verilator/include/verilated_imp.h
2023-01-01 10:18:39 -05:00

586 lines
23 KiB
C++

// -*- mode: C++; c-file-style: "cc-mode" -*-
//*************************************************************************
//
// Code available from: https://verilator.org
//
// Copyright 2009-2023 by Wilson Snyder. This program is free software; you can
// redistribute it and/or modify it under the terms of either the GNU
// Lesser General Public License Version 3 or the Perl Artistic License
// Version 2.0.
// SPDX-License-Identifier: LGPL-3.0-only OR Artistic-2.0
//
//=========================================================================
///
/// \file
/// \brief Verilated implementation Header, only for verilated.cpp internals.
///
/// This file is not part of the Verilated public-facing API.
/// It is only for internal use by the Verilated libraries.
///
//=========================================================================
#ifndef VERILATOR_VERILATED_IMP_H_
#define VERILATOR_VERILATED_IMP_H_
// clang-format off
#if !defined(VERILATOR_VERILATED_CPP_) && !defined(VERILATOR_VERILATED_DPI_CPP_) \
&& !defined(VERILATOR_VERILATED_VPI_CPP_) && !defined(VERILATOR_VERILATED_SAVE_CPP_)
# error "verilated_imp.h only to be included by verilated*.cpp internals"
#endif
#include "verilatedos.h"
#include "verilated.h"
#include "verilated_syms.h"
#include <algorithm>
#include <deque>
#include <limits>
#include <map>
#include <numeric>
#include <set>
#include <string>
#include <utility>
#include <vector>
#include <functional>
#include <queue>
// clang-format on
class VerilatedScope;
//======================================================================
// Threaded message passing
// Message, enqueued on an mtask, and consumed on the main eval thread
class VerilatedMsg final {
public:
// TYPES
struct Cmp {
bool operator()(const VerilatedMsg& a, const VerilatedMsg& b) const {
return a.mtaskId() < b.mtaskId();
}
};
private:
// MEMBERS
uint32_t m_mtaskId; // MTask that did enqueue
std::function<void()> m_cb; // Lambda to execute when message received
public:
// CONSTRUCTORS
explicit VerilatedMsg(const std::function<void()>& cb)
: m_mtaskId{Verilated::mtaskId()}
, m_cb{cb} {}
~VerilatedMsg() = default;
VerilatedMsg(const VerilatedMsg&) = default;
VerilatedMsg(VerilatedMsg&&) = default;
VerilatedMsg& operator=(const VerilatedMsg&) = default;
VerilatedMsg& operator=(VerilatedMsg&&) = default;
// METHODS
uint32_t mtaskId() const { return m_mtaskId; }
// Execute the lambda function
void run() const { m_cb(); }
};
// Each thread has a queue it pushes to
// This assumes no thread starts pushing the next tick until the previous has drained.
// If more aggressiveness is needed, a double-buffered scheme might work well.
