grt: add updateNet for topology refresh in CUGR incremental routing by sparsh-karna · Pull Request #9671 · The-OpenROAD-Project/OpenROAD

sparsh-karna · 2026-03-06T06:45:26Z

Summary

Extends the incremental CUGR routing API (introduced in #9645) with
topology refresh support. When buffer insertion changes a net's pin
list or adds new nets, the existing GRNet objects are stale. This PR
re-reads net topology from ODB before rerouting so the incremental
path operates on correct data.

Depends On

#9645 — must be reviewed/merged first.

Changes

Design::updateNet(dbNet*) — re-reads pins, BTerms, ITerms, and
layer range from ODB in-place; handles brand-new nets created by
buffer insertion (net splits)
CUGR::updateNet(dbNet*) — rips up old routing tree, rebuilds
CUGRNet/GRNet from scratch, registers new nets in gr_nets_,
net_indices_, and db_net_map_
Null guard in GridGraph::commitTree() to prevent crash on freshly
allocated nets with no existing routing tree
Wired into endIncremental() so topology refresh happens before
rerouteNets() runs
grt::update_net Tcl command exposed via GlobalRouter.i SWIG binding
Regression test: src/grt/test/incremental_update_net.tcl

Testing

Regression test splits a net, calls update_net, runs end_incremental,
asserts guides exist for the new net
All existing grt regression tests unaffected

Add addDirtyNet(), startIncremental(), and endIncremental() methods to support incremental global routing in the CUGR engine, along with dirty_net_indices_ and incremental_mode_ state tracking. Signed-off-by: TheUnnamedOne-design <aditya07.11.04.12@gmail.com> Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

Signed-off-by: TheUnnamedOne-design <aditya07.11.04.12@gmail.com> Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

…rtion When repair_timing inserts a buffer, a net is split into two. The previous incremental flow rerouted the stale GRNet objects from init() — pin list, access points, and bounding box all predated the insertion. Implement Design::updateNet() and CUGR::updateNet() to refresh per-net CUGRNet/GRNet state from the database before rerouting. Wire into endIncremental() and expose as grt::update_net Tcl command. Also add null guard in GridGraph::commitTree() for freshly rebuilt nets and a regression test exercising the full incremental update path. Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

gemini-code-assist

Code Review

This pull request introduces support for topology refresh in the incremental CUGR router, which is crucial for handling netlist changes from buffer insertion. The changes include new APIs to update net information from the database, integration into the incremental routing flow, and a new regression test. My review focuses on potential performance improvements and code maintainability. I've identified a few areas with redundant operations and code duplication that could be refactored for better efficiency and clarity.

_{Note: Security Review did not run due to the size of the PR.}

src/grt/src/cugr/src/CUGR.cpp

src/grt/src/cugr/src/Design.cpp

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

github-actions

clang-tidy made some suggestions

github-actions · 2026-03-06T12:48:04Z

src/grt/src/cugr/src/CUGR.cpp

+  auto it = db_net_map_.find(db_net);
+  if (it != db_net_map_.end()) {
+    GRNet* gr_net = it->second;
+    grid_graph_->commitTree(gr_net->getRoutingTree(), /*ripup=*/true);


warning: argument name 'ripup' in comment does not match parameter name 'rip_up' [bugprone-argument-comment]

Suggested change

grid_graph_->commitTree(gr_net->getRoutingTree(), /*ripup=*/true);

grid_graph_->commitTree(gr_net->getRoutingTree(), /*rip_up=*/true);

Additional context

src/grt/src/cugr/src/GridGraph.h:117: 'rip_up' declared here

void commitTree(const std::shared_ptr<GRTreeNode>& tree, bool rip_up = false); ^

github-actions · 2026-03-06T12:48:04Z

src/grt/src/cugr/src/Design.cpp

+
+  auto it = db_net_to_design_idx_.find(db_net);
+  if (it != db_net_to_design_idx_.end()) {
+    nets_[it->second].setPins(std::move(pins));


warning: no header providing "std::move" is directly included [misc-include-cleaner]

src/grt/src/cugr/src/Design.cpp:5:

- #include <vector> + #include <utility> + #include <vector>

github-actions · 2026-03-06T12:48:04Z

src/grt/src/cugr/src/Netlist.h

  int getNumPins() const { return pins_.size(); }
  std::string getName() const { return db_net_->getName(); }
  LayerRange getLayerRange() const { return layer_range_; }
+  void setPins(std::vector<CUGRPin> pins) { pins_ = std::move(pins); }


warning: no header providing "std::move" is directly included [misc-include-cleaner]

src/grt/src/cugr/src/Netlist.h:6:

