mpl: enumerate tilings for macro clusters instead of using simulated annealing #9247

openroad-ci · 2026-01-13T00:58:01Z

This PR changes the behavior of the calculaMacroTilings function. Now, the function enumerates the possible tilings for a given cluster instead of using SA to find the tilings. This enables tighter results, since the generated options are guaranteed to be minimum area.

This PR also removes the size_tolerance variable. It was used to change the maximum and minimum values for the thresholds, reducing the minimum and increasing the maximum. In practice, it made the use o MPL thresholds variables unreliable, with the only justification being a comment saying that it made the macro placer "more robust".

Signed-off-by: João Mai <jmai@precisioninno.com>

gemini-code-assist

Code Review

This pull request is a great simplification, replacing the Simulated Annealing approach for macro tiling with a more direct and deterministic enumeration method. This should lead to more predictable and optimal results for macro cluster shapes. The removal of the size_tolerance variable is also a good cleanup, making the clustering thresholds more reliable. I have a few suggestions to improve the implementation further, mainly focusing on performance, robustness, and code clarity.

gemini-code-assist · 2026-01-13T01:00:49Z

src/mpl/src/hier_rtlmp.cpp

-    return;
-  }
+  int number_of_macros = hard_macros.size();
+  int macro_width = hard_macros.front()->getWidth();


There's a potential for a crash here if cluster->getHardMacros() returns an empty vector, as hard_macros.front() would be undefined behavior. Although the call chain seems to prevent this, adding a defensive check for an empty vector would make the code more robust. For example:

std::vector<HardMacro*> hard_macros = cluster->getHardMacros(); if (hard_macros.empty()) { return; } int number_of_macros = hard_macros.size(); // ...

gemini-code-assist · 2026-01-13T01:00:49Z

src/mpl/src/hier_rtlmp.cpp

+  if (tilings.size() == 0) {
+    TilingList extra_tilings = generateTilingsForMacroCluster(
+        macro_width, macro_height, number_of_macros + 1);
+    tilings.insert(tilings.end(), extra_tilings.begin(), extra_tilings.end());
  }


This logic of trying to generate tilings for number_of_macros + 1 is a clever heuristic to find more aspect ratios, especially when number_of_macros is a prime number. However, this intent is not immediately obvious. Adding a comment would greatly improve code readability and maintainability.

github-actions

clang-tidy made some suggestions

github-actions · 2026-01-13T01:04:03Z

src/mpl/src/hier_rtlmp.cpp

-      }
-    }
-    remaining_runs -= run_thread;
+  if (tilings.size() == 0) {


warning: the 'empty' method should be used to check for emptiness instead of 'size' [readability-container-size-empty]

Suggested change

if (tilings.size() == 0) {

if (tilings.empty()) {

Additional context

/usr/include/c++/13/bits/stl_vector.h:1087: method 'vector'::empty() defined here

empty() const _GLIBCXX_NOEXCEPT ^

github-actions · 2026-01-13T01:04:03Z

src/mpl/src/hier_rtlmp.cpp

  }

-  if (tilings_set.empty()) {
+  if (tilings.size() == 0) {


warning: the 'empty' method should be used to check for emptiness instead of 'size' [readability-container-size-empty]

Suggested change

if (tilings.size() == 0) {

if (tilings.empty()) {

Additional context

/usr/include/c++/13/bits/stl_vector.h:1087: method 'vector'::empty() defined here

empty() const _GLIBCXX_NOEXCEPT ^

Signed-off-by: João Mai <jmai@precisioninno.com>

github-actions · 2026-01-13T01:22:49Z

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

joaomai · 2026-01-13T14:32:37Z

Paired with PR #3781 on ORFS.

joaomai · 2026-01-13T14:33:28Z

Secure CI is ok, public CI has some fails that are addressed in the ORFS PR.

AcKoucher

Nice!

AcKoucher · 2026-01-13T15:35:20Z

src/mpl/src/hier_rtlmp.cpp

+    TilingList extra_tilings = generateTilingsForMacroCluster(
+        macro_width, macro_height, number_of_macros + 1);
+    tilings.insert(tilings.end(), extra_tilings.begin(), extra_tilings.end());


No need for std::vector::insert here. Just assign tilings directly.

