X-Git-Url: https://git.sesse.net/?p=stockfish;a=blobdiff_plain;f=src%2Fmisc.cpp;h=2bcc7c1382de2514bf34c200106860824c6da33b;hp=08df524bf26beb5bed4453910cc161b1ed8364a3;hb=ae5d2c38e1b9f6a990c29d31a37d6555d197f939;hpb=126036abb093df88593ef9db67ac0b9a07f411a6

diff --git a/src/misc.cpp b/src/misc.cpp
index 08df524b..2bcc7c13 100644
--- a/src/misc.cpp
+++ b/src/misc.cpp
@@ -2,7 +2,7 @@
   Stockfish, a UCI chess playing engine derived from Glaurung 2.1
   Copyright (C) 2004-2008 Tord Romstad (Glaurung author)
   Copyright (C) 2008-2015 Marco Costalba, Joona Kiiski, Tord Romstad
-  Copyright (C) 2015-2016 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
+  Copyright (C) 2015-2019 Marco Costalba, Joona Kiiski, Gary Linscott, Tord Romstad
 
   Stockfish is free software: you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
@@ -18,10 +18,34 @@
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */
 
+#ifdef _WIN32
+#if _WIN32_WINNT < 0x0601
+#undef  _WIN32_WINNT
+#define _WIN32_WINNT 0x0601 // Force to include needed API prototypes
+#endif
+
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif
+
+#include <windows.h>
+// The needed Windows API for processor groups could be missed from old Windows
+// versions, so instead of calling them directly (forcing the linker to resolve
+// the calls at compile time), try to load them at runtime. To do this we need
+// first to define the corresponding function pointers.
+extern "C" {
+typedef bool(*fun1_t)(LOGICAL_PROCESSOR_RELATIONSHIP,
+                      PSYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX, PDWORD);
+typedef bool(*fun2_t)(USHORT, PGROUP_AFFINITY);
+typedef bool(*fun3_t)(HANDLE, CONST GROUP_AFFINITY*, PGROUP_AFFINITY);
+}
+#endif
+
 #include <fstream>
 #include <iomanip>
 #include <iostream>
 #include <sstream>
+#include <vector>
 
 #include "misc.h"
 #include "thread.h"
@@ -44,10 +68,10 @@ struct Tie: public streambuf { // MSVC requires split streambuf for cin and cout
 
   Tie(streambuf* b, streambuf* l) : buf(b), logBuf(l) {}
 
-  int sync() { return logBuf->pubsync(), buf->pubsync(); }
-  int overflow(int c) { return log(buf->sputc((char)c), "<< "); }
-  int underflow() { return buf->sgetc(); }
-  int uflow() { return log(buf->sbumpc(), ">> "); }
+  int sync() override { return logBuf->pubsync(), buf->pubsync(); }
+  int overflow(int c) override { return log(buf->sputc((char)c), "<< "); }
+  int underflow() override { return buf->sgetc(); }
+  int uflow() override { return log(buf->sbumpc(), ">> "); }
 
   streambuf *buf, *logBuf;
 
@@ -185,3 +209,114 @@ void prefetch(void* addr) {
 }
 
 #endif
+
+void prefetch2(void* addr) {
+
+  prefetch(addr);
+  prefetch((uint8_t*)addr + 64);
+}
+
+namespace WinProcGroup {
+
+#ifndef _WIN32
+
+void bindThisThread(size_t) {}
+
+#else
+
+/// best_group() retrieves logical processor information using Windows specific
+/// API and returns the best group id for the thread with index idx. Original
+/// code from Texel by Peter Österlund.
+
+int best_group(size_t idx) {
+
+  int threads = 0;
+  int nodes = 0;
+  int cores = 0;
+  DWORD returnLength = 0;
+  DWORD byteOffset = 0;
+
+  // Early exit if the needed API is not available at runtime
+  HMODULE k32 = GetModuleHandle("Kernel32.dll");
+  auto fun1 = (fun1_t)(void(*)())GetProcAddress(k32, "GetLogicalProcessorInformationEx");
+  if (!fun1)
+      return -1;
+
+  // First call to get returnLength. We expect it to fail due to null buffer
+  if (fun1(RelationAll, nullptr, &returnLength))
+      return -1;
+
+  // Once we know returnLength, allocate the buffer
+  SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *buffer, *ptr;
+  ptr = buffer = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)malloc(returnLength);
+
+  // Second call, now we expect to succeed
+  if (!fun1(RelationAll, buffer, &returnLength))
+  {
+      free(buffer);
+      return -1;
+  }
+
+  while (ptr->Size > 0 && byteOffset + ptr->Size <= returnLength)
+  {
+      if (ptr->Relationship == RelationNumaNode)
+          nodes++;
+
+      else if (ptr->Relationship == RelationProcessorCore)
+      {
+          cores++;
+          threads += (ptr->Processor.Flags == LTP_PC_SMT) ? 2 : 1;
+      }
+
+      byteOffset += ptr->Size;
+      ptr = (SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX*)(((char*)ptr) + ptr->Size);
+  }
+
+  free(buffer);
+
+  std::vector<int> groups;
+
+  // Run as many threads as possible on the same node until core limit is
+  // reached, then move on filling the next node.
+  for (int n = 0; n < nodes; n++)
+      for (int i = 0; i < cores / nodes; i++)
+          groups.push_back(n);
+
+  // In case a core has more than one logical processor (we assume 2) and we
+  // have still threads to allocate, then spread them evenly across available
+  // nodes.
+  for (int t = 0; t < threads - cores; t++)
+      groups.push_back(t % nodes);
+
+  // If we still have more threads than the total number of logical processors
+  // then return -1 and let the OS to decide what to do.
+  return idx < groups.size() ? groups[idx] : -1;
+}
+
+
+/// bindThisThread() set the group affinity of the current thread
+
+void bindThisThread(size_t idx) {
+
+  // Use only local variables to be thread-safe
+  int group = best_group(idx);
+
+  if (group == -1)
+      return;
+
+  // Early exit if the needed API are not available at runtime
+  HMODULE k32 = GetModuleHandle("Kernel32.dll");
+  auto fun2 = (fun2_t)(void(*)())GetProcAddress(k32, "GetNumaNodeProcessorMaskEx");
+  auto fun3 = (fun3_t)(void(*)())GetProcAddress(k32, "SetThreadGroupAffinity");
+
+  if (!fun2 || !fun3)
+      return;
+
+  GROUP_AFFINITY affinity;
+  if (fun2(group, &affinity))
+      fun3(GetCurrentThread(), &affinity, nullptr);
+}
+
+#endif
+
+} // namespace WinProcGroup