mirror of https://github.com/YosysHQ/yosys.git
929 lines
22 KiB
C++
929 lines
22 KiB
C++
/*
|
|
* yosys -- Yosys Open SYnthesis Suite
|
|
*
|
|
* Copyright (C) 2012 Claire Xenia Wolf <claire@yosyshq.com>
|
|
*
|
|
* Permission to use, copy, modify, and/or distribute this software for any
|
|
* purpose with or without fee is hereby granted, provided that the above
|
|
* copyright notice and this permission notice appear in all copies.
|
|
*
|
|
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
|
|
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
|
|
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
|
|
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
|
|
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
|
|
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
|
|
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
|
|
*
|
|
*/
|
|
|
|
#include "libparse.h"
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
|
|
#include <istream>
|
|
#include <fstream>
|
|
#include <iostream>
|
|
#include <sstream>
|
|
|
|
#ifndef FILTERLIB
|
|
#include "kernel/log.h"
|
|
#endif
|
|
|
|
using namespace Yosys;
|
|
|
|
LibertyAst::~LibertyAst()
|
|
{
|
|
for (auto child : children)
|
|
delete child;
|
|
children.clear();
|
|
}
|
|
|
|
const LibertyAst *LibertyAst::find(std::string name) const
|
|
{
|
|
for (auto child : children)
|
|
if (child->id == name)
|
|
return child;
|
|
return NULL;
|
|
}
|
|
|
|
void LibertyAst::dump(FILE *f, sieve &blacklist, sieve &whitelist, std::string indent, std::string path, bool path_ok) const
|
|
{
|
|
if (whitelist.count(path + "/*") > 0)
|
|
path_ok = true;
|
|
|
|
path += "/" + id;
|
|
|
|
if (blacklist.count(id) > 0 || blacklist.count(path) > 0)
|
|
return;
|
|
if (whitelist.size() > 0 && whitelist.count(id) == 0 && whitelist.count(path) == 0 && !path_ok) {
|
|
fprintf(stderr, "Automatically added to blacklist: %s\n", path.c_str());
|
|
blacklist.insert(id);
|
|
return;
|
|
}
|
|
|
|
fprintf(f, "%s%s", indent.c_str(), id.c_str());
|
|
if (!args.empty() || !children.empty()) {
|
|
fprintf(f, "(");
|
|
for (size_t i = 0; i < args.size(); i++)
|
|
fprintf(f, "%s%s", i > 0 ? ", " : "", args[i].c_str());
|
|
fprintf(f, ")");
|
|
}
|
|
if (!value.empty())
|
|
fprintf(f, " : %s", value.c_str());
|
|
if (!children.empty()) {
|
|
fprintf(f, " {\n");
|
|
for (size_t i = 0; i < children.size(); i++)
|
|
children[i]->dump(f, blacklist, whitelist, indent + " ", path, path_ok);
|
|
fprintf(f, "%s}\n", indent.c_str());
|
|
} else
|
|
fprintf(f, " ;\n");
|
|
}
|
|
|
|
#ifndef FILTERLIB
|
|
// https://matklad.github.io/2020/04/13/simple-but-powerful-pratt-parsing.html
|
|
LibertyExpression LibertyExpression::parse(Lexer &s, int min_prio) {
|
|
if (s.empty())
|
|
return LibertyExpression{};
|
|
|
|
char c = s.peek();
|
|
auto lhs = LibertyExpression{};
|
|
|
|
while (isspace(c)) {
|
|
if (s.empty())
|
|
return lhs;
|
|
s.next();
|
|
c = s.peek();
|
|
}
|
|
|
|
if (isalpha(c)) { // pin
|
|
lhs.kind = Kind::PIN;
|
|
lhs.name = s.pin();
|
|
} else if (c == '(') { // parens
|
