On 2/22/21 1:56 PM, David A Case wrote:
> On Tue, Feb 09, 2021, Mudrovcic, Korana wrote:
>
>> In our research group, our calculations get the following note:
>>
>> Note: The following floating-point exceptions are
>> signalling: IEEE_INVALID_FLAG IEEE_OVERFLOW_FLAG IEEE_UNDERFLOW_FLAG
>> IEEE_DENORMAL
>>
>> This note is signalling with AMBER20 as well as with AMBER18 (we are using
>> CUDA 10.2 version) and is also triggered in the tests despite all of them
>> passing. This makes us believe that the results are not affected, and that
>> the warning is harmless (?).
>
> This is what Amber developers assume, but I don't know that it has ever been
> investigated thoroughly.
If this is built against MKL then there is a high likelyhood that it's
coming from there. The only way to really check is to build it against a
"-O0 -mieee-fp" compiled refblas from netlib. But that is a tedious
thing to do.
--
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: ake.hpc2n.umu.se Phone: +46 90 7866134 Fax: +46 90-580 14
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Feb 22 2021 - 06:00:06 PST
