-14208-208/+208
|
* Revert "Update copyright headers"Jon Bratseth2017-06-14208-208/+208
|
* Update copyright headersJon Bratseth2017-06-14208-208/+208
|
* Revert "Copyright header"Jon Bratseth2017-06-13208-208/+208
|
* Copyright headerJon Bratseth2017-06-13208-208/+208
|
* auto-detect appropriate compiled function parameter lazynessHaavard2017-06-076-0/+89
| | | | ... but still allow config override
* Add support for conditionally running available shutdown.Henning Baldersheim2017-06-071-0/+5
|