A newer version of this document is available. Customers should click here to go to the newest version.
ignore
Specify runtimes or libraries to ignore time spent in these regions when calculating per-program speedup.
Syntax
--ignore=<string> |
Arguments
<string> is a comma-separated list of runtimes and libraries to ignore:
Argument |
Description |
---|---|
OMP |
Ignore time spent in OpenMP* code regions. |
MPI |
Ignore time spent in MPI code regions. |
TBB |
Ignore time spent in Intel® oneAPI Threading Building Blocks code regions. |
MKL |
Ignore time spent in Intel® Math Kernel Library code regions. |
DAAL |
Ignore time spent in Intel® Data Analytics Acceleration Library code regions. |
OMPTARGET |
Ignore time spent in code regions with OpenMP target construct. |
OCL |
Ignore time spent in OpenCL™ code regions. |
L0 |
Ignore time spent in Intel® oneAPI Level Zero code regions. |
SVML |
Ignore time spent in Short Vector Math Library (SVML) code regions. |
Usage
This option does not affect individual offloads, only the per-program metrics.
Example
Model your application performance on a target device and ignore time spent in OpenMP and Intel oneAPI Threading Building Blocks regions.
advisor --collect=projection --ignore=OMP,TBB --project-dir=./advi_results