Skip to content

Actions: NVIDIA/cuDecomp

Actions

Build

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
build.yml
127 workflow runs
127 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Add automatic workspace management feature. Build #128: Issue comment #159 (comment) created by romerojosh
9m 12s
9m 12s
Add automatic workspace management feature. Build #127: Issue comment #159 (comment) created by romerojosh
3m 31s
3m 31s
Fix potential deadlock due to mismatched NCCL local communicator handling. Build #126: Issue comment #158 (comment) created by romerojosh
9m 53s
9m 53s
Remove redundant NCCL AlltoAll runtime check Build #125: Issue comment #157 (comment) created by romerojosh
6m 10s
6m 10s
Harden public struct ABI for future extensions. Build #124: Issue comment #155 (comment) created by romerojosh
10m 29s
10m 29s
Harden public struct ABI for future extensions. Build #123: Issue comment #155 (comment) created by romerojosh
3m 4s
3m 4s
Add autotuning candidate filtering controls. Add MPI backend disablement option to autotune options struct. Build #122: Issue comment #154 (comment) created by romerojosh
6m 22s
6m 22s
Fix erroneous MPI_Bcast sizing in autotune.cc and remove O1 optimization cap. Build #121: Issue comment #153 (comment) created by romerojosh
9m 33s
9m 33s
Improve NCCL UBR support. Extend cuMem/VMM allocation feature to platforms without fabric allocation support. Build #120: Issue comment #152 (comment) created by romerojosh
7m 2s
7m 2s
Address iso_c_binding warning in Fortran Poisson example. Build #119: Issue comment #151 (comment) created by romerojosh
5m 56s
5m 56s
Add some missing CUDA launch checks to benchmark and examples. Build #118: Issue comment #150 (comment) created by romerojosh
6m 6s
6m 6s
Fix halo workspace API docs and benchmark sample link Build #117: Issue comment #148 (comment) created by romerojosh
1s
1s
Avoid NVSHMEM device header parsing in non-RDC CUDA translation units. Build #116: Issue comment #149 (comment) created by romerojosh
5m 53s
5m 53s
Fix pipelined self-copy ranks Build #115: Issue comment #147 (comment) created by romerojosh
10m 10s
10m 10s
Validate pencil halo and padding extents Build #114: Issue comment #145 (comment) created by romerojosh
5m 53s
5m 53s
Fix pipelined transpose empty rank skips Build #113: Issue comment #144 (comment) created by romerojosh
10m 3s
10m 3s
Fix direct transpose output offset Build #112: Issue comment #146 (comment) created by romerojosh
6m 47s
6m 47s
Allow null halo periods for non-periodic halos. Build #111: Issue comment #143 (comment) created by romerojosh
6m 27s
6m 27s
Reject negative process grid dimensions Build #110: Issue comment #141 (comment) created by romerojosh
5m 58s
5m 58s
Use matching MPI datatype for size_t reduction Build #109: Issue comment #142 (comment) created by romerojosh
1s
1s
Use matching MPI datatype for size_t reduction Build #108: Issue comment #142 (comment) created by romerojosh
8m 26s
8m 26s
Support multiple cuDecomp handles. Build #107: Issue comment #139 (comment) created by romerojosh
6m 36s
6m 36s
Use native NCCL Alltoall API when able. Build #106: Issue comment #136 (comment) created by romerojosh
5m 23s
5m 23s
Add CTest testing framework. Build #105: Issue comment #135 (comment) created by romerojosh
6m 3s
6m 3s
Fix benign typo in cudecompUpdateHaloZ Fortran interface. Build #104: Issue comment #134 (comment) created by romerojosh
7m 0s
7m 0s