synopsys-flow

Synopsys Flow

Synopsys ASIC front-end tools: DC, VCS, SpyGlass, DFT Compiler, VC Formal.

---

When to use

• Running Design Compiler (compile_ultra) for ASIC synthesis.
• Compiling and running simulation in VCS.
• SpyGlass lint, CDC, or low-power rule checking.
• Inserting scan chains with DFT Compiler (insert_dft).
• Property proving in VC Formal.

Not for: Cadence equivalents (use cadence-flow); FPGA flows (use vivado-flow / quartus-flow); SDC/XDC authoring (use timing-constraints); RTL coding rules (use synthesis-guidelines).

---

Design Compiler (synthesis)

# Setup
set TOP my_design
set TECH_LIB /libs/tech.db
set TARGET_LIB /libs/std_cell.db

set link_library "* $TECH_LIB $TARGET_LIB"
set target_library $TARGET_LIB

# Read RTL
analyze -format sverilog [glob rtl/*.sv]
elaborate $TOP
link

# Constraints
source constraints/timing.sdc

# Compile
compile_ultra -timing_high_effort_script

# Reports
report_timing > rpt/timing.rpt
report_area   > rpt/area.rpt
report_power  > rpt/power.rpt
report_qor    > rpt/qor.rpt

# Output
write -format verilog -hierarchy -output netlist/${TOP}.v
write_sdc out/${TOP}.sdc
write_sdf out/${TOP}.sdf

---

VCS (simulation)

# Compile
vcs -full64 -sverilog \
    -f filelist.f \
    -timescale=1ns/1ps \
    +define+SIM \
    -o simv

# Run
./simv +fsdbDumpfile=dump.fsdb

---

SpyGlass (lint + CDC)

# Project setup
new_project lint_proj
read_file -type verilog [glob rtl/*.sv]
set_option top top_module

# Lint
current_goal lint/lint_rtl
run_goal
write_report lint_report.rpt

# CDC
current_goal cdc/cdc_verify
run_goal
write_report cdc_report.rpt

---

DFT Compiler (scan insertion)

# Pre-DFT
set_scan_configuration -chain_count 4
set_scan_configuration -clock_mixing no_mix

# Check DFT rules
create_test_protocol
dft_drc

# Insert scan
insert_dft
write -format verilog -output scan_netlist.v

# Report
report_scan_path > scan_path.rpt

---

VC Formal (property proof)

# Setup
read_file -type verilog netlist/design.v
read_file -type sva assertions.sva

# Run formal
set_engine_mode {Hp Ht}
prove -all

# Report
report_property -all

---

File Organization

project/
├── rtl/              # RTL source
├── tb/               # Testbenches
├── constraints/      # SDC files
├── scripts/
│   ├── syn/          # DC scripts
│   ├── sim/          # VCS scripts
│   ├── lint/         # SpyGlass scripts
│   ├── dft/          # DFT Compiler scripts
│   └── formal/       # VC Formal scripts
├── work/             # Tool outputs
└── reports/          # Reports

---

Anti-patterns (do NOT do this)

1. Library setup hardcoded in scripts. Source a setup.tcl from environment so the same flow runs at multiple sites/PDKs.
2. Skipping link / check_design after analyze + elaborate. Unresolved references and partial elaboration silently produce wrong netlists.
3. Running compile_ultra without an SDC. DC defaults to no clock → trivially-met timing → garbage QoR.
4. Lint and CDC reports not gated in CI. Reports nobody reads = bugs nobody catches.
5. Hand-edited netlists. Always re-run synth from RTL; manual netlist edits don't survive the next ECO.
6. compile instead of compile_ultra. Legacy command — compile_ultra is the modern flow with better timing closure.

---

Validation gates

• analyze + elaborate + link clean (no unresolved references).
• check_design reports zero blocking errors.
• DC WNS ≥ 0 across all clock groups, or every violator documented and waivered.
• SpyGlass lint + CDC waiver lists reviewed; no new unwaived violations.
• VCS regression PASS rate = 100% on the gate-level netlist (GLS sanity).
• DFT Compiler stuck-at coverage ≥ 99% (or project target).
• VC Formal proofs all converge (proven, not bounded), or boundedness justified.
• Tool versions and library setup recorded with each release.