 Minjie Wang
|
870da747ea
|
[CUDA][Kernel] More CUDA kernels; Standardize the behavior for sorted COO/CSR (#1704)
* add cub; array cumsum
* CSRSliceRows
* fix warning
* operator << for ndarray; CSRSliceRows
* add CSRIsSorted
* add csr_sort
* inplace coosort and outplace csrsort
* WIP: coo is sorted
* mv cuda_utils
* add AllTrue utility
* csr sort
* coo sort
* coo2csr for sorted coo arrays
* CSRToCOO from sorted
* pass tests for the new kernel changes
* cannot use inplace sort
* lint
* try fix msvc error
* Fix g.copy_to and g.asnumbits; ToBlock no longer uses CSC
* stash
* revert some hack
* revert some changes
* address comments
* fix
* fix to_block unittest
* add todo note
|
2020-06-28 18:37:28 +08:00 |
|