From 061b68b43a9eb53866b8890dd88904b08de50f08 Mon Sep 17 00:00:00 2001
From: Matthijs Douze <matthijs@fb.com>
Date: Fri, 30 Apr 2021 00:01:44 -0700
Subject: [PATCH] Fix performance regression in ResultHandler (#1840)

Summary:
Pull Request resolved: https://github.com/facebookresearch/faiss/pull/1840

This diff is related to

https://github.com/facebookresearch/faiss/issues/1762

The ResultHandler introduced for FlatL2 and FlatIP was not multithreaded. This diff attempts to fix that. To be verified if it is indeed faster.

Reviewed By: wickedfoo

Differential Revision: D27939173

fbshipit-source-id: c85f01a97d4249fe0c6bfb04396b68a7a9fe643d
---
 faiss/impl/ResultHandler.h | 13 ++++++++-----
 faiss/utils/distances.cpp  |  2 +-
 2 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/faiss/impl/ResultHandler.h b/faiss/impl/ResultHandler.h
index ec03f7f656..65790d6011 100644
--- a/faiss/impl/ResultHandler.h
+++ b/faiss/impl/ResultHandler.h
@@ -92,13 +92,14 @@ struct HeapResultHandler {
 
     /// add results for query i0..i1 and j0..j1
     void add_results(size_t j0, size_t j1, const T* dis_tab) {
-        // maybe parallel for
-        for (size_t i = i0; i < i1; i++) {
+#pragma omp parallel for
+        for (int64_t i = i0; i < i1; i++) {
             T* heap_dis = heap_dis_tab + i * k;
             TI* heap_ids = heap_ids_tab + i * k;
+            const T* dis_tab_i = dis_tab + (j1 - j0) * (i - i0) - j0;
             T thresh = heap_dis[0];
             for (size_t j = j0; j < j1; j++) {
-                T dis = *dis_tab++;
+                T dis = dis_tab_i[j];
                 if (C::cmp(thresh, dis)) {
                     heap_replace_top<C>(k, heap_dis, heap_ids, dis, j);
                     thresh = heap_dis[0];
@@ -281,10 +282,12 @@ struct ReservoirResultHandler {
     /// add results for query i0..i1 and j0..j1
     void add_results(size_t j0, size_t j1, const T* dis_tab) {
         // maybe parallel for
-        for (size_t i = i0; i < i1; i++) {
+#pragma omp parallel for
+        for (int64_t i = i0; i < i1; i++) {
             ReservoirTopN<C>& reservoir = reservoirs[i - i0];
+            const T* dis_tab_i = dis_tab + (j1 - j0) * (i - i0) - j0;
             for (size_t j = j0; j < j1; j++) {
-                T dis = *dis_tab++;
+                T dis = dis_tab_i[j];
                 reservoir.add(dis, j);
             }
         }
diff --git a/faiss/utils/distances.cpp b/faiss/utils/distances.cpp
index 60c50ac1e9..e39b55e801 100644
--- a/faiss/utils/distances.cpp
+++ b/faiss/utils/distances.cpp
@@ -286,7 +286,7 @@ void exhaustive_L2sqr_blas(
                        ip_block.get(),
                        &nyi);
             }
-
+#pragma omp parallel for
             for (int64_t i = i0; i < i1; i++) {
                 float* ip_line = ip_block.get() + (i - i0) * (j1 - j0);