class VerilatedEvalMsgQueue final {
using VerilatedThreadQueue = std::multiset<VerilatedMsg, VerilatedMsg::Cmp>;
std::atomic<uint64_t> m_depth; // Current depth of queue (see comments below)
mutable VerilatedMutex m_mutex; // Mutex protecting queue
VerilatedThreadQueue m_queue VL_GUARDED_BY(m_mutex); // Message queue
public:
// CONSTRUCTORS
VerilatedEvalMsgQueue()
: m_depth{0} {
assert(atomic_is_lock_free(&m_depth));
}
~VerilatedEvalMsgQueue() = default;
private:
VL_UNCOPYABLE(VerilatedEvalMsgQueue);
public:
// METHODS
// Add message to queue (called by producer)
void post(const VerilatedMsg& msg) VL_MT_SAFE_EXCLUDES(m_mutex) {
const VerilatedLockGuard lock{m_mutex};
m_queue.insert(msg); // Pass by value to copy the message into queue
++m_depth;
}
// Service queue until completion (called by consumer)
void process() VL_MT_SAFE_EXCLUDES(m_mutex) {
// Tracking m_depth is redundant to e.g. getting the mutex and looking at queue size,
// but on the reader side it's 4x faster to test an atomic then getting a mutex
while (m_depth) {
// Wait for a message to be added to the queue
// We don't use unique_lock as want to unlock with the message copy still in scope
m_mutex.lock();
assert(!m_queue.empty()); // Otherwise m_depth is wrong
// Unfortunately to release the lock we need to copy the message
// (Or have the message be a pointer, but then new/delete cost on each message)
// We assume messages are small, so copy
const auto it = m_queue.begin();
const VerilatedMsg msg = *(it);
m_queue.erase(it);
m_mutex.unlock();
--m_depth; // Ok if outside critical section as only this code checks the value
{
VL_DEBUG_IF(VL_DBG_MSGF("Executing callback from mtaskId=%d\n", msg.mtaskId()););
msg.run();
}
}
}
};
// Each thread has a local queue to build up messages until the end of the eval() call
class VerilatedThreadMsgQueue final {
std::queue<VerilatedMsg> m_queue;
public:
// CONSTRUCTORS
VerilatedThreadMsgQueue() = default;
~VerilatedThreadMsgQueue() = default;
// The only call of destructor with a non-empty queue is a fatal error.
// So this does not flush the queue, as the destination queue is not known to this class.
private:
VL_UNCOPYABLE(VerilatedThreadMsgQueue);
// METHODS
static VerilatedThreadMsgQueue& threadton() {
static thread_local VerilatedThreadMsgQueue t_s;
return t_s;
}
public:
// Add message to queue, called by producer
static void post(const VerilatedMsg& msg) VL_MT_SAFE {
// Handle calls to threaded routines outside
// of any mtask -- if an initial block calls $finish, say.
if (Verilated::mtaskId() == 0) {
// No queueing, just do the action immediately
msg.run();
} else {
Verilated::endOfEvalReqdInc();
threadton().m_queue.push(msg); // Pass by value to copy the message into queue
}
}
// Push all messages to the eval's queue
static void flush(VerilatedEvalMsgQueue* evalMsgQp) VL_MT_SAFE {
while (!threadton().m_queue.empty()) {
evalMsgQp->post(threadton().m_queue.front());
threadton().m_queue.pop();
Verilated::endOfEvalReqdDec();
}
}
};
// FILE* list constructed from a file-descriptor
class VerilatedFpList final {
FILE* m_fp[31] = {};
std::size_t m_sz = 0;
public:
using const_iterator = FILE* const*;
explicit VerilatedFpList() = default;
const_iterator begin() const { return m_fp; }
const_iterator end() const { return m_fp + m_sz; }
std::size_t size() const { return m_sz; }
static std::size_t capacity() { return 31; }
void push_back(FILE* fd) {
if (VL_LIKELY(size() < capacity())) m_fp[m_sz++] = fd;
}
};
//======================================================================
// VerilatedContextImpData
// Class for hidden implementation members inside VerilatedContext
// Avoids needing std::unordered_map inside verilated.h
class VerilatedContextImpData final {
friend class VerilatedContext;
friend class VerilatedContextImp;
protected:
// Map of <scope_name, scope pointer>
// Used by scopeInsert, scopeFind, scopeErase, scopeNameMap
mutable VerilatedMutex m_nameMutex; // Protect m_nameMap
VerilatedScopeNameMap m_nameMap VL_GUARDED_BY(m_nameMutex);
};
//======================================================================
// VerilatedContextImp
// Class to "add" implementation-only methods to VerilatedContext
class VerilatedContextImp final : VerilatedContext {
friend class VerilatedContext;
// MEMBERS - non-static not allowed, use only VerilatedContext
// Select initial value of otherwise uninitialized signals.
