See update 1 below for my guess as to why the error is happening
I'm trying to develop an application with some C#/WPF and C++. I am having a problem on the C++ side on a part of the code that involves optimizing an object using GNU Scientific Library (GSL) optimization functions. I will avoid including any of the C#/WPF/GSL code in order to keep this question more generic and because the problem is within my C++ code.
For the minimal, complete and verifiable example below, here is what I have. I have a class Foo. And a class Optimizer. An object of class Optimizer is a member of class Foo, so that objects of Foo can optimize themselves when it is required.
The way GSL optimization functions take in external parameters is through a void pointer. I first define a struct Params to hold all the required parameters. Then I define an object of Params and convert it into a void pointer. A copy of this data is made with memcpy_s and a member void pointer optimParamsPtr of Optimizer class points to it so it can access the parameters when the optimizer is called to run later in time. When optimParamsPtr is accessed by CostFn(), I get the following error.
Managed Debugging Assistant 'FatalExecutionEngineError' : 'The runtime
has encountered a fatal error. The address of the error was at
0x6f25e01e, on thread 0x431c. The error code is 0xc0000005. This error
may be a bug in the CLR or in the unsafe or non-verifiable portions of
user code. Common sources of this bug include user marshaling errors
for COM-interop or PInvoke, which may corrupt the stack.'
Just to ensure the validity of the void pointer I made, I call CostFn() at line 81 with the void * pointer passed as an argument to InitOptimizer() and everything works. But in line 85 when the same CostFn() is called with the optimParamsPtr pointing to data copied by memcpy_s, I get the error. So I am guessing something is going wrong with the memcpy_s step. Anyone have any ideas as to what?
#include "pch.h"
#include <iostream>
using namespace System;
using namespace System::Runtime::InteropServices;
using namespace std;
// An optimizer for various kinds of objects
class Optimizer // GSL requires this to be an unmanaged class
{
public:
double InitOptimizer(int ptrID, void *optimParams, size_t optimParamsSize);
void FreeOptimizer();
void * optimParamsPtr;
private:
double cost = 0;
};
ref class Foo // A class whose objects can be optimized
{
private:
int a; // An internal variable that can be changed to optimize the object
Optimizer *fooOptimizer; // Optimizer for a Foo object
public:
Foo(int val) // Constructor
{
a = val;
fooOptimizer = new Optimizer;
}
~Foo()
{
if (fooOptimizer != NULL)
{
delete fooOptimizer;
}
}
void SetA(int val) // Mutator
{
a = val;
}
int GetA() // Accessor
{
return a;
}
double Optimize(int ptrID); // Optimize object
// ptrID is a variable just to change behavior of Optimize() and show what works and what doesn't
};
ref struct Params // Parameters required by the cost function
{
int cost_scaling;
Foo ^ FooObj;
};
double CostFn(void *params) // GSL requires cost function to be of this type and cannot be a member of a class
{
// Cast void * to Params type
GCHandle h = GCHandle::FromIntPtr(IntPtr(params));
Params ^ paramsArg = safe_cast<Params^>(h.Target);
h.Free(); // Deallocate
// Return the cost
int val = paramsArg->FooObj->GetA();
return (double)(paramsArg->cost_scaling * val);
}
double Optimizer::InitOptimizer(int ptrID, void *optimParamsArg, size_t optimParamsSizeArg)
{
optimParamsPtr = ::operator new(optimParamsSizeArg);
memcpy_s(optimParamsPtr, optimParamsSizeArg, optimParamsArg, optimParamsSizeArg);
double ret_val;
// Here is where the GSL stuff would be. But I replace that with a call to CostFn to show the error
if (ptrID == 1)
{
ret_val = CostFn(optimParamsArg); // Works
}
else
{
ret_val = CostFn(optimParamsPtr); // Doesn't work
}
return ret_val;
}
// Release memory used by unmanaged variables in Optimizer
void Optimizer::FreeOptimizer()
{
if (optimParamsPtr != NULL)
{
delete optimParamsPtr;
}
}
double Foo::Optimize(int ptrID)
{
// Create and initialize params object
Params^ paramsArg = gcnew Params;
paramsArg->cost_scaling = 11;
paramsArg->FooObj = this;
// Convert Params type object to void *
void * paramsArgVPtr = GCHandle::ToIntPtr(GCHandle::Alloc(paramsArg)).ToPointer();
size_t paramsArgSize = sizeof(paramsArg); // size of memory block in bytes pointed to by void pointer
double result = 0;
// Initialize optimizer
result = fooOptimizer->InitOptimizer(ptrID, paramsArgVPtr, paramsArgSize);
// Here is where the loop that does the optimization will be. Removed from this example for simplicity.
