svn commit: r279197 - head

[Garrett Cooper]

On Feb 23, 2015, at 11:42, Rui Paulo <rpaulo at me.com> wrote:

> How much time did this save?

Test profile:

- 4 x CPU VMware Fusion hosted amd64 VM with 4GB allocated to it and with open-vm-tools installed
- Kernel with WITNESS watch turned on.
- Multiuser mode (because I’m lazy)

The worst case scenario was tested:
- All BOOTSTRAPPING checks were removed
- The default SRCCONF values were set
- I built with i386 instead of amd64 (to force elftoolchain and the dtrace tools to always be built).

Observations:

-j1: standard deviation was huge (5.75 is not statistically accurate), so I threw out a few outliers and redid the numbers.
-j1 vs -j4: ~3 minutes faster -> ~3x speedup
-j4 vs -j8:  -> no real noticeable change, but the outliers were statistically larger.

Cheers!

========================

-j1:

$ grep ' 1:.*real' ts | awk 'NR > 1 { print $3 }' | ministat 
x <stdin>
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|                  x                                                                                                                                                                                       |
|       x          x    x                   x       x        x                                          x                                                                                           x     x|
||__________________________________________________M____________________A_______________________________________________________________________|                                                         |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10        259.57        275.06        263.04       264.747     5.7545094

-j1 (statistically more accurate sample, but smaller sample size):

$ grep ' 1:.*real' ts | awk 'NR > 1 && NR < 8 { print $3 }' | ministat
x <stdin>
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|    xx                  x                                                                                                 x                                 x                                            x|
||____________________________________________________________________________________A____________________________________M________________________________________________|                              |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x   6        260.45         263.8        262.46     261.83667     1.4530336

-j4:

$ grep ' 4:.*real' ts | awk 'NR > 1 { print $3 }' | ministat
x <stdin>
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|x    x                x             x                                                           x  x                                                x           x                                       xx|
|                  |______________________________________________________________________________A_M___________________________________________________________________________|                          |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10         87.78         90.34         89.04        89.011    0.99822565

$ grep ‘ 8:.*real' ts | awk 'NR > 1 { print $3 }' | ministat                                                                                                       
x <stdin>
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
|x                                                 x     x  x                     x                  x           x                     x        x                                                         x|
|                                     |________________________________________________________A_____M__________________________________________________|                                                  |
+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  10         87.66         93.91         90.77        90.572     1.7769249

-j1 (statistically inaccurate) vs -j4:

$ python2.7 -c 'm1 = 264.747; m4 = 89.011; print (m1/m4), (m1-m4)'                                                                                                   
2.97431778095 175.736

-j1 (statistically accurate) vs -j4:

$ python2.7 -c 'm1 = 261.83667; m4 = 89.011; print (m1/m4), (m1-m4)'                                                                                                  
2.94162148499 172.82567
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 496 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.freebsd.org/pipermail/svn-src-all/attachments/20150226/769ebdfc/attachment.sig>