|
s.next();
|
|
lhs = parse(s);
|
|
if (s.peek() != ')') {
|
|
log_warning("expected ')' instead of '%c' while parsing Liberty expression '%s'\n", s.peek(), s.full_expr().c_str());
|
|
return lhs;
|
|
}
|
|
s.next();
|
|
} else if (c == '!') { // prefix NOT
|
|
s.next();
|
|
lhs.kind = Kind::NOT;
|
|
lhs.children.push_back(parse(s, 7));
|
|
} else {
|
|
log_warning("unrecognised character '%c' while parsing Liberty expression '%s'\n", c, s.full_expr().c_str());
|
|
return lhs;
|
|
}
|
|
|
|
while (true) {
|
|
if (s.empty())
|
|
break;
|
|
|
|
c = s.peek();
|
|
|
|
while (isspace(c)) {
|
|
if (s.empty())
|
|
return lhs;
|
|
s.next();
|
|
c = s.peek();
|
|
}
|
|
|
|
if (c == '\'') { // postfix NOT
|
|
if (min_prio > 7)
|
|
break;
|
|
s.next();
|
|
|
|
auto n = LibertyExpression{};
|
|
n.kind = Kind::NOT;
|
|
n.children.push_back(lhs);
|
|
lhs = std::move(n);
|
|
|
|
continue;
|
|
} else if (c == '^') { // infix XOR
|
|
if (min_prio > 5)
|
|
break;
|
|
s.next();
|
|
|
|
auto rhs = parse(s, 6);
|
|
auto n = LibertyExpression{};
|
|
n.kind = Kind::XOR;
|
|
n.children.push_back(lhs);
|
|
n.children.push_back(rhs);
|
|
lhs = std::move(n);
|
|
|
|
continue;
|
|
} else if (c == '&' || c == '*') { // infix AND
|
|
// technically space should be considered infix AND. it seems rare in practice.
|
|
if (min_prio > 3)
|
|
break;
|
|
s.next();
|
|
|
|
auto rhs = parse(s, 4);
|
|
auto n = LibertyExpression{};
|
|
n.kind = Kind::AND;
|
|
n.children.push_back(lhs);
|
|
n.children.push_back(rhs);
|
|
lhs = std::move(n);
|
|
|
|
continue;
|
|
} else if (c == '+' || c == '|') { // infix OR
|
|
if (min_prio > 1)
|
|
break;
|
|
s.next();
|
|
|
|
auto rhs = parse(s, 2);
|
|
auto n = LibertyExpression{};
|
|
n.kind = Kind::OR;
|
|
n.children.push_back(lhs);
|
|
n.children.push_back(rhs);
|
|
lhs = std::move(n);
|
|
|
|
continue;
|
|
}
|
|
break;
|
|
}
|
|
|
|
return lhs;
|
|
}
|
|
|
|
void LibertyExpression::get_pin_names(pool<std::string>& names) {
|
|
if (kind == Kind::PIN) {
|
|
names.insert(name);
|
|
} else {
|
|
for (auto& child : children)
|
|
child.get_pin_names(names);
|
|
}
|
|
}
|
|
|
|
bool LibertyExpression::eval(dict<std::string, bool>& values) {
|
|
bool result = false;
|
|
switch (kind) {
|
|
case Kind::AND:
|
|
result = true;
|
|
for (auto& child : children)
|
|
result &= child.eval(values);
|
|
return result;
|
|
case Kind::OR:
|
|
result = false;
|
|
for (auto& child : children)
|
|
result |= child.eval(values);
|
|
return result;
|
|
case Kind::NOT:
|
|
log_assert(children.size() == 1);
|
|
return !children[0].eval(values);
|
|
case Kind::XOR:
|
|
result = false;
|
|
for (auto& child : children)
|
|
result ^= child.eval(values);
|
|
return result;
|
|
case Kind::PIN:
|
|
return values.at(name);
|
|
case Kind::EMPTY:
|
|
log_assert(false);
|
|
}
|
|
return false;
|
|
}
|
|
#endif
|
|
|
|
int LibertyParser::lexer(std::string &str)
|
|
{
|
|
int c;
|
|
|
|
// eat whitespace
|
|
do {
|
|
c = f.get();
|
|
} while (c == ' ' || c == '\t' || c == '\r');
|
|
|
|
// search for identifiers, numbers, plus or minus.
