refactor: move parallel code into a separate function

fastmorph/fastmorph.hpp

@@ -4,10 +4,47 @@
 #include <vector>
 #include <cstdlib>
 #include <cmath>
+#include <functional>
 #include "threadpool.h"
 
 namespace fastmorph {
 
+void parallelize_blocks(
+	const std::function<void(
+		const uint64_t, const uint64_t, 
+		const uint64_t, const uint64_t, 
+		const uint64_t, const uint64_t
+	)> &process_block,
+	const uint64_t sx, const uint64_t sy, const uint64_t sz,
+	const uint64_t threads, const uint64_t offset
+) {
+	const uint64_t block_size = 64;
+
+	const uint64_t grid_x = std::max(static_cast<uint64_t>((sx + block_size/2) / block_size), static_cast<uint64_t>(1));
+	const uint64_t grid_y = std::max(static_cast<uint64_t>((sy + block_size/2) / block_size), static_cast<uint64_t>(1));
+	const uint64_t grid_z = std::max(static_cast<uint64_t>((sz + block_size/2) / block_size), static_cast<uint64_t>(1));
+
+	const int real_threads = std::max(std::min(threads, grid_x * grid_y * grid_z), static_cast<uint64_t>(0));
+
+	ThreadPool pool(real_threads);
+
+	for (uint64_t gz = 0; gz < grid_z; gz++) {
+		for (uint64_t gy = 0; gy < grid_y; gy++) {
+			for (uint64_t gx = 0; gx < grid_x; gx++) {
+				pool.enqueue([=]() {
+					process_block(
+						std::max(offset, gx * block_size), std::min((gx+1) * block_size, sx - offset),
+						std::max(offset, gy * block_size), std::min((gy+1) * block_size, sy - offset),
+						std::max(offset, gz * block_size), std::min((gz+1) * block_size, sz - offset)
+					);
+				});
+			}
+		}
+	}
+
+	pool.join();
+}
+
 
 template <typename LABEL>
 void multilabel_dilate(
@@ -251,31 +288,13 @@ void multilabel_dilate(
 		}
 	};
 
-	const uint64_t block_size = 64;
-
-	const uint64_t grid_x = std::max(static_cast<uint64_t>((sx + block_size/2) / block_size), static_cast<uint64_t>(1));
-	const uint64_t grid_y = std::max(static_cast<uint64_t>((sy + block_size/2) / block_size), static_cast<uint64_t>(1));
-	const uint64_t grid_z = std::max(static_cast<uint64_t>((sz + block_size/2) / block_size), static_cast<uint64_t>(1));
-
-	const int real_threads = std::max(std::min(threads, grid_x * grid_y * grid_z), static_cast<uint64_t>(0));
-
-	ThreadPool pool(real_threads);
-
-	for (uint64_t gz = 0; gz < grid_z; gz++) {
-		for (uint64_t gy = 0; gy < grid_y; gy++) {
-			for (uint64_t gx = 0; gx < grid_x; gx++) {
-				pool.enqueue([=]() {
-					process_block(
-						gx * block_size, std::min((gx+1) * block_size, sx),
-						gy * block_size, std::min((gy+1) * block_size, sy),
-						gz * block_size, std::min((gz+1) * block_size, sz)
-					);
-				});
-			}
-		}
-	}
-
-	pool.join();
+	parallelize_blocks(
+		std::function<void(
+			const uint64_t,const uint64_t,const uint64_t,
+			const uint64_t,const uint64_t,const uint64_t
+		)>(process_block), 
+		sx, sy, sz, threads, /*offset=*/0
+	);
 }
 
 template <typename LABEL>
@@ -423,32 +442,13 @@ void multilabel_erode(
 
 #undef FILL_STENCIL
 
-	const uint64_t block_size = 64;
-
-	const uint64_t grid_x = std::max(static_cast<uint64_t>((sx + block_size/2) / block_size), static_cast<uint64_t>(1));
-	const uint64_t grid_y = std::max(static_cast<uint64_t>((sy + block_size/2) / block_size), static_cast<uint64_t>(1));
-	const uint64_t grid_z = std::max(static_cast<uint64_t>((sz + block_size/2) / block_size), static_cast<uint64_t>(1));
-
-	const int real_threads = std::max(std::min(threads, grid_x * grid_y * grid_z), static_cast<uint64_t>(0));
-
-	ThreadPool pool(real_threads);
-
-	for (uint64_t gz = 0; gz < grid_z; gz++) {
-		for (uint64_t gy = 0; gy < grid_y; gy++) {
-			for (uint64_t gx = 0; gx < grid_x; gx++) {
-				pool.enqueue([=]() {
-					const uint64_t one = 1;
-					process_block(
-						std::max(one, gx * block_size), std::min((gx+1) * block_size, sx - 1),
-						std::max(one, gy * block_size), std::min((gy+1) * block_size, sy - 1),
-						std::max(one, gz * block_size), std::min((gz+1) * block_size, sz - 1)
-					);
-				});
-			}
-		}
-	}
-
-	pool.join();
+	parallelize_blocks(
+		std::function<void(
+			const uint64_t,const uint64_t,const uint64_t,
+			const uint64_t,const uint64_t,const uint64_t
+		)>(process_block), 
+		sx, sy, sz, threads, /*offset=*/1
+	);
 }
 
 };