2002-11-24 13:10:10

by Nero

Subject: [BENCHMARK] rmap15, rmap14c and rc1aa1 with contest

noload:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 84.6 95 0 0 1.00
2.4.19-rmap14c [1] 84.2 96 0 0 1.00
2.4.20-rc1aa1 [2] 41.0 49 0 0 inf

process_load:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 140.8 55 133 45 1.66
2.4.19-rmap14c [1] 139.4 56 126 45 1.65
2.4.20-rc1aa1 [2] 133.9 14 211 84 inf

io_load:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 164.8 51 33 19 1.95
2.4.19-rmap14c [1] 160.1 53 33 21 1.90
2.4.20-rc1aa1 [1] 166.2 49 44 23 inf

read_load:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 0.0 0 0 7 0.00
2.4.19-rmap14c [1] 121.0 72 20 7 1.44
2.4.20-rc1aa1 [1] 111.2 79 34 14 inf

rmap15 OOM'd the cc1 process twice and fscked this run up.

list_load:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 99.2 84 0 7 1.17
2.4.19-rmap14c [1] 100.7 83 0 7 1.20
2.4.20-rc1aa1 [1] 96.7 85 0 8 inf

mem_load:
Kernel [runs] Time CPU% Loads LCPU% Ratio
2.4.19-rmap15 [1] 111.2 80 96 8 1.31
2.4.19-rmap14c [1] 106.1 82 96 9 1.26
2.4.20-rc1aa1 [1] 126.3 66 77 3 inf

2002-11-24 13:19:52

by Nero

Subject: Re: [BENCHMARK] rmap15, rmap14c and rc1aa1 with contest

Nero wrote:

> noload:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 84.6 95 0 0 1.00
> 2.4.19-rmap14c [1] 84.2 96 0 0 1.00
> 2.4.20-rc1aa1 [2] 41.0 49 0 0 inf

The time for rc1aa1 here is obviously wrong - this is a bug in contest,
according to Dr. Kolivas :-)

>
> process_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 140.8 55 133 45 1.66
> 2.4.19-rmap14c [1] 139.4 56 126 45 1.65
> 2.4.20-rc1aa1 [2] 133.9 14 211 84 inf
>
> io_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 164.8 51 33 19 1.95
> 2.4.19-rmap14c [1] 160.1 53 33 21 1.90
> 2.4.20-rc1aa1 [1] 166.2 49 44 23 inf
>
> read_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 0.0 0 0 7 0.00
> 2.4.19-rmap14c [1] 121.0 72 20 7 1.44
> 2.4.20-rc1aa1 [1] 111.2 79 34 14 inf
>
> rmap15 OOM'd the cc1 process twice and fscked this run up.
>
> list_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 99.2 84 0 7 1.17
> 2.4.19-rmap14c [1] 100.7 83 0 7 1.20
> 2.4.20-rc1aa1 [1] 96.7 85 0 8 inf
>
> mem_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 111.2 80 96 8 1.31
> 2.4.19-rmap14c [1] 106.1 82 96 9 1.26
> 2.4.20-rc1aa1 [1] 126.3 66 77 3 inf
>

2002-11-24 13:23:10

by Con Kolivas

Subject: Re: [BENCHMARK] rmap15, rmap14c and rc1aa1 with contest

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Thanks for this interesting comparison

>noload:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 84.6 95 0 0 1.00
>2.4.19-rmap14c [1] 84.2 96 0 0 1.00
>2.4.20-rc1aa1 [2] 41.0 49 0 0 inf

Clearly my braindead parser failed here. The time of aa1 should probably be 82
seconds (41.0 x 2) Can you check the full log of 2.4.20-rc1aa1.log ? It has
raw results.

>process_load:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 140.8 55 133 45 1.66
>2.4.19-rmap14c [1] 139.4 56 126 45 1.65
>2.4.20-rc1aa1 [2] 133.9 14 211 84 inf

If the error continued here (2 runs of process_load) then the time of aa1
should be 133.9 x 2. Can you check the full log of 2.4.20-rc1aa1.log here
too?

