tqchen commented on a change in pull request #9727:
URL: https://github.com/apache/tvm/pull/9727#discussion_r818889855
##########
File path: src/tir/ir/expr.cc
##########
@@ -1056,12 +1058,28 @@ TVM_STATIC_IR_FUNCTOR(ReprPrinter, vtable)
.set_dispatch<AnyNode>([](const ObjectRef& node, ReprPrinter* p) {
p->stream << "?"; });
// BufferLoad
+void BufferLoadNode::LegalizeDtype() {
+ int index_lanes = 1;
Review comment:
I agree that could be an useful generalization.
The main rational of the suggestion is that we might want to start
small(that matches the original behavior) then generally analyze the effect and
generalize to cases(like at most one vector indices) after thinking about all
the consequences. This is mainly to avoid possibly bugs on unsupported cases
that we do not yet anticipated
Assuming right now we still restrict vectorize to happen after flattening,
the last dimension restriction could serve the current purose, then we can
discuss for more generlaization.
--
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]
