The OpenCL backend test for Min1D0 to Min4D0 and Max1D0 to Max4D0 in sort_by_key.cpp fails for half (fp16) data type due to incorrect results from a call sort0ByKey operation done in opencl/sort_index.cpp . The library used for this operation is boost compute which does not officially support fp16 operations (see issue)

Description

• Built Arrayfire commit eaa49ca
• Which backend is experiencing this issue? OpenCL
• Do you have a workaround? Forcing internatlly cast to float before executing this operation
• Can the bug be reproduced reliably on your system? Yes

Reproducible Code and/or Steps

Run test_topk_opencl. Reproducible with CPU, Nvidia, and Intel GPUs