tqchen commented on pull request #6394:
URL: https://github.com/apache/incubator-tvm/pull/6394#issuecomment-686664859
It is a bit confusing to do `runtime.enabled("llvm -mtriple=nonexistant")`
test.
Because `llvm -mtriple=nonexistant` is a target(whether we can generate code
for the target), but not necessarily mean that the particular runtime is
enabled. e.g. we can generate ARM code on an x86 machine, or opencl code in a
machine without opencl devices. But it does not mean that we can run the code
on the same env. We might want to think about target enabled for the later case.
In most of the testcases we are interested right now, we are asking about
whether a runtime is enabled(so we can run the generated code). I agree that we
might want to improve the logic further. Or modify the testing logic, to use
TVM_TESTING_RUNTIME and only include cpu/gpu/opencl/metal/ variations for now.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