|
|
if (('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || ('0' <= c && c <= '9') || c == '_' || c == '-' || c == '+' || c == '.') {
|
|
str = static_cast<char>(c);
|
|
while (1) {
|
|
c = f.get();
|
|
if (('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || ('0' <= c && c <= '9') || c == '_' || c == '-' || c == '+' || c == '.')
|
|
str += c;
|
|
else
|
|
break;
|
|
}
|
|
f.unget();
|
|
if (str == "+" || str == "-") {
|
|
/* Single operator is not an identifier */
|
|
// fprintf(stderr, "LEX: char >>%s<<\n", str.c_str());
|
|
return str[0];
|
|
}
|
|
else {
|
|
// fprintf(stderr, "LEX: identifier >>%s<<\n", str.c_str());
|
|
return 'v';
|
|
}
|
|
}
|
|
|
|
// if it wasn't an identifer, number of array range,
|
|
// maybe it's a string?
|
|
if (c == '"') {
|
|
str = "";
|
|
#ifdef FILTERLIB
|
|
str += c;
|
|
#endif
|
|
while (1) {
|
|
c = f.get();
|
|
if (c == '\n')
|
|
line++;
|
|
if (c == '"') {
|
|
#ifdef FILTERLIB
|
|
str += c;
|
|
#endif
|
|
break;
|
|
}
|
|
str += c;
|
|
}
|
|
// fprintf(stderr, "LEX: string >>%s<<\n", str.c_str());
|
|
return 'v';
|
|
}
|
|
|
|
// if it wasn't a string, perhaps it's a comment or a forward slash?
|
|
if (c == '/') {
|
|
c = f.get();
|
|
if (c == '*') { // start of '/*' block comment
|
|
int last_c = 0;
|
|
while (c > 0 && (last_c != '*' || c != '/')) {
|
|
last_c = c;
|
|
c = f.get();
|
|
if (c == '\n')
|
|
line++;
|
|
}
|
|
return lexer(str);
|
|
} else if (c == '/') { // start of '//' line comment
|
|
while (c > 0 && c != '\n')
|
|
c = f.get();
|
|
line++;
|
|
return lexer(str);
|
|
}
|
|
f.unget();
|
|
// fprintf(stderr, "LEX: char >>/<<\n");
|
|
return '/'; // a single '/' charater.
|
|
}
|
|
|
|
// check for a backslash
|
|
if (c == '\\') {
|
|
c = f.get();
|
|
if (c == '\r')
|
|
c = f.get();
|
|
if (c == '\n') {
|
|
line++;
|
|
return lexer(str);
|
|
}
|
|
f.unget();
|
|
return '\\';
|
|
}
|
|
|
|
// check for a new line
|
|
if (c == '\n') {
|
|
line++;
|
|
return 'n';
|
|
}
|
|
|
|
// anything else, such as ';' will get passed
|
|
// through as literal items.
|
|
|
|
// if (c >= 32 && c < 255)
|
|
// fprintf(stderr, "LEX: char >>%c<<\n", c);
|
|
// else
|
|
// fprintf(stderr, "LEX: char %d\n", c);
|
|
return c;
|
|
}
|
|
|
|
LibertyAst *LibertyParser::parse()
|
|
{
|
|
std::string str;
|
|
|
|
int tok = lexer(str);
|
|
|
|
// there are liberty files in the wild that
|
|
// have superfluous ';' at the end of
|
|
// a { ... }. We simply ignore a ';' here.
|
|
// and get to the next statement.
|
|
|
|
while ((tok == 'n') || (tok == ';'))
|
|
tok = lexer(str);
|
|
|
|
if (tok == '}' || tok < 0)
|
|
return NULL;
|
|
|
|
if (tok != 'v') {
|
|
std::string eReport;
|
|
switch(tok)
|
|
{
|
|
case 'n':
|
|
error("Unexpected newline.");
|
|
break;
|
|
case '[':
|
|
case ']':
|
|
case '}':
|
|
case '{':
|
|
case '\"':
|
|
case ':':
|
|
eReport = "Unexpected '";
|
|
eReport += static_cast<char>(tok);
|
|
eReport += "'.";
|
|
error(eReport);
|
|
break;
|
|
default:
|
|
error();
|
|
}
|
|
}
|
|
|
|
LibertyAst *ast = new LibertyAst;
|
|
ast->id = str;
|
|
|
|
while (1)
|
|
{
|
|
tok = lexer(str);
|
|
|
|
// allow both ';' and new lines to
|
|
// terminate a statement.
