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 939b8b9ce7 [Web] Seperate parallel shard download and iterative shard
loading (#16650)
add 9ec72494cf [TIR] Implement max/min_value for fp8 data types (#16723)
add 94866f769a [VM] [Hexagon] Add buffers to `dma_wait` builtin (#16706)
add 45df1247c6 [Web] Implement linear congruential generator, make runtime
seedable (#16722)
add feb104393c [TIR][CUDA] Add native FP8 support to codegen (#16548)
No new revisions were added by this update.
Summary of changes:
include/tvm/tir/transform.h | 6 +-
python/tvm/contrib/nvcc.py | 3 +
src/driver/driver_api.cc | 5 +-
src/runtime/relax_vm/hexagon/builtin.cc | 12 +-
src/target/llvm/codegen_llvm.cc | 2 +
src/target/source/codegen_cuda.cc | 113 ++-
src/target/source/literal/cuda_half_t.h | 42 ++
src/tir/op/op.cc | 12 +
src/tir/transforms/unsupported_dtype_legalize.cc | 28 +-
.../python/codegen/test_target_codegen_cuda_fp8.py | 803 +++++++++++++++++++++
.../contrib/test_hexagon/test_dma_builtin.py | 86 +--
web/src/index.ts | 2 +-
web/src/runtime.ts | 17 +-
web/src/support.ts | 76 ++
web/tests/node/test_random_generator.js | 71 ++
15 files changed, 1169 insertions(+), 109 deletions(-)
create mode 100644 tests/python/codegen/test_target_codegen_cuda_fp8.py
create mode 100644 web/tests/node/test_random_generator.js
