rmlSetMIOpenAutoTuningOn

MIOpen can run in 2 different ways.

It can either search for the best convolution algorithm for a given problem size or avoid that search. This mechanism is called auto tuning. By default it is off as it is a costly operation. But an application may want to use for performance reason. See Auto-tuning the kernels.

rml_status rmlSetMIOpenAutoTuningOn(	rml_context context,
	rml_bool on);

API

MIOpen Interoperation

Parameters

Parameter	Input/Output	Description
context	input	A valid context handle.
on	input	A boolean indicating if the auto tuning state for MIOpen convolution is on.

Return and Status Codes

Returns one of the following status codes:

• RML_OK if the operation is successful

• RML_ERROR_BAD_PARAMETER if context is invalid

To get more details in case of failure, call rmlGetLastError().