HCha/Stockfish
1
0
Fork 0
mirror of https://github.com/HChaZZY/Stockfish.git synced 2025-12-21 09:37:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

Improve smp performance for high number of threads

Browse Source 
Balance threads between split points.

There are huge differences between different machines and autopurging makes it very difficult to measure the improvement in fishtest, but the following was recorded for 16 threads at 15+0.05:

    For Bravone (1000 games): 0 ELO
    For Glinscott (1000 games): +20 ELO
    For bKingUs (1000 games): +50 ELO
    For fastGM (1500 games): +50 ELO

The change was regression for no one, and a big improvement for some, so it should be fine to commit it.
Also for 8 threads at 15+0.05 we measured a statistically significant improvement:
ELO: 6.19 +-3.9 (95%) LOS: 99.9%
Total: 10325 W: 1824 L: 1640 D: 6861

Finally it was verified that there was no (significant) regression for

4 threads:
ELO: 0.09 +-2.8 (95%) LOS: 52.4%
Total: 19908 W: 3422 L: 3417 D: 13069

2 threads:
ELO: 0.38 +-3.0 (95%) LOS: 60.0%
Total: 19044 W: 3480 L: 3459 D: 12105

1 thread:
ELO: -1.27 +-2.1 (95%) LOS: 12.3%
Total: 40000 W: 7829 L: 7975 D: 24196

Resolves #258
This commit is contained in:
Joona Kiiski
2015-02-14 20:46:00 +00:00
parent f8f5dcbb68
commit d65f75c153
3 changed files with 45 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
src/search.cpp
View File

@@ -1042,7 +1042,9 @@ moves_loop: // When in check and at SpNode search starts from here
&& Threads.size() >= 2
&& depth >= Threads.minimumSplitDepth
&& ( !thisThread->activeSplitPoint
|| !thisThread->activeSplitPoint->allSlavesSearching)
|| !thisThread->activeSplitPoint->allSlavesSearching
|| ( int(Threads.size()) > MAX_SLAVES_PER_SPLITPOINT
&& thisThread->activeSplitPoint->slavesCount == MAX_SLAVES_PER_SPLITPOINT))
&& thisThread->splitPointsSize < MAX_SPLITPOINTS_PER_THREAD)
{
assert(bestValue > -VALUE_INFINITE && bestValue < beta);
@@ -1587,6 +1589,11 @@ void Thread::idle_loop() {
// Try to late join to another split point if none of its slaves has
// already finished.
if (Threads.size() > 2)
{
SplitPoint *bestSp = NULL;
int bestThread = 0;
int bestScore = INT_MAX;
for (size_t i = 0; i < Threads.size(); ++i)
{
const int size = Threads[i]->splitPointsSize; // Local copy
@@ -1594,26 +1601,42 @@ void Thread::idle_loop() {
if ( sp
&& sp->allSlavesSearching
&& sp->slavesCount < MAX_SLAVES_PER_SPLITPOINT
&& available_to(Threads[i]))
{
// Recheck the conditions under lock protection
Threads.mutex.lock();
sp->mutex.lock();
int score = sp->spLevel * 256 * 256 + sp->slavesCount * 256 - sp->depth * 1;
if ( sp->allSlavesSearching
&& available_to(Threads[i]))
if (score < bestScore)
{
sp->slavesMask.set(idx);
activeSplitPoint = sp;
searching = true;
bestSp = sp;
bestThread = i;
bestScore = score;
}
sp->mutex.unlock();
Threads.mutex.unlock();
break; // Just a single attempt
}
}
if (bestSp)
{
sp = bestSp;
// Recheck the conditions under lock protection
Threads.mutex.lock();
sp->mutex.lock();
if ( sp->allSlavesSearching
&& sp->slavesCount < MAX_SLAVES_PER_SPLITPOINT
&& available_to(Threads[bestThread]))
{
sp->slavesMask.set(idx);
sp->slavesCount++;
activeSplitPoint = sp;
searching = true;
}
sp->mutex.unlock();
Threads.mutex.unlock();
}
}
}
// Grab the lock to avoid races with Thread::notify_one()