|
|
if ((tok == ';') || (tok == 'n'))
|
|
break;
|
|
|
|
if (tok == ':' && ast->value.empty()) {
|
|
tok = lexer(ast->value);
|
|
if (tok == 'v') {
|
|
tok = lexer(str);
|
|
}
|
|
while (tok == '+' || tok == '-' || tok == '*' || tok == '/' || tok == '!') {
|
|
ast->value += tok;
|
|
tok = lexer(str);
|
|
if (tok != 'v')
|
|
error();
|
|
ast->value += str;
|
|
tok = lexer(str);
|
|
}
|
|
|
|
// In a liberty file, all key : value pairs should end in ';'
|
|
// However, there are some liberty files in the wild that
|
|
// just have a newline. We'll be kind and accept a newline
|
|
// instead of the ';' too..
|
|
if ((tok == ';') || (tok == 'n'))
|
|
break;
|
|
else
|
|
error();
|
|
continue;
|
|
}
|
|
|
|
if (tok == '(') {
|
|
while (1) {
|
|
std::string arg;
|
|
tok = lexer(arg);
|
|
if (tok == ',')
|
|
continue;
|
|
if (tok == ')')
|
|
break;
|
|
|
|
// FIXME: the AST needs to be extended to store
|
|
// these vector ranges.
|
|
if (tok == '[')
|
|
{
|
|
// parse vector range [A] or [A:B]
|
|
std::string arg;
|
|
tok = lexer(arg);
|
|
if (tok != 'v')
|
|
{
|
|
// expected a vector array index
|
|
error("Expected a number.");
|
|
}
|
|
else
|
|
{
|
|
// fixme: check for number A
|
|
}
|
|
tok = lexer(arg);
|
|
// optionally check for : in case of [A:B]
|
|
// if it isn't we just expect ']'
|
|
// as we have [A]
|
|
if (tok == ':')
|
|
{
|
|
tok = lexer(arg);
|
|
if (tok != 'v')
|
|
{
|
|
// expected a vector array index
|
|
error("Expected a number.");
|
|
}
|
|
else
|
|
{
|
|
// fixme: check for number B
|
|
tok = lexer(arg);
|
|
}
|
|
}
|
|
// expect a closing bracket of array range
|
|
if (tok != ']')
|
|
{
|
|
error("Expected ']' on array range.");
|
|
}
|
|
continue;
|
|
}
|
|
if (tok != 'v') {
|
|
std::string eReport;
|
|
switch(tok)
|
|
{
|
|
case 'n':
|
|
continue;
|
|
case '[':
|
|
case ']':
|
|
case '}':
|
|
case '{':
|
|
case '\"':
|
|
case ':':
|
|
eReport = "Unexpected '";
|
|
eReport += static_cast<char>(tok);
|
|
eReport += "'.";
|
|
error(eReport);
|
|
break;
|
|
default:
|
|
error();
|
|
}
|
|
}
|
|
ast->args.push_back(arg);
|
|
}
|
|
continue;
|
|
}
|
|
|
|
if (tok == '{') {
|
|
while (1) {
|
|
LibertyAst *child = parse();
|
|
if (child == NULL)
|
|
break;
|
|
ast->children.push_back(child);
|
|
}
|
|
break;
|
|
}
|
|
|
|
error();
|
|
}
|
|
|
|
return ast;
|
|
}
|
|
|
|
#ifndef FILTERLIB
|
|
|
|
void LibertyParser::error()
|
|
{
|
|
log_error("Syntax error in liberty file on line %d.\n", line);
|
|
}
|
|
|
|
void LibertyParser::error(const std::string &str)
|
|
{
|
|
std::stringstream ss;
|
|
ss << "Syntax error in liberty file on line " << line << ".\n";
|
|
ss << " " << str << "\n";
|
|
log_error("%s", ss.str().c_str());
|
|
}
|
|
|
|
#else
|
|
|
|
void LibertyParser::error()
|
|
{
|
|
fprintf(stderr, "Syntax error in liberty file on line %d.\n", line);
|
|
exit(1);
|
|
}
|
|
|
|
void LibertyParser::error(const std::string &str)
