eyalroz/

cuda-api-wrappers

Thin C++-flavored wrappers for the CUDA Runtime API

Owners:

$0.00

USD raised

Recent activities

eyalroz added the repository

over 2 years ago

Issues

Funded Unfunded Rewarded Closed

cuCtxGetSharedMemConfig is deprecated as of CUDA 12.3, acknowledge this

Unfunded#649created byeyalroz

$0.00

Support graph edge properties

Unfunded#648created byeyalroz

$0.00

Add support for core dump enabling

Unfunded#646created byeyalroz

$0.00

Switch to cuMemAdvise_v2 with CUDA v12 and later

Unfunded#644created byeyalroz

$0.00

Switch to cuMemAdvise_v2 with CUDA v12 and later

Unfunded#644created byeyalroz

$0.00

Use is_trivially_copyable more, and instead of is_trivially_copy_constructible

Unfunded#642created byeyalroz

$0.00

Rethink numerous copy functions vis-a-vis the copy parameters structure

Unfunded#641created byeyalroz

$0.00

Handle -pic default value change

Unfunded#636created byeyalroz

$0.00

Support new JIT compilation flags

Unfunded#635created byeyalroz

$0.00

Support conditional graph nodes

Unfunded#634created byeyalroz

$0.00

Support Green Contexts

Unfunded#633created byeyalroz

$0.00

Use optional's instead of a special bespoke "unused" value in pci_location_t's

Unfunded#626created byeyalroz

$0.00

Discontinue support for set_shared_memory_bank_size and related definitions with CUDA 12?

Unfunded#623created byeyalroz

$0.00

Accept execution affinity parameters when creating a context

Unfunded#616created byeyalroz

$0.00

Drop copy_parameters_t in favor of a nicer copy parameters builder

Unfunded#615created byeyalroz

$0.00

Add support for newer context flags

Unfunded#613created byeyalroz

$0.00

Offer a builder for CUDA arrays and/or array descriptors

Unfunded#611created byeyalroz

$0.00

An invalid file, `name_caching_program.hpp`, has snuck into our code

Unfunded#599created byeyalroz

$0.00

Avoid warnings about narrowing conversion and shadowing

Unfunded#590created byeyalroz

$0.00

Check overall dimensions are not too large to be expressed with our grid and block dimension limits

Unfunded#577created byeyalroz

$0.00

Structure 'attributes' and 'properties' better, using container-like facades and proxies

Unfunded#574created byeyalroz

$0.00

Streamline `cuda-api-wrappers-config.cmake` + no need to use preprocessing for it

Unfunded#562created byeyalroz

$0.00

Work around CMake restriction since 3.25 on setting properties of targets built outside of project

Unfunded#558created byeyalroz

$0.00

Add support for nvJitLink linking

Unfunded#538created byeyalroz

$0.00

Provide API hooking functionality like in libcuhook

Unfunded#533created byeyalroz

$0.00

Support empty nodes as typed_node's

Unfunded#532created byeyalroz

$0.00

Use CMake 3.27's fatbin generation mechanism

Unfunded#531created byeyalroz

$0.00

Getting an unknown CUDA error with cuMemPoolImportFromShareableHandle()

Unfunded#513created byeyalroz

$0.00

Error setting reserved virtual mem region permissions

Unfunded#512created byeyalroz

$0.00

Try to build an executable for each C++ version after 11 supported by the compiler

Unfunded#480created byeyalroz

$0.00

Support texture, surface and tensor "objects"

Unfunded#470created byeyalroz

$0.00

Reconsider accepting pointers without contexts

Unfunded#461created byeyalroz

$0.00

Support execution graph capture via stream_t's

Unfunded#458created byeyalroz

$0.00

Add support for user objects (in graphs etc.)

Unfunded#457created byeyalroz

$0.00

Support CUDA 12.0 graph API changes

Unfunded#456created byeyalroz

$0.00

Support scheduling actual batches of "batch memory operations" on a stream

Unfunded#454created byeyalroz

$0.00

Accept "execution affinity" info when creating a context

Unfunded#447created byeyalroz

$0.00

Support obtaining unique stream and context IDs

Unfunded#446created byeyalroz

$0.00

Support use of external semaphores

Unfunded#445created byeyalroz

$0.00

Should we expose pointer-to-const versions of CUDA driver handles?

Unfunded#444created byeyalroz

$0.00

Should we expose pointer-to-const versions of CUDA driver handles?

Unfunded#444created byeyalroz

$0.00

Should wrapper classes throw when failing on destruction?

Unfunded#440created byeyalroz

$0.00

Pass program name by value instead of const-ref

Unfunded#433created byeyalroz

$0.00

Pass program name by value instead of const-ref

Unfunded#433created byeyalroz

$0.00

overall_dimensions_t should be constructible from a dim3 struct

Unfunded#404created byeyalroz

$0.00

Introduce an event pair class?

Unfunded#402created byeyalroz

$0.00

Add the bandwidthtest CUDA sample as an example program

Unfunded#398created byeyalroz

$0.00

Some NVRTC compilation options are not (directly) supported

Unfunded#397created byeyalroz

$0.00

Some NVRTC compilation options are not (directly) supported

Unfunded#397created byeyalroz

$0.00

scoped_existence_ensurer_t fails if the driver has not been initialized

Unfunded#396created byeyalroz

$0.00

NVTX implicitly depends on libdl (at least on Linux)

Unfunded#393created byeyalroz

$0.00

NVTX implicitly depends on libdl (at least on Linux)

Unfunded#393created byeyalroz

$0.00

Regression: Fix for #304 broke CUDA 10.0 compatibility

Unfunded#391created byeyalroz

$0.00

Regression: Fix for #304 broke CUDA 10.0 compatibility

Unfunded#391created byeyalroz

$0.00

nvrtc.h included in wrong file

Unfunded#387created byeyalroz

$0.00

Make the library usable with FetchContent

Unfunded#382created byeyalroz

$0.00

Compilation log vector<char> contains trailing '\0'

Unfunded#378created byeyalroz

$0.00

When rendering compilation options to a string, we get an extra space

Unfunded#376created byeyalroz

$0.00

If the language dialect is "set" to an empty/null string - unset it

Unfunded#361created byeyalroz

$0.00

Allow enqueueing kernel launches using type-erased marshalled arguments

Unfunded#359created byeyalroz

$0.00

Allow enqueueing kernel launches using type-erased marshalled arguments

Unfunded#359created byeyalroz

$0.00

link_t missing a device_id_ member

Unfunded#358created byeyalroz

$0.00

Some std::string have creeped in

Unfunded#347created byeyalroz

$0.00

Some std::string have creeped in

Unfunded#347created byeyalroz

$0.00

marshalled_options.hpp errors with C++17

Unfunded#329created byeyalroz

$0.00

marshalled_options.hpp errors with C++17

Unfunded#329created byeyalroz

$0.00

Switch memory functions to only work with regions (and spans)

Unfunded#322created byeyalroz

$0.00

Switch memory functions to only work with regions (and spans)

Unfunded#322created byeyalroz

$0.00

std::trivially-copyable requirement for kernel arguments too strong

Unfunded#321created byeyalroz

$0.00

std::trivially-copyable requirement for kernel arguments too strong

Unfunded#321created byeyalroz

$0.00

Failure to set context when recording an event

Unfunded#314created byeyalroz

$0.00

Add NVIDIA's separate-compilation-linking example

Unfunded#312created byeyalroz

$0.00

Switch the NVTX wrapper to use driver-API-style calls

Unfunded#310created byeyalroz

$0.00

Add support for naming streams, events and devices in NVTX

Unfunded#309created byeyalroz

$0.00

Add support for naming streams, events and devices in NVTX

Unfunded#309created byeyalroz

$0.00

Add support for naming streams, events and devices in NVTX

Unfunded#309created byeyalroz

$0.00

Make the nvtx library header-only

Unfunded#307created byeyalroz

$0.00

Make the nvtx library header-only

Unfunded#307created byeyalroz

$0.00

Use CUDA_VERSION instead of CUDART_VERSION

Unfunded#303created byeyalroz

$0.00

Should we include support for GPUDirect/RDMA as in NVIDIA's gdrcopy?

Unfunded#294created byeyalroz

$0.00

Don't use cooperative groups with CUDA 8...

Unfunded#286created byeyalroz

$0.00

Remove more CUDA runtime API uses when using the driver API mostly

Unfunded#279created byeyalroz

$0.00

Should we replace memory::XXXXX::allocate(...) with memory::allocate(cuda::memory::type_t, ...) ?

Unfunded#278created byeyalroz

$0.00

Consider dropping device_t::create_event and device_t::create_stream

Unfunded#277created byeyalroz

$0.00

One status_t to rule them all

Unfunded#276created byeyalroz

$0.00

Support offsets in 2D, 3D copying

Unfunded#274created byeyalroz

$0.00

Support for NVRTC diag-suppress/error/warn

Unfunded#265created byeyalroz

$0.00

Support CUDA 11.5 block-compressed array types

Unfunded#261created byeyalroz

$0.00

Support CUDA 11.5 new arrays formats

Unfunded#260created byeyalroz

$0.00

Support CUDA execution graphs

Unfunded#175created byneoblizz

$0.00

Should we move `detail_::` code and/or implementations to separate files?

Unfunded#166created byeyalroz

$0.00

Integrate or supersede the functionality in nVIDIA's jitify

Unfunded#128created byeyalroz

$0.00

Recent activities

eyalroz added the repository

over 2 years ago

cuda-api-wrappers

$0.00

Issues

Company

Community

Support

Connect