#37 Switched from multiprocessing.Pool to multiprocessing.ThreadPool

src/processing_steps/0600_segment_implant_cc.py

@@ -8,6 +8,7 @@
 from lib.cpp.cpu.io import load_slice
 from lib.cpp.cpu.connected_components import largest_connected_component, connected_components
 import multiprocessing as mp
+from multiprocessing.pool import ThreadPool
 import matplotlib
 matplotlib.use('Agg')
 import matplotlib.pyplot as plt
@@ -100,7 +101,7 @@ def label_chunk(i, chunk_size, chunk_prefix, implant_threshold_u16, global_shape
             return n_features
 
         start = datetime.datetime.now()
-        with mp.Pool(n_cores) as pool:
+        with ThreadPool(n_cores) as pool:
             label_chunk_partial = partial(label_chunk, chunk_size=layers_per_chunk, chunk_prefix=f"{intermediate_folder}/{sample}_", implant_threshold_u16=implant_threshold_u16, global_shape=(nz,ny,nx))
             n_labels = pool.map(label_chunk_partial, range(n_chunks))
         end = datetime.datetime.now()

src/processing_steps/0750_bone_region.py

@@ -17,6 +17,7 @@
 from scipy.ndimage import gaussian_filter1d
 import datetime
 import multiprocessing as mp
+from multiprocessing.pool import ThreadPool
 from functools import partial
 
 # close = dilate then erode
@@ -94,7 +95,7 @@ def largest_cc_of(mask, mask_name):
         return (label == largest_cc_ix)
     else:
         start = datetime.datetime.now()
-        with mp.Pool(n_cores) as pool:
+        with ThreadPool(n_cores) as pool:
             label_chunk_partial = partial(label_chunk, chunk_prefix=f"{intermediate_folder}/{sample}_")
             chunks = [mask[i*layers_per_chunk:(i+1)*layers_per_chunk] for i in range(n_chunks-1)]
             chunks.append(mask[(n_chunks-1)*layers_per_chunk:])

src/processing_steps/1500_segment_blood_cc.py

@@ -11,6 +11,7 @@
 from lib.py.resample import downsample2x, downsample3x
 import cupy as cp
 import multiprocessing as mp
+from multiprocessing.pool import ThreadPool
 import datetime
 from functools import partial
 
@@ -78,7 +79,7 @@ def label_chunk(i, chunk_size, chunk_prefix, global_shape):
             return n_features
 
         start = datetime.datetime.now()
-        with mp.Pool(n_cores) as pool:
+        with ThreadPool(n_cores) as pool:
             label_chunk_partial = partial(label_chunk, chunk_size=layers_per_chunk, chunk_prefix=f"{intermediate_folder}/{sample}_", global_shape=(nz,ny,nx))
             n_labels = pool.map(label_chunk_partial, range(n_chunks))
         end = datetime.datetime.now()