Narrow down by specifying conditions
There are currently 4292 hits.
Intel: I am using the Intel® FPGA SDK for OpenCL™ environment to create and verify an OpenCL™ host application. will decrease. (It takes a long time to process one OpenCL™ kernel.) What could be the cause of the performance degradation?
Category: OpenCL™
Tools: Intel® FPGA SDK for OpenCL™
device:-
I assume that you are calling the OpenCL™ API: clEnqueueNDRangeKernel when running the OpenCL™ kernel, but are you using the event object that can be obtained with the argument "cl_event *event" of this API?
If the event object is used, it is necessary to release the event object by calling the OpenCL™ API: clRelaseEvent after the OpenCL™ kernel execution is completed.
In the OpenCL™ Runtime Environment (RTE), the OpenCL™ specification states that when an event object is created with clEnqueueNDRangeKernel, it must manage all events until the object is released.
Therefore, if clReleaseEvent is not executed, the number of events to be managed will continue to increase, leading to a drop in processing performance.
Experienced FAE
Free consultation is available.
From specific product specifications to parts selection, the Company FAE will answer your technical concerns free of charge. Please feel free to contact us.