Merge pull request #80 from pierotofy/parallel

Parallel image loading

opensplat.cpp

@@ -94,9 +94,9 @@ int main(int argc, char *argv[]){
 
     try{
         InputData inputData = inputDataFromX(projectRoot);
-        for (Camera &cam : inputData.cameras){
+        parallel_for(inputData.cameras.begin(), inputData.cameras.end(), [&downScaleFactor](Camera &cam){
             cam.loadImage(downScaleFactor);
-        }
+        });
 
         // Withhold a validation camera if necessary
         auto t = inputData.getCameras(validate, valImage);

utils.cpp

@@ -1 +1 @@
-#include "utils.hpp"
+#include "utils.hpp"

utils.hpp

@@ -5,6 +5,8 @@
 #include <algorithm>
 #include <random>
 #include <iostream>
+#include <thread>
+#include <functional>
 
 template <typename T>
 class InfiniteRandomIterator
@@ -32,4 +34,28 @@ class InfiniteRandomIterator
     std::default_random_engine engine;
 };
 
+template <typename IndexType, typename FuncType>
+void parallel_for(IndexType begin, IndexType end, FuncType func) {
+    size_t range = end - begin;
+    if (range <= 0) return;
+    size_t numThreads = (std::min)(static_cast<size_t>(std::thread::hardware_concurrency()), range);
+    size_t chunkSize = (range + numThreads - 1) / numThreads;
+    std::vector<std::thread> threads;
+
+    for (unsigned int i = 0; i < numThreads; i++) {
+        IndexType chunkBegin = begin + i * chunkSize;
+        IndexType chunkEnd = (std::min)(chunkBegin + chunkSize, end);
+
+        threads.emplace_back([chunkBegin, chunkEnd, &func]() {
+            for (IndexType item = chunkBegin; item < chunkEnd; item++) {
+                func(*item);
+            }
+        });
+    }
+
+    for (std::thread& t : threads) {
+        t.join();
+    }
+}
+
 #endif