Update topi/cuda schedules to use target.max_num_threads #577
Conversation
python/tvm/target.py
Outdated
|
|
||
|
|
||
| def get_max_num_threads(): | ||
| """Returns the maximum number of threads under current target. |
There was a problem hiding this comment.
Can you remove this function and simply use tvm.target.current_target().max_num_threads?
There was a problem hiding this comment.
Can tvm.target.current_target() be None?
There was a problem hiding this comment.
I did this following your usage of max_num_threads here
There was a problem hiding this comment.
use tvm.target.current_target(allow_missing=False)
There was a problem hiding this comment.
I just think we should remove this function and inline its body to the callers
There was a problem hiding this comment.
since the function itself is small(one line)
There was a problem hiding this comment.
ok
tvm.target.current_target(allow_none=False).max_num_threads
right?
python/tvm/target.py
Outdated
|
|
||
|
|
||
| def get_max_num_threads(): | ||
| """Returns the maximum number of threads under current target. |
There was a problem hiding this comment.
since the function itself is small(one line)
Update topi/cuda schedules to be consistent with #556.
With this commit, all unit tests in topi should pass out of the box with ROCm and OpenCL backend.
For OpenCL backend, I only checked it with AMD device (R9 Nano).