v2.11.0

CHANGELOG.md

@@ -1,3 +1,12 @@
+# v2.11.0
+- [#928](https://github.com/xmrig/xmrig/issues/928) Added support for new algorithm `cryptonight/gpu`, short alias `cn/gpu` (original name `cryptonight-gpu`), for upcoming [Ryo currency](https://ryo-currency.com) fork on February 14.
+- [#749](https://github.com/xmrig/xmrig/issues/749) Added support for detect hardware AES in runtime on ARMv8 platforms.
+- [#292](https://github.com/xmrig/xmrig/issues/292) Fixed build on ARMv8 platforms if compiler not support hardware AES.
+
+# v2.10.0
+- [#904](https://github.com/xmrig/xmrig/issues/904) Added new algorithm `cn-pico/trtl` (aliases `cryptonight-turtle`, `cn-trtl`) for upcoming TurtleCoin (TRTL) fork.
+- Default value for option `max-cpu-usage` changed to `100` also this option now deprecated.
+
 # v2.9.4
 - [#913](https://github.com/xmrig/xmrig/issues/913) Fixed Masari (MSR) support (this update required for upcoming fork).
 - [#915](https://github.com/xmrig/xmrig/pull/915) Improved security, JIT memory now read-only after patching.

CMakeLists.txt

@@ -4,11 +4,14 @@ project(xmrig)
 option(WITH_LIBCPUID  "Use Libcpuid" ON)
 option(WITH_AEON      "CryptoNight-Lite support" ON)
 option(WITH_SUMO      "CryptoNight-Heavy support" ON)
+option(WITH_CN_PICO   "CryptoNight-Pico support" ON)
+option(WITH_CN_GPU    "CryptoNight-GPU support" ON)
 option(WITH_HTTPD     "HTTP REST API" ON)
 option(WITH_DEBUG_LOG "Enable debug log output" OFF)
 option(WITH_TLS       "Enable OpenSSL support" ON)
 option(WITH_ASM       "Enable ASM PoW implementations" ON)
 option(BUILD_STATIC   "Build static binary" OFF)
+option(ARM_TARGET     "Force use specific ARM target 8 or 7" 0)
 
 include (CheckIncludeFile)
 include (cmake/cpu.cmake)
@@ -199,6 +202,7 @@ endif()
 
 include(cmake/OpenSSL.cmake)
 include(cmake/asm.cmake)
+include(cmake/cn-gpu.cmake)
 
 CHECK_INCLUDE_FILE (syslog.h HAVE_SYSLOG_H)
 if (HAVE_SYSLOG_H)
@@ -218,6 +222,10 @@ if (NOT WITH_IPBC)
     add_definitions(/DXMRIG_NO_IPBC)
 endif()
 
+if (NOT WITH_CN_PICO)
+    add_definitions(/DXMRIG_NO_CN_PICO)
+endif()
+
 if (WITH_HTTPD)
     find_package(MHD)
 
@@ -257,5 +265,5 @@ if (WITH_DEBUG_LOG)
     add_definitions(/DAPP_DEBUG)
 endif()
 
-add_executable(${CMAKE_PROJECT_NAME} ${HEADERS} ${SOURCES} ${SOURCES_OS} ${SOURCES_CPUID} ${HEADERS_CRYPTO} ${SOURCES_CRYPTO} ${SOURCES_SYSLOG} ${HTTPD_SOURCES} ${TLS_SOURCES} ${XMRIG_ASM_SOURCES})
+add_executable(${CMAKE_PROJECT_NAME} ${HEADERS} ${SOURCES} ${SOURCES_OS} ${SOURCES_CPUID} ${HEADERS_CRYPTO} ${SOURCES_CRYPTO} ${SOURCES_SYSLOG} ${HTTPD_SOURCES} ${TLS_SOURCES} ${XMRIG_ASM_SOURCES} ${CN_GPU_SOURCES})
 target_link_libraries(${CMAKE_PROJECT_NAME} ${XMRIG_ASM_LIBRARY} ${OPENSSL_LIBRARIES} ${UV_LIBRARIES} ${MHD_LIBRARY} ${EXTRA_LIBS} ${CPUID_LIB})

cmake/cn-gpu.cmake

@@ -0,0 +1,23 @@
+if (WITH_CN_GPU AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+
+    if (XMRIG_ARM)
+        set(CN_GPU_SOURCES src/crypto/cn_gpu_arm.cpp)
+
+        if (CMAKE_CXX_COMPILER_ID MATCHES GNU OR CMAKE_CXX_COMPILER_ID MATCHES Clang)
+            set_source_files_properties(src/crypto/cn_gpu_arm.cpp PROPERTIES COMPILE_FLAGS "-O3")
+        endif()
+    else()
+        set(CN_GPU_SOURCES src/crypto/cn_gpu_avx.cpp src/crypto/cn_gpu_ssse3.cpp)
+
+        if (CMAKE_CXX_COMPILER_ID MATCHES GNU OR CMAKE_CXX_COMPILER_ID MATCHES Clang)
+            set_source_files_properties(src/crypto/cn_gpu_avx.cpp PROPERTIES COMPILE_FLAGS "-O3 -mavx2")
+            set_source_files_properties(src/crypto/cn_gpu_ssse3.cpp PROPERTIES COMPILE_FLAGS "-O3")
+        elseif (CMAKE_CXX_COMPILER_ID MATCHES MSVC)
+            set_source_files_properties(src/crypto/cn_gpu_avx.cpp PROPERTIES COMPILE_FLAGS "/arch:AVX")
+        endif()
+    endif()
+else()
+    set(CN_GPU_SOURCES "")
+
+    add_definitions(/DXMRIG_NO_CN_GPU)
+endif()

cmake/cpu.cmake

@@ -7,19 +7,37 @@ if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|AMD64)$")
     add_definitions(/DRAPIDJSON_SSE2)
 endif()
 
