I'm working with some existing code which is deserializing objects stored in text files (I potentially need to read tens of millions of these). The contents of the file are first read into a wstring and then it makes a wistringstream from that. Running the Very Sleepy profiler on the program shows that it is spending about 20% of its time in the following call stacks:
Mtxlock or RtlEnterCritialSection
std::_Mutex::_Lock
std::flush
std::basic_istream<wchar_t, std::char_traits<wchar_t> >::get
<rest of my program>
and similar ones with std::_Mutex::_Unlock. I'm using Visual C++ 2008.
Looking in istream, I see that it constructs a sentry object which calls _Lock and _Unlock methods on the underlying basic_streambuf. This in turn just call _Lock and _Unlock on a _Mutex associated with that buffer. These are then defined as follows:
#if _MULTI_THREAD
// actually defines non-empty _Lock() and _Unlock() methods
#else /* _MULTI_THREAD */
void _Lock()
{ // do nothing
}
void _Unlock()
{ // do nothing
}
#endif /* _MULTI_THREAD */
It looks like _MULTI_THREAD is set in yvals.h as
#define _MULTI_THREAD 1 /* nontrivial locks if multithreaded */
Now, I know there will never be another thread trying to access this buffer, but it looks to me like there's no way around this locking while using the standard iostreams, which seems both odd and frustrating. Am I missing something? Is there a workaround for this?
Check the value for Runtime Library in Project properties, C/C++, Code Generation. If it's multi-threaded, change it to a non-multithreaded version.
In any version after Visual C++ 7.1 (!), you are out of luck as it's been removed, and you are stuck with the multithreaded CRT.
The std::flush seems senseless in your case. I can't see how you'd flush an istream, so I suspect it's a result of a tie. You may want to un-tie, i.e. call tie(NULL) on your wistringstream. That should also reduce the number of locks taken.
It turned out accessing the underlying buffer directly by replacing things like
c = _text_in->get();
with things like this
c = _text_in->rdbuf()->sbumpc();
fixed the problem and provided a big boost to performance.
Related
Using Visual Studio 2019 Professional on Windows 10 x64. I have several C++ DLL projects, some of which are multi-threaded. I'm using CRITICAL_SECTION objects for thread safety.
In DLL1:
CRITICAL_SECTION critDLL1;
InitializeCriticalSection(&critDLL1);
In DLL2:
CRITICAL_SECTION critDLL2;
InitializeCriticalSection(&critDLL2);
When I use critDLL1 with EnterCriticalSection or LeaveCriticalSection everything is fine in both _DEBUG or NDEBUG mode. But when I use critDLL2, I get an access violation in 'ntdll.dll' in NDEBUG (though not in _DEBUG).
After popping up message boxes in NDEBUG mode, I was eventually able to track the problem down to the first use of EnterCriticalSection.
What might be causing the CRITICAL_SECTION to fail in one project but work in others? The MSDN page was not helpful.
UPDATE 1
After comparing project settings of DLL1 (working) and DLL2 (not working), I've accidentally got DLL2 working. I've confirmed this by reverting to an earlier version (which crashes) and then making the project changes (no crash!).
This is the setting:
Project Properties > C/C++ > Optimization > Whole Program Optimization
Set this to Yes (/GL) and my program crashes. Change that to No and it works fine. What does the /GL switch do and why might it cause this crash?
UPDATE 2
The excellent answer from #Acorn and comment from #RaymondChen, provided the clues to track down and then resolve the issue. There were two problems (both programmer errors).
PROBLEM 1
The assumption of Whole Program Optimzation (wPO) is the MSVC compiler is compiling "the whole program". This is an incorrect assumption for my DLL project which internally consumes a 3rd party library and is in turn consumed by an external application written in Delphi. This setting is set to Yes (/GL) by default but should be No. This feels like a bug in Visual Studio, but in any case, the programmer needs to be aware of this. I don't know all the details of what WPO is meant to do, but at least for DLLs meant to be consumed by other applications, the default should be changed.
PROBLEM 2
Serious programmer error. It was a call into a 3rd party library, which returned a 128-byte ASCII code which was the error:
// Before
// m_config::acSerial defined as "char acSerial[21]"
(void) m_pLib->GetPara(XPARA_PRODUCT_INFO, &m_config.acSerial[0]);
EnterCriticalSection(&crit); // Crash!
// After
#define SERIAL_LEN 20
// m_config::acSerial defined as "char acSerial[SERIAL_LEN+1]"
//...
char acSerial[128];
(void) m_pLib->GetPara(XPARA_PRODUCT_INFO, &acSerial[0]);
strncpy(m_config.acSerial, acSerial, max(SERIAL_LEN, strlen(acSerial)));
EnterCriticalSection(&crit); // Works!
The error, now obvious, is that the 3rd party library did not copy the serial number of the device into the char* I provided...it copied 128 bytes into my char* stomping over everything contiguous in memory after acSerial. This wasn't noticed before because m_pLib->GetPara(XPARA_PRODUCT_INFO, ...) was one of the first calls into the 3rd party library and the rest of the contiguous data was mostly NULL at that point.
The problem was never to do with the CRITICAL_SECTION. My thanks for Acorn and RaymondChen ... sanity has been restored to this corner of the universe.
If your program crashes under WPO (an optimization that assumes that whatever you are compiling is the entire program), it means that either the assumption is incorrect or that the optimizer ends up exploiting some undefined behavior that previously didn't (without the optimization applied), even if the assumption is correct.
In general, avoid enabling optimizations unless you are really sure you know you meet their requirements.
For further analysis, please provide a MRE.
I am writing a scientific computation code in C++. There are outputs that I want to write in a console and outputs that I write into a file. However, when debugging after implementing a new feature, it is useful to print out much more information than usual. So far I was just sending more information to std::cout/clog and commented these lines out when not needed.
What I want is like std::clog, which would go into a file when needed, or not do anything at all, when not needed. It is ok, if I need to recompile the code to switch between the two regimes. It is important that nothing happens when not needed, because for a real large calculation the log file would be enormous (or the the console full of rubbish) and all the writing would slow the calculation down.
I am looking for the smallest possible implementation, ideally using only standard libraries for portability.
The obvious solution is to have a global variable, redirect clog to a file and then use an if statement.
bool DEBUG = true;
std::ofstream out("logfile");
std::clog.rdbuf(out.rdbuf());
...
if (DEBUG) std::clog << "my message" << std::endl;
...
Is there a more elegant way of doing this?
Edit:
I would like to avoid using non-standard libraries and preprocessor macros (program is spread across many files and also a bad programming habit in general). One way I could imagine this working, but I don't know how to do it, is to create a globally accessible object that would be able to accept messages using << and would save them to a file. Then I could just comment out the line inside this object class that saves it to a file. However, I don't know how much performance impact may result from passing messages to such a disfunctional object.
You may use any external logging library for C/C++.
Or create your own small implementation with only utilities what you need.
A traditional logging mechanism is build on macros (or inline functions) and looks like:
#define LOG_MESSAGE(msg) \
{
#ifdef DEBUG
// your debug logging
#else
// your release logging, may be leaved empty
#endif // DEBUG
}
It's also useful to add different logging levels: Error, Warning, Info, etc.
I'm fairly new to c++ and am really interested in learning more. Have been reading quite a bit. Recently discovered the init/fini elf sections.
I started to wonder if & how one would use the init section to prepopulate objects that would be used at runtime. Say for example you wanted
to add performance measurements to your code, recording the time, filename, linenumber, and maybe some ID (monotonic increasing int for ex) or name.
You would place for example:
PROBE(0,"EventProcessing",__FILE__,__LINE__)
...... //process event
PROBE(1,"EventProcessing",__FILE__,__LINE__)
......//different processing on same event
PROBE(2,"EventProcessing",__FILE__,__LINE__)
The PROBE could be some macro that populates a struct containing this data (maybe on an array/list, etc using the id as an indexer).
Would it be possible to have code in the init section that could prepopulate all of this data for each PROBE (except for the time of course), so only the time would need to be retrieved/copied at runtime?
As far as I know the __attribute__((constructor)) can not be applied to member functions?
My initial idea was to create some kind of
linked list with each node pointing to each probe and code in the init secction could iterate it populating the id, file, line, etc, but
that idea assumed I could use a member function that could run in the "init" section, but that does not seem possible. Any tips appreciated!