return result;
}
int main()
{
Foo Foo1(2);
std::cout << Foo1.Optimize(1) << endl; // Use orig void * arg in line 81 and it works
std::cout << Foo1.Optimize(2) << endl; // Use memcpy_s-ed new void * public member of Optimizer in line 85 and it doesn't work
}
Just to reiterate I need to copy the params to a member in the optimizer because the optimizer will run all through the lifetime of the Foo object. So it needs to exist as long as the Optimizer object exist and not just in the scope of Foo::Optimize()
/clr support need to be selected in project properties for the code to compile. Running on an x64 solution platform.
Update 1: While trying to debug this, I got suspicious of the way I get the size of paramsArg at line 109. Looks like I am getting the size of paramsArg as size of int cost_scaling plus size of the memory storing the address to FooObj instead of the size of memory storing FooObj itself. I realized this after stumbling across this answer to another post. I confirmed this by checking the value of paramsArg after adding some new dummy double members to Foo class. As expected the value of paramsArg doesn't change. I suppose this explains why I get the error. A solution would be to write code to correctly calculate the size of a Foo class object and set that to paramsArg instead of using sizeof. But that is turning out to be too complicated and probably another question in itself. For example, how to get size of a ref class object? Anyways hopefully someone will find this helpful.
Related
Using C++ 17, I'm looking for a way to store a lambda that captures the this pointer, without using std::function<>. The reason to not using std::function<> is that I need the guaranty that no dynamic memory allocations are used. The purpose of this, is to be able to define some asynchronous program flow. Example:
class foo {
public:
void start() {
timer(1ms, [this](){
set_pin(1,2);
timer(1ms, [this](){
set_pin(2,1);
}
}
}
private:
template < class Timeout, class Callback >
void timer( Timeout to, Callback&& cb ) {
cb_ = cb;
// setup timer and call cb_ one timeout reached
...
}
??? cb_;
};
Edit: Maybe it's not really clear: std::function<void()> would do the job, but I need / like to have the guaranty, that no dynamic allocations happens as the project is in the embedded field. In practice std::function<void()> seems to not require dynamic memory allocation, if the lambda just captures this. I guess this is due to some small object optimizations, but I would like to not rely on that.
You can write your own function_lite to store the lambda, then you can use static_assert to check the size and alignment requirements are satisfied:
#include <cstddef>
#include <new>
#include <type_traits>
class function_lite {
static constexpr unsigned buffer_size = 16;
using trampoline_type = void (function_lite::*)() const;
trampoline_type trampoline;
trampoline_type cleanup;
alignas(std::max_align_t) char buffer[buffer_size];
template <typename T>
void trampoline_func() const {
auto const obj =
std::launder(static_cast<const T*>(static_cast<const void*>(buffer)));
(*obj)();
}
template <typename T>
void cleanup_func() const {
auto const obj =
std::launder(static_cast<const T*>(static_cast<const void*>(buffer)));
obj->~T();
}
public:
template <typename T>
function_lite(T t)
: trampoline(&function_lite::trampoline_func<T>),
cleanup(&function_lite::cleanup_func<T>) {
static_assert(sizeof(T) <= buffer_size);
static_assert(alignof(T) <= alignof(std::max_align_t));
new (static_cast<void*>(buffer)) T(t);
}
~function_lite() { (this->*cleanup)(); }
function_lite(function_lite const&) = delete;
function_lite& operator=(function_lite const&) = delete;
void operator()() const { (this->*trampoline)(); }
};
int main() {
int x = 0;
function_lite f([x] {});
}
Note: this is not copyable; to add copy or move semantics you will need to add new members like trampoline and cleanup which can properly copy the stored object.