|
|
{
|
|
std::stringstream ss;
|
|
ss << "Syntax error in liberty file on line " << line << ".\n";
|
|
ss << " " << str << "\n";
|
|
printf("%s", ss.str().c_str());
|
|
exit(1);
|
|
}
|
|
|
|
/**** BEGIN: http://svn.clairexen.net/tools/trunk/examples/check.h ****/
|
|
|
|
#define CHECK_NV(result, check) \
|
|
do { \
|
|
auto _R = (result); \
|
|
if (!(_R check)) { \
|
|
fprintf(stderr, "Error from '%s' (%ld %s) in %s:%d.\n", \
|
|
#result, (long int)_R, #check, __FILE__, __LINE__); \
|
|
abort(); \
|
|
} \
|
|
} while(0)
|
|
|
|
#define CHECK_COND(result) \
|
|
do { \
|
|
if (!(result)) { \
|
|
fprintf(stderr, "Error from '%s' in %s:%d.\n", \
|
|
#result, __FILE__, __LINE__); \
|
|
abort(); \
|
|
} \
|
|
} while(0)
|
|
|
|
/**** END: http://svn.clairexen.net/tools/trunk/examples/check.h ****/
|
|
|
|
const LibertyAst *find_non_null(const LibertyAst *node, const char *name)
|
|
{
|
|
const LibertyAst *ret = node->find(name);
|
|
if (ret == NULL)
|
|
fprintf(stderr, "Error: expected to find `%s' node.\n", name);
|
|
return ret;
|
|
}
|
|
|
|
std::string func2vl(std::string str)
|
|
{
|
|
for (size_t pos = str.find_first_of("\" \t"); pos != std::string::npos; pos = str.find_first_of("\" \t")) {
|
|
char c_left = pos > 0 ? str[pos-1] : ' ';
|
|
char c_right = pos+1 < str.size() ? str[pos+1] : ' ';
|
|
if (std::string("\" \t*+").find(c_left) != std::string::npos)
|
|
str.erase(pos, 1);
|
|
else if (std::string("\" \t*+").find(c_right) != std::string::npos)
|
|
str.erase(pos, 1);
|
|
else
|
|
str[pos] = '*';
|
|
}
|
|
|
|
std::vector<size_t> group_start;
|
|
for (size_t pos = 0; pos < str.size(); pos++) {
|
|
if (str[pos] == '(')
|
|
group_start.push_back(pos);
|
|
if (str[pos] == ')' && group_start.size() > 0) {
|
|
if (pos+1 < str.size() && str[pos+1] == '\'') {
|
|
std::string group = str.substr(group_start.back(), pos-group_start.back()+1);
|
|
str[group_start.back()] = '~';
|
|
str.replace(group_start.back()+1, group.size(), group);
|
|
pos++;
|
|
}
|
|
group_start.pop_back();
|
|
}
|
|
if (str[pos] == '\'' && pos > 0) {
|
|
size_t start = str.find_last_of("()'*+^&| ", pos-1)+1;
|
|
std::string group = str.substr(start, pos-start);
|
|
str[start] = '~';
|
|
str.replace(start+1, group.size(), group);
|
|
}
|
|
if (str[pos] == '*')
|
|
str[pos] = '&';
|
|
if (str[pos] == '+')
|
|
str[pos] = '|';
|
|
}
|
|
|
|
return str;
|
|
}
|
|
|
|
void event2vl(const LibertyAst *ast, std::string &edge, std::string &expr)
|
|
{
|
|
edge.clear();
|
|
expr.clear();
|
|
|
|
if (ast != NULL) {
|
|
expr = func2vl(ast->value);
|
|
if (expr.size() > 0 && expr[0] == '~')
|
|
edge = "negedge " + expr.substr(1);
|
|
else
|
|
edge = "posedge " + expr;
|
|
}
|
|
}
|
|
|
|
void clear_preset_var(std::string var, std::string type)
|
|
{
|
|
if (type.find('L') != std::string::npos) {
|
|
printf(" %s <= 0;\n", var.c_str());
|
|
return;
|
|
}
|
|
if (type.find('H') != std::string::npos) {
|
|
printf(" %s <= 1;\n", var.c_str());
|
|
return;
|
|
}
|
|
if (type.find('T') != std::string::npos) {
|
|
printf(" %s <= ~%s;\n", var.c_str(), var.c_str());
|
|
return;
|
|
}
|
|
if (type.find('X') != std::string::npos) {
|
|
printf(" %s <= 'bx;\n", var.c_str());
|
|
return;
|
|
}
|
|
}
|
|
|
|
void gen_verilogsim_cell(const LibertyAst *ast)
|
|
{
|
|
if (ast->find("statetable") != NULL)
|
|
return;
|
|
|
|
CHECK_NV(ast->args.size(), == 1);
|
|
printf("module %s (", ast->args[0].c_str());
|
|
bool first = true;
|
|
for (auto child : ast->children) {
|
|
if (child->id != "pin")
|
|
continue;
|
|
CHECK_NV(child->args.size(), == 1);
|
|
printf("%s%s", first ? "" : ", ", child->args[0].c_str());
|
|
first = false;
|
|
}
|
|
printf(");\n");
|
|
|
|
for (auto child : ast->children) {
|
|
if (child->id != "ff" && child->id != "latch")
|
|
continue;
|
|
printf(" reg ");
|
|
first = true;
|
|
for (auto arg : child->args) {
|
|
printf("%s%s", first ? "" : ", ", arg.c_str());
|
|
first = false;
|
|
}
|
|
printf(";\n");
|
|
}
|
|
|
|
for (auto child : ast->children) {
|
|
if (child->id != "pin")
|
|
continue;
|
|
CHECK_NV(child->args.size(), == 1);
|
|
const LibertyAst *dir = find_non_null(child, "direction");
|
|
const LibertyAst *func = child->find("function");
|
|
printf(" %s %s;\n", dir->value.c_str(), child->args[0].c_str());
|
|
if (func != NULL)
|
|
printf(" assign %s = %s; // %s\n", child->args[0].c_str(), func2vl(func->value).c_str(), func->value.c_str());
|
|
}
|
|
|
|
for (auto child : ast->children)
|
|
{
|
|
if (child->id != "ff" || child->args.size() != 2)
|
|
continue;
|
|
|
|
std::string iq_var = child->args[0];
|
|
std::string iqn_var = child->args[1];
|
|
|
|
std::string clock_edge, clock_expr;
|
|
event2vl(child->find("clocked_on"), clock_edge, clock_expr);
|
|
|
|
std::string clear_edge, clear_expr;
|
|
event2vl(child->find("clear"), clear_edge, clear_expr);
|
|
|
|
std::string preset_edge, preset_expr;
|
|
event2vl(child->find("preset"), preset_edge, preset_expr);
|
|
|
|
std::string edge = "";
|
|
if (!clock_edge.empty())
|
|
edge += (edge.empty() ? "" : ", ") + clock_edge;
|
|
if (!clear_edge.empty())
|
|
edge += (edge.empty() ? "" : ", ") + clear_edge;
|
|
if (!preset_edge.empty())
|
|
edge += (edge.empty() ? "" : ", ") + preset_edge;
|
|
|
|
if (edge.empty())
|
|
continue;
|
|
|
|
printf(" always @(%s) begin\n", edge.c_str());
|
|
|
|
const char *else_prefix = "";
|
|
if (!clear_expr.empty() && !preset_expr.empty()) {
|
|
printf(" %sif ((%s) && (%s)) begin\n", else_prefix, clear_expr.c_str(), preset_expr.c_str());
|
|
clear_preset_var(iq_var, find_non_null(child, "clear_preset_var1")->value);
|
|
clear_preset_var(iqn_var, find_non_null(child, "clear_preset_var2")->value);
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (!clear_expr.empty()) {
|
|
printf(" %sif (%s) begin\n", else_prefix, clear_expr.c_str());