As far as I understand it, you do not actually need an ELF constructor here. Instead, you could emit descriptors for your probes using extended asm statements (using data, instead of code). This also involves switching to a dedicated ELF section for the probe descriptors, say __probes.
The linker will concatenate all the probes and in an array, and generate special symbols __start___probes and __stop___probes, which you can use from your program to access thes probes. See the last paragraph in Input Section Example.
Systemtap implements something quite similar for its userspace probes:
User Space Probe Implementation
Adding User Space Probing to an Application (heapsort example)
Similar constructs are also used within the Linux kernel for its self-patching mechanism.
There's a pretty simple way to have code run on module load time: Use the constructor of a global variable:
struct RunMeSomeCode
{
RunMeSomeCode()
{
// your code goes here
}
} do_it;
The .init/.fini sections basically exist to implement global constructors/destructors as part of the ABI on some platforms. Other platforms may use different mechanisms such as _start and _init functions or .init_array/.deinit_array and .preinit_array. There are lots of subtle differences between all these methods and which one to use for what is a question that can really only be answered by the documentation of your target platform. Not all platforms use ELF to begin with…
The main point to understand is that things like the .init/.fini sections in an ELF binary happen way below the level of C++ as a language. A C++ compiler may use these things to implement certain behavior on a certain target platform. On a different platform, a C++ compiler will probably have to use different mechanisms to implement that same behavior. Many compilers will give you tools in the form of language extensions like __attributes__ or #pragmas to control such platform-specific details. But those generally only make sense and will only work with that particular compiler on that particular platform.
You don't need a member function (which gets a this pointer passed as an arg); instead you can simply create constructor-like functions that reference a global array, like
#define PROBE(id, stuff, more_stuff) \
__attribute__((constructor)) void \
probeinit##id(){ probes[id] = {id, stuff, 0/*to be written later*/, more_stuff}; }
The trick is having this macro work in the middle of another function. GNU C / C++ allows nested functions, but IDK if you can make them constructors.
You don't want to declare a static int dummy#id = something because then you're adding overhead to the function you profile. (gcc has to emit a thread-safe run-once locking mechanism.)
Really what you'd like is some kind of separate pass over the source that identifies all the PROBE macros and collects up their args to declare
struct probe global_probes[] = {
{0, "EventName", 0 /*placeholder*/, filename, linenum},
{1, "EventName", 0 /*placeholder*/, filename, linenum},
...
};
I'm not confident you can make that happen with CPP macros; I don't think it's possible to #define PROBE such that every time it expands, it redefines another macro to tack on more stuff.
But you could easily do that with an awk/perl/python / your fave scripting language program that scans your program and constructs a .c that declares an array with static storage.
Or better (for a single-threaded program): keep the runtime timestamps in one array, and the names and stuff in a separate array. So the cache footprint of the probes is smaller. For a multi-threaded program, stores to the same cache line from different threads is called false sharing, and creates cache-line ping-pong.
So you'd have #define PROBE(id, evname, blah blah) do { probe_times[id] = now(); }while(0)
and leave the handling of the later args to your separate preprocessing.
Is it possible to have a function like this:
const char* load(const char* filename_){
return
#include filename_
;
};
so you wouldn't have to hardcode the #include file?
Maybe with a some macro?
I'm drawing a blank, guys. I can't tell if it's flat out not possible or if it just has a weird solution.
EDIT:
Also, the ideal is to have this as a compile time operation, otherwise I know there's more standard ways to read a file. Hence thinking about #include in the first place.
This is absolutely impossible.
The reason is - as Justin already said in a comment - that #include is evaluated at compile time.
To include files during run time would require a complete compiler "on board" of the program. A lot of script languages support things like that, but C++ is a compiled language and works different: Compile and run time are strictly separated.
You cannot use #include to do what you want to do.
The C++ way of implementing such a function is:
Find out the size of the file.
Allocate memory for the contents of the file.
Read the contents of the file into the allocated memory.
Return the contents of the file to the calling function.
It will better to change the return type to std::string to ease the burden of dealing with dynamically allocated memory.
std::string load(const char* filename)
{
std::string contents;
// Open the file
std::ifstream in(filename);
// If there is a problem in opening the file, deal with it.
if ( !in )
{
// Problem. Figure out what to do with it.
}
// Move to the end of the file.
in.seekg(0, std::ifstream::end);
auto size = in.tellg();
// Allocate memory for the contents.
// Add an additional character for the terminating null character.
contents.resize(size+1);
// Rewind the file.
in.seekg(0);
// Read the contents
auto n = in.read(contents.data(), size);
if ( n != size )
{
// Problem. Figure out what to do with it.
}
contents[size] = '\0';
return contents;
};
PS
Using a terminating null character in the returned object is necessary only if you need to treat the contents of the returned object as a null terminated string for some reason. Otherwise, it maybe omitted.
I can't tell if it's flat out not possible
I can. It's flat out not possible.
Contents of the filename_ string are not determined until runtime - the content is unknown when the pre processor is run. Pre-processor macros are processed before compilation (or as first step of compilation depending on your perspective).
When the choice of the filename is determined at runtime, the file must also be read at runtime (for example using a fstream).
Also, the ideal is to have this as a compile time operation
The latest time you can affect the choice of included file is when the preprocessor runs. What you can use to affect the file is a pre-processor macro:
#define filename_ "path/to/file"
// ...
return
#include filename_
;
it is theoretically possible.
In practice, you're asking to write a PHP construct using C++. It can be done, as too many things can, but you need some awkward prerequisites.
a compiler has to be linked into your executable. Because the operation you call "hardcoding" is essential for the code to be executed.
a (probably very fussy) linker again into your executable, to merge the new code and resolve any function calls etc. in both directions.
Also, the newly imported code would not be reachable by the rest of the program which was not written (and certainly not compiled!) with that information in mind. So you would need an entry point and a means of exchanging information. Then in this block of information you could even put pointers to code to be called.
Not all architectures and OSes will support this, because "data" and "code" are two concerns best left separate. Code is potentially harmful; think of it as nitric acid. External data is fluid and slippery, like glycerine. And handling nitroglycerine is, as I said, possible. Practical and safe are something completely different.
Once the prerequisites were met, you would have two or three nice extra functions and could write:
void *load(const char* filename, void *data) {
// some "don't load twice" functionality is probably needed
void *code = compile_source(filename);
if (NULL == code) {
// a get_last_compiler_error() would be useful
return NULL;
}
if (EXIT_SUCCESS != invoke_code(code, data)) {
// a get_last_runtime_error() would also be useful
release_code(code);
return NULL;
}
// it is now the caller's responsibility to release the code.
return code;
}
And of course it would be a security nightmare, with source code left lying around and being imported into a running application.
Maintaining the code would be a different, but equally scary nightmare, because you'd be needing two toolchains - one to build the executable, one embedded inside said executable - and they wouldn't necessarily be automatically compatible. You'd be crying loud for all the bugs of the realm to come and rejoice.
What problem would be solved?
Implementing require_once in C++ might be fun, but you thought it could answer a problem you have. Which is it exactly? Maybe it can be solved in a more C++ish way.
A better alternative, considering also performances etc., to compile a loadable module beforehand, and load it at runtime.
If you need to perform small tunings to the executable, place parameters into an external configuration file and provide a mechanism to reload it. Once the modules conform to a fixed specification, you can even provide "plugins" that weren't available when the executable was first developed.
Having used gprof and callgrind many times, I have reached the (obvious) conclusion that I cannot use them efficiently when dealing with large (as in a CAD program that loads a whole car) programs. I was thinking that maybe, I could use some C/C++ MACRO magic and somehow build a simple (but nice) logging mechanism. For example, one can call a function using the following macro:
#define CALL_FUN(fun_name, ...) \
fun_name (__VA_ARGS__);
We could add some clocking/timing stuff before and after the function call, so that every function called with CALL_FUN gets timed, e.g
#define CALL_FUN(fun_name, ...) \
time_t(&t0); \
fun_name (__VA_ARGS__); \
time_t(&t1);
The variables t0, t1 could be found in a global logging object. That logging object can also hold the calling graph for each function called through CALL_FUN. Afterwards, that object can be written in a (specifically formatted) file, and be parsed from some other program.
