lergre.blogg.se - New tvmc download

RFC: clarifying buffer declaration and access (#63) ( de4fe97).

RFC-BYOC-DNNL-Integration (#73) ( 7aed0ca).

DietCode: An Auto-Scheduler for Dynamic Tensor Programs (#72) ( a518000).

UMA Universal Modular Accelerator Interface (#60) ( 6990e13).

Model Library Format with Multiple Modules (#76) ( f47c6ad).

Introducing DeclBuffer (#70) ( 87ff1fa).

TUNIP: TVMScript Unified Printer (#74) ( 48d47c5).

These RFCs have been merged in apache/tvm-rfcs since the last release.

And many more! See the list of RFCs and PRs included in v0.9.0 for a complete list, as well as the full change list.

transform.LowerCrossThreadReduction - Enables support for reductions across threads on GPUs.

transform.InjectPTXAsyncCopy - Rewrites global to shared memory copies in CUDA with async copy when annotated tir::attr::async_scope.

transform.CommonSubexprElimTIR - Implements common-subexpression elimination for TIR.

transform.InjectSoftwarePipeline - Transforms annotated loop nest into a pipeline prologue, body and epilogue where producers and consumers are overlapped.

t_axis_separators - Sets axis separators in a buffer to lower to multi-dimensional memory (e.g.ansform_block_layout - Applies a schedule transformation to a block as specified by an IndexMap.ansform_layout - Applies a layout transformation to a buffer as specified by an IndexMap.Several TIR schedule primitives and transforms including (abridged):.MLF - Support for multiple modules in a single MLF artifact.

Testing via Hexagon SDK simulator and on device via Snapdragon-based HDK boards and phones.

AOT - Support for the C++ runtime (with llvm and c targets only) and support for host-driven AOT in the C runtime.

New parameters - runtime=, executor=,.

ARM cascading scheduler for Arm Ethos(TM)-U NPUs.

v0.9.0 is the first release on the new quarterly release schedule and includes many highlights, such as: The TVM community has worked since the v0.8 release to deliver many exciting features and improvements.