|
|
printf(" %s <= 0;\n", iq_var.c_str());
|
|
printf(" %s <= 1;\n", iqn_var.c_str());
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (!preset_expr.empty()) {
|
|
printf(" %sif (%s) begin\n", else_prefix, preset_expr.c_str());
|
|
printf(" %s <= 1;\n", iq_var.c_str());
|
|
printf(" %s <= 0;\n", iqn_var.c_str());
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (*else_prefix)
|
|
printf(" %sbegin\n", else_prefix);
|
|
std::string expr = find_non_null(child, "next_state")->value;
|
|
printf(" // %s\n", expr.c_str());
|
|
printf(" %s <= %s;\n", iq_var.c_str(), func2vl(expr).c_str());
|
|
printf(" %s <= ~(%s);\n", iqn_var.c_str(), func2vl(expr).c_str());
|
|
if (*else_prefix)
|
|
printf(" end\n");
|
|
|
|
printf(" end\n");
|
|
}
|
|
|
|
for (auto child : ast->children)
|
|
{
|
|
if (child->id != "latch" || child->args.size() != 2)
|
|
continue;
|
|
|
|
std::string iq_var = child->args[0];
|
|
std::string iqn_var = child->args[1];
|
|
|
|
std::string enable_edge, enable_expr;
|
|
event2vl(child->find("enable"), enable_edge, enable_expr);
|
|
|
|
std::string clear_edge, clear_expr;
|
|
event2vl(child->find("clear"), clear_edge, clear_expr);
|
|
|
|
std::string preset_edge, preset_expr;
|
|
event2vl(child->find("preset"), preset_edge, preset_expr);
|
|
|
|
printf(" always @* begin\n");
|
|
|
|
const char *else_prefix = "";
|
|
if (!clear_expr.empty() && !preset_expr.empty()) {
|
|
printf(" %sif ((%s) && (%s)) begin\n", else_prefix, clear_expr.c_str(), preset_expr.c_str());
|
|
clear_preset_var(iq_var, find_non_null(child, "clear_preset_var1")->value);
|
|
clear_preset_var(iqn_var, find_non_null(child, "clear_preset_var2")->value);
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (!clear_expr.empty()) {
|
|
printf(" %sif (%s) begin\n", else_prefix, clear_expr.c_str());
|
|
printf(" %s <= 0;\n", iq_var.c_str());
|
|
printf(" %s <= 1;\n", iqn_var.c_str());
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (!preset_expr.empty()) {
|
|
printf(" %sif (%s) begin\n", else_prefix, preset_expr.c_str());
|
|
printf(" %s <= 1;\n", iq_var.c_str());
|
|
printf(" %s <= 0;\n", iqn_var.c_str());
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
if (!enable_expr.empty()) {
|
|
printf(" %sif (%s) begin\n", else_prefix, enable_expr.c_str());
|
|
std::string expr = find_non_null(child, "data_in")->value;
|
|
printf(" %s <= %s;\n", iq_var.c_str(), func2vl(expr).c_str());
|
|
printf(" %s <= ~(%s);\n", iqn_var.c_str(), func2vl(expr).c_str());
|
|
printf(" end\n");
|
|
else_prefix = "else ";
|
|
}
|
|
|
|
printf(" end\n");
|
|
}
|
|
|
|
printf("endmodule\n");
|
|
}
|
|
|
|
void gen_verilogsim(const LibertyAst *ast)
|
|
{
|
|
CHECK_COND(ast->id == "library");
|
|
|
|
for (auto child : ast->children)
|
|
if (child->id == "cell" && !child->find("dont_use"))
|
|
gen_verilogsim_cell(child);
|
|
}
|
|
|
|
void usage()
|
|
{
|