So here comes my (first) question: Do you find this approach tractable ? If yes, how can it be enhanced, and if not, can you propose a better way to measure time and log callgraphs ?
A collegue proposed another approach to deal with this problem, which is annotating with a specific comment each function (that we care to log). Then, during the make process, a special preprocessor must be run, parse each source file, add logging logic for each function we care to log, create a new source file with the newly added (parsing) code, and build that code instead. I guess that reading CALL_FUN... macros (my proposal) all over the place is not the best approach, and his approach would solve this problem. So what is your opinion about this approach?
PS: I am not well versed in the pitfalls of C/C++ MACROs, so if this can be developed using another approach, please say it so.
Thank you.
Well you could do some C++ magic to embed a logging object. something like
class CDebug
{
CDebug() { ... log somehow ... }
~CDebug() { ... log somehow ... }
};
in your functions then you simply write
void foo()
{
CDebug dbg;
...
you could add some debug info
dbg.heythishappened()
...
} // not dtor is called or if function is interrupted called from elsewhere.
I am bit late, but here is what I am doing for this:
On Windows there is a /Gh compiler switch which makes the compiler to insert a hidden _penter function at the start of each function. There is also a switch for getting a _pexit call at the end of each function.
You can utilizes this to get callbacks on each function call. Here is an article with more details and sample source code:
http://www.johnpanzer.com/aci_cuj/index.html
I am using this approach in my custom logging system for storing the last few thousand function calls in a ring buffer. This turned out to be useful for crash debugging (in combination with MiniDumps).
Some notes on this:
The performance impact very much depends on your callback code. You need to keep it as simple as possible.
You just need to store the function address and module base address in the log file. You can then later use the Debug Interface Access SDK to get the function name from the address (via the PDB file).
All this works suprisingly well for me.
Many nice industrial libraries have functions' declarations and definitions wrapped into void macros, just in case. If your code is already like that -- go ahead and debug your performance problems with some simple asynchronous trace logger. If no -- post-insertion of such macros can be an unacceptably time-consuming.
I can understand the pain of running an 1Mx1M matrix solver under valgrind, so I would suggest starting with so called "Monte Carlo profiling method" -- start the process and in parallel run pstack repeatedly, say each second. As a result you will have N stack dumps (N can be quite significant). Then, the mathematical approach would be to count relative frequencies of each stack and make a conclusion about the ones most frequent. In practice you either immediately see the bottleneck or, if no, you switch to bisection, gprof, and finally to valgrind's toolset.
Let me assume the reason you are doing this is you want to locate any performance problems (bottlenecks) so you can fix them to get higher performance.
As opposed to measuring speed or getting coverage info.
It seems you're thinking the way to do this is to log the history of function calls and measure how long each call takes.
There's a different approach.
It's based on the idea that mainly the program walks a big call tree.
If time is being wasted it is because the call tree is more bushy than necessary,
and during the time that's being wasted, the code that's doing the wasting is visible on the stack.
It can be terminal instructions, but more likely function calls, at almost any level of the stack.
Simply pausing the program under a debugger a few times will eventually display it.
Anything you see it doing, on more than one stack sample, if you can improve it, will speed up the program.
It works whether or not the time is being spent in CPU, I/O or anything else that consumes wall clock time.
What it doesn't show you is tons of stuff you don't need to know.
The only way it can not show you bottlenecks is if they are very small,
in which case the code is pretty near optimal.
Here's more of an explanation.
Although I think it will be hard to do anything better than gprof, you can create a special class LOG for instance and instantiate it in the beginning of each function you want to log.
class LOG {
LOG(const char* ...) {
// log time_t of the beginning of the call
}
~LOG(const char* ...) {
// calculate the total time spent,
//by difference between current time and that saved in the constructor
}
};
void somefunction() {
LOG log(__FUNCTION__, __FILE__, ...);
.. do other things
}
Now you can integrate this approach with the preprocessing one you mentioned. Just add something like this in the beginning of each function you want to log:
// ### LOG
and then you replace the string automatically in debug builds (shoudn't be hard).
May be you should use a profiler. AQTime is a relatively good one for Visual Studio. (If you have VS2010 Ultimate, you already have a profiler.)