dsbarinov1 opened a new pull request, #13849:
URL: https://github.com/apache/tvm/pull/13849

   **Motivation:** Assertion failed during tuning
   
   **Error message from _thread_pool.cc:295_:**
   `Check failed: num_task <= num_workers_used_ (8 vs. 1) : Request parallel 
sync task larger than number of threads used  workers=1 request=8`
   
   **Main problem description:**
   Tuning of the _ARM Snapdragon 888 CPU_ architecture ends with an error above.
   
   **Suspected reason:**
   Incorrect (manual) determination of the number of threads. The number of 
threads is determined using _MaxConcurrency_ and returns 8 threads for this 
architecture, but the number of actually used threads is 4. This fix urges to 
use automatic determination of the number of threads by passing 'zero' as 
'num_threads' attribute in 'TVMBackendParallelLaunch' to avoid the 
abovementioned discrepancy.
   
   **Appropriate preventive fix:**
   A problem with incorrect rounding of integers (in context, num_threads) due 
to the default rounding down (div_down) has also been prevented. We now use up 
rounding (div_up) to make sure that the whole tensor is filled in parallel 
(without missing the end of the tensor due to down rounding).
   


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to