Copy only the search stack tail in split()

Only the previous, the current and the next ply SearchStack
are copied.

This reduces split overhead especially at low depth (high ply)
and with many threads.

Possibly no functional change (it is not easy to prove in SMP)

Signed-off-by: Marco Costalba <mcostalba@gmail.com>

diff --git a/src/search.cpp b/src/search.cpp
index df74d58..f205774 100644 (file)
--- a/src/search.cpp
+++ b/src/search.cpp
@@ -2986,15 +2986,15 @@ namespace {
     for (i = 0; i < ActiveThreads; i++)
         splitPoint->slaves[i] = 0;
 
-    // Copy the current search stack to the master thread
-    memcpy(splitPoint->sstack[master], sstck, (ply+1) * sizeof(SearchStack));
+    // Copy the tail of current search stack to the master thread
+    memcpy(splitPoint->sstack[master] + ply - 1, sstck + ply - 1, 3 * sizeof(SearchStack));
     Threads[master].splitPoint = splitPoint;
 
     // Make copies of the current position and search stack for each thread
     for (i = 0; i < ActiveThreads && splitPoint->cpus < MaxThreadsPerSplitPoint; i++)
         if (thread_is_available(i, master))
         {
-            memcpy(splitPoint->sstack[i], sstck, (ply+1) * sizeof(SearchStack));
+            memcpy(splitPoint->sstack[i] + ply - 1, sstck + ply - 1, 3 * sizeof(SearchStack));
             Threads[i].splitPoint = splitPoint;
             splitPoint->slaves[i] = 1;
             splitPoint->cpus++;