This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a change to branch nightly
in repository https://gitbox.apache.org/repos/asf/tvm.git
from d9ee6377cd [Relax][PyTorch] Support neural network ops for
ExportedProgram importer (#17426)
add e80801030e [DLIGHT][GPU] Improve matmul schedule for adreno (#17430)
add 7569148c3c [Relax] Introduce static shape tuning pipeline (#17428)
add 4f94890112 [NVSHMEM] Enable nvshmem memory allocation (#17415)
No new revisions were added by this update.
Summary of changes:
docs/how_to/tutorials/e2e_opt_model.py | 16 +-
python/tvm/dlight/gpu/matmul.py | 108 +++++++-----
python/tvm/relax/pipeline.py | 39 ++++
python/tvm/relax/transform/transform.py | 5 +-
.../contrib/nvshmem/{nvshmem.cc => init.cc} | 2 +
src/runtime/contrib/nvshmem/memory_allocator.cc | 104 +++++++++++
tests/python/disco/test_nvshmem.py | 45 ++++-
tests/python/dlight/test_gpu_matmul.py | 196 ++++++++++++---------
8 files changed, 365 insertions(+), 150 deletions(-)
rename src/runtime/contrib/nvshmem/{nvshmem.cc => init.cc} (96%)
create mode 100644 src/runtime/contrib/nvshmem/memory_allocator.cc
