Lunderberg commented on PR #16361:
URL: https://github.com/apache/tvm/pull/16361#issuecomment-1880158323

   Oh, interesting.  I initially added the check due to encountering [this 
error](https://github.com/apache/tvm/blob/main/src/tir/transforms/lower_warp_memory.cc#L209)
 in `tir.transform.LowerWarpMemory`, concluding that the requirement for a 
single TIR variable per thread index was a TIR limitation and wanting to raise 
the error as early as possible.
   
   Can we add a test case that includes multiple `T.env_thread` assignments 
within the same kernel, verifying that it can be lowered/built without error?  
I'm thinking something like the following:
   
   ```python
   @T.prim_func
   def func(A: T.Buffer(16, "float32"), B: T.Buffer(16, "float32")):
       tx_A = T.launch_thread("threadIdx.x", 16)
       tx_B = T.launch_thread("threadIdx.x", 16)
   
       B[tx_B] = A[tx_A]
   
   
   tvm.build(func, target="cuda")
   ```


-- 
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