Closed kohr-h closed 8 years ago
I can reduce the sizes of the examples. What would be the "optimal rate of reduce" (..on my machine the wavelet tests take about 0.7 seconds..)
Make them as small as possible such that they still "make sense". Reduce the smallest size in the test to (1, 1)
or whatever is the smallest possible, instead of (8, 8)
. Another quite remarkable observation from the test results
test_DiscreteWaveletTrafo 0.28 sec
test_coeff_size_list 0.57 sec
test_pywt_coeff_to_array_and_array_to_pywt_coeff2d 5 ms
test_pywt_coeff_to_array_and_array_to_pywt_coeff3d 0.13 sec
test_wavelet_decomposition3d_and_reconstruction3d 0.36 sec
is that the test_coeff_size_list
takes longer time than all the other tests, even longer than the decomposition and reconstruction test. This is a red flag since this function is only supposed to be a helper and, if possible, instantaneous.
A large portion of this is probably due to #153.
The open pull request contains some reduction in test problem sizes. If we get down to acceptable times by that, we can close.
But the test_coeff_size_list
is no basically instantaneous.
Close this?
Last time we had 0.23 seconds, I guess we're good with that.
Good enough for now I guess.
The wavelet tests take about 1.5 seconds now, which is too much in relation to the rest of the library. We should make smaller examples and add some larger test cases to our "largescale" test suite once it's there.