OFFLOAD_REPORT explanation

[Glover_G_]

Hi, Where can I get an explanation of the OFFLOAD_REPORT values? I using OFFLOAD_REPORT=2 and getting: [MKL] [MIC --] [AO Function] DPOTRF [MKL] [MIC --] [AO DPOTRF Workdivision] -1.00 -1.00 [MKL] [MIC 00] [AO DPOTRF CPU Time] 2.950591 seconds [MKL] [MIC 00] [AO DPOTRF MIC Time] 0.404681 seconds [MKL] [MIC 00] [AO DPOTRF CPU->MIC Data] 276480112 bytes [MKL] [MIC 00] [AO DPOTRF MIC->CPU Data] 199680000 bytes I don't understand the workdivision values of -1 and -1. I have specified MKL_MIC_WORKDIVISION=0.7 MKL_HOST_WORKDIVISION=0.3 yet I don't see those values reflected in the Workdivison line. Cheers.

[Kevin_D_Intel]

I found in newer MKL documentation in our upcoming PSXE 2016 release, some additional information about the OFFLOAD_REPORT=1 setting. It says:

1 - The report includes:
      -    The name of the function called in the AO mode.
      -    Effective work division. The value of -1 indicates that the hint, that is, the work division specified by the mkl_mic_set_workdivison function or the appropriate MKL_*_WORKDIVISION environment variable was ignored in this function call.
      -    The time spent on the host CPU during the call.
      -    The time spent on each available Intel Xeon Phi coprocessor during the call.

Additionally, there is a new special Note stating:

Note
      -    Settings specified by the functions take precedence over the settings specified by the respective environment variables.
      -    Intel MKL interprets the values of MKL_HOST_WORKDIVISION, MKL_MIC_WORKDIVISION, and MKL_MIC_<number>_WORKDIVISION as guidance toward dividing work between coprocessors, but the library may choose a different work division if necessary.
      -    For LAPACK routines, setting the fraction of work to any value other than 0.0 enables the specified processor for AO mode. However Intel MKL LAPACK does not use the value specified to divide the workload. For example, setting the fraction to 0.5 has the same effect as setting the fraction to 1.0.

So it appears -1 may indicate specific values you provided were ignored; however, the NOTE suggests the fraction should be interpreted as 1.0 for the LAPACK function chosen. Perhaps the -1 says the specific fractional value was ignored but given the NOTE that the AO was still active. Since your CPU and MIC times are non-zero, the work appears to have been completed across both the host and coprocessor.

I asked someone from our MKL team to provide some further interpretation/assistance in understanding your findings/question. Please stand-by.