[Lldb-commits] [lldb] DynamicLoaderDarwin load images in parallel with preload (PR #110646)

Dmitrii Galimzianov via lldb-commits lldb-commits at lists.llvm.org
Tue Oct 15 10:57:08 PDT 2024


================
@@ -642,26 +652,86 @@ ModuleSP DynamicLoaderDarwin::GetDYLDModule() {
 
 void DynamicLoaderDarwin::ClearDYLDModule() { m_dyld_module_wp.reset(); }
 
+template <typename InputIterator, typename ResultType>
+static std::vector<ResultType> parallel_map(
+    llvm::ThreadPoolInterface &threadPool, InputIterator first,
+    InputIterator last,
+    llvm::function_ref<ResultType(
+        const typename std::iterator_traits<InputIterator>::value_type &)>
+        transform) {
+  const auto size = std::distance(first, last);
+  std::vector<ResultType> results(size);
+  if (size > 0) {
+    llvm::ThreadPoolTaskGroup taskGroup(threadPool);
+    auto it = first;
+    for (ssize_t i = 0; i < size; ++i, ++it) {
+      taskGroup.async([&, i, it]() { results[i] = transform(*it); });
+    }
+    taskGroup.wait();
+  }
+  return results;
+}
----------------
DmT021 wrote:

@jasonmolenda @JDevlieghere  I decided to inline the helpers. I still think we are mixing important and utilitary code here and it's less than ideal. But at the same time, I fully understand the readability issue. What's also important here the signatures of the map functions aren't ideal to be considered "general purpose helpers". Namely, they expect `std::distance` to be O(1) which isn't always the case with the current template constraints. So I think if we are going to use the same technique in other places (e.g. `Target::SetExecutableModule` and other dynamic loader plugins) - we may want to write better (more general) implementations for these helpers in some common place.

PS Sorry for the long answer, it took me a while to come up with the decision here.

https://github.com/llvm/llvm-project/pull/110646


More information about the lldb-commits mailing list