Added cl_half support for test_select

[shajder]

According to issue description (select section):

#142

[EwanC]

Tried this PR on Codeplay's implementation and can see the new tests pass (and skip correctly against an implementation without cl_khr_fp16), but need to review the PR in more detail.

[EwanC]

Tried this PR on Codeplay's implementation and can see the new tests pass (and skip correctly against an implementation without cl_khr_fp16), but need to review the PR in more detail.

To follow up, I've looked through this again, and happy that how it adds half support is consistent with the other tests.

How/when to do the refactoring to make it nicer is another question.

[shajder]

@EwanC due to the fact my project comes to an end and I can't see any progress around PR I am blocked with I decided to move on.

Except of corrections for above requests I've added optimization which seems to speed up the test ~18% at my machine. Moreover, I've replaced CL objects with specific wrappers.

[EwanC]

@EwanC due to the fact my project comes to an end and I can't see any progress around #1665 I am blocked with I decided to move on.

Except of corrections for above requests I've added optimization which seems to speed up the test ~18% at my machine. Moreover, I've replaced CL objects with specific wrappers.

Had a look through this again, and happy with where we've got to with this change.

[svenvh]

LGTM, but please remove the excess semicolon that Ewan pointed out.

[bashbaug]

LGTM, aside from one minor comment.

I didn't do a super detailed code review but I did verify that the additions pass on our implementation and I did not see any change in coverage (aside from the added half tests). Thanks!

[shajder]

Close/open to trigger CI