mpl: blockages as fixed macros #8438

openroad-ci · 2025-09-25T16:53:09Z

This PR changes how blockages are treated inside MPL simulated annealing, going from special objects to fixed macros. This changes the behavior of blockages from a soft constraint to a hard constraint, but simplifies the use of blockages in general, since they are objects within the sequence pair of the annealer.

The change from soft to hard constraint required changes in the creation of pin access blockages too, since it was causing troubles in design such as mock-array due to lack of space to fit everything.

Signed-off-by: Arthur Koucher <[email protected]>

Signed-off-by: João Mai <[email protected]>

github-actions · 2025-09-25T16:59:57Z

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

AcKoucher · 2025-09-26T14:10:37Z

src/mpl/src/graphics.cpp

+  return macro.getCluster() != nullptr
+         && macro.getCluster()->isClusterOfUnplacedIOPins();


This way we're not skipping fixed terminals.

I think we should use the is_blockage_ flag and have something like:

if (macro.isBlockage()) { return false; } Cluster* cluster = macro.getCluster(); return !cluster || cluster->isClusterOfUnplacedIOPins();

AcKoucher · 2025-09-26T14:11:05Z

src/mpl/src/graphics.cpp

 void Graphics::setSoftMacroBrush(gui::Painter& painter,
                                 const SoftMacro& soft_macro)
 {
+  if (soft_macro.getCluster() == nullptr) {


soft_macro.isBlockage()

AcKoucher · 2025-09-26T14:13:53Z

src/mpl/src/hier_rtlmp.cpp

-    if (macro_array->getNumMacro() % divider == 0) {
-      columns = macro_array->getNumMacro() / divider;
-      rows = divider;
+  const auto outline = tree_->root->getBBox();


Avoid using auto.

AcKoucher · 2025-09-26T14:14:00Z

src/mpl/src/hier_rtlmp.cpp

+  const float proportional_min_width = min_depth_proportion * die.getWidth();
+  const float proportional_min_height = min_depth_proportion * die.getHeight();
+
+  const auto tiling = tree_->root->getTilings().front();


Avoid using auto.

AcKoucher · 2025-09-26T14:22:54Z

src/mpl/src/hier_rtlmp.cpp

+// We model blockages as macro clusters (SoftMacros) constrained to fences.
+// The rationale of this model comes from the fact that the area occupied
+// by blockages should be empty, i.e., with neither macros or std cells.
+void HierRTLMP::blockagesAsFixedSoftMacros(
+    const std::vector<Rect>& blockages,
+    std::map<std::string, int>& soft_macro_id_map,
+    std::vector<SoftMacro>& macros)
+{
+  for (const Rect& blockage : blockages) {
+    const int macro_id = static_cast<int>(macros.size());
+    std::string macro_name = fmt::format("blockage_{}", macro_id);
+    soft_macro_id_map[macro_name] = macro_id;
+
+    macros.emplace_back(blockage, macro_name);
+  }
+}
+


The comment is not accurate. We should remove it as the approach is reasonably simpler now.

Blockages don't need to be part of soft_macro_id_map.

I think that this function would be clearer if named as createSoftMacrosForBlockages()

src/mpl/src/object.cpp

AcKoucher · 2025-09-26T14:23:42Z

src/mpl/src/object.h

  // Interfaces with hard macro
  Cluster* cluster_ = nullptr;
  bool fixed_ = false;  // if the macro is fixed
+  bool is_blockage_ = false;


This is useful for debugging. See other comments.

src/mpl/src/hier_rtlmp.cpp

AcKoucher · 2025-09-26T14:29:43Z

Subsequent work:

Treat placement blockages from odb as soft macros.
Remove support for macro blockages overlap penalty when these changes are stable.

Signed-off-by: João Mai <[email protected]>

github-actions · 2025-09-27T19:12:48Z

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

Signed-off-by: João Mai <[email protected]>

github-actions · 2025-09-29T18:14:15Z

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

Signed-off-by: João Mai <[email protected]>

github-actions · 2025-10-01T16:05:03Z

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

joaomai · 2025-10-01T18:08:17Z

Requires PR #3516 from ORFS.
CIs are green except GCDs from secure CI, failing with the same errors as master.

AcKoucher · 2025-10-01T19:57:58Z

src/mpl/src/hier_rtlmp.cpp

+// We model blockages as macro clusters (SoftMacros) constrained to fences.
+// The rationale of this model comes from the fact that the area occupied
+// by blockages should be empty, i.e., with neither macros or std cells.


It looks like this comment is a leftover from the previous attempted approach. It is not accurate and should be removed.

Please, add the rationale to the PR description.

github-actions · 2025-10-02T01:35:46Z

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

Signed-off-by: João Mai <[email protected]>

AcKoucher · 2025-10-09T12:36:08Z

I'm not sure I understand the problem with clang-tidy.

maliberty · 2025-10-09T20:04:02Z

Holding pending other metric updates in the queue. Will pair with The-OpenROAD-Project/OpenROAD-flow-scripts#3516

github-actions · 2025-10-15T18:29:28Z

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

joaomai · 2025-10-15T22:54:33Z

Should be paired with PR #3583 from ORFS.

