Script check netlist

hw/ip/prim/README.md

@@ -366,3 +366,7 @@ This can be checked by the synthesis tool, e.g. `check_design -unloaded_comb/-un
 5. `lc_en_i`, `mubi_i` signals can only be connected to variables, or legal values (`MuBi4True`, `MuBi4False`, `On`, `Off`)
 
 If all checks are successful, the same constraints can be applied to the full design.
+The script `utils/design/check-netlist.py` [check-netlist] can be used to report a summary of size_only cells in a netlist.
+It can also automate an initial version of checks (4) and (5) but it does **not** replace a final manual inspection of the netlist.
+
+[check-netlist]: https://github.com/lowRISC/opentitan/tree/master/util/design#netlist-checker-script

util/design/README.md

@@ -297,3 +297,45 @@ TODO
 ## KECCAK Coefficient Generation Tool
 
 TODO
+
+## Netlist checker script
+
+`check-netlist.py netlist.v` will report the number of preserved, `size_only` cells.
+In addition, it will parse the netlist for suspicious synthesis optimizations
+such as constant propagation accross preserved instances.
+
+On the `prim_sdc_example.sv` design, the script produces the following output:
+
+```
+================================================================================
+DISCLAIMER:
+This script parses a synthesized netlist for suspicious constructs.
+It does not guarantee that there are no issues in the netlist(!)
+================================================================================
+
+================================================================================
+Final Summary:
+--------------------------------------------------------------------------------
+Found the following size_only instances:
+--------------------------------------------------------------------------------
+u_size_only_xor               120
+u_size_only_xnor              56
+u_size_only_and               56
+u_size_only_mux               0
+u_size_only_flop              252
+u_size_only_buf               328
+u_size_only_clock_gate        2
+others                        2
+--------------------------------------------------------------------------------
+Total                         816
+================================================================================
+Found 0 potential netlist problems in 0 modules!
+================================================================================
+```
+
+If the script reports potential issues, it is likely because the synthesis constraints are not set
+correctly.
+Please refer to the sections [creating-a-technology-library] and [synthesis-constraints].
+
+[creating-a-technology-library]: https://github.com/lowRISC/opentitan/tree/master/hw/ip/prim#creating-a-technology-library
+[synthesis-constraints]: https://github.com/lowRISC/opentitan/tree/master/hw/ip/prim#important-synthesis-constraints-to-keep-important-redundant-constructs

