oneDNN/tests/benchdnn/doc/driver_resampling.md

Resampling Driver

Usage

    ./benchdnn --resampling [benchdnn-knobs] [resampling-knobs] [resampling-desc] ...

where resampling-knobs are:

• --dir={FWD_D [default], BWD_D} -- dnnl_prop_kind_t. Refer to direction for details.
• --sdt={f32 [default], ...} -- src data type. Refer to data types for details.
• --ddt={f32 [default], ...} -- dst data type. Refer to data types for details.
• --tag={nchw [default], ...} -- physical src and dst memory layout. Refer to tags for details.
• --alg={nearest [default], linear} -- resampling algorithm. nearest or resampling_nearest is dnnl_resampling_nearest; linear or resampling_nearest is dnnl_resampling_linear; Refer to resampling primitive for details.
• --mb=INT -- override minibatch size specified in the problem description. When set to 0, use minibatch size as defined by the individual problem descriptor. The default is 0.
• --match=REGEX -- skip problems not matching the regular expression in REGEX. By default no pattern is applied (run everything). Note: Windows may interpret only string arguments surrounded by double quotation marks.
• Any attributes options. Refer to attributes for details.

and resampling-desc is a problem descriptor. The canonical form is:

    mbXicX_idXihXiwX_odXohXowX_nS

Refer to descriptor for details.

Essence of Testing

nearest: Fill input data with integers and expect an integer answer. linear: Fill input data with integers and expect a float answer.

Examples

Run a set of resamplings from an input file with the default settings:

    ./benchdnn --resampling --batch=inputs/resampling/shapes_2d

Run a named problem with single precision src/dst, iterating by:

1. both blocked memory layouts, where channel blocking equals 8 and 16,
2. both forward and backward prop_kind,
3. all algorithm combinations,
4. using default minibatch of 96 and 5:

    ./benchdnn --resampling --sdt=f32 --ddt=f32 --tag=nChw8c,nChw16c \
               --dir=FWD_D,BWD_D --alg=nearest,linear --mb=0,5 \
               mb96ic768_ih17oh34

More examples with different driver options can be found at inputs/resampling/test_*. Examples with different problem descriptors can be found at inputs/resampling/shapes_*. Examples with different benchdnn common options can be found at driver_conv.md.