AcKoucher and others added 7 commits September 25, 2025 14:52

mpl: model blockages as macro clusters constrained by fences

b50a9e0

Signed-off-by: Arthur Koucher <[email protected]>

mpl: blockages as fixed soft macros

49961c0

Signed-off-by: João Mai <[email protected]>

mpl: revert fence weight

9ddbaaa

Signed-off-by: João Mai <[email protected]>

mpl: generate only valid tilings for interconnected arrays

b215fe7

Signed-off-by: João Mai <[email protected]>

mpl: compute min depth based on smallest valid tiling

afe7aa4

Signed-off-by: João Mai <[email protected]>

clang-format

719e326

Signed-off-by: João Mai <[email protected]>

clang-format

507c379

Signed-off-by: João Mai <[email protected]>

openroad-ci assigned joaomai Sep 25, 2025

joaomai requested a review from AcKoucher September 25, 2025 17:11

AcKoucher suggested changes Sep 26, 2025

View reviewed changes

joaomai added 5 commits September 26, 2025 17:30

mpl: use is_blockage_ flag

d082005

Signed-off-by: João Mai <[email protected]>

mpl: remove auto and use isBlockage

40c1818

Signed-off-by: João Mai <[email protected]>

mpl: remove blockages from soft macro map

84caee9

Signed-off-by: João Mai <[email protected]>

Merge branch 'master' into mpl-new-blockages

89e81a2

clang-format

d0f97f6

Signed-off-by: João Mai <[email protected]>

joaomai added 2 commits September 29, 2025 15:14

Merge branch 'master' into mpl-new-blockages

13b56b3

mpl: update tests

da1006e

Signed-off-by: João Mai <[email protected]>

joaomai added 3 commits September 30, 2025 21:31

Merge branch 'master' into mpl-new-blockages

116aca0

mpl: add boost/polygon to bazel deps

22cf64b

Signed-off-by: João Mai <[email protected]>

buildifier

dd38f2b

Signed-off-by: João Mai <[email protected]>

joaomai requested a review from AcKoucher October 1, 2025 16:05

joaomai mentioned this pull request Oct 1, 2025

mpl: update config for bp_fe_top and metrics for bp_multi_top The-OpenROAD-Project/OpenROAD-flow-scripts#3516

Closed

AcKoucher suggested changes Oct 1, 2025

View reviewed changes

Merge branch 'master' into mpl-new-blockages

051f2cd

joaomai added 6 commits October 2, 2025 14:04

mpl: remove outdated comment

b173d4c

Signed-off-by: João Mai <[email protected]>

mpl: floor and comment in PA depth limit

cd0dc9e

Signed-off-by: João Mai <[email protected]>

Merge branch 'master' into mpl-new-blockages

6650dda

mpl: remove comment

82b488b

Signed-off-by: João Mai <[email protected]>

Merge branch 'master' into mpl-new-blockages

d754e9e

clang-format

58bd700

Signed-off-by: João Mai <[email protected]>

joaomai requested a review from AcKoucher October 8, 2025 17:47

AcKoucher approved these changes Oct 9, 2025

View reviewed changes

AcKoucher requested a review from maliberty October 9, 2025 12:35

AcKoucher mentioned this pull request Oct 9, 2025

mpl: evaluate connection between clusters based on connections' ratio #8375

Open

maliberty approved these changes Oct 9, 2025

View reviewed changes

joaomai added 6 commits October 9, 2025 21:25

Merge branch 'master' into mpl-new-blockages

2cb77c5

Merge branch 'master' into mpl-new-blockages

bb6c011

Merge branch 'master' into mpl-new-blockages

af4219e

Merge branch 'master' into mpl-new-blockages

2b24960

Merge branch 'master' into mpl-new-blockages

b953cc8

Merge branch 'master' into mpl-new-blockages

94d3e38

maliberty approved these changes Oct 15, 2025

View reviewed changes

joaomai mentioned this pull request Oct 15, 2025

update configs and metrics for blockages as fixed macros in MPL The-OpenROAD-Project/OpenROAD-flow-scripts#3583

Merged

maliberty merged commit ca004a3 into The-OpenROAD-Project:master Oct 16, 2025
13 checks passed

maliberty deleted the mpl-new-blockages branch October 16, 2025 04:10

This was referenced Oct 16, 2025

Placement blockages should be treated as hard constraints in MPL #8481

Closed

MPL should treat blockages as fixed blocks in SA #8155

Closed

		return macro.getCluster() != nullptr
		&& macro.getCluster()->isClusterOfUnplacedIOPins();

mpl: blockages as fixed macros #8438

mpl: blockages as fixed macros #8438

Uh oh!

Conversation

openroad-ci commented Sep 25, 2025

Uh oh!

github-actions bot commented Sep 25, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

AcKoucher commented Sep 26, 2025

Uh oh!

github-actions bot commented Sep 27, 2025

Uh oh!

github-actions bot commented Sep 29, 2025

Uh oh!

github-actions bot commented Oct 1, 2025

Uh oh!

joaomai commented Oct 1, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Oct 2, 2025

Uh oh!

AcKoucher commented Oct 9, 2025

Uh oh!

maliberty commented Oct 9, 2025

Uh oh!

github-actions bot commented Oct 15, 2025

Uh oh!

joaomai commented Oct 15, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants