[ofa-general] Re: [PATCHv2] opensm: Parallelize (Stripe) LFT sets across switches

[Yevgeny Kliteynik]

Sasha Khapyorsky wrote:
> On 10:25 Wed 05 Aug     , Yevgeny Kliteynik wrote:
>>>      >
>>>      > No noticeable difference between various max_smps_per_node values
>>>      > was observed.
>>>
>>>     What is the reason?
>>>
>>>  
>>> I think the reason was max_wire_smps may have kicked in but Yevgeny is 
>>> best to elaborate on this.
>>>  
>> Correct, this was because of max_wire_smps
> 
> What was 'max_wire_smps' value used in the tests?

The numbers that I posted refer to default max_wire_smps, which is 4.
I didn't try to bump it up, though I guess that it might improve LFT
config time.

-- Yevgeny
 
> Sasha
>