***SPAM*** Re: [ofa-general] atomic operations on ppc64

[Dotan Barak]

>> The problem that you describes is pretty basic and even an RC shouldn't have
>> this issue.
>>
>>     
> Agreed!
>
>   
>> I think that you should upgrade the HCA's Firmware. as Ronni suggested.
>>
>>     
>
> But I'm not sure about the fw version. As I mentioned, on that
> Mellanox page the latest firwmare for the IBM version is 2.3.00 which
> is the one I have. Or am I wrong?
>   
I checked Mellanox's site and i here is the URL for the FW of ConnectX:
http://www.mellanox.com/support/firmware_table_ConnectXIB.php

In this page, i can see that the FW version is 2.5.000.
>   
>> I have a feeling that the problem is in your code:
>> You should access the buffer that the HCA read/write as volatile, to "tip"
>> the compiler
>> that this memory will be modified by other components and he shouldn't do
>> any optimization
>> when you want to read data from it and actually do the reading ...
>>
>>     
> I tried that as you said before but didn't help.
> And the RDMA read works fine.
> Of course, it is possible that the problem is with my code. In fact it
> looks every time closer to this possibility. But can the code be in
> such a way wrong that it works on x86 but not on ppc. That is what
> intrigues me.
>   
Pay attension that the changes from PPC64 to x86 is not only the pointer 
size (64 vs. 32) and the
endianess issue. It is a different CPU architecture which may cause to 
many differences.

If you wish, i can review your code that handles IB offline...

Dotan
> Cheers
>