There is no drop in replacement in the language or the standard library.
Every lambda is a unique type in the typesystem. Technically you may have a lambda as a member, but then its type is fixed. You may not assign other lambdas to it.
If you really want to have an owning function wrapper like std::function, you need to write your own. Actually you want a std::function with a big enough small-buffer-optimization buffer.
Another approach would be to omit the this capture and pass it to the function when doing the call. So you have a captureless lambda, which is convertible to a function pointer which you can easily store. I would take this route and adapter complexer ways if really nessessary.
it would look like this (i trimmed down the code a bit):
class foo
{
public:
void start()
{
timer(1, [](foo* instance)
{
instance->set_pin(1,2);
});
}
private:
template < class Timeout, class Callback >
void timer( Timeout to, Callback&& cb )
{
cb_ = cb;
cb_(this); // call the callback like this
}
void set_pin(int, int)
{
std::cout << "pin set\n";
}
void(*cb_)(foo*);
};
I am using Xtext 2.15 to generate a language that, among other things, processes asynchronous calls in a way they look synchronous.
For instance, the following code in my language:
int a = 1;
int b = 2;
boolean sleepSuccess = doSleep(2000); // sleep two seconds
int c = 3;
int d = 4;
would generate the following Java code:
int a = 1;
int b = 2;
doSleep(2000, new DoSleepCallback() {
public void onTrigger(boolean rc) {
boolean sleepSuccess = rc;
int c = 3;
int d = 4;
}
});
To achieve it, I defined the grammar this way:
grammar org.qedlang.qed.QED with jbase.Jbase // Jbase inherits Xbase
...
FunctionDeclaration return XExpression:
=>({FunctionDeclaration} type=JvmTypeReference name=ValidID '(')
(params+=FullJvmFormalParameter (',' params+=FullJvmFormalParameter)*)?
')' block=XBlockExpression
;
The FunctionDeclaration rule is used to define asynchronous calls. In my language library, I would have as system call:
boolean doSleep(int millis) {} // async FunctionDeclaration element stub
The underlying Java implementation would be:
public abstract class DoSleepCallback {
public abstract void onTrigger(boolean rc);
}
public void doSleep(int millis, DoSleepCallback callback) {
<perform sleep and call callback.onTrigger(<success>)>
}
So, using the inferrer, type computer and compiler, how to identify calls to FunctionDeclaration elements, add a callback parameter and process the rest of the body in an inner class?
I could, for instance, override appendFeatureCall in the language compiler, would it work? There is still a part I don't know how to do...
override appendFeatureCall(XAbstractFeatureCall call, ITreeAppendable b) {
...
val feature = call.feature
...
if (feature instanceof JvmExecutable) {
b.append('(')
val arguments = call.actualArguments
if (!arguments.isEmpty) {
...
arguments.appendArguments(b, shouldBreakFirstArgument)
// HERE IS THE PART I DON'T KNOW HOW TO DO
<IF feature IS A FunctionDeclaration>
<argument.appendArgument(NEW GENERATED CALLBACK PARAMETER)>
<INSERT REST OF XBlockExpression body INSIDE CALLBACK INSTANCE>
<ENDIF>
}
b.append(');')
}
}
So basically, how to tell if "feature" points to FunctionDeclaration? The rest, I may be able to do it...
