GGitHubAdds peer memcpy (#520)
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
PushKernelArg for scalar now takes reference | 1 年前 | |
PushKernelArg for scalar now takes reference | 1 年前 | |
Add CUPTI bindings (#458) * Initial setup for cupti activity api * Add all other activity-namespaced APIs * Add cupti vars * Split out result module and add all other activity apis * Clippy * Fix example * Add cupti to keywords, doc features, remove from default features * Add cuda 13 support * Add missing cuda 13 functions * Prevent generating duplicate driver type definitions * Disable activity in example cleanup * Add remaining callback APIs * Add types from CUDA meta headers * Add types from driver meta headers * Add missing obsolete types used in meta headers * Add CUPTI module to CI checks * Add CUPTI module to CI clippy * Block types with missing declarations in cuda 11040 meta headers * Update Cargo.toml --------- Co-authored-by: Corey Lowman <clowman1993@gmail.com> | 10 个月前 | |
[Breaking] get_global returns CudaViewMut (#517) * #514 get_global returns CudaViewMut * Fix no-std example | 7 个月前 | |
Add safe API for querying CUDA function attributes (#479) * Add safe API for querying CUDA function attributes * get_attribute now binds to context before taking action Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Formatting --------- Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> | 9 个月前 | |
Fixing example compilatin in no-std | 9 个月前 | |
Adds peer memcpy (#520) * fixes a few issues with multi gpu usage in both candle and mistralrs * remove get ctx and add tests * use println and add tst for ctx-thread assocation * Update src/driver/safe/core.rs Co-authored-by: Chelsea Lowman <clowman1993@gmail.com> * Update src/driver/safe/core.rs * moved peer copy into memcpy_dtod * added peer-to-peer test for same device. ensures that cudaslice cloning still works with single card. --------- Co-authored-by: Chelsea Lowman <clowman1993@gmail.com> | 7 个月前 | |
Add safe API for CUDA constant memory operations (#478) * Add safe API for CUDA constant memory operations * Don't panic on invalid name Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Moved get_global to return CudaSlice<u8> --------- Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Deprecate memcpy_stod,memcpy_dtov -> clone_htod,clone_dtoh (#493) | 9 个月前 | |
Fixing example imports | 3 年前 | |
Initial commit of src | 3 年前 | |
Initial commit of src | 3 年前 |