mbs-octoml commented on a change in pull request #9483:
URL: https://github.com/apache/tvm/pull/9483#discussion_r759751938
##########
File path: src/relay/backend/vm/compiler.cc
##########
@@ -481,50 +484,12 @@ class VMFunctionCompiler :
DeviceAwareExprFunctor<void(const Expr& n)> {
this->last_register_ = merge_register;
}
- void EmitShapeFunc(Function func, Array<Expr> inputs, Array<Expr> outputs) {
- // Lower shape function
- tec::CCacheKey key(func, host_se_scope_->target);
- auto cfunc = context_->compiler->LowerShapeFunc(key);
- int op_index = -1;
- // pick the only function inside the context
- ICHECK_EQ(cfunc->funcs->functions.size(), 1);
- auto pfunc =
Downcast<tir::PrimFunc>((*cfunc->funcs->functions.begin()).second);
- if (context_->seen_funcs.count(pfunc) == 0) {
- op_index = context_->cached_funcs.size();
- context_->cached_funcs.push_back(cfunc);
- context_->seen_funcs[pfunc] = op_index;
- } else {
- op_index = context_->seen_funcs[pfunc];
- }
-
- // Prepare input and output registers
+ void EmitInvokeTVMOp(const Expr& func, const Expr& inputs, const Expr&
outputs,
Review comment:
Because it was only needed to figure out which Target to use for the
function-at-a-time lowering, but that has all been done already by the
LowerTEPass.
--
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]