Related to another StackOverflow entry, I had the idea of implementing FunctionDeclaration in the inferrer as a class instead of as a method:
def void inferExpressions(JvmDeclaredType it, FunctionDeclaration function) {
// now let's go over the features
for ( f : (function.block as XBlockExpression).expressions ) {
if (f instanceof FunctionDeclaration) {
members += f.toClass(f.fullyQualifiedName) [
inferVariables(f)
superTypes += typeRef(FunctionDeclarationObject)
// let's add a default constructor
members += f.toConstructor [
for (p : f.params)
parameters += p.toParameter(p.name, p.parameterType)
body = f.block
]
inferExpressions(f)
]
}
}
}
The generated class would extend FunctionDeclarationObject, so I thought there was a way to identify FunctionDeclaration as FunctionDeclarationObject subclasses. But then, I would need to extend the XFeatureCall default scoping to include classes in order to making it work...
I fully realize the question is not obvious, sorry...
Thanks,
Martin
EDIT: modified DoSleepCallback declaration from static to abstract (was erroneous)
I don't think you can generate what you need using the jvm model inferrer.
You should provide your own subclass of the XbaseCompiler (or JBaseCompiler, if any... and don't forget to register with guice in your runtime module), and override doInternalToJavaStatement(XExpression expr, ITreeAppendable it, boolean isReferenced) to manage how your FunctionDeclaration should be generated.
Suppose I'm working with the following C snippet:
void inc(int *num) {*num++;}
void dec(int *num) {*num--;}
void f(int var) {
inc(&var);
dec(&var);
}
By using a static analyzer, I want to be able to tell if the value of var didn't change during the function's execution. I know I have to keep its state on my own (that's the point of writing a Clang checker), but I'm having troubles getting a unique reference of this variable.
For example: if I use the following API
void MySimpleChecker::checkPostCall(const CallEvent &Call,
CheckerContext &C) const {
SymbolRef MyArg = Call.getArgSVal(0).getAsSymbol();
}
I'd expect it to return a pointer to this symbol's representation in my checker's context. However, I always get 0 into MyArg by using it this way. This happens for both inc and dec functions in the pre and post callbacks.
What am I missing here? What concepts did I get wrong?
Note: I'm currently reading the Clang CFE Internals Manual and I've read the excellent How to Write a Checker in 24 Hours material. I still couldn't find my answer so far.
Interpretation of question
Specifically, you want to count the calls to inc and dec applied to each variable and report when they do not balance for some path in a function.
Generally, you want to know how to associate an abstract value, here a number, with a program variable, and be able to update and query that value along each execution path.
High-level answer
Whereas the tutorial checker SimpleStreamChecker.cpp associates an abstract value with the value stored in a variable, here we want associate an abstract value with the variable itself. That is what IteratorChecker.cpp does when tracking containers, so I based my solution on it.
Within the static analyzer's abstract state, each variable is represented by a MemRegion object. So the first step is to make a map where MemRegion is the key:
REGISTER_MAP_WITH_PROGRAMSTATE(TrackVarMap, MemRegion const *, int)
Next, when we have an SVal that corresponds to a pointer to a variable, we can use SVal::getAsRegion to get the corresponding MemRegion. For instance, given a CallEvent, call, with a first argument that is a pointer, we can do:
if (MemRegion const *region = call.getArgSVal(0).getAsRegion()) {
to get the region that the pointer points at.
Then, we can access our map using that region as its key:
state = state->set<TrackVarMap>(region, newValue);
Finally, in checkDeadSymbols, we use SymbolReaper::isLiveRegion to detect when a region (variable) is going out of scope:
const TrackVarMapTy &Map = state->get<TrackVarMap>();
for (auto const &I : Map) {
MemRegion const *region = I.first;
int delta = I.second;
if (SymReaper.isLiveRegion(region) || (delta==0))
continue; // Not dead, or unchanged; skip.
