From tests on Travis-CI it seems that on Apple OpenCL platform
clCreatBuffer() does not return NULL and does not set error to
CL_INVALID_BUFFER_SIZE when size of the buffer memory object is
greater than CL_DEVICE_MAX_MEM_ALLOC_SIZE.
Tests for inplace_reduce, radix_sort and radix_sort_by_key are disabled
for CPU devices on Apple platform because for CPU devices on Apple
platform when local memory is used in a kernel, local work group size
must be [1;1;1]. Those algorithms are designed for GPU devices,
therefore we do no lose any functionality.
clEnqueueSVMMemcpy() operation does not work on AMD devices due to a bug
in drivers (https://community.amd.com/thread/190585). This affects copy()
algorithm (when SVM is used) and results in copy_svm_ptr failing. Now this
test case is skipped on AMD devices.
This adds a couple new functions for checking the vendor of a
compute device. This is useful for algorithms which specialize
based on the type of the underlying hardware.
This adds a check to skip tests which use fill() with pair and
tuple types on AMD platforms. There is a bug which crashes the
OpenCL compiler with an "UNREACHABLE executed!" message on AMD
platforms when using struct assignment in kernel code.
See: http://devgurus.amd.com/thread/166622
