RE: Re: Keeping track of the CPU time for a thread
Thank you for your reply. Just to test the functionality, I tried a very simple thread. It computes a factorial for some number of iterations. void entry_point(void * param) { int num_run = 1000000; int number = 15; int fact; thread_t self = env()->self(); for(int i=0;i<num_run;i++) { fact = 1; for(int j = 2; j<= number; j++) { fact = fact * j; } } printf("Factorial of %d is : %d\n", number, fact); l4_thread_stats_time(self.cap()); printf("Consumed time: %llu\n",l4_utcb_mr()->mr[0]); } What I found is - the consumed time does not depend on the number of iterations (num_run) - I tried from 1 to a very large number. It does vary with the number. Am I missing something? Thanks again. ---------------------------------------------------------------------- Date: Sun, 24 Jun 2012 23:29:25 +0200 From: Adam Lackorzynski <adam@os.inf.tu-dresden.de> To: l4-hackers@os.inf.tu-dresden.de Subject: Re: Keeping track of the CPU time for a thread Message-ID: <20120624212925.GC1246@os.inf.tu-dresden.de> Content-Type: text/plain; charset=iso-8859-1 Hi, On Thu Jun 21, 2012 at 17:44:25 +0000, Mohammad A Haque wrote:
I want to measure how much CPU time a thread has consumed. When the CPU time consumption exceeds a certain limit (specific to that thread), I want to raise an interrupt. Please suggest me how can I do it efficiently?
The L4::Thread interface has the stats_time function to get the consumed time of a thread (FINE_GRAINED_CPUTIME in the kernel might be of interest here). Triggering an interrupt goes with its trigger function. Then it's just about plugging this together. Adam -- Adam adam@os.inf.tu-dresden.de Lackorzynski http://os.inf.tu-dresden.de/~adam/
On Tue Jun 26, 2012 at 18:52:44 +0000, Mohammad A Haque wrote:
Thank you for your reply.
Just to test the functionality, I tried a very simple thread. It computes a factorial for some number of iterations.
void entry_point(void * param) { int num_run = 1000000; int number = 15; int fact;
thread_t self = env()->self(); for(int i=0;i<num_run;i++) { fact = 1; for(int j = 2; j<= number; j++) { fact = fact * j; } } printf("Factorial of %d is : %d\n", number, fact); l4_thread_stats_time(self.cap()); printf("Consumed time: %llu\n",l4_utcb_mr()->mr[0]); }
What I found is - the consumed time does not depend on the number of iterations (num_run) - I tried from 1 to a very large number. It does vary with the number. Am I missing something?
Your examples works for me (in a slightly modified version): int num_run = 1000000; int number = 15; int fact; while (1) { for(int i=0;i<num_run;i++) { fact = 1; for(int j = 2; j<= number; j++) { fact = fact * j; } } printf("Factorial of %d is : %d\n", number, fact); l4_thread_stats_time(l4re_env()->main_thread); printf("Consumed time: %llu\n", *(l4_cpu_time_t *)&l4_utcb_mr()->mr[0]); } The consumed time increases as the program runs. The granularity is 1ms, except you enable "fine-grained cpu time" in the Fiasco kernel, then it is more exact. Adam -- Adam adam@os.inf.tu-dresden.de Lackorzynski http://os.inf.tu-dresden.de/~adam/
participants (2)
-
Adam Lackorzynski -
Mohammad A Haque