[GitHub] [tvm] yzh119 commented on a diff in pull request #15333: [Unity][Dlight] Rule matmul avoiding blockIdx.z

["yzh119 (via GitHub)" <gi...@apache.org>]

yzh119 commented on code in PR #15333:
URL: https://github.com/apache/tvm/pull/15333#discussion_r1264760827


##########
python/tvm/dlight/gpu/matmul.py:
##########
@@ -328,7 +328,7 @@ def is_spatial(block: BlockRV) -> bool:
         by, vy, ty, yi = sch.split(y, [None, vthread_y, block_size_y, micro_size_y])
         ko, ki = sch.split(k, factors=[None, micro_size_k])
         sch.reorder(bx, by, vy, vx, ty, tx, ko, ki, yi, xi)
-        sch.bind(batch, "blockIdx.z")
+        bx = sch.fuse(batch, bx)

Review Comment:
   I suppose fusing `batch` with `by` will more faithfully inherit the semantics of `sch.bind(batch, "blockIdx.z")` because the physical id is `blockIdx.z * (gridDim.x * gridDim.y) + blockIdx.y * (gridDim.x) + blockIdx.x`.



-- 
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: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org