Todo List

Member cuda::context_t::maximum_depth_of_child_grid_synch_calls () const

Is this really a feature of the context? Not of the device?

Member cuda::context_t::maximum_outstanding_kernel_launches () const

Is this really a feature of the context? Not of the device?

Member cuda::context_t::memory_allocation_heap_size () const

Is this really a feature of the context? Not of the device?

Member cuda::context_t::printf_buffer_size () const

Is this really a feature of the context? Not of the device?

Member cuda::context_t::stack_size () const

Is this really a feature of the context? Not of the device?

Member cuda::context_t::stream_priority_range () const

isn't this a feature of devices?

Member cuda::device::peer_to_peer::disable_access (device_t accessor, device_t peer)

Consider disabling this, given that access is context-specific

Member cuda::device::peer_to_peer::enable_access (device_t accessor, device_t peer)

Consider disabling this, given that access is context-specific

Member cuda::grid::block_dimensions_t

Consider having both grid and block dims inhert from the same dimensions_t structure, but be incompatible, to prevent mis-casting one as the other.

Member cuda::link::options_t::fallback_strategy

Use an optional.

Member cuda::link::options_t::max_num_registers_per_thread

Use an optional.

Member cuda::link::options_t::min_num_threads_per_block

Use an optional.

Member cuda::link::options_t::optimization_level

Use an optional.

Member cuda::link::options_t::target

Use a variant or optional+variant.

Member cuda::memory::host::allocate (size_t size_in_bytes, allocation_options options)

Consider a variant of this supporting the cudaHostAlloc flags

Member cuda::memory::mapped::is_part_of_a_region_pair (const void *ptr)

What if it's a managed pointer?

Member cuda::memory::shared::size_t

consider using uint32_t.

Member cuda::module::load_from_file (const context_t &context, const char *path, link::options_t link_options={})

: consider adding load_module methods to context_t

: When switching to the C++17 standard, use string_view's instead of the const char* and ::std::string reference

Member cuda::rtc::compilation_options_t::additional_include_paths

In C++17, consider making the elements std::filesystem::path's.

Member cuda::rtc::compilation_options_t::compile_extensible_whole_program

explain what that is.

Member cuda::rtc::compilation_options_t::maximum_register_count

Use ::std::optional

Member cuda::rtc::compilation_options_t::preinclude_files

In C++17, consider making the elements std::filesystem::path's.

Check how these strings are interpreted. Do they need quotation marks? brackets? full paths?

Member cuda::rtc::compilation_options_t::targets_

Use something less fancy than ::std::unordered_set, e.g. a vector-backed ordered set or a dynamic bit-vector for membership.

Class cuda::runtime_error

Consider renaming this to avoid confusion with the CUDA Runtime.

Class cuda::span< T >

: Replace this with a more proper span, possibly in a file of its own.

Member cuda::stream_t::has_work_remaining () const

What if there are incomplete operations, but they're all waiting on something on another queue? Should the queue count as "busy" then?

Member cuda::synchronize (const event_t &event)

Determine how this waiting takes place (as opposed to stream synchronization).

Class cuda::texture::descriptor_t

Could be expanded into a richer wrapper class allowing actual settings of the various fields.

Member cuda::version_numbers::driver ()

In future CUDA versions which support C++17 - return an optional

File pointer.hpp

Consider allowing for storing attributes within the class, lazily (e.g. with an ::std::optional).