Skip to content

dbSta: read cross-die parasitics on 3D-IC designs - #12089

Merged
jhkim-pii merged 3 commits into
The-OpenROAD-Project:masterfrom
The-OpenROAD-Project-staging:3dic-spef-stitch
Nov 2, 2026
Merged

jhkim-pii merged 3 commits into
The-OpenROAD-Project:masterfrom
The-OpenROAD-Project-staging:3dic-spef-stitch

Conversation

@dsengupta0628

Copy link
Copy Markdown
Contributor

Summary

On 3D-IC designs, read_spef does not merge the parasitics of a cross-die net. 3D RCX writes such a net as three SPEF pieces:

  • each die's wiring, read with read_spef -path <chip_inst>;
  • the die-to-die bond, read at top scope.

Each piece lands on a different net, so a driver sees only its own die's wire. On master the bond resistance has no effect on timing, and cross-die paths come out optimistic.

This supersedes #11189, rebased onto current master.

Fix

All three pieces belong to the dbChipNet above a bonded boundary, which is also the net the bonds SPEF is written against. Three dbNetwork overrides return that chip net:

function before after
net(const Pin*) on a chiplet boundary bterm nullptr the dbChipNet above
highestConnectedNet() on a bonded die net the die net the dbChipNet above
highestNetAbove() on a bonded die net the die net the dbChipNet above
  • net() is how SpefReader::dspfBegin picks the owning net of a -path read, so it is what merges the die pieces into one network.
  • highestConnectedNet() is how Parasitics::findParasiticNet finds a driver's parasitics, so a driver on either die finds the merged network.
  • highestNetAbove() has to agree with it. Otherwise ensureParasiticNode marks every node of the merged network as external.

Supporting changes:

  • A die-net → chip-net index, built with the existing unfolded maps.
  • Helpers chipNetAbove(), chipNetAboveNet() and chipNetOfPadITerm().
  • staToDb(const Net*) now returns null for a chip net instead of casting it to the wrong type.

There is no new command, no Tcl or SWIG change, and no src/sta change.

Usage

read_3dbx top.3dbx
read_spef top.bonds.spef                        ;# top scope, read FIRST
read_spef -path die_a_0 top.die_a.spef
read_spef -path die_b_0 top.die_b.spef

The bonds SPEF must be read first. A top-scope read replaces a net's parasitic network, while a -path read merges into it, so the other order silently discards the die pieces.

Changes since #11189

  • Rebased onto current master, which now uses OpenSTA's const Net* API.
  • The test's bonds SPEF uses the 3D RCX writer's current pad-pin node names (chipA/bump_q:PAD), not the old chipA:q. Timing is the same either way.
  • visitConnectedPins checks that the 3D-IC maps are up to date once per net, not once per iterm.

Type of Change

  • Bug fix
  • New feature

Impact

3dic_spef fixture (two dies; the bond resistance is swept):

cap at chipA/buf/Z arrival, R = 0.5 Ω / 2 kΩ / 50 kΩ
master 2.00 fF 0.15 / 0.15 / 0.15 ns
this PR 102.97 fF (102.88 at 50 kΩ) 0.39 / 0.43 / 4.63 ns

This matches an independent user report on #11189: 102.88 fF and 4.63 ns at 50 kΩ with the PR applied.

Non-3D-IC designs are unaffected, because has3DicChip() is false for them.

Known limitations

  • report_net on a bonded die net shows only part of the cross-die net. Tcl net reports go through OpenSTA's generic net walk. That walk now climbs to the chip net but does not descend into the other die, so the report lists the far die's bump pad but not its loads. 3dic_cross.ok is updated for this. Timing and SPEF reading use dbNetwork directly and are not affected.
  • Estimated parasitics are not covered. estimate_parasitics and rsz build parasitics on the die net, and delay calculation no longer looks there for a bonded net on a 3D-IC design. This PR covers parasitics read from SPEF only.
  • Read order matters, and nothing enforces it. Reading the bonds file after the die files, or reading a die file a second time, silently gives wrong parasitics.
  • Netlist edits on a chiplet invalidate the die-net → chip-net index. For example, buffering a bump's net leaves the index stale. It is rebuilt only when the number of chip nets changes.
  • Die nets whose boundary port has no bond are not handled. One example is a package-facing input with no bump. Their -path parasitics are not stitched by this PR.

