eyalroz / cuda-api-wrappers

Thin, unified, C++-flavored wrappers for the CUDA APIs

eyalroz/cuda-api-wrappers Issues

Widen support for textures, arrays and surfaces
Updated 3 days ago
Switch to cuMemAdvise_v2 with CUDA v12 and later
Updated 3 days ago
Accept execution affinity parameters when creating a context
Updated 3 days ago
cuCtxGetSharedMemConfig is deprecated as of CUDA 12.3, acknowledge this
Closed 3 days ago
Support conditional graph nodes
Updated a month ago
Support graph edge properties
Updated a month ago
Use our unique_span's instead of the faux-dynarray class
Closed a month ago
Support for constructing const_region_t's from region_t&&
Closed a month ago
Drop cuda::dynarray in favor of cuda::unique_span
Closed a month ago
Simply access-permissions-related code
Closed a month ago
Hide "flags" from the interface of devices and contexts
Closed a month ago
Rework the range_attribute_t and advice code
Closed a month ago
Move memory::pool::ipc::ptr_handle_t out of ipc.hpp
Closed a month ago
Managed memory deleter uses wrong free() function leaks PC reference
Closed a month ago
Let a mapped pair of spans "decay" into a pair of regions
Closed a month ago
Move `rtc::marshalled_options_t` and `rtc::marshal()` into the `detail_` subnamespace
Closed a month ago
Drop the `native_word_t` type
Closed a month ago
Move program_base_t into the detail namespace
Closed a month ago
Add a device ID field to the texture_view object
Closed a month ago
Drop `examples/rtc_common.hpp`, use cuda::rtc::render to print compilation options
Closed a month ago
Use optional's instead of a special bespoke "unused" value in pci_location_t's
Closed a month ago
Drop context-specifying functions for allocation host memory
Closed a month ago
Correct signatures of context::current::peer_to_peer functions
Closed a month ago
link::marshalled_options_t and link::option_t should be in namespace detail_
Closed a month ago
Collect the log-related link options into a sub-structure of `link::options_t`
Closed a month ago
Drop specify_default_load_caching_mode from link::options_t
Closed a month ago
Use defaulted comparison operator for launch configurations
Closed a month ago
context::current::scoped_override_t should be declared in current_context.hpp
Closed a month ago
Improve naming of fallback_strategy_t and its values
Closed a month ago
Reduce code repetition between context_t and primary_context_t
Closed a month ago
Replace memory::external::mapped_region_t with unique_region, or rebase the former on the latter
Closed a month ago1
Two bugs with CUDA-12 related launch configuration fields
Closed a month ago
Split off a type_traits.hpp header from types.hpp
Closed a month ago
"Robustify" the RTC buffer getters (log, ptx, lto_ir)
Closed a month ago
Support copying from standard-library containers with contiguous storage
Closed a month ago
Stick to an underscore suffix for proxy class field name
Closed a month ago
Add an empty() method to span()
Closed a month ago
Make `cuda::memory::region_t` more constexpr-friendly
Closed a month ago
Make memory::region_t CUDA-independent
Closed a month ago
Add support for core dump enabling
Updated 2 months ago
Add support for newer context flags
Updated 2 months ago
Support new JIT compilation flags
Updated 2 months ago
Support Green Contexts
Updated 2 months ago
Use is_trivially_copyable more, and instead of is_trivially_copy_constructible
Updated 2 months ago
Rethink numerous copy functions vis-a-vis the copy parameters structure
Updated 2 months ago
Handle -pic default value change
Updated 2 months ago
Discontinue support for set_shared_memory_bank_size and related definitions with CUDA 12?
Updated 3 months ago
Drop copy_parameters_t in favor of a nicer copy parameters builder
Updated 3 months ago
We don't need a context flags type, but do need a primary context flags type
Closed 3 months ago
Offer a builder for CUDA arrays and/or array descriptors
Updated 3 months ago