diff --git a/src/process.cpp b/src/process.cpp
index b05b6f25..107712f9 100644
--- a/src/process.cpp
+++ b/src/process.cpp
@@ -20,6 +20,7 @@
 #include "LAScatalog.h"
 #include "openmp.h"
 
+// Parallel processing strategies
 #define SEQUENTIAL 1
 #define CONCURRENTPOINTS 2
 #define CONCURRENTFILES 3
@@ -126,6 +127,12 @@ SEXP process(SEXP sexppipeline, SEXP sexpprogrss, SEXP sexpncpu, SEXP sexpmode,
 
     bool failure = false;
     int k = 0;
+
+    // Records a pointer to the private pipeline of each threads
+    // This allows to know the order of the results
+    std::vector<Pipeline*> private_pipelines_ptr;
+    private_pipelines_ptr.reserve(ncpu_outer_loop);
+
     #pragma omp parallel num_threads(ncpu_outer_loop)
     {
       try
@@ -134,6 +141,7 @@ SEXP process(SEXP sexppipeline, SEXP sexpprogrss, SEXP sexpncpu, SEXP sexpmode,
         // ensure that shared resources are protected (such as connection to output files)
         // and private data are copied.
         Pipeline private_pipeline(pipeline);
+        private_pipelines_ptr[omp_get_thread_num()] = &private_pipeline;
 
         #pragma omp for
         for (int i = 0 ; i < n ; ++i)
@@ -200,13 +208,16 @@ SEXP process(SEXP sexppipeline, SEXP sexpprogrss, SEXP sexpncpu, SEXP sexpmode,
         // We are outside the main loop. We can clear the pipeline with last = true;
         private_pipeline.clear(true);
 
-        // We have multiple pipelines that each process some chunks and each have a partial
-        // output. We reduce in the main pipeline
-        #pragma omp critical
+        // We have multiple pipelines and each processed some chunks and each have a partial
+        // output. We reduce in the main pipeline. To preserve the ordering of the output, first
+        // we put a barrier, then each thread is merged in order
+        #pragma omp barrier
+        #pragma omp single
         {
           if (!failure)
           {
-            pipeline.merge(private_pipeline);
+            for (int i = 0 ; i < ncpu_outer_loop ; i++)
+              pipeline.merge(*private_pipelines_ptr[i]);
           }
         }
       }