Your First Place and Route Run

This article is a guided first run of place and route in ChipVerify Lab. You take the 8-bit counter from RTL through synthesis and the full OpenROAD flow, then read every section of the report: area and floorplan, timing, clock tree, power, IR drop and the checks. Along the way you will see which numbers change between synthesis and layout, and what the two settings of the PnR panel, the clock period and the utilization, actually do.

9 min read | Beginner

What You'll Learn

• Run synthesis and place and route in ChipVerify Lab
• Explain what the clock period and core utilization settings control
• Read the area, timing, CTS, power and IR drop sections of a PnR report
• Compare post-layout results with the synthesis estimates

Step 1: Synthesize

Open the example (link at the end of this article). The design is the counter from The RTL-to-GDSII Flow. In the synthesis panel, keep the hd library and the typical corner, and run synthesis. The result:

The flip-flops take 57 % of the area: a dfxtp_1 is 7.36 um wide, more than five inverters.

Step 2: Set Up Place and Route

The PnR tab has two main settings:

• Clock Period: the target for the clock clk, here 5 ns (200 MHz). The flow creates the clock with create_clock, builds the clock tree and repairs setup timing against it.
• Core Utilization: how full the core should be, here 50 %. It sets the target placement density. The flow sizes the floorplan for at most 40 % utilization, so that small designs get enough rows, and very small designs get a fixed 60 um die.

The library and corner are taken from the synthesis run. PRO users can also set a die area, use TinyTapeout mode, or edit the OpenROAD script directly.

Step 3: Run and Read the Report

Click Run Place & Route. The counter finishes in well under a minute. Here is what each section shows for this run, and what it means.

Area and Floorplan

The design area is 298 um² at 46 % utilization. The extra 18 um² over synthesis are three clock buffers and five tap cells. The core is 26.2 um by 24.5 um with 9 rows, and the die, with a 10 um margin for the pins, is 46.5 um square. All 11 pins ended up on the bottom edge; I/O Pin Placement explains how the pin placer decides.

Static Timing (Post-Route)

The worst setup path runs from flip-flop _40_ through an and4_1, two and3_1 and an o21a_1 to flip-flop _47_, the carry chain of the counter:

  0.2669 ^ _40_/CLK (sky130_fd_sc_hd__dfxtp_1)
  0.6752 ^ _40_/Q (sky130_fd_sc_hd__dfxtp_1)
  0.9607 ^ _21_/X (sky130_fd_sc_hd__and4_1)
  1.2469 ^ _22_/X (sky130_fd_sc_hd__and3_1)
  1.4572 ^ _23_/X (sky130_fd_sc_hd__and3_1)
  1.5758 ^ _39_/X (sky130_fd_sc_hd__o21a_1)
  1.5758   data arrival time
  5.2246   data required time
  3.6488   slack (MET)

Two things are new compared with synthesis timing. The clock now arrives through the clock tree, 0.27 ns after the clock pin, at both flip-flops, so the launch and capture edges both move by the same amount. And the wires add capacitance, so each gate is a little slower than with ideal wires. With 3.65 ns of slack, this design could run much faster.

CTS

The clock tree has 8 sinks (the flip-flops) and 3 buffers, all clkbuf_1: one root and two leaves, each leaf driving four flip-flops. The path depth is 2 buffers to every flip-flop, which keeps the skew small. See CTS in OpenROAD.

Power

Group                  Internal  Switching    Leakage      Total
Sequential             6.50e-05   1.12e-06   6.49e-11   6.62e-05  67.3%
Combinational          1.60e-06   1.27e-06   4.58e-11   2.87e-06   2.9%
Clock                  9.87e-06   1.94e-05   3.54e-12   2.93e-05  29.8%
Total                  7.65e-05   2.18e-05   1.14e-10   9.83e-05 100.0%

The total is 98 uW at 200 MHz with the default switching activity. The flip-flops and the clock network use 97 % of it: they toggle on every clock edge, while the counter logic switches far less. Leakage is negligible at the typical corner.

IR Drop

The IR drop analysis solves the resistance of the power grid with the current drawn by every cell. For the counter, the worst drop is 1.87e-05 V on VDD and 1.83e-05 V on VSS, against a 1.80 V supply: about 0.001 %. A design this small hardly loads the grid. The IR Drop article shows a design where it matters.

Antenna, DRC and LVS

The antenna check finds 0 net and 0 pin violations, and the detailed router reports no routing violations. The DRC section (KLayout, PRO) and the LVS section then check the GDSII against the full manufacturing rules and against the netlist; DRC and LVS explain how to read them.

Step 4: Push the Clock

With 3.65 ns of slack at 5 ns, the critical path takes about 1.35 ns. What happens at 1.5 ns and 1.2 ns?

• 1.5 ns: the flow finds no setup violation during placement, and the routed design has +0.15 ns of slack. The layout is the same as at 5 ns.
• 1.2 ns: after CTS, repair_timing finds 3 failing endpoints and upsizes 4 cells. It reaches +0.025 ns with estimated wires, and the area grows to 307 um². But after routing, with the real wires, the worst path ends at -0.0049 ns: violated by 5 picoseconds.

Running It in ChipVerify Lab

The example is the counter. Synthesize it, run Place and Route at 5 ns and 50 %, and check your numbers against this article. Then try the clock periods of Step 4.

Common Mistakes