Possible evidence of performance regression for 8.1-S (vs. 7.1)

[Julian Elischer]

On 10/25/10 9:55 AM, David Wolfskill wrote:
> On Wed, Oct 20, 2010 at 04:50:10PM -0700, Julian Elischer wrote:
>> ...
>> try the 7.x machine but running the 8.x kernel.. i.e. change nothing,
>> but boot the new kernel.
>> ...
> OK; here are results of previous tests, along with the above.  As noted
> earlier, I needed to set the UNAME_r environmant variable in order for
> the build to succeed with the 7.x userland&  8.x kernel:
>
>       start        stop      real      user       sys  host    os
> 1287436357  1287461948  25590.99  81502.22  18115.07  ref-8x  8.1-S
> 1287462797  1287488766  25969.26  81452.14  17920.14  ref-8x  8.1-S
> 1287489641  1287515287  25645.84  81548.40  18256.52  ref-8x  8.1-S
> 1287516151  1287541481  25329.64  81546.23  18294.10  ref-8x  8.1-S
> 1287542355  1287568599  26244.59  81431.47  17902.39  ref-8x  8.1-S
>
> 1287525363  1287546846  21483.13  82628.20  21703.09  ref-7x  7.1-R+
> 1287548005  1287569100  21094.63  82853.19  22185.02  ref-7x  7.1-R+
> 1287570300  1287591371  21071.33  82756.81  21943.22  ref-7x  7.1-R+
> 1287592592  1287614103  21511.23  82637.30  21849.90  ref-7x  7.1-R+
> 1287615323  1287636770  21446.42  82715.81  21708.97  ref-7x  7.1-R+
>
> 1287710312  1287732046  21733.20  82688.01  22108.95  ref-8x  7.1-R+
> 1287733360  1287754549  21188.88  82869.09  21890.83  ref-8x  7.1-R+
> 1287755881  1287777566  21684.09  82772.50  21933.74  ref-8x  7.1-R+
>
> 1287879508  1287905173  25665.03  81696.22  18278.42  ref-8x  8.1-S[7.1-R+ user]
> 1287906039  1287931709  25669.48  81735.00  18265.78  ref-8x  8.1-S[7.1-R+ user]
> 1287932573  1287958275  25700.99  81700.40  18246.04  ref-8x  8.1-S[7.1-R+ user]
> 1287959144  1287984859  25714.74  81800.40  18346.57  ref-8x  8.1-S[7.1-R+ user]
> 1287985748  1288011752  26004.33  81569.28  17967.68  ref-8x  8.1-S[7.1-R+ user]
>
>
> The stanzas depict resource usage during each iteration of the workload
> under test (building software) under various conditions:
>
> * First, building on the ref-8x machine, running FreeBSD 8.1-STABLE r214029.
>
> * For comparison, building on the ref-7.x machine, running a lightly-
>    patched 7.1-R.
>
> * As a reality check, building on the ref-8.x machine, running the
>    above-cited 7.1-R (+patches) -- just to verify that there wasn't
>    something obviously different in the hardware configurations or
>    connections.
>
> * Finally, the requested 8.1-S kernel with 7.1-R+ userland (though I ran
>    it on the ref-8x machine, as I had some other things to do with the
>    ref-7.x machine).
>
> It appears to me that the last test runs show results that are just
> about identical to the "native" 8.1-S kernel+userland, so if I
> understand the logic correctly, that appears to implicate something in
> the 8.1-S kernel (or the default configuration for same).

yes,  exactly.
however the interesting thing is that while it took more wall-clock time,
it took less system and user time.

you might try the 4bsd scheduler to see what that does..
also, compare the configs of the two kernels


> Peace,
> david