InputImageType, OutputImageType, RealOutputPixelType]
GPUNeighborhoodFilterType = itk.GPUNeighborhoodOperatorImageFilter[
InputGPUImageType, OutputGPUImageType, RealOutputPixelType]
# Create 1D Gaussian operator
OperatorType = itk.GaussianOperator[RealOutputPixelType, Dimension]
oper = OperatorType()
oper.SetDirection(0)
oper.SetVariance(8.0)
oper.CreateDirectional()
# test 1~8 work units for CPU
for number_of_work_units in range(1, 9):
cpu_filter = NeighborhoodFilterType.New()
cpu_timer = itk.TimeProbe()
cpu_timer.Start()
cpu_filter.SetNumberOfWorkUnits(number_of_work_units)
cpu_filter.SetInput(input_image)
cpu_filter.SetOperator(oper)
cpu_filter.Update()
cpu_timer.Stop()
print(
"CPU NeighborhoodFilter took {} seconds with {} work units.\n".format(
cpu_timer.GetMean(), cpu_filter.GetNumberOfWorkUnits()))
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import itk
def LongFunction():
# CPython loops are much slower than C++,
# so a smaller range is used in this case.
for i in range(int(1e5)):
a = 0.0 # noqa: F841
clock = itk.TimeProbe()
clock.Start()
LongFunction()
clock.Stop()
print('Mean: ' + str(clock.GetMean()))
print('Total: ' + str(clock.GetTotal()))
clock.Start()
LongFunction()
clock.Stop()
print('Mean: ' + str(clock.GetMean()))
print('Total: ' + str(clock.GetTotal()))