dpnp
dpnp copied to clipboard
Performance testing.
Need to implement performance testing.
- It should be easy to use for developer - like running regular test system.
- Output must be readable and ready to be pasted to a spreadsheet for further analysis.
Example
unning on: Intel(R) Core(TM) i5-4570 CPU @ 3.20GHz
queue initialization time:0.000660675 (sec.)
...Test data type is <class 'numpy.float64'>, each test repetitions 5
type:float64:N: 16:Python:9.775e-06:MKL:1.054e-04:ratio: 0.09:verification:True
type:float64:N: 32:Python:3.409e-05:MKL:3.567e-03:ratio: 0.01:verification:True
type:float64:N: 64:Python:7.796e-05:MKL:5.098e-03:ratio: 0.02:verification:True
type:float64:N: 128:Python:3.386e-04:MKL:8.929e-04:ratio: 0.38:verification:True
type:float64:N: 256:Python:1.574e-03:MKL:6.076e-03:ratio: 0.26:verification:True
type:float64:N: 512:Python:7.821e-03:MKL:6.583e-03:ratio: 1.19:verification:True
type:float64:N:1024:Python:3.879e-02:MKL:5.012e-02:ratio: 0.77:verification:True
...Test data type is <class 'numpy.float32'>, each test repetitions 5
type:float32:N: 16:Python:5.007e-06:MKL:5.565e-04:ratio: 0.01:verification:True
type:float32:N: 32:Python:5.722e-06:MKL:5.448e-04:ratio: 0.01:verification:True
type:float32:N: 64:Python:7.653e-05:MKL:8.216e-04:ratio: 0.09:verification:True
type:float32:N: 128:Python:2.210e-04:MKL:4.613e-03:ratio: 0.05:verification:True
fields delimiter is ":" by default