You are viewing a plain text version of this content. The canonical link for it is here.
Posted to discuss-archive@tvm.apache.org by "akaashrp (via GitHub)" <gi...@apache.org> on 2026/07/30 22:45:56 UTC
Re: [PR] [Web] Batch GPU-to-GPU copies, fix WebGPU synchronization, and add tests for command batching [tvm]
akaashrp commented on code in PR #20059:
URL: https://github.com/apache/tvm/pull/20059#discussion_r3686771878
##########
web/src/webgpu.ts:
##########
@@ -470,7 +469,7 @@ export class WebGPUContext {
this.device.queue.submit([this.pendingEncoder.finish()]);
this.pendingEncoder = null;
this.pendingDispatchCount = 0;
- // A compute submission is now the last queue operation, so the
+ // A command submission is now the last queue operation, so the
// GPU→CPU copy fast path in sync() is no longer valid.
this.pendingGPUToCPUCopy = null;
Review Comment:
Good catch. I've updated `flushCommands()` to preserve the pending readback and only invalidate the queue tail fast path. When later work exists, `sync()` now awaits both the readback and queue completion.
--
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: discuss-archive-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: discuss-archive-unsubscribe@tvm.apache.org
For additional commands, e-mail: discuss-archive-help@tvm.apache.org