// Internal note: Globals may multi-construct, see verilated.cpp top.
// Medium speed, so uses singleton accessing
struct Statics {
VerilatedMutex s_randMutex; // Mutex protecting s_randSeedEpoch
// Number incrementing on each reseed, 0=illegal
int s_randSeedEpoch = 1; // Reads ok, wish had a VL_WRITE_GUARDED_BY(s_randMutex)
};
static Statics& s() VL_MT_SAFE {
static Statics s_s;
return s_s;
}
public: // But only for verilated*.cpp
// CONSTRUCTORS - no data can live here, use only VerilatedContext
VerilatedContextImp() = delete;
~VerilatedContextImp() = delete;
// METHODS - extending into VerilatedContext, call via impp()->
// Random seed handling
uint64_t randSeedDefault64() const VL_MT_SAFE;
static uint32_t randSeedEpoch() VL_MT_SAFE { return s().s_randSeedEpoch; }
// METHODS - timeformat
int timeFormatUnits() const VL_MT_SAFE {
if (m_s.m_timeFormatUnits == VerilatedContext::Serialized::UNITS_NONE)
return timeprecision();
return m_s.m_timeFormatUnits;
}
void timeFormatUnits(int value) VL_MT_SAFE { m_s.m_timeFormatUnits = value; }
int timeFormatPrecision() const VL_MT_SAFE { return m_s.m_timeFormatPrecision; }
void timeFormatPrecision(int value) VL_MT_SAFE { m_s.m_timeFormatPrecision = value; }
int timeFormatWidth() const VL_MT_SAFE { return m_s.m_timeFormatWidth; }
void timeFormatWidth(int value) VL_MT_SAFE { m_s.m_timeFormatWidth = value; }
std::string timeFormatSuffix() const VL_MT_SAFE_EXCLUDES(m_timeDumpMutex) {
const VerilatedLockGuard lock{m_timeDumpMutex};
return m_timeFormatSuffix;
}
void timeFormatSuffix(const std::string& value) VL_MT_SAFE_EXCLUDES(m_timeDumpMutex) {
const VerilatedLockGuard lock{m_timeDumpMutex};
m_timeFormatSuffix = value;
}
// METHODS - arguments
std::string argPlusMatch(const char* prefixp) VL_MT_SAFE_EXCLUDES(m_argMutex);
std::pair<int, char**> argc_argv() VL_MT_SAFE_EXCLUDES(m_argMutex);
// METHODS - scope name - INTERNAL only for verilated*.cpp
void scopeInsert(const VerilatedScope* scopep) VL_MT_SAFE;
void scopeErase(const VerilatedScope* scopep) VL_MT_SAFE;
// METHODS - file IO - INTERNAL only for verilated*.cpp
IData fdNewMcd(const char* filenamep) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
if (m_fdFreeMct.empty()) return 0;
const IData idx = m_fdFreeMct.back();
m_fdFreeMct.pop_back();
m_fdps[idx] = std::fopen(filenamep, "w");
if (VL_UNLIKELY(!m_fdps[idx])) return 0;
return (1 << idx);
}
IData fdNew(const char* filenamep, const char* modep) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
FILE* const fp = std::fopen(filenamep, modep);
if (VL_UNLIKELY(!fp)) return 0;
// Bit 31 indicates it's a descriptor not a MCD
const VerilatedLockGuard lock{m_fdMutex};
if (m_fdFree.empty()) {
// Need to create more space in m_fdps and m_fdFree
const std::size_t start = std::max<std::size_t>(31UL + 1UL + 3UL, m_fdps.size());
const std::size_t excess = 10;
m_fdps.resize(start + excess);
std::fill(m_fdps.begin() + start, m_fdps.end(), static_cast<FILE*>(nullptr));
m_fdFree.resize(excess);
for (std::size_t i = 0, id = start; i < m_fdFree.size(); ++i, ++id) {
m_fdFree[i] = id;
}
}
const IData idx = m_fdFree.back();
m_fdFree.pop_back();
m_fdps[idx] = fp;
return (idx | (1UL << 31)); // bit 31 indicates not MCD
}
void fdFlush(IData fdi) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
