COMPMID-2043: Add support for "dummy threads" in CLGEMMReshaped
Change-Id: I89403b97503fbb99f6a32f5d62b8c535ab26a7be
Signed-off-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Reviewed-on: https://review.mlplatform.org/c/877
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/arm_compute/core/CL/CLHelpers.h b/arm_compute/core/CL/CLHelpers.h
index 78427c3..16fe09f 100644
--- a/arm_compute/core/CL/CLHelpers.h
+++ b/arm_compute/core/CL/CLHelpers.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2016-2018 ARM Limited.
+ * Copyright (c) 2016-2019 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -144,5 +144,14 @@
* @return preferred vector width
*/
size_t preferred_vector_width(const cl::Device &device, DataType dt);
+
+/** Helper function to check if "dummy work-items" are preferred to have a power of two NDRange
+ * In case dummy work-items is enabled, it is OpenCL kernel responsibility to check if the work-item is out-of range or not
+ *
+ * @param[in] device A CL device
+ *
+ * @return True if dummy work-items should be preferred to dispatch the NDRange
+ */
+bool preferred_dummy_work_items_support(const cl::Device &device);
}
#endif /* __ARM_COMPUTE_CLHELPERS_H__ */