Now all tests work even if Context::context, Context::device,
Context::queue are not default context, device and queue. This
is required for developing better tests in the future.
Note: Some tests may work only for default context/queue/device
since classes that they test work only for default context/q/d.
There are two solutions for this problem: either those tests run
on default queue (no matter what) or they does not run when
Context::context is not the default context. See test_string.cpp.
When InputIterator (host) is a non-contiguous iterator we don't
need a separate algorithm for cases when value_types of InputIterator
and OutputIterator (device) do not match and cases when they do
match.
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.
libc++ fails to extract a valid float/double literal when it is followed
by any special characters that can also be found in a float value,
in this case, f (https://llvm.org/bugs/show_bug.cgi?id=17782).
This commit works around the problem by not putting the "f" suffix into
the string for extraction.
The tests are failing under OS X/clang, but it isn't obvious to me why.
This commit simplifies the test code by ignoring the stringstream I/O
status, and reorganizes it to be clearer whether it's the initial
conversion to string that is failing, or the subsequent conversion back
from string.
make_literal is losing some precision when making literal floating
point values because it uses digits10, but that only gives us the
number of decimal digits that will survive a decimal->native>decimal
conversion; what we are doing is a native->decimal->native conversion,
which requires the user of ::max_digits10 (which is 2 (double) or 3
(float) larger than ::digits10).
max_digits10 is a c++11 feature, however, so this commit uses
digits10 + 3 when the c++11 numeric_limits isn't available.