|
fprintf(stderr, "Usage: filterlib [rules-file [liberty-file]]\n");
|
|
fprintf(stderr, " or: filterlib -verilogsim [liberty-file]\n");
|
|
exit(1);
|
|
}
|
|
|
|
int main(int argc, char **argv)
|
|
{
|
|
bool flag_verilogsim = false;
|
|
std::set<std::string> whitelist, blacklist;
|
|
|
|
if (argc > 3)
|
|
usage();
|
|
|
|
if (argc > 1)
|
|
{
|
|
if (!strcmp(argv[1], "-verilogsim"))
|
|
flag_verilogsim = true;
|
|
if (!strcmp(argv[1], "-") || !strcmp(argv[1], "-verilogsim"))
|
|
{
|
|
whitelist.insert("/library");
|
|
whitelist.insert("/library/cell");
|
|
whitelist.insert("/library/cell/area");
|
|
whitelist.insert("/library/cell/cell_footprint");
|
|
whitelist.insert("/library/cell/dont_touch");
|
|
whitelist.insert("/library/cell/dont_use");
|
|
whitelist.insert("/library/cell/ff");
|
|
whitelist.insert("/library/cell/ff/*");
|
|
whitelist.insert("/library/cell/latch");
|
|
whitelist.insert("/library/cell/latch/*");
|
|
whitelist.insert("/library/cell/pin");
|
|
whitelist.insert("/library/cell/pin/clock");
|
|
whitelist.insert("/library/cell/pin/direction");
|
|
whitelist.insert("/library/cell/pin/driver_type");
|
|
whitelist.insert("/library/cell/pin/function");
|
|
whitelist.insert("/library/cell/pin_opposite");
|
|
whitelist.insert("/library/cell/pin/state_function");
|
|
whitelist.insert("/library/cell/pin/three_state");
|
|
whitelist.insert("/library/cell/statetable");
|
|
whitelist.insert("/library/cell/statetable/*");
|
|
}
|
|
else
|
|
{
|
|
FILE *f = fopen(argv[1], "r");
|
|
if (f == NULL) {
|
|
fprintf(stderr, "Can't open rules file `%s'.\n", argv[1]);
|
|
usage();
|
|
}
|
|
|
|
char buffer[1024];
|
|
while (fgets(buffer, 1024, f) != NULL)
|
|
{
|
|
char mode = 0;
|
|
std::string id;
|
|
for (char *p = buffer; *p; p++)
|
|
{
|
|
if (*p == '-' || *p == '+') {
|
|
if (mode != 0)
|
|
goto syntax_error;
|
|
mode = *p;
|
|
continue;
|
|
}
|
|
if (*p == ' ' || *p == '\t' || *p == '\r' || *p == '\n' || *p == '#') {
|
|
if (!id.empty()) {
|
|
if (mode == '-')
|
|
blacklist.insert(id);
|
|
else
|
|
if (mode == '+')
|
|
whitelist.insert(id);
|
|
else
|
|
goto syntax_error;
|
|
}
|
|
id.clear();
|
|
if (*p == '#')
|
|
break;
|
|
continue;
|
|
}
|
|
id += *p;
|
|
continue;
|
|
|
|
syntax_error:
|
|
fprintf(stderr, "Syntax error in rules file:\n%s", buffer);
|
|
exit(1);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
std::istream *f = &std::cin;
|
|
|
|
if (argc == 3) {
|
|
std::ifstream *ff = new std::ifstream;
|
|
ff->open(argv[2]);
|
|
if (ff->fail()) {
|
|
delete ff;
|
|
fprintf(stderr, "Can't open liberty file `%s'.\n", argv[2]);
|
|
usage();
|
|
}
|
|
f = ff;
|
|
}
|
|
|
|
LibertyParser parser(*f);
|
|
if (parser.ast) {
|
|
if (flag_verilogsim)
|
|
gen_verilogsim(parser.ast);
|
|
else
|
|
parser.ast->dump(stdout, blacklist, whitelist);
|
|
}
|
|
|
|
if (argc == 3)
|
|
delete f;
|
|
|
|
return 0;
|
|
}
|
|
|
|
#endif
|
|
|