-
+if (NOT ARM_TARGET)
-if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64)$")
+    if (CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64|arm64|armv8-a)$")
-    set(XMRIG_ARM     ON)
+        set(ARM_TARGET 8)
-    set(XMRIG_ARMv8   ON)
+    elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "^(armv7|armv7f|armv7s|armv7k|armv7-a|armv7l)$")
-    set(WITH_LIBCPUID OFF)
+        set(ARM_TARGET 7)
-
+    endif()
-    add_definitions(/DXMRIG_ARM)
+endif()
-    add_definitions(/DXMRIG_ARMv8)
+
-elseif (CMAKE_SYSTEM_PROCESSOR MATCHES "^(armv7|armv7f|armv7s|armv7k|armv7-a|armv7l)$")
+if (ARM_TARGET AND ARM_TARGET GREATER 6)
-    set(XMRIG_ARM     ON)
+    set(XMRIG_ARM     ON)
-    set(XMRIG_ARMv7   ON)
+    set(WITH_LIBCPUID OFF)
-    set(WITH_LIBCPUID OFF)
+    add_definitions(/DXMRIG_ARM)
-
+
-    add_definitions(/DXMRIG_ARM)
+    message(STATUS "Use ARM_TARGET=${ARM_TARGET} (${CMAKE_SYSTEM_PROCESSOR})")
-    add_definitions(/DXMRIG_ARMv7)
+
+    include(CheckCXXCompilerFlag)
+
+    if (ARM_TARGET EQUAL 8)
+        set(XMRIG_ARMv8 ON)
+        add_definitions(/DXMRIG_ARMv8)
+
+        CHECK_CXX_COMPILER_FLAG(-march=armv8-a+crypto XMRIG_ARM_CRYPTO)
+
+        if (XMRIG_ARM_CRYPTO)
+            add_definitions(/DXMRIG_ARM_CRYPTO)
+            set(ARM8_CXX_FLAGS "-march=armv8-a+crypto")
+        else()
+            set(ARM8_CXX_FLAGS "-march=armv8-a")
+        endif()
+    elseif (ARM_TARGET EQUAL 7)
+        set(XMRIG_ARMv7 ON)
+        add_definitions(/DXMRIG_ARMv7)
+    endif()
 endif()

cmake/flags.cmake

@@ -19,8 +19,8 @@ if (CMAKE_CXX_COMPILER_ID MATCHES GNU)
     set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -s")
 
     if (XMRIG_ARMv8)
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crypto")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARM8_CXX_FLAGS}")
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crypto -flax-vector-conversions")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${ARM8_CXX_FLAGS} -flax-vector-conversions")
     elseif (XMRIG_ARMv7)
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfpu=neon")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mfpu=neon -flax-vector-conversions")
@@ -60,8 +60,8 @@ elseif (CMAKE_CXX_COMPILER_ID MATCHES Clang)
     set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast -funroll-loops -fmerge-all-constants")
 
     if (XMRIG_ARMv8)
-        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crypto")
+        set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARM8_CXX_FLAGS}")
-        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crypto")
+        set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${ARM8_CXX_FLAGS}")
     elseif (XMRIG_ARMv7)
         set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfpu=neon -march=${CMAKE_SYSTEM_PROCESSOR}")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mfpu=neon -march=${CMAKE_SYSTEM_PROCESSOR}")

src/Mem.cpp

@@ -34,7 +34,6 @@ bool Mem::m_enabled = true;
 int Mem::m_flags    = 0;
 
 
-
 MemInfo Mem::create(cryptonight_ctx **ctx, xmrig::Algo algorithm, size_t count)
 {
     using namespace xmrig;
@@ -42,11 +41,9 @@ MemInfo Mem::create(cryptonight_ctx **ctx, xmrig::Algo algorithm, size_t count)
     MemInfo info;
     info.size = cn_select_memory(algorithm) * count;
 
-#   ifndef XMRIG_NO_AEON
+    constexpr const size_t align_size = 2 * 1024 * 1024;
-    info.size += info.size % cn_select_memory<CRYPTONIGHT>();
+    info.size  = ((info.size + align_size - 1) / align_size) * align_size;
-#   endif
+    info.pages = info.size / align_size;
-
-    info.pages = info.size / cn_select_memory<CRYPTONIGHT>();
 
     allocate(info, m_enabled);
 

src/Summary.cpp

@@ -4,9 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2019 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
@@ -74,18 +74,25 @@ static void print_cpu(xmrig::Config *config)
     using namespace xmrig;
 
     if (config->isColors()) {
-        Log::i()->text(GREEN_BOLD(" * ") WHITE_BOLD("%-13s%s (%d)") " %sx64 %sAES",
+        Log::i()->text(GREEN_BOLD(" * ") WHITE_BOLD("%-13s%s (%d)") " %sx64 %sAES %sAVX2",
                        "CPU",
                        Cpu::info()->brand(),
                        Cpu::info()->sockets(),
-                       Cpu::info()->isX64() ? "\x1B[1;32m" : "\x1B[1;31m-",
+                       Cpu::info()->isX64()   ? "\x1B[1;32m" : "\x1B[1;31m-",
-                       Cpu::info()->hasAES() ? "\x1B[1;32m" : "\x1B[1;31m-");
+                       Cpu::info()->hasAES()  ? "\x1B[1;32m" : "\x1B[1;31m-",
+                       Cpu::info()->hasAVX2() ? "\x1B[1;32m" : "\x1B[1;31m-");
 #       ifndef XMRIG_NO_LIBCPUID
         Log::i()->text(GREEN_BOLD(" * ") WHITE_BOLD("%-13s%.1f MB/%.1f MB"), "CPU L2/L3", Cpu::info()->L2() / 1024.0, Cpu::info()->L3() / 1024.0);
 #       endif
     }
     else {
-        Log::i()->text(" * %-13s%s (%d) %sx64 %sAES", "CPU", Cpu::info()->brand(), Cpu::info()->sockets(), Cpu::info()->isX64() ? "" : "-", Cpu::info()->hasAES() ? "" : "-");
+        Log::i()->text(" * %-13s%s (%d) %sx64 %sAES %sAVX2",
+                       "CPU",
+                       Cpu::info()->brand(),
+                       Cpu::info()->sockets(),
+                       Cpu::info()->isX64()   ? "" : "-",
+                       Cpu::info()->hasAES()  ? "" : "-",
+                       Cpu::info()->hasAVX2() ? "" : "-");
 #       ifndef XMRIG_NO_LIBCPUID
         Log::i()->text(" * %-13s%.1f MB/%.1f MB", "CPU L2/L3", Cpu::info()->L2() / 1024.0, Cpu::info()->L3() / 1024.0);
 #       endif

src/common/cpu/BasicCpuInfo.cpp

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -35,6 +36,10 @@
 #   define bit_AES (1 << 25)
 #endif
 
+#ifndef bit_AVX2
+#   define bit_AVX2 (1 << 5)
+#endif
+
 
 #include "common/cpu/BasicCpuInfo.h"
 
@@ -93,9 +98,19 @@ static inline bool has_aes_ni()
 }
 
 
+static inline bool has_avx2()
+{
+    int32_t cpu_info[4] = { 0 };
+    cpuid(EXTENDED_FEATURES, cpu_info);
+
+    return (cpu_info[EBX_Reg] & bit_AVX2) != 0;
+}
+
+
 xmrig::BasicCpuInfo::BasicCpuInfo() :
     m_assembly(ASM_NONE),
     m_aes(has_aes_ni()),
+    m_avx2(has_avx2()),
     m_brand(),
     m_threads(std::thread::hardware_concurrency())
 {

src/common/cpu/BasicCpuInfo.h

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -41,6 +42,7 @@ protected:
 
     inline Assembly assembly() const override       { return m_assembly; }
     inline bool hasAES() const override             { return m_aes; }
+    inline bool hasAVX2() const override            { return m_avx2; }
     inline bool isSupported() const override        { return true; }
     inline const char *brand() const override       { return m_brand; }
     inline int32_t cores() const override           { return -1; }
@@ -59,6 +61,7 @@ protected:
 private:
     Assembly m_assembly;
     bool m_aes;
+    bool m_avx2;
     char m_brand[64];
     int32_t m_threads;
 };

src/common/cpu/BasicCpuInfo_arm.cpp

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -24,19 +25,29 @@
 #include <string.h>
 #include <thread>
 
+#if __ARM_FEATURE_CRYPTO
+#   include <sys/auxv.h>
+#   include <asm/hwcap.h>
+#endif
+
 
 #include "common/cpu/BasicCpuInfo.h"
 
 
 xmrig::BasicCpuInfo::BasicCpuInfo() :
     m_aes(false),
+    m_avx2(false),
     m_brand(),
     m_threads(std::thread::hardware_concurrency())
 {
-    memcpy(m_brand, "Unknown", 7);
+#   ifdef XMRIG_ARMv8
+    memcpy(m_brand, "ARMv8", 5);
+#   else
+    memcpy(m_brand, "ARMv7", 5);
+#   endif
 
 #   if __ARM_FEATURE_CRYPTO
-    m_aes = true;
+    m_aes = getauxval(AT_HWCAP) & HWCAP_AES;
 #   endif
 }
 

src/common/crypto/Algorithm.cpp

@@ -6,7 +6,7 @@
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -78,6 +78,18 @@ static AlgoData const algorithms[] = {
     { "cryptonight-heavy/xhv",  "cn-heavy/xhv",  xmrig::CRYPTONIGHT_HEAVY, xmrig::VARIANT_XHV  },
     { "cryptonight-heavy/tube", "cn-heavy/tube", xmrig::CRYPTONIGHT_HEAVY, xmrig::VARIANT_TUBE },
 #   endif
+
+#   ifndef XMRIG_NO_CN_PICO
+    { "cryptonight-pico/trtl",  "cn-pico/trtl",  xmrig::CRYPTONIGHT_PICO, xmrig::VARIANT_TRTL },
+    { "cryptonight-pico",       "cn-pico",       xmrig::CRYPTONIGHT_PICO, xmrig::VARIANT_TRTL },
+    { "cryptonight-turtle",     "cn-trtl",       xmrig::CRYPTONIGHT_PICO, xmrig::VARIANT_TRTL },
+    { "cryptonight-ultralite",  "cn-ultralite",  xmrig::CRYPTONIGHT_PICO, xmrig::VARIANT_TRTL },
+    { "cryptonight_turtle",     "cn_turtle",     xmrig::CRYPTONIGHT_PICO, xmrig::VARIANT_TRTL },
+#   endif
+
+#   ifndef XMRIG_NO_CN_GPU
+    { "cryptonight/gpu",        "cn/gpu",  xmrig::CRYPTONIGHT, xmrig::VARIANT_GPU },
+#   endif
 };
 
 
@@ -97,6 +109,8 @@ static AlgoData const xmrStakAlgorithms[] = {
     { "cryptonight_masari",      nullptr, xmrig::CRYPTONIGHT,       xmrig::VARIANT_MSR  },
     { "cryptonight-bittube2",    nullptr, xmrig::CRYPTONIGHT_HEAVY, xmrig::VARIANT_TUBE }, // bittube-miner
     { "cryptonight_alloy",       nullptr, xmrig::CRYPTONIGHT,       xmrig::VARIANT_XAO  }, // xmr-stak-alloy
+    { "cryptonight_turtle",      nullptr, xmrig::CRYPTONIGHT_PICO,  xmrig::VARIANT_TRTL },
+    { "cryptonight_gpu",         nullptr, xmrig::CRYPTONIGHT,       xmrig::VARIANT_GPU  },
 };
 #endif
 
@@ -111,7 +125,9 @@ static const char *variants[] = {
     "xao",
     "rto",
     "2",
-    "half"
+    "half",
+    "trtl",
+    "gpu"
 };
 
 
@@ -225,6 +241,10 @@ void xmrig::Algorithm::parseVariant(int variant)
 void xmrig::Algorithm::setAlgo(Algo algo)
 {
     m_algo = algo;
+
+    if (m_algo == CRYPTONIGHT_PICO && m_variant == VARIANT_AUTO) {
+        m_variant = xmrig::VARIANT_TRTL;
+    }
 }
 
 

src/common/interfaces/ICpuInfo.h

@@ -4,7 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2016-2018 XMRig       <support@xmrig.com>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -40,6 +42,7 @@ public:
     virtual ~ICpuInfo() {}
 
     virtual bool hasAES() const                                               = 0;
+    virtual bool hasAVX2() const                                              = 0;
     virtual bool isSupported() const                                          = 0;
     virtual bool isX64() const                                                = 0;
     virtual const char *brand() const                                         = 0;

src/common/interfaces/IStrategy.h

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -21,8 +22,8 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef __ISTRATEGY_H__
+#ifndef XMRIG_ISTRATEGY_H
-#define __ISTRATEGY_H__
+#define XMRIG_ISTRATEGY_H
 
 
 #include <stdint.h>
@@ -31,18 +32,24 @@
 class JobResult;
 
 
+namespace xmrig {
+    class Algorithm;
+}
+
+
 class IStrategy
 {
 public:
     virtual ~IStrategy() {}
 
-    virtual bool isActive() const                   = 0;
+    virtual bool isActive() const                      = 0;
-    virtual int64_t submit(const JobResult &result) = 0;
+    virtual int64_t submit(const JobResult &result)    = 0;
-    virtual void connect()                          = 0;
+    virtual void connect()                             = 0;
-    virtual void resume()                           = 0;
+    virtual void resume()                              = 0;
-    virtual void stop()                             = 0;
+    virtual void setAlgo(const xmrig::Algorithm &algo) = 0;
-    virtual void tick(uint64_t now)                 = 0;
+    virtual void stop()                                = 0;
+    virtual void tick(uint64_t now)                    = 0;
 };
 
 
-#endif // __ISTRATEGY_H__
+#endif // XMRIG_ISTRATEGY_H

src/common/log/Log.h

@@ -77,6 +77,7 @@ private:
 #define CYAN(x)         "\x1B[0;36m" x "\x1B[0m"
 #define WHITE_BOLD(x)   "\x1B[1;37m" x "\x1B[0m"
 #define WHITE(x)        "\x1B[0;37m" x "\x1B[0m"
+#define GRAY(x)         "\x1B[1;30m" x "\x1B[0m"
 
 
 #define LOG_ERR(x, ...)    Log::i()->message(ILogBackend::ERR,     x, ##__VA_ARGS__)

src/common/net/Job.h

@@ -105,7 +105,7 @@ private:
     xmrig::Id m_id;
 
 #   ifdef XMRIG_PROXY_PROJECT
-    char m_rawBlob[176];
+    char m_rawBlob[kMaxBlobSize * 2 + 8];
     char m_rawTarget[24];
 #   endif
 };

src/common/net/Pool.cpp

@@ -4,8 +4,8 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -106,8 +106,8 @@ bool Pool::isCompatible(const xmrig::Algorithm &algorithm) const
     }
 
 #   ifdef XMRIG_PROXY_PROJECT
-    if (m_algorithm.algo() == xmrig::CRYPTONIGHT && algorithm.algo() == xmrig::CRYPTONIGHT && m_algorithm.variant() == xmrig::VARIANT_XTL) {
+    if (m_algorithm.algo() == xmrig::CRYPTONIGHT && algorithm.algo() == xmrig::CRYPTONIGHT) {
-        return true;
+        return m_algorithm.variant() == xmrig::VARIANT_XTL || m_algorithm.variant() == xmrig::VARIANT_MSR;
     }
 #   endif
 
@@ -203,10 +203,10 @@ rapidjson::Value Pool::toJSON(rapidjson::Document &doc) const
 
     Value obj(kObjectType);
 
-    obj.AddMember("url",    StringRef(url()), allocator);
+    obj.AddMember("url",    m_url.toJSON(), allocator);
-    obj.AddMember("user",   StringRef(user()), allocator);
+    obj.AddMember("user",   m_user.toJSON(), allocator);
-    obj.AddMember("pass",   StringRef(password()), allocator);
+    obj.AddMember("pass",   m_password.toJSON(), allocator);
-    obj.AddMember("rig-id", rigId() ? Value(StringRef(rigId())).Move() : Value(kNullType).Move(), allocator);
+    obj.AddMember("rig-id", m_rigId.toJSON(), allocator);
 
 #   ifndef XMRIG_PROXY_PROJECT
     obj.AddMember("nicehash", isNicehash(), allocator);
@@ -223,17 +223,20 @@ rapidjson::Value Pool::toJSON(rapidjson::Document &doc) const
     case xmrig::VARIANT_AUTO:
     case xmrig::VARIANT_0:
     case xmrig::VARIANT_1:
-    case xmrig::VARIANT_2:
         obj.AddMember("variant", m_algorithm.variant(), allocator);
         break;
 
+    case xmrig::VARIANT_2:
+        obj.AddMember("variant", 2, allocator);
+        break;
+
     default:
         obj.AddMember("variant", StringRef(m_algorithm.variantName()), allocator);
         break;
     }
 
     obj.AddMember("tls",             isTLS(), allocator);
-    obj.AddMember("tls-fingerprint", fingerprint() ? Value(StringRef(fingerprint())).Move() : Value(kNullType).Move(), allocator);
+    obj.AddMember("tls-fingerprint", m_fingerprint.toJSON(), allocator);
 
     return obj;
 }
@@ -419,6 +422,7 @@ void Pool::rebuild()
     addVariant(xmrig::VARIANT_XHV);
     addVariant(xmrig::VARIANT_XAO);
     addVariant(xmrig::VARIANT_RTO);
+    addVariant(xmrig::VARIANT_GPU);
     addVariant(xmrig::VARIANT_AUTO);
 #   endif
 }

src/common/net/strategies/FailoverStrategy.cpp

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -76,6 +77,14 @@ void FailoverStrategy::resume()
 }
 
 
+void FailoverStrategy::setAlgo(const xmrig::Algorithm &algo)
+{
+    for (Client *client : m_pools) {
+        client->setAlgo(algo);
+    }
+}
+
+
 void FailoverStrategy::stop()
 {
     for (size_t i = 0; i < m_pools.size(); ++i) {

src/common/net/strategies/FailoverStrategy.h

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -21,8 +22,8 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef __FAILOVERSTRATEGY_H__
+#ifndef XMRIG_FAILOVERSTRATEGY_H
-#define __FAILOVERSTRATEGY_H__
+#define XMRIG_FAILOVERSTRATEGY_H
 
 
 #include <vector>
@@ -42,7 +43,7 @@ class FailoverStrategy : public IStrategy, public IClientListener
 {
 public:
     FailoverStrategy(const std::vector<Pool> &urls, int retryPause, int retries, IStrategyListener *listener, bool quiet = false);
-    ~FailoverStrategy();
+    ~FailoverStrategy() override;
 
 public:
     inline bool isActive() const override  { return m_active >= 0; }
@@ -50,6 +51,7 @@ public:
     int64_t submit(const JobResult &result) override;
     void connect() override;
     void resume() override;
+    void setAlgo(const xmrig::Algorithm &algo) override;
     void stop() override;
     void tick(uint64_t now) override;
 
@@ -71,4 +73,4 @@ private:
     std::vector<Client*> m_pools;
 };
 
-#endif /* __FAILOVERSTRATEGY_H__ */
+#endif /* XMRIG_FAILOVERSTRATEGY_H */

src/common/net/strategies/SinglePoolStrategy.cpp

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -68,6 +69,12 @@ void SinglePoolStrategy::resume()
 }
 
 
+void SinglePoolStrategy::setAlgo(const xmrig::Algorithm &algo)
+{
+    m_client->setAlgo(algo);
+}
+
+
 void SinglePoolStrategy::stop()
 {
     m_client->disconnect();

src/common/net/strategies/SinglePoolStrategy.h

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -21,8 +22,8 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef __SINGLEPOOLSTRATEGY_H__
+#ifndef XMRIG_SINGLEPOOLSTRATEGY_H
-#define __SINGLEPOOLSTRATEGY_H__
+#define XMRIG_SINGLEPOOLSTRATEGY_H
 
 
 #include "common/interfaces/IClientListener.h"
@@ -31,14 +32,14 @@
 
 class Client;
 class IStrategyListener;
-class Url;
+class Pool;
 
 
 class SinglePoolStrategy : public IStrategy, public IClientListener
 {
 public:
     SinglePoolStrategy(const Pool &pool, int retryPause, int retries, IStrategyListener *listener, bool quiet = false);
-    ~SinglePoolStrategy();
+    ~SinglePoolStrategy() override;
 
 public:
     inline bool isActive() const override  { return m_active; }
@@ -46,6 +47,7 @@ public:
     int64_t submit(const JobResult &result) override;
     void connect() override;
     void resume() override;
+    void setAlgo(const xmrig::Algorithm &algo) override;
     void stop() override;
     void tick(uint64_t now) override;
 
@@ -61,4 +63,4 @@ private:
     IStrategyListener *m_listener;
 };
 
-#endif /* __SINGLEPOOLSTRATEGY_H__ */
+#endif /* XMRIG_SINGLEPOOLSTRATEGY_H */

src/common/xmrig.h

@@ -5,7 +5,7 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -32,9 +32,11 @@ namespace xmrig
 
 enum Algo {
     INVALID_ALGO = -1,
-    CRYPTONIGHT,       /* CryptoNight (Monero) */
+    CRYPTONIGHT,        /* CryptoNight (2 MB) */
-    CRYPTONIGHT_LITE,  /* CryptoNight-Lite (AEON) */
+    CRYPTONIGHT_LITE,   /* CryptoNight (1 MB) */
-    CRYPTONIGHT_HEAVY  /* CryptoNight-Heavy (RYO) */
+    CRYPTONIGHT_HEAVY,  /* CryptoNight (4 MB) */
+    CRYPTONIGHT_PICO,   /* CryptoNight (256 KB) */
+    ALGO_MAX
 };
 
 
@@ -70,6 +72,8 @@ enum Variant {
     VARIANT_RTO  = 7,  // Modified CryptoNight variant 1 (Arto only)
     VARIANT_2    = 8,  // CryptoNight variant 2
     VARIANT_HALF = 9,  // CryptoNight variant 2 with half iterations (Masari/Stellite)
+    VARIANT_TRTL = 10, // CryptoNight Turtle (TRTL)
+    VARIANT_GPU  = 11, // CryptoNight-GPU (Ryo)
     VARIANT_MAX
 };
 

src/config.json

@@ -19,7 +19,7 @@
     "huge-pages": true,
     "hw-aes": null,
     "log-file": null,
-    "max-cpu-usage": 75,
+    "max-cpu-usage": 100,
     "pools": [
         {
             "url": "donate.v2.xmrig.com:3333",

src/core/Config.cpp

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -48,7 +49,7 @@ xmrig::Config::Config() : xmrig::CommonConfig(),
     m_hugePages(true),
     m_safe(false),
     m_shouldSave(false),
-    m_maxCpuUsage(75),
+    m_maxCpuUsage(100),
     m_priority(-1)
 {
 }

src/core/cpu/AdvancedCpuInfo.cpp

@@ -4,9 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2016-2019 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
@@ -33,6 +33,7 @@
 xmrig::AdvancedCpuInfo::AdvancedCpuInfo() :
     m_assembly(ASM_NONE),
     m_aes(false),
+    m_avx2(false),
     m_L2_exclusive(false),
     m_brand(),
     m_cores(0),
@@ -83,6 +84,8 @@ xmrig::AdvancedCpuInfo::AdvancedCpuInfo() :
             m_assembly = ASM_INTEL;
         }
     }
+
+    m_avx2 = data.flags[CPU_FEATURE_AVX2];
 }
 
 

src/core/cpu/AdvancedCpuInfo.h

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -41,6 +42,7 @@ protected:
 
     inline Assembly assembly() const override       { return m_assembly; }
     inline bool hasAES() const override             { return m_aes; }
+    inline bool hasAVX2() const override            { return m_avx2; }
     inline bool isSupported() const override        { return true; }
     inline const char *brand() const override       { return m_brand; }
     inline int32_t cores() const override           { return m_cores; }
@@ -59,6 +61,7 @@ protected:
 private:
     Assembly m_assembly;
     bool m_aes;
+    bool m_avx2;
     bool m_L2_exclusive;
     char m_brand[64];
     int32_t m_cores;

src/crypto/CryptoNight_arm.h

@@ -5,9 +5,9 @@
  * Copyright 2014-2016 Wolf9466     <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee    <jayddee246@gmail.com>
  * Copyright 2016      Imran Yusuff <https://github.com/imranyusuff>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
- * Copyright 2018      SChernykh   <https://github.com/SChernykh>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -284,6 +284,34 @@ static inline void cn_explode_scratchpad(const __m128i *input, __m128i *output)
 }
 
 
+#ifndef XMRIG_NO_CN_GPU
+template<xmrig::Algo ALGO, size_t MEM>
+void cn_explode_scratchpad_gpu(const uint8_t *input, uint8_t *output)
+{
+    constexpr size_t hash_size = 200; // 25x8 bytes
+    alignas(16) uint64_t hash[25];
+
+    for (uint64_t i = 0; i < MEM / 512; i++)
+    {
+        memcpy(hash, input, hash_size);
+        hash[0] ^= i;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 160);
+        output += 160;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 176);
+        output += 176;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 176);
+        output += 176;
+    }
+}
+#endif
+
+
 template<xmrig::Algo ALGO, size_t MEM, bool SOFT_AES>
 static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
 {
@@ -541,6 +569,35 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 }
 
 
+#ifndef XMRIG_NO_CN_GPU
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_arm(const uint8_t *spad, uint8_t *lpad);
+
+
+template<xmrig::Algo ALGO, bool SOFT_AES, xmrig::Variant VARIANT>
+inline void cryptonight_single_hash_gpu(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx)
+{
+    constexpr size_t MASK         = xmrig::CRYPTONIGHT_GPU_MASK;
+    constexpr size_t ITERATIONS   = xmrig::cn_select_iter<ALGO, VARIANT>();
+    constexpr size_t MEM          = xmrig::cn_select_memory<ALGO>();
+
+    static_assert(MASK > 0 && ITERATIONS > 0 && MEM > 0, "unsupported algorithm/variant");
+
+    xmrig::keccak(input, size, ctx[0]->state);
+    cn_explode_scratchpad_gpu<ALGO, MEM>(ctx[0]->state, ctx[0]->memory);
+
+    fesetround(FE_TONEAREST);
+
+    cn_gpu_inner_arm<ITERATIONS, MASK>(ctx[0]->state, ctx[0]->memory);
+
+    cn_implode_scratchpad<xmrig::CRYPTONIGHT_HEAVY, MEM, SOFT_AES>((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
+
+    xmrig::keccakf((uint64_t*) ctx[0]->state, 24);
+    memcpy(output, ctx[0]->state, 32);
+}
+#endif
+
+
 template<xmrig::Algo ALGO, bool SOFT_AES, xmrig::Variant VARIANT>
 inline void cryptonight_double_hash(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, struct cryptonight_ctx **__restrict__ ctx)
 {

src/crypto/CryptoNight_constants.h

@@ -4,8 +4,9 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -22,10 +23,11 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef __CRYPTONIGHT_CONSTANTS_H__
+#ifndef XMRIG_CRYPTONIGHT_CONSTANTS_H
-#define __CRYPTONIGHT_CONSTANTS_H__
+#define XMRIG_CRYPTONIGHT_CONSTANTS_H
 
 
+#include <stddef.h>
 #include <stdint.h>
 
 
@@ -41,6 +43,9 @@ constexpr const uint32_t CRYPTONIGHT_ITER         = 0x80000;
 constexpr const uint32_t CRYPTONIGHT_HALF_ITER    = 0x40000;
 constexpr const uint32_t CRYPTONIGHT_XAO_ITER     = 0x100000;
 
+constexpr const uint32_t CRYPTONIGHT_GPU_ITER     = 0xC000;
+constexpr const uint32_t CRYPTONIGHT_GPU_MASK     = 0x1FFFC0;
+
 constexpr const size_t   CRYPTONIGHT_LITE_MEMORY  = 1 * 1024 * 1024;
 constexpr const uint32_t CRYPTONIGHT_LITE_MASK    = 0xFFFF0;
 constexpr const uint32_t CRYPTONIGHT_LITE_ITER    = 0x40000;
@@ -49,11 +54,17 @@ constexpr const size_t   CRYPTONIGHT_HEAVY_MEMORY = 4 * 1024 * 1024;
 constexpr const uint32_t CRYPTONIGHT_HEAVY_MASK   = 0x3FFFF0;
 constexpr const uint32_t CRYPTONIGHT_HEAVY_ITER   = 0x40000;
 
+constexpr const size_t   CRYPTONIGHT_PICO_MEMORY = 256 * 1024;
+constexpr const uint32_t CRYPTONIGHT_PICO_MASK   = 0x1FFF0;
+constexpr const uint32_t CRYPTONIGHT_PICO_ITER   = 0x40000;
+constexpr const uint32_t CRYPTONIGHT_TRTL_ITER   = 0x10000;
+
 
 template<Algo ALGO> inline constexpr size_t cn_select_memory()           { return 0; }
 template<> inline constexpr size_t cn_select_memory<CRYPTONIGHT>()       { return CRYPTONIGHT_MEMORY; }
 template<> inline constexpr size_t cn_select_memory<CRYPTONIGHT_LITE>()  { return CRYPTONIGHT_LITE_MEMORY; }
 template<> inline constexpr size_t cn_select_memory<CRYPTONIGHT_HEAVY>() { return CRYPTONIGHT_HEAVY_MEMORY; }
+template<> inline constexpr size_t cn_select_memory<CRYPTONIGHT_PICO>()  { return CRYPTONIGHT_PICO_MEMORY; }
 
 
 inline size_t cn_select_memory(Algo algorithm)
@@ -69,6 +80,9 @@ inline size_t cn_select_memory(Algo algorithm)
     case CRYPTONIGHT_HEAVY:
         return CRYPTONIGHT_HEAVY_MEMORY;
 
+    case CRYPTONIGHT_PICO:
+        return CRYPTONIGHT_PICO_MEMORY;
+
     default:
         break;
     }
@@ -81,6 +95,7 @@ template<Algo ALGO> inline constexpr uint32_t cn_select_mask()           { retur
 template<> inline constexpr uint32_t cn_select_mask<CRYPTONIGHT>()       { return CRYPTONIGHT_MASK; }
 template<> inline constexpr uint32_t cn_select_mask<CRYPTONIGHT_LITE>()  { return CRYPTONIGHT_LITE_MASK; }
 template<> inline constexpr uint32_t cn_select_mask<CRYPTONIGHT_HEAVY>() { return CRYPTONIGHT_HEAVY_MASK; }
+template<> inline constexpr uint32_t cn_select_mask<CRYPTONIGHT_PICO>()  { return CRYPTONIGHT_PICO_MASK; }
 
 
 inline uint32_t cn_select_mask(Algo algorithm)
@@ -96,6 +111,9 @@ inline uint32_t cn_select_mask(Algo algorithm)
     case CRYPTONIGHT_HEAVY:
         return CRYPTONIGHT_HEAVY_MASK;
 
+    case CRYPTONIGHT_PICO:
+        return CRYPTONIGHT_PICO_MASK;
+
     default:
         break;
     }
@@ -113,11 +131,13 @@ template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT, VARIANT_HALF>()
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT, VARIANT_MSR>()        { return CRYPTONIGHT_HALF_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT, VARIANT_XAO>()        { return CRYPTONIGHT_XAO_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT, VARIANT_RTO>()        { return CRYPTONIGHT_ITER; }
+template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT, VARIANT_GPU>()        { return CRYPTONIGHT_GPU_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_LITE, VARIANT_0>()     { return CRYPTONIGHT_LITE_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_LITE, VARIANT_1>()     { return CRYPTONIGHT_LITE_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_HEAVY, VARIANT_0>()    { return CRYPTONIGHT_HEAVY_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_HEAVY, VARIANT_XHV>()  { return CRYPTONIGHT_HEAVY_ITER; }
 template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_HEAVY, VARIANT_TUBE>() { return CRYPTONIGHT_HEAVY_ITER; }
+template<> inline constexpr uint32_t cn_select_iter<CRYPTONIGHT_PICO, VARIANT_TRTL>()  { return CRYPTONIGHT_TRTL_ITER; }
 
 
 inline uint32_t cn_select_iter(Algo algorithm, Variant variant)
@@ -127,9 +147,15 @@ inline uint32_t cn_select_iter(Algo algorithm, Variant variant)
     case VARIANT_HALF:
         return CRYPTONIGHT_HALF_ITER;
 
+    case VARIANT_GPU:
+        return CRYPTONIGHT_GPU_ITER;
+
     case VARIANT_RTO:
         return CRYPTONIGHT_XAO_ITER;
 
+    case VARIANT_TRTL:
+        return CRYPTONIGHT_TRTL_ITER;
+
     default:
         break;
     }
@@ -164,9 +190,11 @@ template<> inline constexpr Variant cn_base_variant<VARIANT_XAO>()   { return VA
 template<> inline constexpr Variant cn_base_variant<VARIANT_RTO>()   { return VARIANT_1; }
 template<> inline constexpr Variant cn_base_variant<VARIANT_2>()     { return VARIANT_2; }
 template<> inline constexpr Variant cn_base_variant<VARIANT_HALF>()  { return VARIANT_2; }
+template<> inline constexpr Variant cn_base_variant<VARIANT_TRTL>()  { return VARIANT_2; }
+template<> inline constexpr Variant cn_base_variant<VARIANT_GPU>()   { return VARIANT_GPU; }
 
 
 } /* namespace xmrig */
 
 
-#endif /* __CRYPTONIGHT_CONSTANTS_H__ */
+#endif /* XMRIG_CRYPTONIGHT_CONSTANTS_H */

src/crypto/CryptoNight_test.h

@@ -6,6 +6,7 @@
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
  *
  *   This program is free software: you can redistribute it and/or modify
@@ -26,6 +27,9 @@
 #define XMRIG_CRYPTONIGHT_TEST_H
 
 
+#include <stdint.h>
+
+
 const static uint8_t test_input[380] = {
     0x03, 0x05, 0xA0, 0xDB, 0xD6, 0xBF, 0x05, 0xCF, 0x16, 0xE5, 0x03, 0xF3, 0xA6, 0x6F, 0x78, 0x00,
     0x7C, 0xBF, 0x34, 0x14, 0x43, 0x32, 0xEC, 0xBF, 0xC2, 0x2E, 0xD9, 0x5C, 0x87, 0x00, 0x38, 0x3B,
@@ -254,4 +258,38 @@ const static uint8_t test_output_tube_heavy[160] = {
 #endif
 
 
+#ifndef XMRIG_NO_CN_PICO
+// "cn-pico/trtl"
+const static uint8_t test_output_pico_trtl[160] = {
+    0x08, 0xF4, 0x21, 0xD7, 0x83, 0x31, 0x17, 0x30, 0x0E, 0xDA, 0x66, 0xE9, 0x8F, 0x4A, 0x25, 0x69,
+    0x09, 0x3D, 0xF3, 0x00, 0x50, 0x01, 0x73, 0x94, 0x4E, 0xFC, 0x40, 0x1E, 0x9A, 0x4A, 0x17, 0xAF,
+    0xB2, 0x17, 0x2E, 0xC9, 0x46, 0x6E, 0x1A, 0xEE, 0x70, 0xEC, 0x85, 0x72, 0xA1, 0x4C, 0x23, 0x3E,
+    0xE3, 0x54, 0x58, 0x2B, 0xCB, 0x93, 0xF8, 0x69, 0xD4, 0x29, 0x74, 0x4D, 0xE5, 0x72, 0x6A, 0x26,
+    0x4E, 0xFD, 0x28, 0xFC, 0xD3, 0x74, 0x8A, 0x83, 0xF3, 0xCA, 0x92, 0x84, 0xE7, 0x4E, 0x10, 0xC2,
+    0x05, 0x62, 0xC7, 0xBE, 0x99, 0x73, 0xED, 0x90, 0xB5, 0x6F, 0xDA, 0x64, 0x71, 0x2D, 0x99, 0x39,
+    0x29, 0xDB, 0x22, 0x2B, 0x97, 0xB6, 0x37, 0x0E, 0x9A, 0x03, 0x65, 0xCC, 0xF7, 0xD0, 0x9A, 0xB7,
+    0x68, 0xCE, 0x07, 0x3E, 0x15, 0x40, 0x3C, 0xCE, 0x8C, 0x63, 0x16, 0x72, 0xB5, 0x74, 0x84, 0xF4,
+    0xA1, 0xE7, 0x53, 0x85, 0xFB, 0x72, 0xDD, 0x75, 0x90, 0x39, 0xB2, 0x3D, 0xC3, 0x08, 0x2C, 0xD5,
+    0x01, 0x08, 0x27, 0x75, 0x86, 0xB9, 0xBB, 0x9B, 0xDF, 0xEA, 0x49, 0xDE, 0x46, 0xCB, 0x83, 0x45
+};
+#endif
+
+
+#ifndef XMRIG_NO_CN_GPU
+// "cn/gpu"
+const static uint8_t test_output_gpu[160] = {
+    0xE5, 0x5C, 0xB2, 0x3E, 0x51, 0x64, 0x9A, 0x59, 0xB1, 0x27, 0xB9, 0x6B, 0x51, 0x5F, 0x2B, 0xF7,
+    0xBF, 0xEA, 0x19, 0x97, 0x41, 0xA0, 0x21, 0x6C, 0xF8, 0x38, 0xDE, 0xD0, 0x6E, 0xFF, 0x82, 0xDF,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+};
+#endif
+
+
 #endif /* XMRIG_CRYPTONIGHT_TEST_H */

src/crypto/CryptoNight_x86.h

@@ -4,7 +4,7 @@
  * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
- * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
  * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
  * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
  * Copyright 2016-2019 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
@@ -35,6 +35,7 @@
 #endif
 
 
+#include "common/cpu/Cpu.h"
 #include "common/crypto/keccak.h"
 #include "crypto/CryptoNight.h"
 #include "crypto/CryptoNight_constants.h"
@@ -289,6 +290,34 @@ static inline void cn_explode_scratchpad(const __m128i *input, __m128i *output)
 }
 
 
+#ifndef XMRIG_NO_CN_GPU
+template<xmrig::Algo ALGO, size_t MEM>
+void cn_explode_scratchpad_gpu(const uint8_t *input, uint8_t *output)
+{
+    constexpr size_t hash_size = 200; // 25x8 bytes
+    alignas(16) uint64_t hash[25];
+
+    for (uint64_t i = 0; i < MEM / 512; i++)
+    {
+        memcpy(hash, input, hash_size);
+        hash[0] ^= i;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 160);
+        output += 160;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 176);
+        output += 176;
+
+        xmrig::keccakf(hash, 24);
+        memcpy(output, hash, 176);
+        output += 176;
+    }
+}
+#endif
+
+
 template<xmrig::Algo ALGO, size_t MEM, bool SOFT_AES>
 static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
 {
@@ -458,6 +487,8 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
     constexpr size_t MEM          = xmrig::cn_select_memory<ALGO>();
     constexpr xmrig::Variant BASE = xmrig::cn_base_variant<VARIANT>();
 
+    static_assert(MASK > 0 && ITERATIONS > 0 && MEM > 0, "unsupported algorithm/variant");
+
     if (BASE == xmrig::VARIANT_1 && size < 43) {
         memset(output, 0, 32);
         return;
@@ -564,17 +595,63 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 }
 
 
+#ifndef XMRIG_NO_CN_GPU
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_avx(const uint8_t *spad, uint8_t *lpad);
+
+
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_ssse3(const uint8_t *spad, uint8_t *lpad);
+
+
+template<xmrig::Algo ALGO, bool SOFT_AES, xmrig::Variant VARIANT>
+inline void cryptonight_single_hash_gpu(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx)
+{
+    constexpr size_t MASK         = xmrig::CRYPTONIGHT_GPU_MASK;
+    constexpr size_t ITERATIONS   = xmrig::cn_select_iter<ALGO, VARIANT>();
+    constexpr size_t MEM          = xmrig::cn_select_memory<ALGO>();
+
+    static_assert(MASK > 0 && ITERATIONS > 0 && MEM > 0, "unsupported algorithm/variant");
+
+    xmrig::keccak(input, size, ctx[0]->state);
+    cn_explode_scratchpad_gpu<ALGO, MEM>(ctx[0]->state, ctx[0]->memory);
+
+#   ifdef _MSC_VER
+    _control87(RC_NEAR, MCW_RC);
+#   else
+    fesetround(FE_TONEAREST);
+#   endif
+
+    if (xmrig::Cpu::info()->hasAVX2()) {
+        cn_gpu_inner_avx<ITERATIONS, MASK>(ctx[0]->state, ctx[0]->memory);
+    } else {
+        cn_gpu_inner_ssse3<ITERATIONS, MASK>(ctx[0]->state, ctx[0]->memory);
+    }
+
+    cn_implode_scratchpad<xmrig::CRYPTONIGHT_HEAVY, MEM, SOFT_AES>((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
+
+    xmrig::keccakf((uint64_t*) ctx[0]->state, 24);
+    memcpy(output, ctx[0]->state, 32);
+}
+#endif
+
+
 #ifndef XMRIG_NO_ASM
 extern "C" void cnv2_mainloop_ivybridge_asm(cryptonight_ctx *ctx);
 extern "C" void cnv2_mainloop_ryzen_asm(cryptonight_ctx *ctx);
 extern "C" void cnv2_mainloop_bulldozer_asm(cryptonight_ctx *ctx);
 extern "C" void cnv2_double_mainloop_sandybridge_asm(cryptonight_ctx *ctx0, cryptonight_ctx *ctx1);
 
