vyges-dpl

Detailed placement over the OpenDB design database: legality checking and legalization.

vyges physical dpl check-placement    design.odb
vyges physical dpl detailed-placement design.odb --out-odb out.odb

detailed-placement takes upstream's own tunables — --max-displacement, --site-search-window, --row-search-window, --drc-penalty, --disable-window-extension, --use-diamond-legalizer — with upstream's defaults. Run --help for the full list, and --describe for the machine-readable contract.

ℹ️ Two of upstream's options are deliberately absent. -disallow_one_site_gaps is deprecated there (it warns DPL-3 and ignores the flag, because the setting is derived from hasOneSiteMaster()), so accepting it would promise a control that cannot change the answer; this engine refuses it with that explanation rather than accepting it silently. -incremental is unimplemented and named in not_done.

Scope — read this before the correctness section

⛔ Every run names what it did NOT do. check-placement reports not_checked (families it did not evaluate) and limitations (families it did evaluate, but could not see everything); detailed-placement reports not_done. A clean verdict from a partial tool must not read as a complete one, so those fields are emitted whether or not they are empty.

Legalization

Two legalizers, and the default is the one upstream defaults to: negotiated congestion, in which cells are allowed to overlap, contested sites accumulate a history cost, and each iteration rips up and re-places until nobody overlaps. --use-diamond-legalizer selects the other — a diamond search outward from each cell's own position — mirroring upstream's flag of the same name.

⚠️ The two produce different placements, so the report says which one ran. Comparing one legalizer's output against the other's expected result measures nothing.

⬜ Not implemented, and named in not_done on every run: groups and regions, placement padding values, incremental placement, and two of countDRCViolations' four terms — checkEdgeSpacing (needs each master's LEF58 cell-edge list) and checkBlockedLayers.

⚠️ Nothing in the comparable corpus exercises those two DRC terms, so their absence is invisible to the score rather than shown to be harmless. checkPadding was in exactly that position until it was built, and wiring it moved aes by 5,700 cells.

Every run also names, in filtered_out, each instance the model filter excluded — a filter that drops instances silently is indistinguishable from a design that has none of them.

Checking

Seven of upstream's nine check families are evaluated: site alignment, placed, overlap, in rows, padding, blocked layers and one-site gaps.

family	status
site alignment · placed · overlap · in rows	✅ evaluated
padding	✅ evaluated, with zero padding — no set_placement_padding model, so only class-pair conflicts are caught
blocked layers	✅ evaluated where the design has a vertical M2/M3 special wire; where it has none the mask is empty and a pass would be vacuous, so it is reported as unchecked instead
one-site gaps	✅ evaluated when the technology has no one-site master — the condition upstream derives it from, not a user option
region placement	⬜ needs a region model
edge spacing	⬜ the LEF58 cell-edge rule is implemented and tested; what is missing is each master's edge list

🔑 The checker came first on purpose. It is the oracle — it needs no legalizer to be correct in order to be useful — and upstream's own suite leans on it harder: 77 of 92 .tcl tests call check_placement, 68 call detailed_placement.

Status

✅ Legalization matches the reference on every comparable case in its own regression suite — 28 of 28, at pin 7d490b8ecd357199c0c0e9f3e32becd5eb507c34.

🔑 The agreement is SWEEP-LEVEL, not final-placement only. Upstream's own per-iteration debug trace and this engine's match line for line — same cell, same order, same chosen position, every iteration — on simple05, simple07, gcd (574 lines) and hybrid_cells.

ℹ️ It read 18 of 28 when the pin first moved from 945a9f48dc6e5cc91d865daa92c45a1094cb682c, because upstream reworked the negotiation legalizer's initial snapping across seven commits and refreshed 24 of its own goldens. Four mechanisms recovered it:

mechanism	what it changed
Grid::gridRoundX	the initial x rounds to the nearest site rather than truncating
displacementInSites / rowDispInSites	displacement is site widths on both axes, so a row jump is no longer priced as one step — and the search wavefront reorders with it
initialSnap()'s diamond search	replaces a four-direction scan that could not see a diagonal
the snap runs as its own pass	after every fixed cell is blockaded, testing grid capacity rather than merely whether a site exists

⚠️ The last one is a CALL SEQUENCE fix, not a logic one — the function was right and was being called on inputs that were not ready. It took the gate from 23/28 to 28/28 on its own.

The three large designs, at the current pin:

design	components	result at 7d490b8
aes	21,340	identical
ibex	34,184	identical
gcd	549	identical
the other 25 cases	—	identical

⚠️ And the denominator is short. Upstream ships 69 .tcl cases calling detailed_placement, not 63: five (report_failures, fragmented_row03, pad02, fillers8, obstruction2) wrap it as catch { detailed_placement }, so the harness's command filter does not see them and they are scored by nothing. All five are error-path cases.

🔑 The agreement is sweep-level, not final-placement only. Upstream's own per-iteration debug trace and this engine's match line for line — same cell, same order, same chosen position, every iteration. A matching output can be coincidence; a matching decision sequence is the algorithm.

⛔ That is a claim about what this corpus asks, not about every design. 35 of upstream's 63 detailed_placement cases are outside it and are not scored at all: 12 ship no golden, 8 need filler placement, 7 declare regions or groups, 7 need placement padding values, 1 needs both.

⚠️ Every score is scoped to one upstream commit. A score quoted without its pin says nothing: the reference moves.

Correctness

Correlated against OpenROAD at pin 945a9f48dc6e5cc91d865daa92c45a1094cb682c, in both directions — a checker that cannot fail proves nothing:

design	reference	this engine
aes.defok (legal)	clean	21,340 cells, 0 violations
cell_on_block1.def (illegal)	"Site aligned check failed (4)"	4 site-align failures

Four behaviours that are easy to get wrong and are pinned as tests at their sites:

• Site alignment is core-relative. Upstream compares cell->getLeft() % siteWidth, and getLeft() is relative to core_.xMin(). Read as an absolute coordinate it reports every cell of a clean design as misaligned — measured, 21,340 of 21,340.
• A site-alignment failure removes the cell from the overlap comparison entirely. That is a side effect of upstream's continue, not a separate rule: checkOverlap is what paints a cell into its pixels, so a skipped cell is never there to collide with.
• A cell's height in ROWS comes from its master, not from where the cell currently sits. A cell resting a few database units above its row covers two row bands; it is still a single-height cell, and every rule that treats multi-row cells specially depends on the difference.
• A hard placement blockage reaches the legality test only through pixel CAPACITY. Nothing in the earlier guards — in-die, valid row, site orientation — consults pixel validity, so a legality test that stops before the final footprint loop calls a cell inside a blockage perfectly legal.

ℹ️ The overlap acceleration differs deliberately — a rectangle sweep here, a pixel walk upstream. The predicate is identical and the failing set matches; which partner is reported can differ, because upstream names whichever cell already owns the pixel.

Status

--describe reports maturity: partial, and a test fails if that word is left behind when the unimplemented families listed above are filled in.

⚠️ Correlation against upstream's corpus is ongoing and the engine is not yet a drop-in replacement. The scope section above is the honest boundary; not_done, not_checked and limitations in the run output are the machine-readable version of it, and they are what to trust over any prose here.

License

Apache-2.0. See LICENSE and NOTICE.

ℹ️ This engine reads the design database through vyges-opendb, which binds OpenROAD's OpenDB (libodb) — BSD 3-Clause, Copyright (c) 2019-2026 The Regents of the University of California. The attribution is in NOTICE.