Age | Commit message (Collapse) | Author |
|
|
|
times; requires CLTune 2.5.0
|
|
|
|
|
|
reproducability of results
|
|
NWGD and KWGD into one WGD parameter
|
|
indirect version
|
|
|
|
|
|
environmental variable
|
|
can't handle long strings
|
|
now automatically taken from 32-bit if there are no entries at all
|
|
and convienient plain JSON/dict data-type
|
|
M370X GPU
|
|
best-performing cases for a specific parameters combination
|
|
explored exhaustively and a larger set which is explored randomly
|
|
problems if C contains NaNs
|
|
|
|
handle duplicate entries of different runs
|
|
Support passing environment variables CLBLAST_PLATFORM and CLBLAST_DEVICE
instead of -platform and -device arguments to test executables.
This is for `ctest`.
|
|
|
|
method as for known device groups
|
|
search space to have a better chance to evaluate more likely parameter combinations
|
|
dvasschemacq-master
Conflicts:
src/kernels/level1/xaxpy.opencl
src/kernels/level2/xgemv.opencl
src/kernels/level2/xgemv_fast.opencl
src/kernels/level2/xger.opencl
src/kernels/level2/xher.opencl
src/kernels/level2/xher2.opencl
src/kernels/level3/xgemm_part2.opencl
|
|
In OpenCL 1.1 __kernel has to be before __attribute__, at least with
Vivante compiler.
|
|
tuning results common for a device/vendor type
|
|
|
|
|
|
couple of devices
|
|
|
|
|
|
and kernels
|
|
|
|
|
|
enabling better memory performance
|
|
by routine implementation
|
|
support empty LWS
|
|
|
|
|
|
Do not hardcode the knowledge about "A and C col-major, B row-major".
This allows for easier reuse of the DoGemm() routine with different
kernels.
|
|
though
|
|
slow reference kernel as a place-holder
|
|
|
|
Make sure the passed types are large enough.
|
|
|
|
Make sure all out parameters that are passed to functions such
as clGetDeviceInfo are large enough to contain the replies.
|
|
case of fp16 arguments are cast on host and in kernel
|
|
|
|
OpenCL context
|
|
verbose print statements to the cache
|