summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2017-02-13Added first version of the OverrideParameters functionCedric Nugteren
2017-02-13Fixed a small bug in GEMV: unused kernel in parameter listCedric Nugteren
2017-02-12Split the database into several smaller cached per-kernel databases (in ↵Cedric Nugteren
preparation of per-kernel database overrides)
2017-02-12Made RemoveBySubset from the cache work with references to keysCedric Nugteren
2017-02-11Added an option to remove items from the caches, optionally by a subset of 2 ↵Cedric Nugteren
specific key-values only
2017-02-08Added tuning results for Titan X (Pascal version)Cedric Nugteren
2017-01-24Updated the changelog for PR131 and PR132Cedric Nugteren
2017-01-24Merge pull request #132 from intelfx/cacheCedric Nugteren
Refactor cache subsystem
2017-01-24Database: pass Device instead of Queue for clarityIvan Shapovalov
2017-01-24Routine: cache the database instance as wellIvan Shapovalov
This does not change much, but will become useful in next commits when plugin support is introduced.
2017-01-24Database: ref-count the internal map for cachingIvan Shapovalov
2017-01-24Routine, Cache: generalize, reduce amount of copying in fast pathIvan Shapovalov
Implement a generalized Cache<K, V>. Two variants are provided: the first one is based on std::map, using C++14-specific transparent std::less<> and generalized std::map::find() to allow searching by tuple of references. The second one is based on std::vector and O(n) lookup, but remains C++11-compliant.
2017-01-24Merge pull request #131 from intelfx/miscCedric Nugteren
Assorted minor fixes
2017-01-24.travis.yml: do not build for osx twice, there's no gcc thereIvan Shapovalov
2017-01-24treewide: silence type mismatch warnings in *printf()Ivan Shapovalov
2017-01-24Tester: always fail on OpenCL and CLBlast internal errorsIvan Shapovalov
These errors are self-evident and enough to fail the test even if there is no clBLAS reference to compare error codes with.
2017-01-24FillCache: perform compilation for each precision separatelyIvan Shapovalov
Thus do not prevent filling cache for float if the device does not support e. g. double.
2017-01-24Routine: fix semi-warm routine construction (when binary is in cache)Ivan Shapovalov
There was a missing return statement in the semi-warm path that made CLBlast to continue to cold path after a cache hit.
2017-01-24src/clpp11.hpp: check pointers before clRelease*()Ivan Shapovalov
This is to avoid spurious "induced" errors on destruction, if construction failed for some reason.
2017-01-24src/clpp11.hpp: do not store program source/binary in ProgramIvan Shapovalov
The stored source/binary does not seem to serve any purpose, yet its presence makes Program a heavy (not pure refcounted) object, which is undesired esp. because it is copied from the cache in the hot path.
2017-01-24samples: add CL_USE_DEPRECATED_OPENCL_1_*_APIS where neededIvan Shapovalov
2017-01-20treewide: include clpp11.hpp first to silence deprecation warningsIvan Shapovalov
Otherwise, cl.h gets included through clblast.h before clpp11.hpp.
2017-01-20Routine: use PrecisionSupported<>() instead of duplicating the checkIvan Shapovalov
2017-01-19Added tuning results for NVIDIA GTX 1080 and Intel Core i7-4790KCedric Nugteren
2017-01-07Updated the link to cl.hpp in the Khronos registry for the samplesCedric Nugteren
2017-01-07Always enables cl_khr_fp64 when running double-precision, not just for ↵Cedric Nugteren
OpenCL 1.1 or lower
2017-01-03Added tuning results for the AMD Turks GPU and the Intel Core i7-2670QM CPUCedric Nugteren
2016-12-18Fixed a bug when using offsets in the direct GEMM kernelsCedric Nugteren
2016-11-29Made Intel GPUs always use the indirect version of the GEMM kernelCedric Nugteren
2016-11-27Updated to version 0.10.0Cedric Nugteren
2016-11-27Made it possible to use the command-line environmental variables for each ↵Cedric Nugteren
executable and without re-running CMake
2016-11-27Merge branch 'better_defaults' into developmentCedric Nugteren
2016-11-26Improved the default parameters for cases with non-common parameters across ↵Cedric Nugteren
all devices
2016-11-24Merge pull request #125 from CNugteren/netlib_blas_apiCedric Nugteren
Netlib CBLAS API for CLBlast
2016-11-23Made the Netlib SGEMM example also optionally compiledCedric Nugteren
2016-11-23Fixed a vector-size related bug in the CLBlast Netlib APICedric Nugteren
2016-11-23Made compilation of the Netlib CBLAS API conditionalCedric Nugteren
2016-11-23Fixed a bug in the HSCAL routineCedric Nugteren
2016-11-22Minor changes to ensure full compatibility with the Netlib CBLAS APICedric Nugteren
2016-11-20Made functions with scalar-buffers as output properly return valuesCedric Nugteren
2016-11-20Added performance results for the Skylake ULT GT2 GPUCedric Nugteren
2016-11-20Now correctly tests for validaty of the B matrix in the TRMM routineCedric Nugteren
2016-11-20Forced OpenCL 1.1 compilation and disabled a deprecation warningCedric Nugteren
2016-11-20Fixed a bug in the TRMM routine caused by overwriting input data before ↵Cedric Nugteren
consuming everything
2016-11-19Generating FP16 performance graphs now uses FP32 as a reference for comparisonCedric Nugteren
2016-11-19Changed the GEMM kernel selection parameters for Skylake GPUs to always ↵Cedric Nugteren
favour the regular kernel
2016-11-17Added a proper half-precision reference for testing of xomatcopyCedric Nugteren
2016-11-17Fixed a bug in the error margins; relaxed the error margins for half-precisionCedric Nugteren
2016-11-15Updated the tuning results for the Intel Skylake ULT GT2 GPUCedric Nugteren
2016-10-25Added an example and documentation for the Netlib CBLAS APICedric Nugteren