Re: [myrinet] Problems sending 1k and 2k messages

These are the results we obtain from gm_allsize. The test runs ok.

What can be the problem?

Best regards
Jose Flich and Salva Coll

> # Opening board 0  port 2. Sending to port 2.
> # Not verifying contents of messages.
> remote GM node id itb (3).
> # Reliable Point-To-Point Half-Round-Trip Latency (usec)
> # size 20
> # gm_allsize -h itb --size=20 -lat -b --min-len=1020 --max-len=1040
> # (reporting half round trip time)
> 1020 44.09
> 1021 43.60
> 1022 43.60
> 1023 43.62
> 1024 43.42
> 1025 43.70
> 1026 43.74
> 1027 43.33
> 1028 43.66
> 1029 43.91
> 1030 43.74
> 1031 43.87
> 1032 43.64
> 1033 43.91
> 1034 43.74
> 1035 44.46
> 1036 44.20
> 1037 44.26
> 1038 43.73
> 1039 43.97
> 1040 43.94



Bob Felderman wrote:

> => We are runing GM version 1.2pre16 on a SuSE 6.1 Linux distribution with
> => kernel 2.0.36 with PII processors.
> =>
> => When we test the Myrinet network with the gm_basic (or gm_allsize)
> => program we obtain good results but when it reaches 1k message size the
> => test usually fails (i.e. for gm_basic the first strcpy call fails). We
> => have also experienced the same problem for 2k message sizes. For higher
> => messages the system runs perfectly.
> =>
> => Is there any problem related to operating system virtual memory ? What
> => can be the problem? How can we fix that?
>
> What happens when you run
> host1: gm_allsize --slave --size=20
> host2: gm_allsize -h host1 --size=20 -lat -b --min-len=1020 --max-len=1040
>
> does the test complete?