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 4d3cf08a2c [REFACTOR] Transition VisitAttrs to new reflection
mechanism (#18098)
add b6db2ec89f [Runtime] CutensorMap support (#18097)
add 9eb8b3004b Add support for bucketize (#18040)
add 910aeafb84 [RELAX] Fix rotary embedding buffer size calculation
(#18102)
No new revisions were added by this update.
Summary of changes:
include/tvm/ir/type.h | 29 +++
include/tvm/relax/attrs/search.h | 18 ++
include/tvm/script/ir_builder/tir/ir.h | 2 +
python/tvm/ir/type.py | 16 ++
python/tvm/relax/backend/dispatch_sort_scan.py | 12 ++
.../relax/frontend/nn/llm/position_embedding.py | 2 +-
.../frontend/torch/base_fx_graph_translator.py | 12 ++
.../frontend/torch/exported_program_translator.py | 1 +
python/tvm/relax/frontend/torch/fx_translator.py | 1 +
python/tvm/relax/op/__init__.py | 2 +-
python/tvm/relax/op/search.py | 25 +++
python/tvm/relax/transform/legalize_ops/search.py | 10 +
python/tvm/script/ir_builder/relax/ir.py | 2 +
python/tvm/script/ir_builder/tir/ir.py | 2 +
python/tvm/topi/gpu/sort.py | 89 ++++++++-
src/ir/type.cc | 12 ++
src/relax/op/tensor/search.cc | 53 +++++-
src/relax/op/tensor/search.h | 10 +
src/runtime/cuda/cuda_device_api.cc | 206 +++++++++++++++++++++
src/runtime/cuda/cuda_module.cc | 2 +-
src/runtime/file_utils.cc | 13 ++
src/runtime/meta_data.h | 3 +
src/runtime/pack_args.h | 26 ++-
src/script/ir_builder/tir/ir.cc | 1 +
src/script/printer/tir/expr.cc | 30 +--
src/target/build_common.h | 10 +
src/target/llvm/codegen_cpu.cc | 4 +
src/target/llvm/codegen_llvm.cc | 8 +-
src/target/llvm/codegen_llvm.h | 1 +
src/target/source/codegen_c.cc | 16 +-
src/target/source/codegen_cuda.cc | 3 +-
tests/python/codegen/test_target_codegen_cuda.py | 31 ++++
.../relax/test_frontend_from_exported_program.py | 25 +++
tests/python/relax/test_frontend_from_fx.py | 22 +++
.../test_tir_transform_inject_ptx_async_copy.py | 4 +-
35 files changed, 676 insertions(+), 27 deletions(-)