Complete example
To demonstrate, here is a complete checker that reports unbalanced use of inc and dec:
// TrackVarChecker.cpp
// https://stackoverflow.com/questions/23448540/how-to-keep-track-of-a-variable-with-clangs-static-analyzer
#include "clang/StaticAnalyzer/Checkers/BuiltinCheckerRegistration.h"
#include "clang/StaticAnalyzer/Core/BugReporter/BugType.h"
#include "clang/StaticAnalyzer/Core/Checker.h"
#include "clang/StaticAnalyzer/Core/CheckerManager.h"
#include "clang/StaticAnalyzer/Core/PathSensitive/CallEvent.h"
#include "clang/StaticAnalyzer/Core/PathSensitive/CheckerContext.h"
#include "clang/StaticAnalyzer/Core/PathSensitive/ProgramState.h"
#include "clang/StaticAnalyzer/Core/PathSensitive/ProgramStateTrait.h"
using namespace clang;
using namespace ento;
namespace {
class TrackVarChecker
: public Checker< check::PostCall,
check::DeadSymbols >
{
mutable IdentifierInfo *II_inc, *II_dec;
mutable std::unique_ptr<BuiltinBug> BT_modified;
public:
TrackVarChecker() : II_inc(nullptr), II_dec(nullptr) {}
void checkPostCall(CallEvent const &Call, CheckerContext &C) const;
void checkDeadSymbols(SymbolReaper &SymReaper, CheckerContext &C) const;
};
} // end anonymous namespace
// Map from memory region corresponding to a variable (that is, the
// variable itself, not its current value) to the difference between its
// current and original value.
REGISTER_MAP_WITH_PROGRAMSTATE(TrackVarMap, MemRegion const *, int)
void TrackVarChecker::checkPostCall(CallEvent const &call, CheckerContext &C) const
{
const FunctionDecl *FD = dyn_cast<FunctionDecl>(call.getDecl());
if (!FD || FD->getKind() != Decl::Function) {
return;
}
ASTContext &Ctx = C.getASTContext();
if (!II_inc) {
II_inc = &Ctx.Idents.get("inc");
}
if (!II_dec) {
II_dec = &Ctx.Idents.get("dec");
}
if (FD->getIdentifier() == II_inc || FD->getIdentifier() == II_dec) {
// We expect the argument to be a pointer. Get the memory region
// that the pointer points at.
if (MemRegion const *region = call.getArgSVal(0).getAsRegion()) {
// Increment the associated value, creating it first if needed.
ProgramStateRef state = C.getState();
int delta = (FD->getIdentifier() == II_inc)? +1 : -1;
int const *curp = state->get<TrackVarMap>(region);
int newValue = (curp? *curp : 0) + delta;
state = state->set<TrackVarMap>(region, newValue);
C.addTransition(state);
}
}
}
void TrackVarChecker::checkDeadSymbols(
SymbolReaper &SymReaper, CheckerContext &C) const
{
ProgramStateRef state = C.getState();
const TrackVarMapTy &Map = state->get<TrackVarMap>();
for (auto const &I : Map) {
// Check for a memory region (variable) going out of scope that has
// a non-zero delta.
MemRegion const *region = I.first;
int delta = I.second;
if (SymReaper.isLiveRegion(region) || (delta==0)) {
continue; // Not dead, or unchanged; skip.
}
//llvm::errs() << region << " dead with delta " << delta << "\n";
if (ExplodedNode *N = C.generateNonFatalErrorNode()) {
if (!BT_modified) {
BT_modified.reset(
new BuiltinBug(this, "Delta not zero",
"Variable changed from its original value."));
}
C.emitReport(llvm::make_unique<BugReport>(
*BT_modified, BT_modified->getDescription(), N));
}
}
}
void ento::registerTrackVarChecker(CheckerManager &mgr) {
mgr.registerChecker<TrackVarChecker>();
}
bool ento::shouldRegisterTrackVarChecker(const LangOptions &LO) {
return true;
}
To hook this in to the rest of Clang, add entries to:
clang/include/clang/StaticAnalyzer/Checkers/Checkers.td and
clang/lib/StaticAnalyzer/Checkers/CMakeLists.txt
Example input to test it:
// trackvar.c
// Test for TrackVarChecker.
