RTL-guided functional ECO
Implement revised RTL directly in the gate-level design without waiting for a full re-synthesis cycle.
See the RTL-to-netlist flowFunctional netlist ECO
GOF ECO turns a revised RTL or reference netlist into a controlled implementation-netlist patch. Generate the fix automatically, refine it in a schematic or script, and verify equivalence in the same flow.
Choose the constraint that matters
Choose RTL-guided ECO for speed or reference-netlist-driven ECO for broader changes, then carry the verified patch through post-layout and DFT netlists.
Implement revised RTL directly in the gate-level design without waiting for a full re-synthesis cycle.
See the RTL-to-netlist flowCompare a re-synthesized reference netlist with the implementation netlist. This established flow is slower than RTL-guided ECO, but handles larger and more extensive logic changes.
End-to-end path ~2 days when full reference re-synthesis is required
See the reference-netlist flowMap a verified patch onto available standard or gate-array spare cells while respecting a fixed placement.
See the post-layout flowConstrain test modes, preserve scan connectivity, and explicitly stitch new flops into scan chains when required.
See the DFT-aware flowFrom change to handoff
Each stage produces information you can inspect: mapped differences, the selected patch, equivalence results, and implementation outputs.
Read the implementation, revised RTL or reference netlist, libraries, and optional LEF/DEF physical data.
Map hierarchy and endpoints, then isolate the logic that must change, including synthesis-optimized and boundary-optimized logic.
Evaluate structure, reorder, and cut-point strategies; select a compact patch and refine it with GUI or Perl APIs when needed.
Check equivalence, review the ECO report, and write the updated netlist plus scripts for downstream implementation.
Control when the case demands it
Automatic ECO provides a complete design view and generates candidate patches. For focused or repeated changes, the same design database is available through an incremental schematic and Perl-compatible scripting APIs.
Explore GUI and Script ModeAnalyze non-equivalence and choose a patch strategy across the design.
Trace the affected cone and edit directly on the incremental schematic.
Automate repeated changes and design queries with netlist-processing APIs.
Combine targeted preprocessing with automatic patch generation in one run.
Measured production-scale case
A post-layout SoC state-machine ECO completed with GOF 11.8. The first automatic run generated a reusable patch; the same script was then applied to the post-layout and DFT netlists.
THE CHANGE
xs_control state machine.The 15-clock-domain design initially reported four non-equivalent points in the cur_state flops. The ECO adds hci_en to the transition condition in state NOV2.
NOV2:
- if (noverlap_clr) begin
+ if (noverlap_clr & hci_en) begin
if (vci_en)
nxt_state = FSM_VCI;
else
nxt_state = FSM_VGH;
end
THE RESULT
Compare RTL and generate the patch on the 412 MB pre-layout netlist.
Export the ECO script for review and deterministic replay.
Apply the script to each post-layout or DFT netlist in 5–10 minutes.
Verify equivalence against the synthesized revised RTL with Synopsys Formality.
ECO Report for 'apu_sm_eco'
MODULE xs_control
Added wires 1
Added instances 1 AOI21
Modified instances 1
Deleted instances 0
Failing points 0
read_library("tsmc_5nm_svt.lib");
read_rtl("-ref", "modified_apu_top.sv");
read_rtl("-imp", "original_apu_top.sv");
set_top("apu_top");
rtl_compare();
read_design("-imp", "apu_top.prelayout.gv");
set_top("apu_top");
fix_design();
report_eco();
write_perl("eco_apu_top.pl");
read_design("-imp", "apu_top.postlayout.gv");
set_top("apu_top");
run("eco_apu_top.pl");
write_verilog("eco_apu_top.post.gv");
# The same patch can be replayed on
# the DFT-inserted implementation netlist.
Measured on a 2.3M-instance SoC using 32 CPU cores. The 30-minute result is the first automatic ECO run; 5–10 minutes is the subsequent script-application time per target netlist.
Patch quality starts with mapping
Physical synthesis can clone ports, invert boundary phases, or merge flops. GOF accounts for those transformations when mapping the implementation and reference designs. In the illustrated DCT/DCG case, the correct mapping keeps a one-gate change localized.
Review the detailed comparisonTechnical resources
Use the manual for flow setup, the API reference for automation, and worked cases for concrete examples.
Evaluate GOF on the work that matters
We will review the design stage, constraints, and current blocker, then define a focused technical evaluation.