From: "tmm1 (Aman Gupta)" <ruby@...1.net>
Date: 2012-12-04T12:24:10+09:00
Subject: [ruby-core:50544] [ruby-trunk - Bug #7500] Improve GC profiler timings on linux


Issue #7500 has been updated by tmm1 (Aman Gupta).


Increasing the number of iterations shows the 4ms granularity.

  tmm1@fe19:~$ grep loops timings.c 
  int loops = 1000000;
      for (n=0; n<loops; n++) pow(2, 2048);
      for (n=0; n<loops; n++) pow(2, 2048);

  tmm1@fe19:~$ gcc -o timings timings.c -lrt && ./timings 
  getrusage() before: 0.000000
  getrusage() after: 0.004000
  clock_gettime() before: 0.006966
  clock_gettime() after: 0.010733

----------------------------------------
Bug #7500: Improve GC profiler timings on linux
https://bugs.ruby-lang.org/issues/7500#change-34387

Author: tmm1 (Aman Gupta)
Status: Open
Priority: Normal
Assignee: authorNari (Narihiro Nakamura)
Category: core
Target version: 2.0.0
ruby -v: ruby 2.0.0dev (2012-12-03 trunk 38149) [x86_64-darwin12.2.0]


On linux kernels, getrusage()'s precision depends on the value of HZ when the kernel was compiled. By default, HZ=250 provides a 4ms granularity.

This patch uses clock_gettime() with CLOCK_PROCESS_CPUTIME_ID when available, which provides a 1ns precision on linux.


-- 
http://bugs.ruby-lang.org/