def __init__(self, shape):
super().__init__(shape)
t0 = time.time()
self.filter = fast_fft.SpatialFilter(shape, self.PERIOD_RANGE, precision=32, threads=6, better_plan=True)
if time.time() - t0 > 0.5:
fast_fft.store_plan_hints(FFTW_WISDOM)
logger.debug('FFTW wisdom stored')
def get_fft_filter(image_shape, min_feature, max_feature, microns_per_pixel):
fftw_hints = pathlib.Path(__file__).parent / 'fftw_hints'
if fftw_hints.exists():
fast_fft.load_plan_hints(str(fftw_hints))
fft_filter = fast_fft.SpatialFilter(image_shape, period_range=(min_feature, max_feature),
spacing=microns_per_pixel, order=2, keep_dc=False, threads=1, better_plan=True)
fast_fft.store_plan_hints(str(fftw_hints))
return fft_filter
def __init__(self, shape):
super().__init__(shape)
t0 = time.time()
self.filter = fast_fft.SpatialFilter(
shape, self.PERIOD_RANGE, precision=32, threads=multiprocessing.cpu_count(), better_plan=False
)
if time.time() - t0 > 0.5:
fast_fft.store_plan_hints(FFTW_WISDOM)
def _get_filter(shape, period_range):
timer = threading.Timer(1, logger.warning, ['Slow construction of FFTW filter for image shape {} (likely no cached plan could be found). May take >30 minutes!', shape])
timer.start()
fft_filter = fast_fft.SpatialFilter(shape, period_range, precision=32, threads=6, better_plan=True)
if timer.is_alive():
timer.cancel()
else: # timer went off and warning was issued...
logger.info('FFT filter constructed. Caching plan wisdom for next time.')
fast_fft.store_plan_hints(str(FFTW_WISDOM))
return fft_filter.filter
def _get_filter(shape, period_range):
timer = threading.Timer(1, logger.warning, ['Slow construction of FFTW filter for image shape {} (likely no cached plan could be found). May take >30 minutes!', shape])
timer.start()
fft_filter = fast_fft.SpatialFilter(shape, period_range, precision=32, threads=6, better_plan=True)
if timer.is_alive():
timer.cancel()
else: # timer went off and warning was issued...
logger.info('FFT filter constructed. Caching plan wisdom for next time.')
fast_fft.store_plan_hints(str(FFTW_WISDOM))
return fft_filter.filter