util/design/check-netlist.py

@@ -0,0 +1,205 @@
+#!/usr/bin/env python3
+# Copyright lowRISC contributors (OpenTitan project).
+# Licensed under the Apache License, Version 2.0, see LICENSE for details.
+# SPDX-License-Identifier: Apache-2.0
+r"""Checks the generated netlist for suspicious synthesis optimizations
+
+Specifically, this script parses the netlist for suspicious patterns. These
+checks do *not* replace a final manual inspection of the netlist, but help to
+find problems early in the design!
+"""
+
+import sys
+import re
+import argparse
+
+
+def hr(char: str) -> None:
+    print(char * 80)
+
+
+class Parser:
+    # List of patterns we want to use for finding size_only constraints.
+    size_only_patterns = [
+        "u_size_only",
+        "u_size_only_xor", "u_size_only_xnor", "u_size_only_and",
+        "u_size_only_mux", "u_size_only_flop", "u_size_only_buf",
+        "u_size_only_clock_gate"
+    ]
+
+    # Regex to find ".lc_en_i, ( ... )" and capture the content
+    pattern_lc_in = re.compile(r'\.lc_en_i \((.*?)\)')
+
+    # Regex to find ".mubi_i, ( ... )" and capture the content
+    pattern_mubi_in = re.compile(r'\.mubi_i \((.*?)\)')
+
+    # Regex to find "assign mubi_o[*] = " const"
+    pattern_mubi_const = re.compile(r'assign\s+([^=\s]*mubi_o[^=\s]*)\s*'
+                                    r'=\s*(.*?)\'b[01]+;')
+
+    # Regex to find "assign test*_o[*] = " const"
+    pattern_test_const = re.compile(r'assign\s+(test_.*_o.*?)\s*='
+                                    r'\s*(.*?)\'b[01]+;')
+
+    # tie low/high pattern
+    pattern_tie_low_high = re.compile(r"\'b[01]")
+
+    # pattern to find a module name
+    pattern_module = re.compile(r'^\s*module\s+([a-zA-Z_][a-zA-Z0-9_]*)')
+
+    # valid mubi4, mubi8, mubi12 patterns
+    mubi_allowed_patterns = ["12'b100101101001", "12'b11010010110",
+                             "8'b1101001", "8'b10010110",
+                             "4'b110", "4'b1001"]
+
+    # valid lc patterns
+    lc_allowed_patterns = ["4'b101", "4'b1010"]
+
+    # A count of how many instances we have seen of each of the size_only
+    # patterns (ordered the same as size_only_patterns).
+    size_only_count: list[int] = [0 for pat in size_only_patterns]
+
+    # The name of the module (if known)
+    module_name: str | None = None
+
+    # Number of errors seen so far
+    errors: int = 0
+
+    # A list of module names where errors have been found
+    error_modules: set[str] = set()
+
+    def describe_module(self) -> str:
+        return ('unknown module' if self.module_name is None
+                else f'module {self.module_name}')
+
+    def on_error(self,
+                 linenum: int, line: str,
+                 desc: str, bad_match: str | None) -> None:
+        print(f"Error: {desc} in {self.describe_module()}, line {linenum}:")
+        print(f"  Full line: {line.strip()}")
+        if bad_match is not None:
+            print(f"  Content parsed: {bad_match}")
+
+        hr('-')
+
+        self.errors += 1
+        if self.module_name is not None:
+            self.error_modules.add(self.module_name)
+
+    def take_line(self, line_number: int, line: str) -> None:
+        # Parse for module name: "module xxx(..."
+        match_module = Parser.pattern_module.match(line)
+        if match_module:
+            self.module_name = match_module.group(1)
+
+        # check for constant lc signals
+        lc_match = Parser.pattern_lc_in.search(line)
+        if lc_match:
+            content = lc_match.group(1).strip()
+            if Parser.pattern_tie_low_high.search(content):
+                if content not in Parser.lc_allowed_patterns:
+                    self.on_error(line_number, line,
+                                  'invalid constant lc_en_i', content)
+
+        # check for constant mubi_i signals
+        mubi_i_match = Parser.pattern_mubi_in.search(line)
+        if mubi_i_match:
+            content = mubi_i_match.group(1).strip()
+            if Parser.pattern_tie_low_high.search(content):
+                if content not in Parser.mubi_allowed_patterns:
+                    self.on_error(line_number, line,
+                                  'invalid constant mubi_i', content)
+
+        # check for constant mubi_o signals
+        if Parser.pattern_mubi_const.search(line):
+            self.on_error(line_number, line, 'tied low/high mubi bit', None)
+
+        # check for constant test-outputs in prim_sdc_example
+        if ((self.module_name == "prim_sdc_example" and
+             Parser.pattern_test_const.search(line))):
+            self.on_error(line_number, line, 'tied low/high test*_o*', None)
+
+        # Count size_only instances
+        for i, pattern in enumerate(Parser.size_only_patterns):
+            if pattern in line:
+                self.size_only_count[i] += 1
+
+    def report(self) -> None:
+        hr('=')
+        print("Final Summary:")
+        hr('-')
+        print("Found the following size_only instances:")
+        hr('-')
+
+        # Print the number of specific size_only matches that we have seen
+        # (with a suffix from _xor, _xnor, _only_and, ...)
+        specific_count = 0
+        for i, pattern in enumerate(Parser.size_only_patterns[1:], 1):
+            print(f"{pattern:<30}{self.size_only_count[i]}")
+            specific_count += self.size_only_count[i]
+
+        # Each of those specific hits will also have matched u_size_only (the
+        # first pattern). Subtracting specific_count from size_only_count[0]
+        # gives the number of lines that are match u_size_only but don't match
+        # one of the specific patterns.
+        other_count = self.size_only_count[0] - specific_count
+        print(f"{'others':<30}{other_count}")
+
+        hr('-')
+
+        # Since we know that every match for a specific case will also match
+        # the first pattern, the number of lines that match equals
+        # size_only_count[0]
+        print(f"{'Total':<30}{self.size_only_count[0]}")
+        hr('=')
+
+        print(f"Found {self.errors} potential netlist problems "
+              f"in {len(self.error_modules)} modules!")
+        hr('=')
+
+
+def parse_expressions(filename):
+    try:
+        with open(filename, 'r') as file:
+            lines = list(file)
+    except FileNotFoundError:
+        print(f"Error: The file '{filename}' was not found.")
+        return 1
+
+    parser = Parser()
+    for idx, line in enumerate(lines, 1):
+        parser.take_line(idx, line)
+    parser.report()
+
+    return parser.errors
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser(
+        description = "Parses a synthesized netlist for suspicious constructs and counts all"
+                      "size_only instances",
+        epilog = "Note: This script does not guarantee that there are no issues in the netlist (!)"
+    )
+
+    parser.add_argument(
+        "netlist_filename",
+        type=str,
+        help="The path to the netlist file (e.g., netlist.v) to be parsed."
+    )
+
+    args = parser.parse_args()
+    hr('=')
+    print("DISCLAIMER:\n"
+          "This script parses a synthesized netlist for suspicious constructs.\n"
+          "It does not guarantee that there are no issues in the netlist(!)")
+    hr('=')
+    print('\n')
+
+    file_to_parse = args.netlist_filename
+    errors = parse_expressions(file_to_parse)
+
+    if (errors > 0):
+        sys.exit(1)
+    else:
+        sys.exit(0)