Verification

  • I have verified that the local build succeeds (./etc/Build.sh). Built with Bazel (bazel build //:openroad); ./etc/Build.sh not run.
  • I have run the relevant tests and they pass. bazel test //src/dbSta/test:all: 2555/2555 pass.
  • My code follows the repository's formatting guidelines.
  • I have included tests to prevent regressions. New test 3dic_spef: two dies, a 2 kΩ bond, and checks on the merged driver load and the cross-die arrival.
  • I have signed my commits (DCO).

Related Issues

  • Supersedes #11189.
  • Overlaps #11769 (chip bumps as dbInsts) in dbNetwork::bumpPadITerm; see the comment there.

A cross-die net arrives from 3D RCX as three SPEF pieces: each die's
wiring (read with -path <chip_inst>) and the die-to-die bond (read at
top scope). read_spef left them on three different nets, so a driver
found only its own die's piece. On the 3dic_cross fixture chipA's
driver saw 2 fF instead of ~103 fF, and changing the bond resistance
did not change timing at all.

The dbChipNet above a bonded boundary is the net all three pieces
belong to, and the net the bonds SPEF is written against. Three
dbNetwork overrides ascend to it:

- net(const Pin*) on a chiplet boundary bterm returns the chip net.
  SpefReader::dspfBegin climbs term -> pin -> net to pick the owning
  net of a -path read, so this is what merges the die pieces.
- highestConnectedNet() on a bonded die net returns the chip net, so
  Parasitics::findParasiticNet lets a driver on either die find it.
- highestNetAbove() must agree, or ensureParasiticNode marks every
  node of the merged network external.

A die-net -> chip-net index is built with the existing unfolded maps.
staToDb(const Net*) returns null for a chip net instead of casting it.

Read the bonds SPEF before the die SPEFs: a top-scope read replaces a
net's parasitic network, a -path read merges into it.

Known limitation: Tcl net reports (report_net) walk the network
through OpenSTA's generic traversal, which now climbs from a bonded
die net to the chip net but does not descend into the other die, so
they list the far die's bump pad but not its loads. Timing and SPEF
reading use dbNetwork directly and are unaffected.

Also not covered: estimate_parasitics and rsz build parasitics on the
die net, which delay calculation no longer looks up for a bonded net
on a 3D-IC design. This PR is for parasitics read from SPEF.

New test 3dic_spef. 3dic_cross.ok updated for the report_net change.

Signed-off-by: dsengupta0628 <dsengupta@precisioninno.com>
@dsengupta0628 dsengupta0628 self-assigned this Oct 31, 2026

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request introduces support for cross-die parasitics in 3DIC designs by mapping die-side nets to their corresponding chip-nets above the chiplet boundary. It implements helper methods to retrieve these chip-nets and updates network queries like net(), highestConnectedNet(), and highestNetAbove() to return the chip-net when applicable. A new regression test, 3dic_spef, is also added. The review feedback highlights a potential undefined behavior in staToDb() where a sibling class pointer is downcast directly to dbNet* to check its object type; casting to the common base class odb::dbObject* first is recommended to resolve this safely.

Comment thread src/dbSta/src/dbNetwork.cc Resolved
Calling getObjectType() through a dbNet* that may point at a dbChipNet
is undefined behavior (sibling classes). Cast to the common dbObject
base first, as staToDbChipNet() and staToDb(const Port*) already do.

Signed-off-by: dsengupta0628 <dsengupta@precisioninno.com>
@maliberty
maliberty removed their request for review October 31, 2026 23:10
Signed-off-by: dsengupta0628 <dsengupta@precisioninno.com>

# Conflicts:
#	src/dbSta/test/CMakeLists.txt
@dsengupta0628

Copy link
Copy Markdown
Contributor Author

@osamahammad21 osamahammad21 left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. The logic is accepted with the known limitations listed in the description. This still needs a review from a dbSta code owner.

@jhkim-pii
jhkim-pii merged commit 30e4295 into The-OpenROAD-Project:master Nov 2, 2026
20 checks passed
@jhkim-pii
jhkim-pii deleted the 3dic-spef-stitch branch November 2, 2026 21:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants