Merge xmrig v6.16.2 into master

MoneroOcean · Dec 2, 2021 · 615715e · 615715e
2 parents eab57aa + 718c7e0
commit 615715e
Show file tree

Hide file tree

Showing 30 changed files with 1,506 additions and 339 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,12 +1,22 @@
+# v6.16.2
+- [#2751](https://github.com/xmrig/xmrig/pull/2751) Fixed crash on CPUs supporting VAES and running GCC-compiled xmrig.
+- [#2761](https://github.com/xmrig/xmrig/pull/2761) Fixed broken auto-tuning in GCC Windows build.
+- [#2771](https://github.com/xmrig/xmrig/issues/2771) Fixed environment variables support for GhostRider and KawPow. 
+- [#2769](https://github.com/xmrig/xmrig/pull/2769) Performance fixes:
+  - Fixed several performance bottlenecks introduced in v6.16.1.
+  - Fixed overall GCC-compiled build performance, it's the same speed as MSVC build now.
+  - **Linux builds are up to 10% faster now compared to v6.16.0 GCC build.**
+  - **Windows builds are up to 5% faster now compared to v6.16.0 MSVC build.**
+
 # v6.16.1
 - [#2729](https://github.com/xmrig/xmrig/pull/2729) GhostRider fixes:
-  - Added average hashrate display
-  - Fixed the number of threads shown at startup
-  - Fixed `--threads` or `-t` command line option (but `--cpu-max-threads-hint` is recommended to use)
+  - Added average hashrate display.
+  - Fixed the number of threads shown at startup.
+  - Fixed `--threads` or `-t` command line option (but `--cpu-max-threads-hint` is recommended to use).
 - [#2738](https://github.com/xmrig/xmrig/pull/2738) GhostRider fixes:
-  - Fixed "difficulty is not a number" error when diff is high on some pools
-  - Fixed GhostRider compilation when WITH_KAWPOW=OFF
-- [#2740](https://github.com/xmrig/xmrig/pull/2740) Added VAES support for Cryptonight variants **+4% speedup on Zen3**
+  - Fixed "difficulty is not a number" error when diff is high on some pools.
+  - Fixed GhostRider compilation when `WITH_KAWPOW=OFF`.
+- [#2740](https://github.com/xmrig/xmrig/pull/2740) Added VAES support for Cryptonight variants **+4% speedup on Zen3**.
   - VAES instructions are available on Intel Ice Lake/AMD Zen3 and newer CPUs.
   - +4% speedup on Ryzen 5 5600X.
 

diff --git a/src/backend/cpu/CpuWorker.h b/src/backend/cpu/CpuWorker.h
@@ -55,7 +55,7 @@ class CpuWorker : public Worker
     size_t threads() const override
     {
 #       ifdef XMRIG_ALGO_GHOSTRIDER
-        return m_ghHelper ? 2 : 1;
+        return ((m_algorithm.family() == Algorithm::GHOSTRIDER) && m_ghHelper) ? 2 : 1;
 #       else
         return 1;
 #       endif

diff --git a/src/backend/cpu/platform/BasicCpuInfo.cpp b/src/backend/cpu/platform/BasicCpuInfo.cpp
@@ -30,6 +30,12 @@
 #endif
 
 
+#include "crypto/cn/CryptoNight_monero.h"
+#ifdef XMRIG_VAES
+#   include "crypto/cn/CryptoNight_x86_vaes.h"
+#endif
+
+
 #include "backend/cpu/platform/BasicCpuInfo.h"
 #include "3rdparty/rapidjson/document.h"
 #include "crypto/common/Assembly.h"
@@ -294,6 +300,9 @@ xmrig::BasicCpuInfo::BasicCpuInfo() :
         }
     }
 #   endif
+
+    cn_sse41_enabled = has(FLAG_SSE41);
+    cn_vaes_enabled = has(FLAG_VAES);
 }
 
 

diff --git a/src/base/base.cmake b/src/base/base.cmake
@@ -132,6 +132,7 @@ set(SOURCES_BASE
     src/base/net/tools/LineReader.cpp
     src/base/net/tools/NetBuffer.cpp
     src/base/tools/Arguments.cpp
+    src/base/tools/Chrono.cpp
     src/base/tools/cryptonote/BlockTemplate.cpp
     src/base/tools/cryptonote/crypto-ops-data.c
     src/base/tools/cryptonote/crypto-ops.c

diff --git a/src/base/net/stratum/EthStratumClient.cpp b/src/base/net/stratum/EthStratumClient.cpp
@@ -73,7 +73,7 @@ int64_t xmrig::EthStratumClient::submit(const JobResult& result)
     auto& allocator = doc.GetAllocator();
 
     Value params(kArrayType);
-    params.PushBack(m_pool.user().toJSON(), allocator);
+    params.PushBack(m_user.toJSON(), allocator);
     params.PushBack(result.jobId.toJSON(), allocator);
 
 #   ifdef XMRIG_ALGO_GHOSTRIDER
@@ -471,8 +471,8 @@ void xmrig::EthStratumClient::authorize()
     auto &allocator = doc.GetAllocator();
 
     Value params(kArrayType);
-    params.PushBack(m_pool.user().toJSON(), allocator);
-    params.PushBack(m_pool.password().toJSON(), allocator);
+    params.PushBack(m_user.toJSON(), allocator);
+    params.PushBack(m_password.toJSON(), allocator);
 
     JsonRequest::create(doc, m_sequence, "mining.authorize", params);
 

diff --git a/src/base/net/stratum/benchmark/BenchClient.cpp b/src/base/net/stratum/benchmark/BenchClient.cpp
@@ -50,6 +50,7 @@ xmrig::BenchClient::BenchClient(const std::shared_ptr<BenchConfig> &benchmark, I
 
 #   ifdef XMRIG_ALGO_GHOSTRIDER
     if (m_benchmark->algorithm() == Algorithm::GHOSTRIDER_RTM) {
+        const uint32_t q = (benchmark->rotation() / 20) & 1;
         const uint32_t r = benchmark->rotation() % 20;
 
         static constexpr uint32_t indices[20][3] = {
@@ -75,9 +76,9 @@ xmrig::BenchClient::BenchClient(const std::shared_ptr<BenchConfig> &benchmark, I
              { 3, 4, 5 },
         };
 
-        blob[ 8] = '0' + indices[r][1];
+        blob[ 8] = '0' + indices[r][q ? 2 : 1];
         blob[ 9] = '0' + indices[r][0];
-        blob[11] = '0' + indices[r][2];
+        blob[11] = '0' + indices[r][q ? 1 : 2];
     }
 #   endif
 

diff --git a/src/base/tools/Chrono.cpp b/src/base/tools/Chrono.cpp
@@ -0,0 +1,44 @@
+/* XMRig
+ * Copyright (c) 2018-2021 SChernykh   <https://github.com/SChernykh>
+ * Copyright (c) 2016-2021 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ *
+ *   This program is free software: you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, either version 3 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "Chrono.h"
+
+
+#ifdef XMRIG_OS_WIN
+#   include <Windows.h>
+#endif
+
+
+namespace xmrig {
+
+
+double Chrono::highResolutionMSecs()
+{
+#   ifdef XMRIG_OS_WIN
+    LARGE_INTEGER f, t;
+    QueryPerformanceFrequency(&f);
+    QueryPerformanceCounter(&t);
+    return static_cast<double>(t.QuadPart) * 1e3 / f.QuadPart;
+#   else
+    using namespace std::chrono;
+    return static_cast<uint64_t>(duration_cast<nanoseconds>(high_resolution_clock::now().time_since_epoch()).count()) / 1e6;
+#   endif
+}
+
+
+} /* namespace xmrig */
diff --git a/src/base/tools/Chrono.h b/src/base/tools/Chrono.h
@@ -29,12 +29,7 @@ namespace xmrig {
 class Chrono
 {
 public:
-    static inline uint64_t highResolutionMSecs()
-    {
-        using namespace std::chrono;
-
-        return static_cast<uint64_t>(time_point_cast<milliseconds>(high_resolution_clock::now()).time_since_epoch().count());
-    }
+    static double highResolutionMSecs();
 
 
     static inline uint64_t steadyMSecs()

diff --git a/src/crypto/cn/CnHash.cpp b/src/crypto/cn/CnHash.cpp
@@ -55,6 +55,10 @@
     } while (0)
 
 
+bool cn_sse41_enabled = false;
+bool cn_vaes_enabled = false;
+
+
 #ifdef XMRIG_FEATURE_ASM
 #   define ADD_FN_ASM(algo) do {                                                                                    \
         m_map[algo]->data[AV_SINGLE][Assembly::INTEL]     = cryptonight_single_hash_asm<algo, Assembly::INTEL>;     \
@@ -97,6 +101,27 @@ cn_mainloop_fun        cn_double_double_mainloop_sandybridge_asm  = nullptr;
 cn_mainloop_fun        cn_upx2_mainloop_asm                       = nullptr;
 cn_mainloop_fun        cn_upx2_double_mainloop_asm                = nullptr;
 
+cn_mainloop_fun        cn_gr0_single_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr1_single_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr2_single_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr3_single_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr4_single_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr5_single_mainloop_asm                 = nullptr;
+
+cn_mainloop_fun        cn_gr0_double_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr1_double_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr2_double_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr3_double_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr4_double_mainloop_asm                 = nullptr;
+cn_mainloop_fun        cn_gr5_double_mainloop_asm                 = nullptr;
+
+cn_mainloop_fun        cn_gr0_quad_mainloop_asm                   = nullptr;
+cn_mainloop_fun        cn_gr1_quad_mainloop_asm                   = nullptr;
+cn_mainloop_fun        cn_gr2_quad_mainloop_asm                   = nullptr;
+cn_mainloop_fun        cn_gr3_quad_mainloop_asm                   = nullptr;
+cn_mainloop_fun        cn_gr4_quad_mainloop_asm                   = nullptr;
+cn_mainloop_fun        cn_gr5_quad_mainloop_asm                   = nullptr;
+
 
 template<Algorithm::Id SOURCE_ALGO = Algorithm::CN_2, typename T, typename U>
 static void patchCode(T dst, U src, const uint32_t iterations, const uint32_t mask = CnAlgo<Algorithm::CN_HALF>().mask())
@@ -136,7 +161,7 @@ static void patchCode(T dst, U src, const uint32_t iterations, const uint32_t ma
 
 static void patchAsmVariants()
 {
-    const int allocation_size = 131072;
+    constexpr size_t allocation_size = 0x20000;
     auto base = static_cast<uint8_t *>(VirtualMemory::allocateExecutableMemory(allocation_size, false));
 
     cn_half_mainloop_ivybridge_asm              = reinterpret_cast<cn_mainloop_fun>         (base + 0x0000);
@@ -173,6 +198,29 @@ static void patchAsmVariants()
     cn_upx2_double_mainloop_asm                 = reinterpret_cast<cn_mainloop_fun>         (base + 0x15000);
 #   endif
 
+#   ifdef XMRIG_ALGO_GHOSTRIDER
+    cn_gr0_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x16000);
+    cn_gr1_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x16800);
+    cn_gr2_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x17000);
+    cn_gr3_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x17800);
+    cn_gr4_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x18000);
+    cn_gr5_single_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x18800);
+
+    cn_gr0_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x19000);
+    cn_gr1_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x19800);
+    cn_gr2_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x1A000);
+    cn_gr3_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x1A800);
+    cn_gr4_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x1B000);
+    cn_gr5_double_mainloop_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x1B800);
+
+    cn_gr0_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1C000);
+    cn_gr1_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1C800);
+    cn_gr2_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1D000);
+    cn_gr3_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1D800);
+    cn_gr4_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1E000);
+    cn_gr5_quad_mainloop_asm                    = reinterpret_cast<cn_mainloop_fun>         (base + 0x1E800);
+#   endif
+
     {
         constexpr uint32_t ITER = CnAlgo<Algorithm::CN_HALF>().iterations();
 
@@ -230,7 +278,30 @@ static void patchAsmVariants()
         patchCode<Algorithm::CN_RWZ>(cn_upx2_mainloop_asm,        cnv2_rwz_mainloop_asm,            ITER,   MASK);
         patchCode<Algorithm::CN_RWZ>(cn_upx2_double_mainloop_asm, cnv2_rwz_double_mainloop_asm,     ITER,   MASK);
     }
-#endif
+#   endif
+
+#   ifdef XMRIG_ALGO_GHOSTRIDER
+    patchCode<Algorithm::CN_1>(cn_gr0_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_0>().iterations(), CnAlgo<Algorithm::CN_GR_0>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr1_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_1>().iterations(), CnAlgo<Algorithm::CN_GR_1>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr2_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_2>().iterations(), CnAlgo<Algorithm::CN_GR_2>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr3_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_3>().iterations(), CnAlgo<Algorithm::CN_GR_3>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr4_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_4>().iterations(), CnAlgo<Algorithm::CN_GR_4>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr5_single_mainloop_asm, cnv1_single_mainloop_asm, CnAlgo<Algorithm::CN_GR_5>().iterations(), CnAlgo<Algorithm::CN_GR_5>().mask());
+
+    patchCode<Algorithm::CN_1>(cn_gr0_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_0>().iterations(), CnAlgo<Algorithm::CN_GR_0>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr1_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_1>().iterations(), CnAlgo<Algorithm::CN_GR_1>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr2_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_2>().iterations(), CnAlgo<Algorithm::CN_GR_2>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr3_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_3>().iterations(), CnAlgo<Algorithm::CN_GR_3>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr4_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_4>().iterations(), CnAlgo<Algorithm::CN_GR_4>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr5_double_mainloop_asm, cnv1_double_mainloop_asm, CnAlgo<Algorithm::CN_GR_5>().iterations(), CnAlgo<Algorithm::CN_GR_5>().mask());
+
+    patchCode<Algorithm::CN_1>(cn_gr0_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_0>().iterations(), CnAlgo<Algorithm::CN_GR_0>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr1_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_1>().iterations(), CnAlgo<Algorithm::CN_GR_1>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr2_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_2>().iterations(), CnAlgo<Algorithm::CN_GR_2>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr3_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_3>().iterations(), CnAlgo<Algorithm::CN_GR_3>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr4_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_4>().iterations(), CnAlgo<Algorithm::CN_GR_4>().mask());
+    patchCode<Algorithm::CN_1>(cn_gr5_quad_mainloop_asm, cnv1_quad_mainloop_asm, CnAlgo<Algorithm::CN_GR_5>().iterations(), CnAlgo<Algorithm::CN_GR_5>().mask());
+#   endif
 
     VirtualMemory::protectRX(base, allocation_size);
     VirtualMemory::flushInstructionCache(base, allocation_size);

diff --git a/src/crypto/cn/CryptoNight.h b/src/crypto/cn/CryptoNight.h
@@ -52,8 +52,10 @@ struct cryptonight_r_data {
 struct cryptonight_ctx {
     alignas(16) uint8_t state[224];
     alignas(16) uint8_t *memory;
+    const uint32_t* tweak1_table;
+    uint64_t tweak1_2;
 
-    uint8_t unused[40];
+    uint8_t unused[24];
     const uint32_t *saes_table;
 
     cn_mainloop_fun_ms_abi generated_code;

diff --git a/src/crypto/cn/CryptoNight_monero.h b/src/crypto/cn/CryptoNight_monero.h
@@ -204,4 +204,7 @@
     v4_random_math(code##part, r##part); \
   }
 
+extern bool cn_sse41_enabled;
+extern bool cn_vaes_enabled;
+
 #endif /* XMRIG_CRYPTONIGHT_MONERO_H */