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 8f658cc3c6 [FFI][REFACTOR] Refactor python ffi call mechanism for perf
(#18302)
add 85dc1d7a02 Clear ext_lib_dll_names for macOS platform (#18304)
add 4404334f84 [Relax] Fix RelaxToPyFuncConverter compatibility and
improve fallback handling (#18301)
add 71635d03b6 [FFI][ABI][REFACTOR] Enhance DLPack Exchange Speed and
Behavior (#18306)
add 0c9e7cda7d [FFI] Update `load_inline` interface (#18307)
add 00ae64744e [FFI][ABI] Refactor the naming of DLPack speed converter
(#18308)
No new revisions were added by this update.
Summary of changes:
ffi/CMakeLists.txt | 3 +-
ffi/docs/get_started/quick_start.md | 4 +-
ffi/examples/inline_module/main.py | 2 +-
ffi/examples/quick_start/run_example.py | 2 +-
ffi/examples/quick_start/src/add_one_cuda.cu | 4 +-
ffi/include/tvm/ffi/c_api.h | 15 +
ffi/include/tvm/ffi/container/tensor.h | 56 ++-
ffi/include/tvm/ffi/extra/c_env_api.h | 31 +-
ffi/licenses/LICENSE.pytorch.txt | 84 ++++
ffi/licenses/NOTICE.pytorch.txt | 456 +++++++++++++++++++++
ffi/pyproject.toml | 2 +-
ffi/python/tvm_ffi/__init__.py | 2 +
ffi/python/tvm_ffi/_optional_torch_c_dlpack.py | 405 ++++++++++++++++++
ffi/python/tvm_ffi/cpp/load_inline.py | 20 +-
ffi/python/tvm_ffi/cython/base.pxi | 32 +-
ffi/python/tvm_ffi/cython/function.pxi | 91 ++--
ffi/python/tvm_ffi/cython/tensor.pxi | 70 ++--
ffi/python/tvm_ffi/cython/tvm_ffi_python_helpers.h | 94 ++++-
ffi/python/tvm_ffi/libinfo.py | 23 ++
ffi/scripts/benchmark_dlpack.py | 5 +-
ffi/src/ffi/extra/env_context.cc | 120 ++++++
ffi/src/ffi/extra/stream_context.cc | 81 ----
ffi/tests/cpp/test_tensor.cc | 45 ++
ffi/tests/python/test_load_inline.py | 67 ++-
ffi/tests/python/test_tensor.py | 22 +-
python/tvm/contrib/cutlass/attention_operation.py | 8 +-
python/tvm/contrib/cutlass/conv2d_operation.py | 2 +-
python/tvm/contrib/cutlass/gemm_operation.py | 4 +-
python/tvm/contrib/cutlass/layer_norm_operation.py | 2 +-
python/tvm/contrib/cutlass/rms_norm_operation.py | 2 +-
python/tvm/libinfo.py | 5 +-
python/tvm/relax/base_py_module.py | 23 +-
python/tvm/relax/relax_to_pyfunc_converter.py | 194 +++++++--
src/contrib/msc/plugin/tvm_codegen.cc | 2 +-
src/runtime/contrib/cublas/cublas.cc | 2 +-
src/runtime/contrib/cublas/cublas_json_runtime.cc | 2 +-
src/runtime/contrib/cublas/cublas_utils.cc | 4 +-
src/runtime/contrib/cudnn/cudnn_json_runtime.cc | 3 +-
src/runtime/contrib/cudnn/cudnn_utils.cc | 4 +-
src/runtime/contrib/cutlass/fp16_group_gemm.cuh | 2 +-
src/runtime/contrib/cutlass/fp8_gemm.cu | 3 +-
src/runtime/contrib/cutlass/fp8_group_gemm_sm90.cu | 3 +-
.../contrib/cutlass/fp8_groupwise_scaled_gemm.cuh | 4 +-
.../fp8_groupwise_scaled_group_gemm_sm100.cu | 3 +-
.../contrib/hipblas/hipblas_json_runtime.cc | 2 +-
src/runtime/contrib/hipblas/hipblas_utils.cc | 3 +-
src/runtime/contrib/miopen/miopen_utils.cc | 3 +-
src/runtime/contrib/msc/tensorrt_runtime.cc | 2 +-
src/runtime/contrib/thrust/thrust.cu | 2 +-
src/runtime/cuda/cuda_device_api.cc | 6 +-
src/runtime/cuda/cuda_module.cc | 2 +-
src/runtime/cuda/l2_cache_flush.cc | 2 +-
src/runtime/device_api.cc | 5 +-
src/runtime/rocm/rocm_device_api.cc | 4 +-
src/runtime/rocm/rocm_module.cc | 2 +-
src/runtime/vm/cuda/cuda_graph_builtin.cc | 11 +-
.../python/relax/test_relax_to_pyfunc_converter.py | 178 +++++++-
57 files changed, 1914 insertions(+), 316 deletions(-)
create mode 100644 ffi/licenses/LICENSE.pytorch.txt
create mode 100644 ffi/licenses/NOTICE.pytorch.txt
create mode 100644 ffi/python/tvm_ffi/_optional_torch_c_dlpack.py
create mode 100644 ffi/src/ffi/extra/env_context.cc
delete mode 100644 ffi/src/ffi/extra/stream_context.cc