>io_load:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 164.8 51 33 19 1.95
>2.4.19-rmap14c [1] 160.1 53 33 21 1.90
>2.4.20-rc1aa1 [1] 166.2 49 44 23 inf
>
>read_load:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 0.0 0 0 7 0.00
>2.4.19-rmap14c [1] 121.0 72 20 7 1.44
>2.4.20-rc1aa1 [1] 111.2 79 34 14 inf
>
>rmap15 OOM'd the cc1 process twice and fscked this run up.
>
>list_load:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 99.2 84 0 7 1.17
>2.4.19-rmap14c [1] 100.7 83 0 7 1.20
>2.4.20-rc1aa1 [1] 96.7 85 0 8 inf
>
>mem_load:
>Kernel [runs] Time CPU% Loads LCPU% Ratio
>2.4.19-rmap15 [1] 111.2 80 96 8 1.31
>2.4.19-rmap14c [1] 106.1 82 96 9 1.26
>2.4.20-rc1aa1 [1] 126.3 66 77 3 inf

With only one run these numbers appear not to be greatly statistically
significant, except for the mem_load results.

Con
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.0 (GNU/Linux)

iD8DBQE94NTSF6dfvkL3i1gRAkwXAJ9YJmnxvxdiYyESNYLeQa9sbwR+2wCcCZal
bFKWReK3EP8oayP/GwiIK7c=
=gYaS
-----END PGP SIGNATURE-----

2002-11-24 13:27:52

by Nero

Subject: Re: [BENCHMARK] rmap15, rmap14c and rc1aa1 with contest

Con Kolivas wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Thanks for this interesting comparison
>
>
> >noload:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 84.6 95 0 0 1.00
> >2.4.19-rmap14c [1] 84.2 96 0 0 1.00
> >2.4.20-rc1aa1 [2] 41.0 49 0 0 inf
>
>
> Clearly my braindead parser failed here. The time of aa1 should
> probably be 82
> seconds (41.0 x 2) Can you check the full log of 2.4.20-rc1aa1.log ?
> It has
> raw results.
>
Yep.
noload Time: 81.95 CPU: 98% LoadRuns: 0 LoadCPU%: 0

>
> >process_load:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 140.8 55 133 45 1.66
> >2.4.19-rmap14c [1] 139.4 56 126 45 1.65
> >2.4.20-rc1aa1 [2] 133.9 14 211 84 inf
>
>
> If the error continued here (2 runs of process_load) then the time of aa1
> should be 133.9 x 2. Can you check the full log of 2.4.20-rc1aa1.log here
> too?

process_load Time: 267.85 CPU: 29% LoadRuns: 422 LoadCPU%: 70%

>
>
> >io_load:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 164.8 51 33 19 1.95
> >2.4.19-rmap14c [1] 160.1 53 33 21 1.90
> >2.4.20-rc1aa1 [1] 166.2 49 44 23 inf
> >
> >read_load:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 0.0 0 0 7 0.00
> >2.4.19-rmap14c [1] 121.0 72 20 7 1.44
> >2.4.20-rc1aa1 [1] 111.2 79 34 14 inf
> >
> >rmap15 OOM'd the cc1 process twice and fscked this run up.
> >
> >list_load:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 99.2 84 0 7 1.17
> >2.4.19-rmap14c [1] 100.7 83 0 7 1.20
> >2.4.20-rc1aa1 [1] 96.7 85 0 8 inf
> >
> >mem_load:
> >Kernel [runs] Time CPU% Loads LCPU% Ratio
> >2.4.19-rmap15 [1] 111.2 80 96 8 1.31
> >2.4.19-rmap14c [1] 106.1 82 96 9 1.26
> >2.4.20-rc1aa1 [1] 126.3 66 77 3 inf
>
>
> With only one run these numbers appear not to be greatly statistically
> significant, except for the mem_load results.
>
> Con
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.0 (GNU/Linux)
>
> iD8DBQE94NTSF6dfvkL3i1gRAkwXAJ9YJmnxvxdiYyESNYLeQa9sbwR+2wCcCZal
> bFKWReK3EP8oayP/GwiIK7c=
> =gYaS
> -----END PGP SIGNATURE-----
>
>

2002-11-24 17:32:19

by Rik van Riel

Subject: Re: [BENCHMARK] rmap15, rmap14c and rc1aa1 with contest

On Mon, 25 Nov 2002, Nero wrote:

> read_load:
> Kernel [runs] Time CPU% Loads LCPU% Ratio
> 2.4.19-rmap15 [1] 0.0 0 0 7 0.00
> 2.4.19-rmap14c [1] 121.0 72 20 7 1.44
> 2.4.20-rc1aa1 [1] 111.2 79 34 14 inf
>
> rmap15 OOM'd the cc1 process twice and fscked this run up.

Known bug, should be fixed with the 2.4.19-rmap15-splitactive
patch I posted on friday:

http://surriel.com/patches/2.4/2.4.19-rmap15-splitactive

cheers,

Rik
--
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/ http://guru.conectiva.com/
Current spamtrap: <a href=mailto:"[email protected]">[email protected]</a>