// The behavior of these functions is hardcoded in the checker.
void inc(int *num);
void dec(int *num);
void call_inc(int var) {
inc(&var);
} // reported
void call_inc_dec(int var) {
inc(&var);
dec(&var);
} // NOT reported
void if_inc(int var) {
if (var > 2) {
inc(&var);
}
} // reported
void indirect_inc(int val) {
int *p = &val;
inc(p);
} // reported
Sample run:
$ gcc -E -o trackvar.i trackvar.c
$ ~/bld/llvm-project/build/bin/clang -cc1 -analyze -analyzer-checker=alpha.core.TrackVar trackvar.i
trackvar.c:10:1: warning: Variable changed from its original value
}
^
trackvar.c:21:1: warning: Variable changed from its original value
}
^
trackvar.c:26:1: warning: Variable changed from its original value
}
^
3 warnings generated.
I think you missed the check that this call event is a call to your function inc/dec. You should have something like
void MySimpleChecker::checkPostCall(const CallEvent &Call,
CheckerContext &C) const {
const IdentifierInfo* callee = Call.getCalleeIdentifier();
if (callee->getName().str() == "inc" || callee->getName().str() == "dec")
SymbolRef MyArg = Call.getArgSVal(0).getAsSymbol();
}
I would like to bind the operator() using Boost::Python but I don't really see how to do this. Consider the example:
C++:
class Queuer
{
public:
void Queuer::operator()(const qfc::Queue & iq, const qfc::Message & im) const;
void Queuer::operator()(const qfc::Agent & ia, const qfc::Message & im) const;
// some other overloaded operator() methods
};
So in a Python script, after importing the module I'm using (called qfc), I would like to do:
Python:
>>> queuer = qfc.Queuer()
// instantiating a Message an Agent and a Queue object
>>> queuer(queue,message)
>>> queuer(agent,message)
>>> ...
Would you have any idea on how to do it? maybe with boost::python call<>?
Thank you,
Kevin
When exposing the Queuer class, define a __call__ method for each Queuer::operator() member function. Boost.Python will handle the appropriate dispatching based on types. The only complexity is introduced with pointer-to-member-function syntax, as the caller is required to disambiguate &Queuer::operator().
Additionally, when attempting to pass derived classes in Python to a C++ function with a parameter of the Base class, then some additional information needs to be exposed to Boost.Python:
The base C++ class needs to be exposed with class_. For example, class_<BaseType>("Base").
The derived class needs to explicitly list its base classes when being exposed with bases_. For example, class_<DerivedType, bases<BaseType> >("Derived"). With this information, Boost.Python can do proper casting while dispatching.
Here is a complete example:
#include <iostream>
#include <boost/python.hpp>
// Mockup classes.
struct AgentBase {};
struct MessageBase {};
struct QueueBase {};
struct SpamBase {};
struct Agent: AgentBase {};
struct Message: MessageBase {};
struct Queue: QueueBase {};
struct Spam: SpamBase {};
// Class with overloaded operator().
class Queuer
{
public:
void operator()(const AgentBase&, const MessageBase&) const
{
std::cout << "Queuer::operator() with Agent." << std::endl;
}
void operator()(const QueueBase&, const MessageBase&) const
{
std::cout << "Queuer::operator() with Queue." << std::endl;
}
void operator()(const SpamBase&, const MessageBase&) const
{
std::cout << "Queuer::operator() with Spam." << std::endl;
}
};
/// Depending on the overlaod signatures, helper types may make the
/// code slightly more readable by reducing pointer-to-member-function syntax.
template <typename A1>
struct queuer_overload
{
typedef void (Queuer::*type)(const A1&, const MessageBase&) const;
static type get(type fn) { return fn; }
};
BOOST_PYTHON_MODULE(example)
{
namespace python = boost::python;
// Expose only the base class types. Do not allow the classes to be
// directly initialized in Python.
python::class_<AgentBase >("AgentBase", python::no_init);
python::class_<MessageBase>("MessageBase", python::no_init);
python::class_<QueueBase >("QueueBase", python::no_init);
python::class_<SpamBase >("SpamBase", python::no_init);
// Expose the user types. These classes inerit from their respective
// base classes.
python::class_<Agent, python::bases<AgentBase> >("Agent");
python::class_<Message, python::bases<MessageBase> >("Message");
python::class_<Queue, python::bases<QueueBase> >("Queue");
python::class_<Spam, python::bases<SpamBase> >("Spam");
// Disambiguate via a varaible.
queuer_overload<AgentBase>::type queuer_op_agent = &Queuer::operator();
python::class_<Queuer>("Queuer")
// Disambiguate via a variable.
