-
Notifications
You must be signed in to change notification settings - Fork 3.8k
Closed
Description
Attempting to build on Windows fails with config.cmake options [set(USE_CUDA ON), set(USE_LLVM ON)]. The build process followed instructions @ https://docs.tvm.ai/install/from_source.html
By default the micro runtime is built, however this fails due to incompatibility with Windows.
For example, see unix-specific library export macro definitions:
- "TVM_MICRO_RUNTIME_API_API" - include/tvm/runtime/micro/standalone/utvm_runtime.h
- "TVM_MICRO_RUNTIME_API_BACKEND_API" - src/runtime/micro/standalone/utvm_graph_runtime.h
(Note, further incompatibilities exist after resolving these definitions, however I did not debug further)
I was able to successfully build by disabling compilation of the micro runtime [config.cmake option set(USE_MICRO_STANDALONE_RUNTIME ON->OFF)]. While this resolved the problem for me, I am unsure if this runtime is a general requirement.
Metadata
Metadata
Assignees
Labels
No labels