Merge 3bc2f27b97 into e32731b60b

Detect AMD engineering samples in randomx_boost.sh

scripts/build.hwloc.sh

@@ -1,8 +1,8 @@
 #!/bin/sh -e
 
 HWLOC_VERSION_MAJOR="2"
-HWLOC_VERSION_MINOR="10"
-HWLOC_VERSION_PATCH="0"
+HWLOC_VERSION_MINOR="11"
+HWLOC_VERSION_PATCH="2"
 
 HWLOC_VERSION="${HWLOC_VERSION_MAJOR}.${HWLOC_VERSION_MINOR}.${HWLOC_VERSION_PATCH}"
 

scripts/build.openssl3.sh

@@ -1,6 +1,6 @@
 #!/bin/sh -e
 
-OPENSSL_VERSION="3.0.14"
+OPENSSL_VERSION="3.0.15"
 
 mkdir -p deps
 mkdir -p deps/include

scripts/build.uv.sh

@@ -1,6 +1,6 @@
 #!/bin/sh -e
 
-UV_VERSION="1.48.0"
+UV_VERSION="1.49.2"
 
 mkdir -p deps
 mkdir -p deps/include

scripts/randomx_boost.sh

@@ -8,7 +8,7 @@ else
 	modprobe msr allow_writes=on
 fi
 
-if grep -E 'AMD Ryzen|AMD EPYC' /proc/cpuinfo > /dev/null;
+if grep -E 'AMD Ryzen|AMD EPYC|AuthenticAMD' /proc/cpuinfo > /dev/null;
 	then
 	if grep "cpu family[[:space:]]\{1,\}:[[:space:]]25" /proc/cpuinfo > /dev/null;
 		then
@@ -28,6 +28,14 @@ if grep -E 'AMD Ryzen|AMD EPYC' /proc/cpuinfo > /dev/null;
 					wrmsr -a 0xc001102b 0x2000cc10
 					echo "MSR register values for Zen3 applied"
 				fi
+		elif grep "cpu family[[:space:]]\{1,\}:[[:space:]]26" /proc/cpuinfo > /dev/null;
+			then
+				echo "Detected Zen5 CPU"
+				wrmsr -a 0xc0011020 0x4400000000000
+				wrmsr -a 0xc0011021 0x4000000000040
+				wrmsr -a 0xc0011022 0x8680000401570000
+				wrmsr -a 0xc001102b 0x2040cc10
+				echo "MSR register values for Zen5 applied"
 		else
 			echo "Detected Zen1/Zen2 CPU"
 			wrmsr -a 0xc0011020 0

src/backend/cpu/CpuWorker.cpp

@@ -67,7 +67,8 @@ VirtualMemory* cn_heavyZen3Memory = nullptr;
 template<size_t N>
 xmrig::CpuWorker<N>::CpuWorker(size_t id, const CpuLaunchData &data) :
     Worker(id, data.affinity, data.priority),
-    m_algorithm(data.algorithm),
+    m_algorithm_family(data.algorithm.family()),
+    m_algorithm_l3(data.algorithm.l3()),
     m_assembly(data.assembly),
     m_hwAES(data.hwAES),
     m_yield(data.yield),
@@ -82,24 +83,29 @@ xmrig::CpuWorker<N>::CpuWorker(size_t id, const CpuLaunchData &data) :
     const uint32_t model = Cpu::info()->model();
     const bool is_vermeer = (arch == ICpuInfo::ARCH_ZEN3) && (model == 0x21);
     const bool is_raphael = (arch == ICpuInfo::ARCH_ZEN4) && (model == 0x61);
-    if ((N == 1) && (m_av == CnHash::AV_SINGLE) && (m_algorithm.family() == Algorithm::CN_HEAVY) && (m_assembly != Assembly::NONE) && (is_vermeer || is_raphael)) {
+    if ((N == 1) && (m_av == CnHash::AV_SINGLE) && (m_algorithm_family == Algorithm::CN_HEAVY) && (m_assembly != Assembly::NONE) && (is_vermeer || is_raphael)) {
         std::lock_guard<std::mutex> lock(cn_heavyZen3MemoryMutex);
         if (!cn_heavyZen3Memory) {
             // Round up number of threads to the multiple of 8
             const size_t num_threads = ((m_threads + 7) / 8) * 8;
-            cn_heavyZen3Memory = new VirtualMemory(m_algorithm.l3() * num_threads, data.hugePages, false, false, node());
+            cn_heavyZen3Memory = new VirtualMemory(m_algorithm_l3 * num_threads, data.hugePages, false, false, node());
         }
         m_memory = cn_heavyZen3Memory;
     }
     else
 #   endif
     {
-        m_memory = new VirtualMemory(m_algorithm.l3() * N, data.hugePages, false, true, node());
+        m_memory = new VirtualMemory(m_algorithm_l3 * N, data.hugePages, false, true, node());
     }
 
 #   ifdef XMRIG_ALGO_GHOSTRIDER
     m_ghHelper = ghostrider::create_helper_thread(affinity(), data.priority, data.affinities);
 #   endif
+
+    std::vector<Algorithm> family = Algorithm::all([this](const Algorithm &algo) -> bool { return m_algorithm_family == algo.family(); });
+    for (auto& algorithm : family) {
+        m_hash_fns.emplace(algorithm.id(), CnHash::fn(algorithm, m_av, m_assembly));
+    }
 }
 
 
@@ -124,6 +130,23 @@ xmrig::CpuWorker<N>::~CpuWorker()
 #   endif
 }
 
+template<size_t N>
+xmrig::cn_hash_fun xmrig::CpuWorker<N>::fn(const Algorithm &algorithm) const {
+#if 1
+    if (!algorithm.isValid()) {
+        return nullptr;
+    }
+
+    const auto it = m_hash_fns.find(algorithm.id());
+    if (it != m_hash_fns.end()) {
+        return it->second;
+    }
+    return nullptr;
+#else
+    return CnHash::fn(algorithm, m_av, m_assembly);
+#endif
+}
+
 
 #ifdef XMRIG_ALGO_RANDOMX
 template<size_t N>
@@ -159,7 +182,7 @@ template<size_t N>
 bool xmrig::CpuWorker<N>::selfTest()
 {
 #   ifdef XMRIG_ALGO_RANDOMX
-    if (m_algorithm.family() == Algorithm::RANDOM_X) {
+    if (m_algorithm_family == Algorithm::RANDOM_X) {
         return N == 1;
     }
 #   endif
@@ -167,12 +190,12 @@ bool xmrig::CpuWorker<N>::selfTest()
     allocateCnCtx();
 
 #   ifdef XMRIG_ALGO_GHOSTRIDER
-    if (m_algorithm.family() == Algorithm::GHOSTRIDER) {
+    if (m_algorithm_family == Algorithm::GHOSTRIDER) {
         return (N == 8) && verify(Algorithm::GHOSTRIDER_RTM, test_output_gr);
     }
 #   endif
 
-    if (m_algorithm.family() == Algorithm::CN) {
+    if (m_algorithm_family == Algorithm::CN) {
         const bool rc = verify(Algorithm::CN_0,      test_output_v0)   &&
                         verify(Algorithm::CN_1,      test_output_v1)   &&
                         verify(Algorithm::CN_2,      test_output_v2)   &&
@@ -190,14 +213,14 @@ bool xmrig::CpuWorker<N>::selfTest()
     }
 
 #   ifdef XMRIG_ALGO_CN_LITE
-    if (m_algorithm.family() == Algorithm::CN_LITE) {
+    if (m_algorithm_family == Algorithm::CN_LITE) {
         return verify(Algorithm::CN_LITE_0,    test_output_v0_lite) &&
                verify(Algorithm::CN_LITE_1,    test_output_v1_lite);
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_CN_HEAVY
-    if (m_algorithm.family() == Algorithm::CN_HEAVY) {
+    if (m_algorithm_family == Algorithm::CN_HEAVY) {
         return verify(Algorithm::CN_HEAVY_0,    test_output_v0_heavy)  &&
                verify(Algorithm::CN_HEAVY_XHV,  test_output_xhv_heavy) &&
                verify(Algorithm::CN_HEAVY_TUBE, test_output_tube_heavy);
@@ -205,20 +228,20 @@ bool xmrig::CpuWorker<N>::selfTest()
 #   endif
 
 #   ifdef XMRIG_ALGO_CN_PICO
-    if (m_algorithm.family() == Algorithm::CN_PICO) {
+    if (m_algorithm_family == Algorithm::CN_PICO) {
         return verify(Algorithm::CN_PICO_0, test_output_pico_trtl) &&
                verify(Algorithm::CN_PICO_TLO, test_output_pico_tlo);
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_CN_FEMTO
-    if (m_algorithm.family() == Algorithm::CN_FEMTO) {
+    if (m_algorithm_family == Algorithm::CN_FEMTO) {
         return verify(Algorithm::CN_UPX2, test_output_femto_upx2);
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_ARGON2
-    if (m_algorithm.family() == Algorithm::ARGON2) {
+    if (m_algorithm_family == Algorithm::ARGON2) {
         return verify(Algorithm::AR2_CHUKWA, argon2_chukwa_test_out) &&
                verify(Algorithm::AR2_CHUKWA_V2, argon2_chukwa_v2_test_out) &&
                verify(Algorithm::AR2_WRKZ, argon2_wrkz_test_out);
@@ -262,7 +285,7 @@ void xmrig::CpuWorker<N>::start()
         while (!Nonce::isOutdated(Nonce::CPU, m_job.sequence())) {
             const Job &job = m_job.currentJob();
 
-            if (job.algorithm().l3() != m_algorithm.l3()) {
+            if (job.algorithm().l3() != m_algorithm_l3) {
                 break;
             }
 
@@ -359,7 +382,9 @@ void xmrig::CpuWorker<N>::start()
             }
         }
 
-        consumeJob();
+        if (!Nonce::isPaused()) {
+            consumeJob();
+        }
     }
 }
 
@@ -487,11 +512,11 @@ void xmrig::CpuWorker<N>::allocateCnCtx()
 #       ifdef XMRIG_ALGO_CN_HEAVY
         // cn-heavy optimization for Zen3 CPUs
         if (m_memory == cn_heavyZen3Memory) {
-            shift = (id() / 8) * m_algorithm.l3() * 8 + (id() % 8) * 64;
+            shift = (id() / 8) * m_algorithm_l3 * 8 + (id() % 8) * 64;
         }
 #       endif
 
-        CnCtx::create(m_ctx, m_memory->scratchpad() + shift, m_algorithm.l3(), N);
+        CnCtx::create(m_ctx, m_memory->scratchpad() + shift, m_algorithm_l3, N);
     }
 }
 

src/backend/cpu/CpuWorker.h

@@ -26,6 +26,8 @@
 #include "base/tools/Object.h"
 #include "net/JobResult.h"
 
+#include <unordered_map>
+
 
 #ifdef XMRIG_ALGO_RANDOMX
 class randomx_vm;
@@ -55,7 +57,7 @@ public:
     size_t threads() const override
     {
 #       ifdef XMRIG_ALGO_GHOSTRIDER
-        return ((m_algorithm.family() == Algorithm::GHOSTRIDER) && m_ghHelper) ? 2 : 1;
+        return ((m_algorithm_family == Algorithm::GHOSTRIDER) && m_ghHelper) ? 2 : 1;
 #       else
         return 1;
 #       endif
@@ -71,7 +73,7 @@ protected:
     inline void jobEarlyNotification(const Job&) override   {}
 
 private:
-    inline cn_hash_fun fn(const Algorithm &algorithm) const { return CnHash::fn(algorithm, m_av, m_assembly); }
+    cn_hash_fun fn(const Algorithm &algorithm) const;
 
 #   ifdef XMRIG_ALGO_RANDOMX
     void allocateRandomX_VM();
@@ -84,8 +86,10 @@ private:
     void consumeJob();
 
     alignas(8) uint8_t m_hash[N * 32]{ 0 };
-    const Algorithm m_algorithm;
+    const uint32_t m_algorithm_family;
+    const size_t m_algorithm_l3;
     const Assembly m_assembly;
+    std::unordered_map<Algorithm::Id, cn_hash_fun> m_hash_fns;
     const bool m_hwAES;
     const bool m_yield;
     const CnHash::AlgoVariant m_av;

src/backend/cpu/platform/HwlocCpuInfo.cpp

@@ -326,7 +326,8 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
         }
     }
 
-    if (scratchpad == 2 * oneMiB) {
+    // This code is supposed to run only on Intel CPUs
+    if ((vendor() == VENDOR_INTEL) && (scratchpad == 2 * oneMiB)) {
         if (L2 && (cores.size() * oneMiB) == L2 && L2_associativity == 16 && L3 >= L2) {
             L3    = L2;
             extra = L2;

src/backend/cuda/CudaWorker.cpp

@@ -158,7 +158,7 @@ void xmrig::CudaWorker::start()
             std::this_thread::yield();
         }
 
-        if (!consumeJob()) {
+        if (isReady() && !consumeJob()) {
             return;
         }
     }

src/backend/opencl/OclWorker.cpp

@@ -190,7 +190,7 @@ void xmrig::OclWorker::start()
             std::this_thread::yield();
         }
 
-        if (!consumeJob()) {
+        if (isReady() && !consumeJob()) {
             return;
         }
     }

src/core/Miner.cpp

@@ -576,6 +576,11 @@ void xmrig::Miner::setJob(const Job &job, bool donate)
 
 #   ifdef XMRIG_ALGO_RANDOMX
     const bool ready = d_ptr->initRX();
+
+    // Always reset nonce on RandomX dataset change
+    if (!ready) {
+        d_ptr->reset = true;
+    }
 #   else
     constexpr const bool ready = true;
 #   endif

src/crypto/ghostrider/README.md

@@ -16,7 +16,7 @@ xmrig -a gr -o rtm.suprnova.cc:4273 --tls -u WALLET_ADDRESS -p x
 
 You can use **rtm_ghostrider_example.cmd** as a template and put pool URL and your wallet address there. The general XMRig documentation is available [here](https://xmrig.com/docs/miner).
 
-**Using `--threads` or `-t` option is NOT recommended because it turns off advanced built-in config.** If you want to tweak the nubmer of threads used for GhostRider, it's recommended to start using config.json instead of command line. The best suitable command line option for this is `--cpu-max-threads-hint=N` where N can be between 0 and 100.
+**Using `--threads` or `-t` option is NOT recommended because it turns off advanced built-in config.** If you want to tweak the number of threads used for GhostRider, it's recommended to start using config.json instead of command line. The best suitable command line option for this is `--cpu-max-threads-hint=N` where N can be between 0 and 100.
 
 ## Performance
 

src/crypto/randomx/jit_compiler_x86.cpp

@@ -267,8 +267,8 @@ namespace randomx {
 						initDatasetAVX2 = false;
 						break;
 					case xmrig::ICpuInfo::ARCH_ZEN5:
-						// TODO: test it
-						initDatasetAVX2 = false;
+						// AVX2 init is 49% faster on Zen5
+						initDatasetAVX2 = true;
 						break;
 					}
 				}

src/crypto/rx/RxConfig.cpp

@@ -60,8 +60,6 @@ static const std::array<MsrItems, kMsrArraySize> msrPresets = {
     MsrItems{{ 0xC0011020, 0ULL }, { 0xC0011021, 0x40ULL, ~0x20ULL }, { 0xC0011022, 0x1510000ULL }, { 0xC001102b, 0x2000cc16ULL }},
     MsrItems{{ 0xC0011020, 0x0004480000000000ULL }, { 0xC0011021, 0x001c000200000040ULL, ~0x20ULL }, { 0xC0011022, 0xc000000401570000ULL }, { 0xC001102b, 0x2000cc10ULL }},
     MsrItems{{ 0xC0011020, 0x0004400000000000ULL }, { 0xC0011021, 0x0004000000000040ULL, ~0x20ULL }, { 0xC0011022, 0x8680000401570000ULL }, { 0xC001102b, 0x2040cc10ULL }},
-
-    // TODO: Tune it for Zen5 when it's available
     MsrItems{{ 0xC0011020, 0x0004400000000000ULL }, { 0xC0011021, 0x0004000000000040ULL, ~0x20ULL }, { 0xC0011022, 0x8680000401570000ULL }, { 0xC001102b, 0x2040cc10ULL }},
 
     MsrItems{{ 0x1a4, 0xf }},

src/version.h

@@ -22,7 +22,7 @@
 #define APP_ID        "xmrig"
 #define APP_NAME      "XMRig"
 #define APP_DESC      "XMRig miner"
-#define APP_VERSION   "6.22.0"
+#define APP_VERSION   "6.22.1-dev"
 #define APP_DOMAIN    "xmrig.com"
 #define APP_SITE      "www.xmrig.com"
 #define APP_COPYRIGHT "Copyright (C) 2016-2024 xmrig.com"
@@ -30,7 +30,7 @@
 
 #define APP_VER_MAJOR  6
 #define APP_VER_MINOR  22
-#define APP_VER_PATCH  0
+#define APP_VER_PATCH  1
 
 #ifdef _MSC_VER
 #   if (_MSC_VER >= 1930)