mpl: -write_macro_placement output rejected by place_macro (snap residue past core, MPL-0034)

Describe the bug

The file written by rtl_macro_placer -write_macro_placement does not round-trip through place_macro: the Snapper picks the track/manufacturing-grid position closest to the pin center, rounding up when in doubt, so a macro at the top or right core edge is committed with its bounding box a snap residue past the core boundary. place_macro then rejects the placer's own output with MPL-0034, which breaks every re-injection workflow built on MACRO_PLACEMENT_TCL.

To reproduce

From src/mpl/test (uses only in-repo test files; the pin constraint is the same one orientation_improve1.tcl uses and drives the macro to the boundary):

read_lef Nangate45/Nangate45.lef
read_lef testcases/orientation_improve1.lef
read_def testcases/orientation_improve1.def
set_io_pin_constraint -direction INPUT -region right:10-30*
set_thread_count 0
rtl_macro_placer -report_directory rtlmp_rpt -write_macro_placement mp.tcl
set core [[ord::get_db_block] getCoreArea]
foreach inst [[ord::get_db_block] getInsts] {
  if { [[$inst getMaster] isBlock] } {
    set b [$inst getBBox]
    puts "[$inst getName] bbox ([$b xMin] [$b yMin]) ([$b xMax] [$b yMax]),\
      core ([$core xMin] [$core yMin]) ([$core xMax] [$core yMax])"
  }
}
source mp.tcl

Observed:

MACRO_1 bbox (21680 21330) (221680 221330), core (18000 19600) (221680 221200)
[ERROR MPL-0034] Cannot place MACRO_1 at (10.84, 10.665) (110.84, 110.665), outside of the core (9, 9.8) (110.84, 110.6).

The committed bbox is 130 dbu (65nm) past the core top. Expected: the snap search only considers positions that keep the macro inside the core (nudging inward by manufacturing-grid steps when no aligned position qualifies), so write_macro_placement -> place_macro round-trips exactly.

Several checked-in test goldens encode such out-of-core placements today (boundary_push1, fixed_macros2, halos5, orientation_improve1/3 — verify by comparing each macro bbox against dbBlock::getCoreArea() in the .defok).

A secondary observation while reducing this: when the placement file is re-applied to a design whose macros are still placed, place_macro reports the macro overlapping itself (MPL-0041 ... Found overlap with other macros: MACRO_1) — findOverlappedMacros() does not exclude the instance being placed.

Environment

OpenROAD master 3ca581e384324a1d09a53d29b00d4fad1c365e70, Linux x86_64, bazel build.

We currently carry a containment fix downstream as a patch in bazel-orfs and can upstream it once the ongoing mpl churn settles.

🤖 Generated with Claude Code