- #include <vector> + #include <utility> + #include <vector>

… detours Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

github-actions · 2026-03-06T19:21:01Z

clang-tidy review says "All clean, LGTM! 👍"

eder-matheus · 2026-03-09T13:30:01Z

src/grt/src/cugr/src/Design.cpp

  }
 }

+std::vector<CUGRPin> Design::readNetPins(odb::dbNet* db_net)


Good code refactoring. I'd suggest renaming the function to makeNetPins.

eder-matheus · 2026-03-09T13:31:41Z

src/grt/src/cugr/src/Design.cpp

+  if (it != db_net_to_design_idx_.end()) {
+    nets_[it->second].setPins(std::move(pins));
+    nets_[it->second].setLayerRange(layer_range);
+    return;


Instead of return, you can use an else here. It makes the code clearer and easier to follow.

eder-matheus · 2026-03-09T13:34:12Z

src/grt/src/cugr/src/GridGraph.cpp

+  if (!tree) {
+    return;
+  }


It's odd to call this function for a tree that doesn't exists. I prefer having this check before the function call, only in the context where it makes sense.

eder-matheus · 2026-03-09T13:34:40Z

src/grt/src/cugr/src/GRTree.cpp

+  if (!node) {
+    return;
+  }


Why is it needed?

eder-matheus · 2026-03-09T13:36:13Z

src/grt/src/GlobalRouter.cpp

+void GlobalRouter::updateNet(odb::dbNet* net)
+{
+  if (use_cugr_) {
+    cugr_->updateNet(net);
+  }
+}


This function seems CUGR specific. Make it clear in function name, with something like updateCUGRNet.

eder-matheus · 2026-03-09T13:44:48Z

src/grt/src/GlobalRouter.i

+void
+update_net(odb::dbNet* net)
+{
+  getGlobalRouter()->updateNet(net);
+}


Add a comment expliciting that this function is only used for debug. I don't think we'll need a Tcl interface for the incremental CUGR project.

Also, rename it to be CUGR specific, similar to my other comment.

eder-matheus · 2026-03-09T13:48:37Z

src/grt/src/GlobalRouter.cpp

+  if (use_cugr_) {
+    cugr_->endIncremental();
+    routes_ = cugr_->getRoutes();
+    updatePinAccessPoints();


This is consistent with how end incremental works for FastRouteCore, but you'll also need to look at how the IncrementalGRoute works. I would suggest to leave these changes in startIncremental and endIncremental out for now, and work on it in a separate PR.

sparsh-karna and others added 7 commits March 5, 2026 23:03

grt: refactor incremental CUGR reroute logic into helper method

11a684a

Signed-off-by: TheUnnamedOne-design <aditya07.11.04.12@gmail.com> Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

grt: use std::set_difference for secondary net filtering

7b57166

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

grt: use std::set_difference for secondary net filtering

a568365

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

grt: use std::ranges::set_difference per clang-tidy

f55110d

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

grt: sort dirty nets by slack before rip-up in incremental CUGR

d131a35

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

gemini-code-assist bot reviewed Mar 6, 2026

View reviewed changes

src/grt/src/cugr/src/CUGR.cpp Outdated Show resolved Hide resolved

src/grt/src/cugr/src/CUGR.cpp Show resolved Hide resolved

src/grt/src/cugr/src/Design.cpp Outdated Show resolved Hide resolved

src/grt/src/cugr/src/Design.cpp Outdated Show resolved Hide resolved

grt: fix clang-format

ce4f149

Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

github-actions bot reviewed Mar 6, 2026

View reviewed changes

grt: extract congestion view after rip-up in rerouteNets for accurate…

833cbf9

… detours Signed-off-by: Sparsh Karna <sparsh2005karna@gmail.com>

eder-matheus self-requested a review March 8, 2026 15:16

eder-matheus requested changes Mar 9, 2026

View reviewed changes

	grid_graph_->commitTree(gr_net->getRoutingTree(), /ripup=/true);
	grid_graph_->commitTree(gr_net->getRoutingTree(), /rip_up=/true);

Conversation

sparsh-karna commented Mar 6, 2026

Summary

Depends On

Changes

Testing

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Mar 6, 2026

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

eder-matheus Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants