Merge 6c8098378a into 36fdfa2694

Optimized autoconfig for AMD CPUs with < 2 MB L3 cache per thread

CMakeLists.txt

@@ -32,6 +32,7 @@ option(WITH_VAES            "Enable VAES instructions for Cryptonight" ON)
 option(WITH_BENCHMARK       "Enable builtin RandomX benchmark and stress test" ON)
 option(WITH_SECURE_JIT      "Enable secure access to JIT memory" OFF)
 option(WITH_DMI             "Enable DMI/SMBIOS reader" ON)
+option(WITH_AUTO_HUGEPAGE   "Enable Automatic setting of nr_hugepages (Linux Only)" ON)
 
 option(BUILD_STATIC         "Build static binary" OFF)
 option(ARM_V8               "Force ARMv8 (64 bit) architecture, use with caution if automatic detection fails, but you sure it may work" OFF)
@@ -181,6 +182,10 @@ else()
     if (XMRIG_OS_ANDROID)
         set(EXTRA_LIBS pthread rt dl log)
     elseif (XMRIG_OS_LINUX)
+        if (WITH_AUTO_HUGEPAGE)
+            add_definitions(/DXMRIG_FEATURE_AUTO_HUGEPAGE)
+        endif()
+
         list(APPEND SOURCES_OS
             src/crypto/common/LinuxMemory.h
             src/crypto/common/LinuxMemory.cpp

src/backend/cpu/platform/HwlocCpuInfo.cpp

@@ -320,8 +320,13 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
             L2 += l2->attr->cache.size;
             L2_associativity = l2->attr->cache.associativity;
 
-            if (L3_exclusive && l2->attr->cache.size >= scratchpad) {
+            if (L3_exclusive) {
-                extra += scratchpad;
+                if (vendor() == VENDOR_AMD) {
+                    extra += std::min<size_t>(l2->attr->cache.size, scratchpad);
+                }
+                else if (l2->attr->cache.size >= scratchpad) {
+                    extra += scratchpad;
+                }
             }
         }
     }

src/crypto/common/LinuxMemory.cpp

@@ -60,7 +60,11 @@ bool xmrig::LinuxMemory::reserve(size_t size, uint32_t node, size_t hugePageSize
         return false;
     }
 
+#   ifdef XMRIG_FEATURE_AUTO_HUGEPAGE
     return write_nr_hugepages(node, hugePageSize, std::max<size_t>(nr_hugepages(node, hugePageSize), 0) + (required - available));
+#   else
+    return false;
+#   endif
 }