diff options
author | Vidhya Sudhan Loganathan <vidhyasudhan.loganathan@arm.com> | 2019-02-07 11:14:42 +0000 |
---|---|---|
committer | VidhyaSudhan Loganathan <vidhyasudhan.loganathan@arm.com> | 2019-02-12 14:02:47 +0000 |
commit | ca65af3542134841b79b94060195fd9a65224d6a (patch) | |
tree | b8fce60e6bbd7d4dca300fe90650ea3280186984 /src | |
parent | 581c898aa5612e60392e3b785d1ac216ce7a4650 (diff) | |
download | ComputeLibrary-ca65af3542134841b79b94060195fd9a65224d6a.tar.gz |
COMPMID-1919: Opencl - Tuning with Instrumentation does not work
Change-Id: Iee6a07d5bf6a35af04071865682bcc4a615c14f9
Signed-off-by: Vidhya Sudhan Loganathan <vidhyasudhan.loganathan@arm.com>
Reviewed-on: https://review.mlplatform.org/638
Reviewed-by: Giuseppe Rossini <giuseppe.rossini@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
Diffstat (limited to 'src')
-rw-r--r-- | src/runtime/CL/CLTuner.cpp | 8 |
1 files changed, 6 insertions, 2 deletions
diff --git a/src/runtime/CL/CLTuner.cpp b/src/runtime/CL/CLTuner.cpp index 56baad82c3..a262d6b95c 100644 --- a/src/runtime/CL/CLTuner.cpp +++ b/src/runtime/CL/CLTuner.cpp @@ -161,8 +161,6 @@ cl::NDRange CLTuner::find_optimal_lws(ICLKernel &kernel) auto interceptor = [this](cl_command_queue command_queue, cl_kernel kernel, cl_uint work_dim, const size_t *gwo, const size_t *gws, const size_t *lws, cl_uint num_events_in_wait_list, const cl_event * event_wait_list, cl_event * event) { - ARM_COMPUTE_ERROR_ON_MSG(event != nullptr, "Not supported"); - ARM_COMPUTE_UNUSED(event); if(this->kernel_event_is_set()) { // If the event is already set it means the kernel enqueue is sliced: given that we only time the first slice we can save time by skipping the other enqueues. @@ -174,6 +172,12 @@ cl::NDRange CLTuner::find_optimal_lws(ICLKernel &kernel) // Set OpenCL event this->set_cl_kernel_event(tmp); + if(event != nullptr) + { + //return cl_event from the intercepted call + clRetainEvent(tmp); + *event = tmp; + } return retval; }; CLSymbols::get().clEnqueueNDRangeKernel_ptr = interceptor; |