Different way of using times()

[itpp14]

Hello, I have a bit of linux code that I can't get working under windows, can anyone rewrite this bit so it compiles in c++ ?

if ((start = times(&tmsstart)) == (clock_t)-1)
[...]
if ((end = times(&tmsend)) == (clock_t)-1)

Full source is here;
https://github.com/nbs-system/naxsi/blob/master/naxsi_src/naxsi_skeleton.c
Lines 1097 to 1108

[Duthomhas]

I don't have the time to look through all that (no pun intended), but the equivalent function in Windows is GetProcessTimes()
http://www.google.com/search?btnI=1&q=msdn+GetProcessTimes

Good luck!

[itpp14]

This is what I was looking for;

#ifdef _MSC_VER
    start = clock();   
    ngx_http_dummy_data_parse(ctx, r);
    cf->request_processed++;
    end = clock(); 
#else
    if ((start = times(&tmsstart)) == (clock_t)-1)
      ngx_log_debug0(NGX_LOG_DEBUG_HTTP, r->connection->log, 0,
		    "XX-dummy : Failed to get time");
    ngx_http_dummy_data_parse(ctx, r);
    cf->request_processed++;
    if ((end = times(&tmsend)) == (clock_t)-1)
      ngx_log_debug0(NGX_LOG_DEBUG_HTTP, r->connection->log, 0,
		    "XX-dummy : Failed to get time");
#endif
    if (end - start > 10) // report if it took more than 1/10MS to perform all the checks

Any comments ? it compiles, links and works, but not sure if the 1/10ms test works the same.

[Duthomhas]

I'm not sure.

The clock() function returns the number of ticks that pass only while your process is running.

The times() function returns system time.


What's wrong with using GetProcessTimes()?

[itpp14]

clock() looks simpler, I just googled how to get milliseconds in c++ since the test is about more then 10ms this looks ok to me, what makes GetProcessTimes() work better then clock() in this test ?

It is used in a non-blocking and event driven module so getting time past while the thread is running sounds like its suppose to work.

And is it this simple as;

start = GetProcessTimes(); ??

[itpp14]

Ok, changed it into:

#ifdef _MSC_VER
    start = GetProcessTimes(GetCurrentProcess(), &createTime, &exitTime, &sysTime, &cpuTime);
    ngx_http_dummy_data_parse(ctx, r);
    cf->request_processed++;
    end = GetProcessTimes(GetCurrentProcess(), &createTime2, &exitTime2, &sysTime2, &cpuTime2);
#else
    if ((start = times(&tmsstart)) == (clock_t)-1)
      ngx_log_debug0(NGX_LOG_DEBUG_HTTP, r->connection->log, 0,
		    "XX-dummy : Failed to get time");
    ngx_http_dummy_data_parse(ctx, r);
    cf->request_processed++;
    if ((end = times(&tmsend)) == (clock_t)-1)
      ngx_log_debug0(NGX_LOG_DEBUG_HTTP, r->connection->log, 0,
		    "XX-dummy : Failed to get time");
#endif
    if (end - start > 10)

Compiles and links ok, but still the same question, not sure if the 1/10ms test works the same.
And which codeblock is the best in this case.
Some advice which is better and why would be appreciated.

[Duthomhas]

Did you read the documentation I linked?

Because the function doesn't work that way. Make yourself a helper function. You'll have to parse the time data out of the FILETIME struct.

BTW, I just noticed that your last line:

    if (end - start > 10) // report if it took more than 1/10MS to perform all the checks

isn't quite right. The amount of time in a clock tick isn't necessarily 1/10MS. You should be using the CLOCKS_PER_SEC value to help you here.


If you want another way of looking at simplified precision timing, check out this post:
http://www.cplusplus.com/forum/beginner/28855/3/#msg159698