Signed-off-by: João Mai <jmai@precisioninno.com>

github-actions · 2026-01-13T15:59:05Z

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

AcKoucher

LGTM.

maliberty · 2026-01-13T18:42:54Z

src/mpl/src/hier_rtlmp.cpp

+    if (number_of_macros % cols == 0) {
+      rows = number_of_macros / cols;


If we only allow perfect tiling we will have more limited choices (e.g. for an 9 macros you won't try 2x5, only 1x9 and 3x3). Is this intentional?

Yes. The previous code kept only the smallest area tilings too. Doing it this way also avoids notches.

maliberty · 2026-01-13T18:43:53Z

src/mpl/src/hier_rtlmp.cpp

-  }
+  int number_of_macros = hard_macros.size();
+  int macro_width = hard_macros.front()->getWidth();
+  int macro_height = hard_macros.front()->getHeight();


Where does this check that all macros are of a single master? What if they are not?

The clusters are always made of macros of the same size. The function that creates/merges macro clusters checks for their sizes (ClusteringEngine::groupSingleMacroCluster using the size_class variable).

joaomai added 8 commits January 12, 2026 22:54

mpl: include tight packing tilings for all macro clsuters

5d71d78

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: remove dupes from macro cluster tilings

0c2bb5c

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: remove size_tolerance

bafe6dd

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: remove old comments

e57ea8a

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: minor adjustments to macro cluster coarse shaping

65f03e6

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: remove annealing from macro cluster coarse shaping

7d5e86b

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: error if no tiling is found for macro cluster

01e0f06

Signed-off-by: João Mai <jmai@precisioninno.com>

mpl: update names for clarity

0e67816

Signed-off-by: João Mai <jmai@precisioninno.com>

openroad-ci assigned joaomai Jan 13, 2026

gemini-code-assist bot reviewed Jan 13, 2026

View reviewed changes

github-actions bot reviewed Jan 13, 2026

View reviewed changes

mpl: changes size() == 0 calls to empty()

ed39a9d

Signed-off-by: João Mai <jmai@precisioninno.com>

joaomai requested a review from AcKoucher January 13, 2026 02:07

joaomai mentioned this pull request Jan 13, 2026

update metrics due to MPL's macro tiling change The-OpenROAD-Project/OpenROAD-flow-scripts#3781

Merged

AcKoucher suggested changes Jan 13, 2026

View reviewed changes

mpl: remove uneeded insert call

75221b8

Signed-off-by: João Mai <jmai@precisioninno.com>

AcKoucher self-requested a review January 13, 2026 16:36

AcKoucher approved these changes Jan 13, 2026

View reviewed changes

AcKoucher requested a review from maliberty January 13, 2026 16:37

maliberty reviewed Jan 13, 2026

View reviewed changes

maliberty approved these changes Jan 13, 2026

View reviewed changes

maliberty merged commit b421595 into The-OpenROAD-Project:master Jan 13, 2026
13 checks passed

maliberty deleted the mpl-tight-packing branch January 13, 2026 19:32

joaomai mentioned this pull request Jan 13, 2026

MPL shouldn't anneal when shaping possibilities are evident #8153

Closed

		if (number_of_macros % cols == 0) {
		rows = number_of_macros / cols;

mpl: enumerate tilings for macro clusters instead of using simulated annealing #9247

mpl: enumerate tilings for macro clusters instead of using simulated annealing #9247

Uh oh!

Conversation

openroad-ci commented Jan 13, 2026

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 13, 2026

Uh oh!

joaomai commented Jan 13, 2026

Uh oh!

joaomai commented Jan 13, 2026

Uh oh!

AcKoucher left a comment

Choose a reason for hiding this comment

Uh oh!

AcKoucher Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 13, 2026

Uh oh!

AcKoucher left a comment

Choose a reason for hiding this comment

Uh oh!

maliberty Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

joaomai Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

maliberty Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

joaomai Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants