verilator/src/V3Sched.cpp

1064 lines
46 KiB
C++
Raw Normal View History

// -*- mode: C++; c-file-style: "cc-mode" -*-
//*************************************************************************
// DESCRIPTION: Verilator: Code scheduling
//
// Code available from: https://verilator.org
//
//*************************************************************************
//
// Copyright 2003-2022 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
//
//*************************************************************************
//
// V3Sched::schedule is the top level entry-point to the scheduling algorithm
// at a high level, the process is:
//
// - Gather and classify all logic in the design based on what triggers its execution
// - Schedule static, initial and final logic classes in source order
// - Break combinational cycles by introducing hybrid logic
// - Create 'settle' region that restores the combinational invariant
// - Partition the clocked and combinational (including hybrid) logic into pre/act/nba.
// All clocks (signals referenced in an AstSenTree) generated via a blocking assignment
// (including combinationally generated signals) are computed within the act region.
// - Replicate combinational logic
// - Create input combinational logic loop
// - Create the pre/act/nba triggers
// - Create the 'act' region evaluation function
// - Create the 'nba' region evaluation function
// - Bolt it all together to create the '_eval' function
//
// Details of the algorithm are described in the internals documentation docs/internals.rst
//
//*************************************************************************
#include "config_build.h"
#include "verilatedos.h"
2022-08-05 13:15:59 +02:00
#include "V3Sched.h"
#include "V3Ast.h"
#include "V3EmitCBase.h"
#include "V3EmitV.h"
#include "V3Order.h"
#include "V3SenExprBuilder.h"
#include "V3Stats.h"
2022-09-22 18:28:42 +02:00
VL_DEFINE_DEBUG_FUNCTIONS;
namespace V3Sched {
namespace {
//============================================================================
// Utility functions
AstCFunc* makeSubFunction(AstNetlist* netlistp, const string& name, bool slow) {
AstScope* const scopeTopp = netlistp->topScopep()->scopep();
AstCFunc* const funcp = new AstCFunc{netlistp->fileline(), name, scopeTopp, ""};
funcp->dontCombine(true);
funcp->isStatic(false);
funcp->isLoose(true);
funcp->slow(slow);
funcp->isConst(false);
funcp->declPrivate(true);
scopeTopp->addBlocksp(funcp);
return funcp;
}
AstCFunc* makeTopFunction(AstNetlist* netlistp, const string& name, bool slow) {
AstCFunc* const funcp = makeSubFunction(netlistp, name, slow);
funcp->entryPoint(true);
return funcp;
}
std::vector<const AstSenTree*> getSenTreesUsedBy(const std::vector<const LogicByScope*>& lbsps) {
const VNUser1InUse user1InUse;
std::vector<const AstSenTree*> result;
for (const LogicByScope* const lbsp : lbsps) {
for (const auto& pair : *lbsp) {
AstActive* const activep = pair.second;
AstSenTree* const senTreep = activep->sensesp();
if (senTreep->user1SetOnce()) continue;
if (senTreep->hasClocked() || senTreep->hasHybrid()) result.push_back(senTreep);
}
}
return result;
}
AstAssign* setVar(AstVarScope* vscp, uint32_t val) {
FileLine* const flp = vscp->fileline();
AstVarRef* const refp = new AstVarRef{flp, vscp, VAccess::WRITE};
2022-08-19 14:25:30 +02:00
AstConst* const valp = new AstConst{flp, AstConst::DTyped{}, vscp->dtypep()};
valp->num().setLong(val);
return new AstAssign{flp, refp, valp};
};
void remapSensitivities(const LogicByScope& lbs,
std::unordered_map<const AstSenTree*, AstSenTree*> senTreeMap) {
for (const auto& pair : lbs) {
AstActive* const activep = pair.second;
AstSenTree* const senTreep = activep->sensesp();
if (senTreep->hasCombo()) continue;
activep->sensesp(senTreeMap.at(senTreep));
}
}
void invertAndMergeSenTreeMap(
std::unordered_map<const AstSenItem*, const AstSenTree*>& result,
const std::unordered_map<const AstSenTree*, AstSenTree*>& senTreeMap) {
for (const auto& pair : senTreeMap) {
UASSERT_OBJ(!pair.second->sensesp()->nextp(), pair.second, "Should be single AstSenIem");
result.emplace(pair.second->sensesp(), pair.first);
}
}
//============================================================================
// Split large function according to --output-split-cfuncs
void splitCheck(AstCFunc* ofuncp) {
if (!v3Global.opt.outputSplitCFuncs() || !ofuncp->stmtsp()) return;
if (ofuncp->nodeCount() < v3Global.opt.outputSplitCFuncs()) return;
int funcnum = 0;
int func_stmts = 0;
AstCFunc* funcp = nullptr;
// Unlink all statements, then add item by item to new sub-functions
AstBegin* const tempp = new AstBegin{ofuncp->fileline(), "[EditWrapper]",
ofuncp->stmtsp()->unlinkFrBackWithNext()};
2022-05-16 21:02:49 +02:00
// Currently we do not use finalsp in V3Sched, if we do, it needs to be handled here
UASSERT_OBJ(!ofuncp->finalsp(), ofuncp, "Should not have any finalps");
while (tempp->stmtsp()) {
AstNode* const itemp = tempp->stmtsp()->unlinkFrBack();
const int stmts = itemp->nodeCount();
if (!funcp || (func_stmts + stmts) > v3Global.opt.outputSplitCFuncs()) {
// Make a new function
funcp = new AstCFunc{ofuncp->fileline(), ofuncp->name() + "__" + cvtToStr(funcnum++),
ofuncp->scopep()};
funcp->dontCombine(true);
funcp->isStatic(false);
funcp->isLoose(true);
funcp->slow(ofuncp->slow());
ofuncp->scopep()->addBlocksp(funcp);
//
AstCCall* const callp = new AstCCall{funcp->fileline(), funcp};
ofuncp->addStmtsp(callp);
func_stmts = 0;
}
funcp->addStmtsp(itemp);
func_stmts += stmts;
}
VL_DO_DANGLING(tempp->deleteTree(), tempp);
}
//============================================================================
// Collect and classify all logic in the design
LogicClasses gatherLogicClasses(AstNetlist* netlistp) {
LogicClasses result;
netlistp->foreach<AstScope>([&](AstScope* scopep) {
std::vector<AstActive*> empty;
scopep->foreach<AstActive>([&](AstActive* activep) {
AstSenTree* const senTreep = activep->sensesp();
if (!activep->stmtsp()) {
// Some AstActives might be empty due to previous optimizations
empty.push_back(activep);
} else if (senTreep->hasStatic()) {
UASSERT_OBJ(!senTreep->sensesp()->nextp(), activep,
"static initializer with additional sensitivities");
result.m_static.emplace_back(scopep, activep);
} else if (senTreep->hasInitial()) {
UASSERT_OBJ(!senTreep->sensesp()->nextp(), activep,
"'initial' logic with additional sensitivities");
result.m_initial.emplace_back(scopep, activep);
} else if (senTreep->hasFinal()) {
UASSERT_OBJ(!senTreep->sensesp()->nextp(), activep,
"'final' logic with additional sensitivities");
result.m_final.emplace_back(scopep, activep);
} else if (senTreep->hasCombo()) {
UASSERT_OBJ(!senTreep->sensesp()->nextp(), activep,
"combinational logic with additional sensitivities");
if (VN_IS(activep->stmtsp(), AlwaysPostponed)) {
result.m_postponed.emplace_back(scopep, activep);
} else {
result.m_comb.emplace_back(scopep, activep);
}
} else {
UASSERT_OBJ(senTreep->hasClocked(), activep, "What else could it be?");
result.m_clocked.emplace_back(scopep, activep);
}
});
for (AstActive* const activep : empty) activep->unlinkFrBack()->deleteTree();
});
return result;
}
//============================================================================
// Simple ordering in source order
void orderSequentially(AstCFunc* funcp, const LogicByScope& lbs) {
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Create new subfunc for scope
const auto createNewSubFuncp = [&](AstScope* const scopep) {
const string subName{funcp->name() + "__" + scopep->nameDotless()};
AstCFunc* const subFuncp = new AstCFunc{scopep->fileline(), subName, scopep};
subFuncp->isLoose(true);
subFuncp->isConst(false);
subFuncp->declPrivate(true);
subFuncp->slow(funcp->slow());
scopep->addBlocksp(subFuncp);
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Call it from the top function
funcp->addStmtsp(new AstCCall{scopep->fileline(), subFuncp});
return subFuncp;
};
const VNUser1InUse user1InUse; // AstScope -> AstCFunc: the sub-function for the scope
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
const VNUser2InUse user2InUse; // AstScope -> int: sub-function counter used for names
for (const auto& pair : lbs) {
AstScope* const scopep = pair.first;
AstActive* const activep = pair.second;
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Create a sub-function per scope so we can V3Combine them later
if (!scopep->user1p()) scopep->user1p(createNewSubFuncp(scopep));
// Add statements to sub-function
for (AstNode *logicp = activep->stmtsp(), *nextp; logicp; logicp = nextp) {
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
auto* subFuncp = VN_AS(scopep->user1p(), CFunc);
nextp = logicp->nextp();
if (AstNodeProcedure* const procp = VN_CAST(logicp, NodeProcedure)) {
if (AstNode* bodyp = procp->stmtsp()) {
bodyp->unlinkFrBackWithNext();
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// If the process is suspendable, we need a separate function (a coroutine)
if (procp->isSuspendable()) {
subFuncp = createNewSubFuncp(scopep);
subFuncp->name(subFuncp->name() + "__" + cvtToStr(scopep->user2Inc()));
subFuncp->rtnType("VlCoroutine");
if (VN_IS(procp, Always)) {
subFuncp->slow(false);
FileLine* const flp = procp->fileline();
bodyp
= new AstWhile{flp, new AstConst{flp, AstConst::BitTrue{}}, bodyp};
}
}
subFuncp->addStmtsp(bodyp);
}
} else {
logicp->unlinkFrBack();
subFuncp->addStmtsp(logicp);
}
}
if (activep->backp()) activep->unlinkFrBack();
VL_DO_DANGLING(activep->deleteTree(), activep);
}
}
//============================================================================
// Create simply ordered functions
void createStatic(AstNetlist* netlistp, const LogicClasses& logicClasses) {
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval_static", /* slow: */ true);
orderSequentially(funcp, logicClasses.m_static);
splitCheck(funcp);
}
AstCFunc* createInitial(AstNetlist* netlistp, const LogicClasses& logicClasses) {
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval_initial", /* slow: */ true);
orderSequentially(funcp, logicClasses.m_initial);
return funcp; // Not splitting yet as it is not final
}
AstCFunc* createPostponed(AstNetlist* netlistp, const LogicClasses& logicClasses) {
if (logicClasses.m_postponed.empty()) return nullptr;
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval_postponed", /* slow: */ true);
orderSequentially(funcp, logicClasses.m_postponed);
splitCheck(funcp);
return funcp;
}
void createFinal(AstNetlist* netlistp, const LogicClasses& logicClasses) {
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval_final", /* slow: */ true);
orderSequentially(funcp, logicClasses.m_final);
splitCheck(funcp);
}
//============================================================================
// A TriggerKit holds all the components related to a TRIGGERVEC variable
struct TriggerKit {
// The TRIGGERVEC AstVarScope representing these trigger flags
AstVarScope* const m_vscp;
// The AstCFunc that computes the current active triggers
AstCFunc* const m_funcp;
// The AstCFunc that dumps the current active triggers
AstCFunc* const m_dumpp;
// The map from input sensitivity list to trigger sensitivity list
const std::unordered_map<const AstSenTree*, AstSenTree*> m_map;
VL_UNCOPYABLE(TriggerKit);
// Utility that assigns the given index trigger to fire when the given variable is zero
void addFirstIterationTriggerAssignment(AstVarScope* counterp, uint32_t /*index*/) const {
FileLine* const flp = counterp->fileline();
AstVarRef* const vrefp = new AstVarRef{flp, m_vscp, VAccess::WRITE};
AstCMethodHard* const callp = new AstCMethodHard{flp, vrefp, "at", new AstConst{flp, 0}};
callp->dtypeSetBit();
callp->pure(true);
m_funcp->stmtsp()->addHereThisAsNext(new AstAssign{
flp, callp,
new AstEq{flp, new AstVarRef{flp, counterp, VAccess::READ}, new AstConst{flp, 0}}});
}
// Utility to set then clear the dpiExportTrigger trigger
void addDpiExportTriggerAssignment(AstVarScope* dpiExportTriggerVscp, uint32_t index) const {
FileLine* const flp = dpiExportTriggerVscp->fileline();
AstVarRef* const vrefp = new AstVarRef{flp, m_vscp, VAccess::WRITE};
AstCMethodHard* const callp
= new AstCMethodHard{flp, vrefp, "at", new AstConst{flp, index}};
callp->dtypeSetBit();
callp->pure(true);
AstNode* stmtp
= new AstAssign{flp, callp, new AstVarRef{flp, dpiExportTriggerVscp, VAccess::READ}};
stmtp->addNext(new AstAssign{flp, new AstVarRef{flp, dpiExportTriggerVscp, VAccess::WRITE},
new AstConst{flp, AstConst::BitFalse{}}});
m_funcp->stmtsp()->addHereThisAsNext(stmtp);
}
};
// Create an AstSenTree that is sensitive to the given trigger index. Must not exist yet!
AstSenTree* createTriggerSenTree(AstNetlist* netlistp, AstVarScope* const vscp, uint32_t index) {
AstTopScope* const topScopep = netlistp->topScopep();
FileLine* const flp = topScopep->fileline();
AstVarRef* const vrefp = new AstVarRef{flp, vscp, VAccess::READ};
AstCMethodHard* const callp = new AstCMethodHard{flp, vrefp, "at", new AstConst{flp, index}};
callp->dtypeSetBit();
callp->pure(true);
AstSenItem* const senItemp = new AstSenItem{flp, VEdgeType::ET_TRUE, callp};
AstSenTree* const resultp = new AstSenTree{flp, senItemp};
topScopep->addSenTreesp(resultp);
return resultp;
}
//============================================================================
// Utility for extra trigger allocation
class ExtraTriggers final {
std::vector<string> m_descriptions; // Human readable descirption of extra triggers
public:
ExtraTriggers() = default;
size_t allocate(const string& description) {
const size_t index = m_descriptions.size();
m_descriptions.push_back(description);
return index;
}
size_t size() const { return m_descriptions.size(); }
const string& description(size_t index) const { return m_descriptions[index]; }
};
//============================================================================
// Create a TRIGGERVEC and the related TriggerKit for the given AstSenTree vector
const TriggerKit createTriggers(AstNetlist* netlistp, SenExprBuilder& senExprBuilder,
const std::vector<const AstSenTree*>& senTreeps,
const string& name, const ExtraTriggers& extraTriggers,
bool slow = false) {
AstTopScope* const topScopep = netlistp->topScopep();
AstScope* const scopeTopp = topScopep->scopep();
FileLine* const flp = scopeTopp->fileline();
std::unordered_map<const AstSenTree*, AstSenTree*> map;
const uint32_t nTriggers = senTreeps.size() + extraTriggers.size();
// Create the TRIGGERVEC variable
AstBasicDType* const tDtypep
= new AstBasicDType{flp, VBasicDTypeKwd::TRIGGERVEC, VSigning::UNSIGNED,
static_cast<int>(nTriggers), static_cast<int>(nTriggers)};
netlistp->typeTablep()->addTypesp(tDtypep);
AstVarScope* const vscp = scopeTopp->createTemp("__V" + name + "Triggered", tDtypep);
// Create the trigger computation function
AstCFunc* const funcp = makeSubFunction(netlistp, "_eval_triggers__" + name, slow);
// Create the trigger dump function (for debugging, always 'slow')
AstCFunc* const dumpp = makeSubFunction(netlistp, "_dump_triggers__" + name, true);
dumpp->ifdef("VL_DEBUG");
// Add a print to the dumping function if there are no triggers pending
{
AstCMethodHard* const callp
= new AstCMethodHard{flp, new AstVarRef{flp, vscp, VAccess::READ}, "any"};
callp->dtypeSetBit();
AstIf* const ifp = new AstIf{flp, callp};
dumpp->addStmtsp(ifp);
ifp->addElsesp(
new AstText{flp, "VL_DBG_MSGF(\" No triggers active\\n\");\n", true});
}
// Create a reference to a trigger flag
const auto getTrigRef = [&](uint32_t index, VAccess access) {
AstVarRef* const vrefp = new AstVarRef{flp, vscp, access};
AstConst* const idxp = new AstConst{flp, index};
AstCMethodHard* callp = new AstCMethodHard{flp, vrefp, "at", idxp};
callp->dtypeSetBit();
callp->pure(true);
return callp;
};
// Add a debug dumping statement for this trigger
const auto addDebug = [&](uint32_t index, const string& text = "") {
std::stringstream ss;
ss << "VL_DBG_MSGF(\" '" << name << "' region trigger index " << cvtToStr(index)
<< " is active";
if (!text.empty()) ss << ": " << text;
ss << "\\n\");\n";
const string message{ss.str()};
AstIf* const ifp = new AstIf{flp, getTrigRef(index, VAccess::READ)};
dumpp->addStmtsp(ifp);
ifp->addThensp(new AstText{flp, message, true});
};
// Add a print for each of the extra triggers
for (unsigned i = 0; i < extraTriggers.size(); ++i) {
addDebug(i, "Internal '" + name + "' trigger - " + extraTriggers.description(i));
}
// Add trigger computation
uint32_t triggerNumber = extraTriggers.size();
AstNode* initialTrigsp = nullptr;
for (const AstSenTree* const senTreep : senTreeps) {
UASSERT_OBJ(senTreep->hasClocked() || senTreep->hasHybrid(), senTreep,
"Cannot create trigger expression for non-clocked sensitivity");
// Create the trigger AstSenTrees and associate it with the original AstSenTree
AstCMethodHard* const senp = getTrigRef(triggerNumber, VAccess::READ);
AstSenItem* const senItemp = new AstSenItem{flp, VEdgeType::ET_TRUE, senp};
AstSenTree* const trigpSenp = new AstSenTree{flp, senItemp};
topScopep->addSenTreesp(trigpSenp);
map[senTreep] = trigpSenp;
// Add the trigger computation
const auto& pair = senExprBuilder.build(senTreep);
funcp->addStmtsp(
new AstAssign{flp, getTrigRef(triggerNumber, VAccess::WRITE), pair.first});
// Add initialization time trigger
if (pair.second || v3Global.opt.xInitialEdge()) {
AstNode* const assignp = new AstAssign{flp, getTrigRef(triggerNumber, VAccess::WRITE),
new AstConst{flp, 1}};
initialTrigsp = AstNode::addNext(initialTrigsp, assignp);
}
// Add a debug statement for this trigger
std::stringstream ss;
V3EmitV::verilogForTree(senTreep, ss);
addDebug(triggerNumber, ss.str());
//
++triggerNumber;
}
// Add the update statements
for (AstNodeStmt* const nodep : senExprBuilder.getAndClearPostUpdates()) {
funcp->addStmtsp(nodep);
}
const auto& preUpdates = senExprBuilder.getAndClearPreUpdates();
if (!preUpdates.empty()) {
for (AstNodeStmt* const nodep : vlstd::reverse_view(preUpdates)) {
UASSERT_OBJ(funcp->stmtsp(), funcp,
"No statements in trigger eval function, but there are pre updates");
funcp->stmtsp()->addHereThisAsNext(nodep);
}
}
const auto& locals = senExprBuilder.getAndClearLocals();
if (!locals.empty()) {
UASSERT_OBJ(funcp->stmtsp(), funcp,
"No statements in trigger eval function, but there are locals");
for (AstVar* const nodep : vlstd::reverse_view(locals)) {
funcp->stmtsp()->addHereThisAsNext(nodep);
}
}
// Add the initialization statements
if (initialTrigsp) {
AstVarScope* const vscp = scopeTopp->createTemp("__V" + name + "DidInit", 1);
AstVarRef* const condp = new AstVarRef{flp, vscp, VAccess::READ};
AstIf* const ifp = new AstIf{flp, new AstNot{flp, condp}};
funcp->addStmtsp(ifp);
ifp->branchPred(VBranchPred::BP_UNLIKELY);
ifp->addThensp(setVar(vscp, 1));
ifp->addThensp(initialTrigsp);
}
// Add a call to the dumping function if debug is enabled
{
AstTextBlock* const blockp = new AstTextBlock{flp};
funcp->addStmtsp(blockp);
const auto add = [&](const string& text) { blockp->addText(flp, text, true); };
add("#ifdef VL_DEBUG\n");
add("if (VL_UNLIKELY(vlSymsp->_vm_contextp__->debug())) {\n");
blockp->addNodesp(new AstCCall{flp, dumpp});
add("}\n");
add("#endif\n");
}
// The debug code might leak signal names, so simply delete it when using --protect-ids
if (v3Global.opt.protectIds()) dumpp->stmtsp()->unlinkFrBackWithNext()->deleteTree();
return {vscp, funcp, dumpp, map};
}
//============================================================================
// Helpers to construct an evaluation loop.
AstNode* buildLoop(AstNetlist* netlistp, const string& name,
const std::function<void(AstVarScope*, AstWhile*)>& build) //
{
AstTopScope* const topScopep = netlistp->topScopep();
AstScope* const scopeTopp = topScopep->scopep();
FileLine* const flp = scopeTopp->fileline();
// Create the loop condition variable
AstVarScope* const condp = scopeTopp->createTemp("__V" + name + "Continue", 1);
// Initialize the loop condition variable to true
AstNode* const resp = setVar(condp, 1);
// Add the loop
AstWhile* const loopp = new AstWhile{flp, new AstVarRef{flp, condp, VAccess::READ}};
resp->addNext(loopp);
// Clear the loop condition variable in the loop
loopp->addStmtsp(setVar(condp, 0));
// Build the body
build(condp, loopp);
// Done
return resp;
};
std::pair<AstVarScope*, AstNode*> makeEvalLoop(AstNetlist* netlistp, const string& tag,
const string& name, AstVarScope* trigVscp,
AstCFunc* trigDumpp,
std::function<AstNode*()> computeTriggers,
std::function<AstNode*()> makeBody) {
UASSERT_OBJ(trigVscp->dtypep()->basicp()->isTriggerVec(), trigVscp, "Not TRIGGERVEC");
AstTopScope* const topScopep = netlistp->topScopep();
AstScope* const scopeTopp = topScopep->scopep();
FileLine* const flp = scopeTopp->fileline();
AstVarScope* const counterp = scopeTopp->createTemp("__V" + tag + "IterCount", 32);
AstNode* nodep = setVar(counterp, 0);
nodep->addNext(buildLoop(netlistp, tag, [&](AstVarScope* continuep, AstWhile* loopp) {
// Compute triggers
loopp->addStmtsp(computeTriggers());
// Invoke body if triggered
{
AstVarRef* const refp = new AstVarRef{flp, trigVscp, VAccess::READ};
AstCMethodHard* const callp = new AstCMethodHard{flp, refp, "any"};
callp->dtypeSetBit();
AstIf* const ifp = new AstIf{flp, callp};
loopp->addStmtsp(ifp);
ifp->addThensp(setVar(continuep, 1));
// If we exceeded the iteration limit, die
{
const uint32_t limit = v3Global.opt.convergeLimit();
AstVarRef* const refp = new AstVarRef{flp, counterp, VAccess::READ};
2022-08-19 14:25:30 +02:00
AstConst* const constp = new AstConst{flp, AstConst::DTyped{}, counterp->dtypep()};
constp->num().setLong(limit);
AstNodeMath* const condp = new AstGt{flp, refp, constp};
AstIf* const failp = new AstIf{flp, condp};
ifp->addThensp(failp);
AstTextBlock* const blockp = new AstTextBlock{flp};
failp->addThensp(blockp);
FileLine* const locp = netlistp->topModulep()->fileline();
const string& file = EmitCBaseVisitor::protect(locp->filename());
const string& line = cvtToStr(locp->lineno());
const auto add = [&](const string& text) { blockp->addText(flp, text, true); };
add("#ifdef VL_DEBUG\n");
blockp->addNodesp(new AstCCall{flp, trigDumpp});
add("#endif\n");
add("VL_FATAL_MT(\"" + file + "\", " + line + ", \"\", ");
add("\"" + name + " region did not converge.\");\n");
}
// Increment iteration count
{
AstVarRef* const wrefp = new AstVarRef{flp, counterp, VAccess::WRITE};
AstVarRef* const rrefp = new AstVarRef{flp, counterp, VAccess::READ};
2022-08-19 14:25:30 +02:00
AstConst* const onep = new AstConst{flp, AstConst::DTyped{}, counterp->dtypep()};
onep->num().setLong(1);
ifp->addThensp(new AstAssign{flp, wrefp, new AstAdd{flp, rrefp, onep}});
}
// Add body
ifp->addThensp(makeBody());
}
}));
return {counterp, nodep};
}
//============================================================================
// Order the combinational logic to create the settle loop
void createSettle(AstNetlist* netlistp, SenExprBuilder& senExprBulider,
LogicClasses& logicClasses) {
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval_settle", true);
// Clone, because ordering is destructive, but we still need them for "_eval"
LogicByScope comb = logicClasses.m_comb.clone();
LogicByScope hybrid = logicClasses.m_hybrid.clone();
// Nothing to do if there is no logic.
// While this is rare in real designs, it reduces noise in small tests.
if (comb.empty() && hybrid.empty()) return;
// We have an extra trigger denoting this is the first iteration of the settle loop
ExtraTriggers extraTriggers;
const size_t firstIterationTrigger = extraTriggers.allocate("first iteration");
// Gather the relevant sensitivity expressions and create the trigger kit
const auto& senTreeps = getSenTreesUsedBy({&comb, &hybrid});
const TriggerKit& trig
= createTriggers(netlistp, senExprBulider, senTreeps, "stl", extraTriggers, true);
// Remap sensitivities (comb has none, so only do the hybrid)
remapSensitivities(hybrid, trig.m_map);
// Create the inverse map from trigger ref AstSenTree to original AstSenTree
std::unordered_map<const AstSenItem*, const AstSenTree*> trigToSen;
invertAndMergeSenTreeMap(trigToSen, trig.m_map);
// First trigger is for pure combinational triggers (first iteration)
AstSenTree* const inputChanged
= createTriggerSenTree(netlistp, trig.m_vscp, firstIterationTrigger);
// Create and the body function
AstCFunc* const stlFuncp = V3Order::order(
netlistp, {&comb, &hybrid}, trigToSen, "stl", false, true,
[=](const AstVarScope*, std::vector<AstSenTree*>& out) { out.push_back(inputChanged); });
splitCheck(stlFuncp);
// Create the eval loop
const auto& pair = makeEvalLoop(
netlistp, "stl", "Settle", trig.m_vscp, trig.m_dumpp,
[&]() { // Trigger
return new AstCCall{stlFuncp->fileline(), trig.m_funcp};
},
[&]() { // Body
return new AstCCall{stlFuncp->fileline(), stlFuncp};
});
// Add the first iteration trigger to the trigger computation function
trig.addFirstIterationTriggerAssignment(pair.first, firstIterationTrigger);
// Add the eval loop to the top function
funcp->addStmtsp(pair.second);
}
//============================================================================
// Order the replicated combinational logic to create the 'ico' region
AstNode* createInputCombLoop(AstNetlist* netlistp, SenExprBuilder& senExprBuilder,
LogicByScope& logic) {
// Nothing to do if no combinational logic is sensitive to top level inputs
if (logic.empty()) return nullptr;
// SystemC only: Any top level inputs feeding a combinational logic must be marked,
// so we can make them sc_sensitive
if (v3Global.opt.systemC()) {
logic.foreachLogic([](AstNode* logicp) {
logicp->foreach<AstVarRef>([](AstVarRef* refp) {
if (refp->access().isWriteOnly()) return;
AstVarScope* const vscp = refp->varScopep();
if (vscp->scopep()->isTop() && vscp->varp()->isNonOutput()) {
vscp->varp()->scSensitive(true);
}
});
});
}
// We have some extra trigger denoting external conditions
AstVarScope* const dpiExportTriggerVscp = netlistp->dpiExportTriggerp();
ExtraTriggers extraTriggers;
const size_t firstIterationTrigger = extraTriggers.allocate("first iteration");
const size_t dpiExportTriggerIndex = dpiExportTriggerVscp
? extraTriggers.allocate("DPI export trigger")
: std::numeric_limits<unsigned>::max();
// Gather the relevant sensitivity expressions and create the trigger kit
const auto& senTreeps = getSenTreesUsedBy({&logic});
const TriggerKit& trig
= createTriggers(netlistp, senExprBuilder, senTreeps, "ico", extraTriggers);
if (dpiExportTriggerVscp) {
trig.addDpiExportTriggerAssignment(dpiExportTriggerVscp, dpiExportTriggerIndex);
}
// Remap sensitivities
remapSensitivities(logic, trig.m_map);
// Create the inverse map from trigger ref AstSenTree to original AstSenTree
std::unordered_map<const AstSenItem*, const AstSenTree*> trigToSen;
invertAndMergeSenTreeMap(trigToSen, trig.m_map);
// The trigger top level inputs (first iteration)
AstSenTree* const inputChanged
= createTriggerSenTree(netlistp, trig.m_vscp, firstIterationTrigger);
// The DPI Export trigger
AstSenTree* const dpiExportTriggered
= createTriggerSenTree(netlistp, trig.m_vscp, dpiExportTriggerIndex);
// Create and Order the body function
AstCFunc* const icoFuncp
= V3Order::order(netlistp, {&logic}, trigToSen, "ico", false, false,
[=](const AstVarScope* vscp, std::vector<AstSenTree*>& out) {
2022-07-15 17:18:41 +02:00
AstVar* const varp = vscp->varp();
if (varp->isPrimaryInish() || varp->isSigUserRWPublic()) {
out.push_back(inputChanged);
}
2022-07-15 17:18:41 +02:00
if (varp->isWrittenByDpi()) out.push_back(dpiExportTriggered);
});
splitCheck(icoFuncp);
// Create the eval loop
const auto& pair = makeEvalLoop(
netlistp, "ico", "Input combinational", trig.m_vscp, trig.m_dumpp,
[&]() { // Trigger
return new AstCCall{icoFuncp->fileline(), trig.m_funcp};
},
[&]() { // Body
return new AstCCall{icoFuncp->fileline(), icoFuncp};
});
// Add the first iteration trigger to the trigger computation function
trig.addFirstIterationTriggerAssignment(pair.first, firstIterationTrigger);
// Return the eval loop itself
return pair.second;
}
//============================================================================
// Bold together parts to create the top level _eval function
void createEval(AstNetlist* netlistp, //
AstNode* icoLoop, //
const TriggerKit& actTrig, //
AstVarScope* preTrigsp, //
AstVarScope* nbaTrigsp, //
AstCFunc* actFuncp, //
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
AstCFunc* nbaFuncp, //
AstCFunc* postponedFuncp, //
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
TimingKit& timingKit //
) {
FileLine* const flp = netlistp->fileline();
AstCFunc* const funcp = makeTopFunction(netlistp, "_eval", false);
netlistp->evalp(funcp);
// Start with the ico loop, if any
if (icoLoop) funcp->addStmtsp(icoLoop);
// Create the NBA trigger dumping function, which is the same as act trigger
// dumping function, but referencing the nba trigger vector.
AstCFunc* const nbaDumpp = actTrig.m_dumpp->cloneTree(false);
actTrig.m_dumpp->addNextHere(nbaDumpp);
nbaDumpp->name("_dump_triggers__nba");
nbaDumpp->foreach<AstVarRef>([&](AstVarRef* refp) {
UASSERT_OBJ(refp->access().isReadOnly(), refp, "Should only read state");
if (refp->varScopep() == actTrig.m_vscp) {
refp->replaceWith(new AstVarRef{refp->fileline(), nbaTrigsp, VAccess::READ});
}
});
nbaDumpp->foreach<AstText>([&](AstText* textp) { //
textp->text(VString::replaceWord(textp->text(), "act", "nba"));
});
// Create the active eval loop
AstNode* const activeEvalLoopp
= makeEvalLoop(
netlistp, "act", "Active", actTrig.m_vscp, actTrig.m_dumpp,
[&]() { // Trigger
2022-09-17 17:08:05 +02:00
AstNode* const resultp = new AstCCall{flp, actTrig.m_funcp};
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Commit trigger awaits from the previous iteration
2022-09-17 17:08:05 +02:00
if (AstNode* const commitp = timingKit.createCommit(netlistp)) {
resultp->addNext(commitp);
}
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
return resultp;
},
[&]() { // Body
AstNode* resultp = nullptr;
// Compute the pre triggers
{
AstVarRef* const lhsp = new AstVarRef{flp, preTrigsp, VAccess::WRITE};
AstVarRef* const opap = new AstVarRef{flp, actTrig.m_vscp, VAccess::READ};
AstVarRef* const opbp = new AstVarRef{flp, nbaTrigsp, VAccess::READ};
opap->addNext(opbp);
AstCMethodHard* const callp = new AstCMethodHard{flp, lhsp, "andNot", opap};
callp->statement(true);
callp->dtypeSetVoid();
resultp = AstNode::addNext(resultp, callp);
}
// Latch the active trigger flags under the NBA trigger flags
{
AstVarRef* const lhsp = new AstVarRef{flp, nbaTrigsp, VAccess::WRITE};
AstVarRef* const argp = new AstVarRef{flp, actTrig.m_vscp, VAccess::READ};
AstCMethodHard* const callp = new AstCMethodHard{flp, lhsp, "set", argp};
callp->statement(true);
callp->dtypeSetVoid();
resultp = AstNode::addNext(resultp, callp);
}
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Resume triggered timing schedulers
2022-09-17 17:08:05 +02:00
if (AstNode* const resumep = timingKit.createResume(netlistp)) {
resultp = AstNode::addNext(resultp, resumep);
}
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Invoke body function
return AstNode::addNext(resultp, new AstCCall{flp, actFuncp});
})
.second;
// Create the NBA eval loop. This uses the Active eval loop in the trigger section.
AstNode* const nbaEvalLoopp
= makeEvalLoop(
netlistp, "nba", "NBA", nbaTrigsp, nbaDumpp,
[&]() { // Trigger
AstNode* resultp = nullptr;
// Reset NBA triggers
{
AstVarRef* const refp = new AstVarRef{flp, nbaTrigsp, VAccess::WRITE};
AstCMethodHard* const callp = new AstCMethodHard{flp, refp, "clear"};
callp->statement(true);
callp->dtypeSetVoid();
resultp = AstNode::addNext(resultp, callp);
}
// Run the Active eval loop
return AstNode::addNext(resultp, activeEvalLoopp);
},
[&]() { // Body
return new AstCCall{flp, nbaFuncp};
})
.second;
// Add the NBA eval loop
funcp->addStmtsp(nbaEvalLoopp);
// Add the Postponed eval call
if (postponedFuncp) funcp->addStmtsp(new AstCCall{flp, postponedFuncp});
}
} // namespace
//============================================================================
// Top level entry-point to scheduling
void schedule(AstNetlist* netlistp) {
const auto addSizeStat = [](const string& name, const LogicByScope& lbs) {
uint64_t size = 0;
lbs.foreachLogic([&](AstNode* nodep) { size += nodep->nodeCount(); });
V3Stats::addStat("Scheduling, " + name, size);
};
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
// Step 0. Prepare timing-related logic and external domains
auto timingKit = prepareTiming(netlistp);
// Step 1. Gather and classify all logic in the design
LogicClasses logicClasses = gatherLogicClasses(netlistp);
if (v3Global.opt.stats()) {
V3Stats::statsStage("sched-gather");
addSizeStat("size of class: static", logicClasses.m_static);
addSizeStat("size of class: initial", logicClasses.m_initial);
addSizeStat("size of class: final", logicClasses.m_final);
}
// Step 2. Schedule static, initial and final logic classes in source order
createStatic(netlistp, logicClasses);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-static");
AstCFunc* const initp = createInitial(netlistp, logicClasses);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-initial");
createFinal(netlistp, logicClasses);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-final");
// Step 3: Break combinational cycles by introducing hybrid logic
// Note: breakCycles also removes corresponding logic from logicClasses.m_comb;
logicClasses.m_hybrid = breakCycles(netlistp, logicClasses.m_comb);
if (v3Global.opt.stats()) {
addSizeStat("size of class: clocked", logicClasses.m_clocked);
addSizeStat("size of class: combinational", logicClasses.m_comb);
addSizeStat("size of class: hybrid", logicClasses.m_hybrid);
V3Stats::statsStage("sched-break-cycles");
}
// We pass around a single SenExprBuilder instance, as we only need one set of 'prev' variables
// for edge/change detection in sensitivity expressions, which this keeps track of.
SenExprBuilder senExprBuilder{netlistp, initp};
// Step 4: Create 'settle' region that restores the combinational invariant
createSettle(netlistp, senExprBuilder, logicClasses);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-settle");
// Step 5: Partition the clocked and combinational (including hybrid) logic into pre/act/nba.
// All clocks (signals referenced in an AstSenTree) generated via a blocking assignment
// (including combinationally generated signals) are computed within the act region.
LogicRegions logicRegions
= partition(logicClasses.m_clocked, logicClasses.m_comb, logicClasses.m_hybrid);
if (v3Global.opt.stats()) {
addSizeStat("size of region: Active Pre", logicRegions.m_pre);
addSizeStat("size of region: Active", logicRegions.m_act);
addSizeStat("size of region: NBA", logicRegions.m_nba);
V3Stats::statsStage("sched-partition");
}
// Step 6: Replicate combinational logic
LogicReplicas logicReplicas = replicateLogic(logicRegions);
if (v3Global.opt.stats()) {
addSizeStat("size of replicated logic: Input", logicReplicas.m_ico);
addSizeStat("size of replicated logic: Active", logicReplicas.m_act);
addSizeStat("size of replicated logic: NBA", logicReplicas.m_nba);
V3Stats::statsStage("sched-replicate");
}
// Step 7: Create input combinational logic loop
AstNode* const icoLoopp = createInputCombLoop(netlistp, senExprBuilder, logicReplicas.m_ico);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-create-ico");
// Step 8: Create the pre/act/nba triggers
AstVarScope* const dpiExportTriggerVscp = netlistp->dpiExportTriggerp();
// We may have an extra trigger for variable updated in DPI exports
ExtraTriggers extraTriggers;
const size_t dpiExportTriggerIndex = dpiExportTriggerVscp
? extraTriggers.allocate("DPI export trigger")
: std::numeric_limits<unsigned>::max();
const auto& senTreeps = getSenTreesUsedBy({&logicRegions.m_pre, //
&logicRegions.m_act, //
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
&logicRegions.m_nba, //
&timingKit.m_lbs});
const TriggerKit& actTrig
= createTriggers(netlistp, senExprBuilder, senTreeps, "act", extraTriggers);
if (dpiExportTriggerVscp) {
actTrig.addDpiExportTriggerAssignment(dpiExportTriggerVscp, dpiExportTriggerIndex);
}
AstTopScope* const topScopep = netlistp->topScopep();
AstScope* const scopeTopp = topScopep->scopep();
AstVarScope* const actTrigVscp = actTrig.m_vscp;
AstVarScope* const preTrigVscp = scopeTopp->createTempLike("__VpreTriggered", actTrigVscp);
AstVarScope* const nbaTrigVscp = scopeTopp->createTempLike("__VnbaTriggered", actTrigVscp);
const auto cloneMapWithNewTriggerReferences
= [=](std::unordered_map<const AstSenTree*, AstSenTree*> map, AstVarScope* vscp) {
// Copy map
auto newMap{map};
VNDeleter deleter;
// Replace references in each mapped value with a reference to the given vscp
for (auto& pair : newMap) {
pair.second = pair.second->cloneTree(false);
pair.second->foreach<AstVarRef>([&](AstVarRef* refp) {
UASSERT_OBJ(refp->varScopep() == actTrigVscp, refp, "Unexpected reference");
UASSERT_OBJ(refp->access() == VAccess::READ, refp, "Should be read ref");
refp->replaceWith(new AstVarRef{refp->fileline(), vscp, VAccess::READ});
deleter.pushDeletep(refp);
});
topScopep->addSenTreesp(pair.second);
}
return newMap;
};
const auto& actTrigMap = actTrig.m_map;
const auto preTrigMap = cloneMapWithNewTriggerReferences(actTrigMap, preTrigVscp);
const auto nbaTrigMap = cloneMapWithNewTriggerReferences(actTrigMap, nbaTrigVscp);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-create-triggers");
// Note: Experiments so far show that running the Act (or Ico) regions on
// multiple threads is always a net loss, so only use multi-threading for
// NBA for now. This can be revised if evidence is available that it would
// be beneficial
// Step 9: Create the 'act' region evaluation function
// Remap sensitivities of the input logic to the triggers
remapSensitivities(logicRegions.m_pre, preTrigMap);
remapSensitivities(logicRegions.m_act, actTrigMap);
remapSensitivities(logicReplicas.m_act, actTrigMap);
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
remapSensitivities(timingKit.m_lbs, actTrigMap);
const auto& actTimingDomains = timingKit.remapDomains(actTrigMap);
// Create the inverse map from trigger ref AstSenTree to original AstSenTree
std::unordered_map<const AstSenItem*, const AstSenTree*> trigToSenAct;
invertAndMergeSenTreeMap(trigToSenAct, preTrigMap);
invertAndMergeSenTreeMap(trigToSenAct, actTrigMap);
// The DPI Export trigger AstSenTree
AstSenTree* const dpiExportTriggeredAct
= createTriggerSenTree(netlistp, actTrig.m_vscp, dpiExportTriggerIndex);
AstCFunc* const actFuncp = V3Order::order(
netlistp, {&logicRegions.m_pre, &logicRegions.m_act, &logicReplicas.m_act}, trigToSenAct,
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
"act", false, false, [&](const AstVarScope* vscp, std::vector<AstSenTree*>& out) {
auto it = actTimingDomains.find(vscp);
if (it != actTimingDomains.end()) out = it->second;
if (vscp->varp()->isWrittenByDpi()) out.push_back(dpiExportTriggeredAct);
});
splitCheck(actFuncp);
if (v3Global.opt.stats()) V3Stats::statsStage("sched-create-act");
// Step 10: Create the 'nba' region evaluation function
// Remap sensitivities of the input logic to the triggers
remapSensitivities(logicRegions.m_nba, nbaTrigMap);
remapSensitivities(logicReplicas.m_nba, nbaTrigMap);
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
const auto& nbaTimingDomains = timingKit.remapDomains(nbaTrigMap);
// Create the inverse map from trigger ref AstSenTree to original AstSenTree
std::unordered_map<const AstSenItem*, const AstSenTree*> trigToSenNba;
invertAndMergeSenTreeMap(trigToSenNba, nbaTrigMap);
AstSenTree* const dpiExportTriggeredNba
= createTriggerSenTree(netlistp, nbaTrigVscp, dpiExportTriggerIndex);
AstCFunc* const nbaFuncp = V3Order::order(
netlistp, {&logicRegions.m_nba, &logicReplicas.m_nba}, trigToSenNba, "nba",
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
v3Global.opt.mtasks(), false, [&](const AstVarScope* vscp, std::vector<AstSenTree*>& out) {
auto it = nbaTimingDomains.find(vscp);
if (it != nbaTimingDomains.end()) out = it->second;
if (vscp->varp()->isWrittenByDpi()) out.push_back(dpiExportTriggeredNba);
});
splitCheck(nbaFuncp);
netlistp->evalNbap(nbaFuncp); // Remember for V3LifePost
if (v3Global.opt.stats()) V3Stats::statsStage("sched-create-nba");
// Step 11: Create the 'postponed' region evaluation function
auto* const postponedFuncp = createPostponed(netlistp, logicClasses);
// Step 12: Bolt it all together to create the '_eval' function
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
createEval(netlistp, icoLoopp, actTrig, preTrigVscp, nbaTrigVscp, actFuncp, nbaFuncp,
postponedFuncp, timingKit);
Timing support (#3363) Adds timing support to Verilator. It makes it possible to use delays, event controls within processes (not just at the start), wait statements, and forks. Building a design with those constructs requires a compiler that supports C++20 coroutines (GCC 10, Clang 5). The basic idea is to have processes and tasks with delays/event controls implemented as C++20 coroutines. This allows us to suspend and resume them at any time. There are five main runtime classes responsible for managing suspended coroutines: * `VlCoroutineHandle`, a wrapper over C++20's `std::coroutine_handle` with move semantics and automatic cleanup. * `VlDelayScheduler`, for coroutines suspended by delays. It resumes them at a proper simulation time. * `VlTriggerScheduler`, for coroutines suspended by event controls. It resumes them if its corresponding trigger was set. * `VlForkSync`, used for syncing `fork..join` and `fork..join_any` blocks. * `VlCoroutine`, the return type of all verilated coroutines. It allows for suspending a stack of coroutines (normally, C++ coroutines are stackless). There is a new visitor in `V3Timing.cpp` which: * scales delays according to the timescale, * simplifies intra-assignment timing controls and net delays into regular timing controls and assignments, * simplifies wait statements into loops with event controls, * marks processes and tasks with timing controls in them as suspendable, * creates delay, trigger scheduler, and fork sync variables, * transforms timing controls and fork joins into C++ awaits There are new functions in `V3SchedTiming.cpp` (used by `V3Sched.cpp`) that integrate static scheduling with timing. This involves providing external domains for variables, so that the necessary combinational logic gets triggered after coroutine resumption, as well as statements that need to be injected into the design eval function to perform this resumption at the correct time. There is also a function that transforms forked processes into separate functions. See the comments in `verilated_timing.h`, `verilated_timing.cpp`, `V3Timing.cpp`, and `V3SchedTiming.cpp`, as well as the internals documentation for more details. Signed-off-by: Krzysztof Bieganski <kbieganski@antmicro.com>
2022-08-22 14:26:32 +02:00
transformForks(netlistp);
splitCheck(initp);
netlistp->dpiExportTriggerp(nullptr);
2022-09-22 18:28:42 +02:00
V3Global::dumpCheckGlobalTree("sched", 0, dumpTree() >= 3);
}
} // namespace V3Sched