const VerilatedFpList fdlist = fdToFpList(fdi);
for (const auto& i : fdlist) std::fflush(i);
}
IData fdSeek(IData fdi, IData offset, IData origin) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
const VerilatedFpList fdlist = fdToFpList(fdi);
if (VL_UNLIKELY(fdlist.size() != 1)) return ~0U; // -1
return static_cast<IData>(
std::fseek(*fdlist.begin(), static_cast<long>(offset), static_cast<int>(origin)));
}
IData fdTell(IData fdi) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
const VerilatedFpList fdlist = fdToFpList(fdi);
if (VL_UNLIKELY(fdlist.size() != 1)) return ~0U; // -1
return static_cast<IData>(std::ftell(*fdlist.begin()));
}
void fdWrite(IData fdi, const std::string& output) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
const VerilatedFpList fdlist = fdToFpList(fdi);
for (const auto& i : fdlist) {
if (VL_UNLIKELY(!i)) continue;
(void)fwrite(output.c_str(), 1, output.size(), i);
}
}
void fdClose(IData fdi) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
if (VL_BITISSET_I(fdi, 31)) {
// Non-MCD case
const IData idx = VL_MASK_I(31) & fdi;
if (VL_UNLIKELY(idx >= m_fdps.size())) return;
if (VL_UNLIKELY(idx <= 2)) return; // stdout/stdin/stderr
if (VL_UNLIKELY(!m_fdps[idx])) return; // Already free
std::fclose(m_fdps[idx]);
m_fdps[idx] = nullptr;
m_fdFree.push_back(idx);
} else {
// MCD case
// Starts at 1 to skip stdout
fdi >>= 1;
for (int i = 1; (fdi != 0) && (i < 31); i++, fdi >>= 1) {
if (fdi & VL_MASK_I(1)) {
std::fclose(m_fdps[i]);
m_fdps[i] = nullptr;
m_fdFreeMct.push_back(i);
}
}
}
}
FILE* fdToFp(IData fdi) VL_MT_SAFE_EXCLUDES(m_fdMutex) {
const VerilatedLockGuard lock{m_fdMutex};
const VerilatedFpList fdlist = fdToFpList(fdi);
if (VL_UNLIKELY(fdlist.size() != 1)) return nullptr;
return *fdlist.begin();
}
private:
VerilatedFpList fdToFpList(IData fdi) VL_REQUIRES(m_fdMutex) {
VerilatedFpList fp;
// cppverilator-suppress integerOverflow shiftTooManyBitsSigned
if ((fdi & (1 << 31)) != 0) {
// Non-MCD case
const IData idx = fdi & VL_MASK_I(31);
switch (idx) {
case 0: fp.push_back(stdin); break;
case 1: fp.push_back(stdout); break;
case 2: fp.push_back(stderr); break;
default:
if (VL_LIKELY(idx < m_fdps.size())) fp.push_back(m_fdps[idx]);
break;
}
} else {
// MCD Case
if (fdi & 1) fp.push_back(stdout);
fdi >>= 1;
for (size_t i = 1; (fdi != 0) && (i < fp.capacity()); ++i, fdi >>= 1) {
if (fdi & VL_MASK_I(1)) fp.push_back(m_fdps[i]);
}
}
return fp;
}
protected:
// METHODS - protected
void commandArgsGuts(int argc, const char** argv) VL_MT_SAFE_EXCLUDES(m_argMutex);
void commandArgsAddGutsLock(int argc, const char** argv) VL_MT_SAFE_EXCLUDES(m_argMutex);
void commandArgsAddGuts(int argc, const char** argv) VL_REQUIRES(m_argMutex);
void commandArgVl(const std::string& arg);
bool commandArgVlString(const std::string& arg, const std::string& prefix,
std::string& valuer);
bool commandArgVlUint64(const std::string& arg, const std::string& prefix, uint64_t& valuer,
uint64_t min = std::numeric_limits<uint64_t>::min(),
uint64_t max = std::numeric_limits<uint64_t>::max());
void commandArgDump() const VL_MT_SAFE_EXCLUDES(m_argMutex);
};
//======================================================================
// VerilatedImp
class VerilatedImpData final {
// Whole class is internal use only - Global information shared between verilated*.cpp files.
// All only medium-speed, so use singleton function
protected:
friend class Verilated;
friend class VerilatedImp;
// TYPES
using UserMap = std::map<std::pair<const void*, void*>, void*>;
using ExportNameMap = std::map<const char*, int, VerilatedCStrCmp>;
// MEMBERS
// Nothing below here is save-restored; users expected to re-register appropriately
VerilatedMutex m_userMapMutex; // Protect m_userMap
// For userInsert, userFind. As indexed by pointer is common across contexts.
UserMap m_userMap VL_GUARDED_BY(m_userMapMutex); // Map of <(scope,userkey), userData>
VerilatedMutex m_hierMapMutex; // Protect m_hierMap
// Map that represents scope hierarchy
// Used by hierarchyAdd, hierarchyRemove, hierarchyMap
VerilatedHierarchyMap m_hierMap VL_GUARDED_BY(m_hierMapMutex);
// Slow - somewhat static:
VerilatedMutex m_exportMutex; // Protect m_nameMap
// Map of <export_func_proto, func number>
// Used by exportInsert, exportFind, exportName.
// Export numbers same across all contexts as just a string-to-number conversion
ExportNameMap m_exportMap VL_GUARDED_BY(m_exportMutex);
int m_exportNext VL_GUARDED_BY(m_exportMutex) = 0; // Next export funcnum
// CONSTRUCTORS
VerilatedImpData() = default;
};
class VerilatedImp final {
// Whole class is internal use only - Global information shared between verilated*.cpp files.
protected:
friend class Verilated;
// MEMBERS
static VerilatedImpData& s() VL_MT_SAFE { // Singleton
static VerilatedImpData s_s;
return s_s;
}
public: // But only for verilated*.cpp
// CONSTRUCTORS
VerilatedImp() = default;
~VerilatedImp() = default;
private:
VL_UNCOPYABLE(VerilatedImp);
public:
// METHODS - debug
static void versionDump() VL_MT_SAFE;
// METHODS - user scope tracking
// We implement this as a single large map instead of one map per scope.
// There's often many more scopes than userdata's and thus having a ~48byte
// per map overhead * N scopes would take much more space and cache thrashing.
// As scopep's are pointers, this implicitly handles multiple Context's
static void userInsert(const void* scopep, void* userKey, void* userData) VL_MT_SAFE {
const VerilatedLockGuard lock{s().m_userMapMutex};
const auto it = s().m_userMap.find(std::make_pair(scopep, userKey));
if (it != s().m_userMap.end()) {
it->second = userData;
} else {
s().m_userMap.emplace(std::make_pair(scopep, userKey), userData);
}
}
static void* userFind(const void* scopep, void* userKey) VL_MT_SAFE {
const VerilatedLockGuard lock{s().m_userMapMutex};
const auto& it = vlstd::as_const(s().m_userMap).find(std::make_pair(scopep, userKey));
if (VL_UNLIKELY(it == s().m_userMap.end())) return nullptr;
return it->second;
}
// METHODS - But only for verilated.cpp
// Symbol table destruction cleans up the entries for each scope.
static void userEraseScope(const VerilatedScope* scopep) VL_MT_SAFE {
// Slow ok - called once/scope on destruction, so we only iterate.
const VerilatedLockGuard lock{s().m_userMapMutex};
for (auto it = s().m_userMap.begin(); it != s().m_userMap.end();) {
if (it->first.first == scopep) {
s().m_userMap.erase(it++);
} else {
++it;
}
}
}
static void userDump() VL_MT_SAFE {
const VerilatedLockGuard lock{s().m_userMapMutex}; // Avoid it changing in middle of dump
bool first = true;
for (const auto& i : s().m_userMap) {
if (first) {
VL_PRINTF_MT(" userDump:\n");
first = false;
}
VL_PRINTF_MT(" DPI_USER_DATA scope %p key %p: %p\n", i.first.first, i.first.second,
i.second);
}
}
// METHODS - hierarchy - only for verilated*.cpp
static void hierarchyAdd(const VerilatedScope* fromp, const VerilatedScope* top) VL_MT_SAFE {
// Slow ok - called at construction for VPI accessible elements
const VerilatedLockGuard lock{s().m_hierMapMutex};
s().m_hierMap[fromp].push_back(top);
}
static void hierarchyRemove(const VerilatedScope* fromp,
const VerilatedScope* top) VL_MT_SAFE {
// Slow ok - called at destruction for VPI accessible elements
const VerilatedLockGuard lock{s().m_hierMapMutex};
VerilatedHierarchyMap& map = s().m_hierMap;
if (map.find(fromp) == map.end()) return;
auto& scopes = map[fromp];
const auto it = find(scopes.begin(), scopes.end(), top);
if (it != scopes.end()) scopes.erase(it);
}
static const VerilatedHierarchyMap* hierarchyMap() VL_MT_SAFE_POSTINIT {
// Thread save only assuming this is called only after model construction completed
return &s().m_hierMap;
}
// METHODS - export names - only for verilated*.cpp
// Each function prototype is converted to a function number which we
// then use to index a 2D table also indexed by scope number, because we
// can't know at Verilation time what scopes will exist in other modules
// in the design that also happen to have our same callback function.
// Rather than a 2D map, the integer scheme saves 500ish ns on a likely
// miss at the cost of a multiply, and all lookups move to slowpath.
static int exportInsert(const char* namep) VL_MT_SAFE {
// Slow ok - called once/function at creation
const VerilatedLockGuard lock{s().m_exportMutex};
const auto it = s().m_exportMap.find(namep);
if (it == s().m_exportMap.end()) {
s().m_exportMap.emplace(namep, s().m_exportNext++);
return s().m_exportNext++;
} else {
return it->second;
}
}
static int exportFind(const char* namep) VL_MT_SAFE {
const VerilatedLockGuard lock{s().m_exportMutex};
const auto& it = s().m_exportMap.find(namep);
if (VL_LIKELY(it != s().m_exportMap.end())) return it->second;
const std::string msg = (std::string{"%Error: Testbench C called "} + namep
+ " but no such DPI export function name exists in ANY model");
VL_FATAL_MT("unknown", 0, "", msg.c_str());
return -1;
}
static const char* exportName(int funcnum) VL_MT_SAFE {
// Slowpath; find name for given export; errors only so no map to reverse-map it
const VerilatedLockGuard lock{s().m_exportMutex};
for (const auto& i : s().m_exportMap) {
if (i.second == funcnum) return i.first;
}
return "*UNKNOWN*";
}
static void exportsDump() VL_MT_SAFE {
const VerilatedLockGuard lock{s().m_exportMutex};
bool first = true;
for (const auto& i : s().m_exportMap) {
if (first) {
VL_PRINTF_MT(" exportDump:\n");
first = false;
}
VL_PRINTF_MT(" DPI_EXPORT_NAME %05d: %s\n", i.second, i.first);
}
}
// We don't free up m_exportMap until the end, because we can't be sure
// what other models are using the assigned funcnum's.
};
//======================================================================
#endif // Guard