| 1 |
We've actually noticed the opposite results here with OS X 10.4 vs. |
| 2 |
Linux as far as HPL goes. |
| 3 |
I don't have the numbers on hand, but Linux was notably faster with HPL |
| 4 |
over both GigE and |
| 5 |
Myrinet. In fact the numbers for Colsa (our partner) in the Top500 were |
| 6 |
achieved using Yellow |
| 7 |
Dog instead of OS X. That being said if you have a moment I have a few |
| 8 |
questions: |
| 9 |
|
| 10 |
1.) What libraries were you using? Did you use Goto's? |
| 11 |
2.) Which compiler(s) did you use? |
| 12 |
3.) What sort of setup do you have? Diskless or not and so on. |
| 13 |
|
| 14 |
|
| 15 |
Brady Catherman wrote: |
| 16 |
> Unfortunately BPS has an annoying habit of overwriting the test output |
| 17 |
> files every time it is run regardless of which test you actually run. |
| 18 |
> I didn't notice this when I was doing my initial stuff on RedHat so I |
| 19 |
> lost all the other tests results from BPS. I have better tests of the |
| 20 |
> hardware since then. We ran HPL on the PowerPC cluster running both |
| 21 |
> Gentoo 2005.1 and Mac OS 10.4 (mac os was ~ %5 faster on average). I |
| 22 |
> ran IOZone on our XRaid using HFS+ under Mac OS, and Reiser, xfs, jfs, |
| 23 |
> and ext3 under Gentoo for PPC64. |
| 24 |
> |
| 25 |
> I also ran tests using Mr Bayes, ClustalW, MPIBench, and several other |
| 26 |
> in house tests that I had available but all of these tests where |
| 27 |
> either inconclusive or failed to demonstrate any difference between |
| 28 |
> the platforms. |
| 29 |
> |
| 30 |
> I will be doing a bunch of this again here in a bit in order to get |
| 31 |
> data on our new OS build we are putting into production =) |
| 32 |
> |
| 33 |
> |
| 34 |
> On Apr 11, 2006, at 6:20 PM, Brandon Edens wrote: |
| 35 |
> |
| 36 |
>>> Yeah, from the numbers it looks as if it would be dependent on the |
| 37 |
>>> purpose of the cluster whether OS X or Gentoo would do better. On ppc, |
| 38 |
>>> Gentoo does poorly on the first two benchmarks and also on context |
| 39 |
>>> switching. On x86, the first two are more comparable with RH, but the |
| 40 |
>>> others, Gentoo has a small to large advantage over RH, just as on ppc. |
| 41 |
>> |
| 42 |
>> First and second benchmarks are dhrystone and whetstone, synthetic |
| 43 |
>> benchmarks. |
| 44 |
>> I'd be wary of these types of benchmarks; they've been cheated |
| 45 |
>> before. I'd like |
| 46 |
>> to see some real-world computation benchmarks. |
| 47 |
>> |
| 48 |
>> Brandon Edens |
| 49 |
>> |
| 50 |
> |
| 51 |
> --gentoo-cluster@g.o mailing list |
| 52 |
> |
| 53 |
> |
| 54 |
|
| 55 |
-- |
| 56 |
Justin Bronder |
| 57 |
University of Maine, Orono |
| 58 |
|
| 59 |
Advanced Computing Research Lab |
| 60 |
20 Godfrey Dr |
| 61 |
Orono, ME 04473 |
| 62 |
www.clusters.umaine.edu |
| 63 |
|
| 64 |
Mathematics Department |
| 65 |
425 Neville Hall |
| 66 |
Orono, ME 04469 |
| 67 |
|
| 68 |
|
| 69 |
-- |
| 70 |
gentoo-cluster@g.o mailing list |
Archives