[jira] Commented: (STDCXX-249) std::operator>>(istream, string&) inefficient

["Travis Vitek (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/STDCXX-249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12565855#action_12565855 ] 

Travis Vitek commented on STDCXX-249:
-------------------------------------

Performance test follows...

{noformat}
#include <iostream>
#include <fstream>
#include <string>

#include <stdlib.h>

int main (int argc, char* argv[])
{
    size_t loops = 1000;
    if (1 < argc)
        loops = strtoul (argv [1], 0, 0);

    const char* file = __FILE__;
    if (2 < argc)
        file = argv [2];

    std::ifstream ifs (file);
    if (!ifs.is_open ())
        return 1;

    size_t tokens = 0;

    std::string token;
    token.reserve (100);

    for (size_t i = 0; i < loops; ++i)
    {
        while (ifs >> token)
            ++tokens;

        ifs.clear ();
        ifs.seekg (0);
    }

    std::cout << tokens << std::endl;
    return 0;
}
{noformat}

Results on Linux/GCC 12D build before and after...

{noformat}
$ time ./t 10000 /nfs/devco/vitek/stdcxx/trunk/tests/strings/21.string.io.cpp
77630000

real    0m7.269s
user    0m6.722s
sys     0m0.466s

$ time ./t 10000 /nfs/devco/vitek/stdcxx/trunk/tests/strings/21.string.io.cpp
77630000

real    0m7.148s
user    0m6.667s
sys     0m0.142s
{noformat}

So I'm seeing almost no difference in terms of performance with a release build on Linux. There is a difference for an 11s build...

{noformat}
$ time ./t 1000 /nfs/devco/vitek/stdcxx/trunk/tests/strings/21.string.io.cpp
7763000

real    0m6.753s
user    0m6.521s
sys     0m0.210s

$ time ./t 1000 /nfs/devco/vitek/stdcxx/trunk/tests/strings/21.string.io.cpp
7763000

real    0m3.781s
user    0m3.755s
sys     0m0.017s
{noformat}



> std::operator>>(istream, string&) inefficient
> ---------------------------------------------
>
>                 Key: STDCXX-249
>                 URL: https://issues.apache.org/jira/browse/STDCXX-249
>             Project: C++ Standard Library
>          Issue Type: Improvement
>          Components: 21. Strings
>    Affects Versions: 4.1.2, 4.1.3, 4.1.4, 4.2.0
>         Environment: all
>            Reporter: Martin Sebor
>            Assignee: Travis Vitek
>            Priority: Minor
>             Fix For: 4.2.1
>
>   Original Estimate: 8h
>  Remaining Estimate: 8h
>
> The string extractor reads and appends one character at a time. It could be made more efficient by extracting and appending all non-whitespace characters that are available in the buffer in chunks. See the definition of the operator here:
> http://svn.apache.org/viewvc/stdcxx/trunk/include/istream.cc?revision=383265

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.