HPX_HOST_DEVICE
HPX_FORCEINLINE static
T call(T t, std::size_t max_count, Stride& offset, std::true_type)
{
if (is_negative(offset))
{
offset = Stride(negate(
// NVCC seems to have a bug with std::min...
max_count < std::size_t(negate(offset)) ?
max_count : negate(offset)
));
return T(t + offset);
}
// NVCC seems to have a bug with std::min...
offset = Stride(max_count < std::size_t(offset) ? max_count : offset);
return T(t + offset);
}
HPX_HOST_DEVICE
HPX_FORCEINLINE static
Iter call(Iter iter, std::size_t max_count, Stride& offset, std::false_type)
{
// advance through the end or max number of elements
// NVCC seems to have a bug with std::min...
offset = Stride(max_count < std::size_t(offset) ? max_count : offset);
std::advance(iter, offset);
return iter;
}
HPX_HOST_DEVICE
HPX_FORCEINLINE static
Iter call(Iter iter, std::size_t max_count, Stride& offset, std::true_type)
{
// advance through the end or max number of elements
if (!is_negative(offset))
{
// NVCC seems to have a bug with std::min...
offset = Stride(max_count < std::size_t(offset) ? max_count : offset);
std::advance(iter, offset);
}
else
{
offset = negate(Stride(
// NVCC seems to have a bug with std::min...
max_count < negate(offset) ? max_count : negate(offset)
));
std::advance(iter, offset);
}
return iter;
}
HPX_HOST_DEVICE
HPX_FORCEINLINE static
Iter call(Iter iter, std::size_t max_count, Stride& offset)
{
// anything less than a bidirectional iterator does not support
// negative offsets
HPX_ASSERT(!std::is_signed<Stride>::value || !is_negative(offset));
// NVCC seems to have a bug with std::min...
Stride count =
Stride(max_count < std::size_t(offset) ? max_count : offset);
// advance through the end or max number of elements
for (/**/; count != 0; (void)++iter, --count)
/**/;
offset -= count;
return iter;
}
std::shared_ptr<SL::Remote_Access_Library::Utilities::Image> SL::Remote_Access_Library::Utilities::Image::CreateImage(unsigned int h, unsigned int w, const char * data, size_t pixel_stride)
{
assert(pixel_stride == 3 || pixel_stride == 4);
auto img = CreateImage(h, w);
img->Size = h*w*img->Stride();
if (pixel_stride == 4) {
memcpy(img->data(), data, img->Size);
}
else {
auto dst = (unsigned int*)img->data();
auto src = data;
for (decltype(h*w) i = 0; i < h*w; i++) {
auto srcdata = *((unsigned int*)src);
*dst++ = (srcdata & 0xffffff00) | 0x000000ff;
src += 3;
}
}
return img;
}
inline void createTensor4dDesc(cudnnTensorDescriptor_t* desc, Size size) {
createTensor4dDesc<Dtype>(desc, size, Stride(size));
}
