electriclilies commented on a change in pull request #8926:
URL: https://github.com/apache/tvm/pull/8926#discussion_r702193690
##########
File path: src/relay/backend/aot_executor_codegen.cc
##########
@@ -656,6 +651,20 @@ class AOTExecutorCodegen : public MixedModeVisitor {
auto storage_rewrite = tir::transform::StorageRewrite();
mod_run = storage_rewrite(mod_run);
+ // The workspace for main function should be calculated after performing
storage_rewrite for
+ // the top level TIR function.
+ auto workspace_byte_alignment =
+
target_host_->GetAttr<Integer>("workspace-byte-alignment").value_or(16);
+ Integer main_workspace_size = CalculateWorkspaceBytes(
+
Downcast<tir::PrimFunc>(mod_run->Lookup(::tvm::runtime::symbol::tvm_run_func_suffix)),
+ workspace_byte_alignment);
+
+ Optional<backend::FunctionInfo> main_func_info =
+ lowered_mod->GetAttr<backend::FunctionInfo>("main_func_info");
+ ICHECK(main_func_info) << "The attribute \"main_func_info\" should be set
at this point.";
+ main_func_info.value()->workspace_sizes.Set(target_host_,
main_workspace_size);
+ function_metadata_.Set(runtime::symbol::tvm_module_main,
main_func_info.value());
Review comment:
Right, so we are computing original guesses for the workspace size, then
updating them after the storage rewrite pass runs to fine tune it.
RE the StorageRewrite in CreatePassList -- I took a quick look and it
actually looks like that might only be called through the Python `tvm.lower`
and for the legacy TE schedules. So I'm not sure how much that StorageRewrite
is actually being applied to code
--
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]