.def("__call__", queuer_op_agent)
// Disambiguate via a helper type.
.def("__call__", queuer_overload<QueueBase>::get(&Queuer::operator()))
// Disambiguate via explicit cast.
.def("__call__",
static_cast<void (Queuer::*)(const SpamBase&,
const MessageBase&) const>(
&Queuer::operator()))
;
}
And its usage:
>>> import example
>>> queuer = example.Queuer()
>>> queuer(example.Agent(), example.Message())
Queuer::operator() with Agent.
>>> queuer(example.Queue(), example.Message())
Queuer::operator() with Queue.
>>> queuer(example.Spam(), example.Message())
Queuer::operator() with Spam.
Thanks for your help.
Actually I've already tested the static cast solution. In reality, I need to pass a qfc::lqs::Message or qfc::lqs::Agent or qfc::lqs::Spam when invoking queuer(). qfc::lqs::Message for example, as for qfc::lqs::Agent inherit from qfc::Message and qfc::Agent respectively.
So can I "cast" qfc::lqs::Message, qfc::lqs::Agent and qfc::lqs::Spam to qfc::Message, qfc::Agent and qfc::Spam when invoking the operator() so that the signature corresponds to operator() ?
This to avoid the error shown below:
error: invalid static_cast from type '<unresolved overloaded function type>' to type 'void (qfc::lqs::Queuer::*)(const qfc::lqs::Queue&, const qfc::lqs::Message&)const'
Here is my code:
struct S {
int a, b;
float c, d;
};
class A {
private:
S* d;
S h[3];
public:
A() {
cutilSafeCall(cudaMalloc((void**)&d, sizeof(S)*3));
}
void Init();
};
void A::Init() {
for (int i=0;i<3;i++) {
h[i].a = 0;
h[i].b = 1;
h[i].c = 2;
h[i].d = 3;
}
cutilSafeCall(cudaMemcpy(d, h, 3*sizeof(S), cudaMemcpyHostToDevice));
}
A a;
In fact it is a complex program which contain CUDA and OpenGL. When I debug this program, it fails when running at cudaMemcpy with the error information
cudaSafeCall() Runtime API error 11: invalid argument.
Actually, this program is transformed from another one that can run correctly. But in that one, I used two variables S* d and S h[3] in the main function instead of in the class. What is more weird is that I implement this class A in a small program, it works fine.
And I've updated my driver, error still exists.
Could anyone give me a hint on why this happen and how to solve it. Thanks.
Because the memory operations in CUDA are blocking, they make a synchronization point. So other errors, if not checked with cudaThreadSynchonize, will seem like errors on the memory calls.
So if an error is received on a memory operation, try to place a cudaThreadSynchronize before it and check the result.
Be sure that the first malloc statement is being executed. If it is a problem about initialization of CUDA, like #Harrism indicate, then it would fail in this statement?? Try to place printf statements, and see proper initializations are performed. I think generally invalid argument errors are generated because of using uninitalized memory areas.
Write a printf to your constructor showing the address of the cudaMalloc'ed memory area
A()
{
d = NULL;
cutilSafeCall(cudaMalloc((void**)&d, sizeof(S)*3));
printf("D: %p\n", d);
}
Try to make a memory copy for an area that is locally allocated, namely move the cudaMalloc to above of cudaMemcopy (just for testing).
void A::Init()
{
for (int i=0;i<3;i++)
{
h[i].a = 0;
h[i].b = 1;
h[i].c = 2;
h[i].d = 3;
}
cutilSafeCall(cudaMalloc((void**)&d, sizeof(S)*3)); // here!..
cutilSafeCall(cudaMemcpy(d, h, 3*sizeof(S), cudaMemcpyHostToDevice));
}
Good luck.