-extern xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_ivybridge_asm;
+extern xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_ivybridge_asm;
-extern xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_ryzen_asm;
+extern xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_ryzen_asm;
-extern xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_bulldozer_asm;
+extern xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_bulldozer_asm;
 extern xmrig::CpuThread::cn_mainloop_double_fun cn_half_double_mainloop_sandybridge_asm;
 
+extern xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_ivybridge_asm;
+extern xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_ryzen_asm;
+extern xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_bulldozer_asm;
+extern xmrig::CpuThread::cn_mainloop_double_fun cn_trtl_double_mainloop_sandybridge_asm;
+
 
 template<xmrig::Algo ALGO, xmrig::Variant VARIANT, xmrig::Assembly ASM>
 inline void cryptonight_single_hash_asm(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx)
@@ -606,6 +683,17 @@ inline void cryptonight_single_hash_asm(const uint8_t *__restrict__ input, size_
             cn_half_mainloop_bulldozer_asm(ctx[0]);
         }
     }
+    else if (VARIANT == xmrig::VARIANT_TRTL) {
+        if (ASM == xmrig::ASM_INTEL) {
+            cn_trtl_mainloop_ivybridge_asm(ctx[0]);
+        }
+        else if (ASM == xmrig::ASM_RYZEN) {
+            cn_trtl_mainloop_ryzen_asm(ctx[0]);
+        }
+        else {
+            cn_trtl_mainloop_bulldozer_asm(ctx[0]);
+        }
+    }
 
     cn_implode_scratchpad<ALGO, MEM, false>(reinterpret_cast<__m128i*>(ctx[0]->memory), reinterpret_cast<__m128i*>(ctx[0]->state));
     xmrig::keccakf(reinterpret_cast<uint64_t*>(ctx[0]->state), 24);
@@ -630,6 +718,9 @@ inline void cryptonight_double_hash_asm(const uint8_t *__restrict__ input, size_
     else if (VARIANT == xmrig::VARIANT_HALF) {
         cn_half_double_mainloop_sandybridge_asm(ctx[0], ctx[1]);
     }
+    else if (VARIANT == xmrig::VARIANT_TRTL) {
+        cn_trtl_double_mainloop_sandybridge_asm(ctx[0], ctx[1]);
+    }
 
     cn_implode_scratchpad<ALGO, MEM, false>(reinterpret_cast<__m128i*>(ctx[0]->memory), reinterpret_cast<__m128i*>(ctx[0]->state));
     cn_implode_scratchpad<ALGO, MEM, false>(reinterpret_cast<__m128i*>(ctx[1]->memory), reinterpret_cast<__m128i*>(ctx[1]->state));

src/crypto/cn_gpu_arm.cpp

@@ -0,0 +1,240 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
+ * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
+ * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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 <arm_neon.h>
+
+
+#include "crypto/CryptoNight_constants.h"
+
+
+inline void vandq_f32(float32x4_t &v, uint32_t v2)
+{
+    uint32x4_t vc = vdupq_n_u32(v2);
+    v = (float32x4_t)vandq_u32((uint32x4_t)v, vc);
+}
+
+
+inline void vorq_f32(float32x4_t &v, uint32_t v2)
+{
+    uint32x4_t vc = vdupq_n_u32(v2);
+    v = (float32x4_t)vorrq_u32((uint32x4_t)v, vc);
+}
+
+
+template <size_t v>
+inline void vrot_si32(int32x4_t &r)
+{
+    r = (int32x4_t)vextq_s8((int8x16_t)r, (int8x16_t)r, v);
+}
+
+template <>
+inline void vrot_si32<0>(int32x4_t &r)
+{
+}
+
+
+inline uint32_t vheor_s32(const int32x4_t &v)
+{
+    int32x4_t v0 = veorq_s32(v, vrev64q_s32(v));
+    int32x2_t vf = veor_s32(vget_high_s32(v0), vget_low_s32(v0));
+    return (uint32_t)vget_lane_s32(vf, 0);
+}
+
+
+inline void prep_dv(int32_t *idx, int32x4_t &v, float32x4_t &n)
+{
+    v = vld1q_s32(idx);
+    n = vcvtq_f32_s32(v);
+}
+
+
+inline void sub_round(const float32x4_t &n0, const float32x4_t &n1, const float32x4_t &n2, const float32x4_t &n3, const float32x4_t &rnd_c, float32x4_t &n, float32x4_t &d, float32x4_t &c)
+{
+    float32x4_t ln1 = vaddq_f32(n1, c);
+    float32x4_t nn = vmulq_f32(n0, c);
+    nn = vmulq_f32(ln1, vmulq_f32(nn, nn));
+    vandq_f32(nn, 0xFEFFFFFF);
+    vorq_f32(nn, 0x00800000);
+    n = vaddq_f32(n, nn);
+
+    float32x4_t ln3 = vsubq_f32(n3, c);
+    float32x4_t dd = vmulq_f32(n2, c);
+    dd = vmulq_f32(ln3, vmulq_f32(dd, dd));
+    vandq_f32(dd, 0xFEFFFFFF);
+    vorq_f32(dd, 0x00800000);
+    d = vaddq_f32(d, dd);
+
+    //Constant feedback
+    c = vaddq_f32(c, rnd_c);
+    c = vaddq_f32(c, vdupq_n_f32(0.734375f));
+    float32x4_t r = vaddq_f32(nn, dd);
+    vandq_f32(r, 0x807FFFFF);
+    vorq_f32(r, 0x40000000);
+    c = vaddq_f32(c, r);
+}
+
+
+inline void round_compute(const float32x4_t &n0, const float32x4_t &n1, const float32x4_t &n2, const float32x4_t &n3, const float32x4_t &rnd_c, float32x4_t &c, float32x4_t &r)
+{
+    float32x4_t n = vdupq_n_f32(0.0f), d = vdupq_n_f32(0.0f);
+
+    sub_round(n0, n1, n2, n3, rnd_c, n, d, c);
+    sub_round(n1, n2, n3, n0, rnd_c, n, d, c);
+    sub_round(n2, n3, n0, n1, rnd_c, n, d, c);
+    sub_round(n3, n0, n1, n2, rnd_c, n, d, c);
+    sub_round(n3, n2, n1, n0, rnd_c, n, d, c);
+    sub_round(n2, n1, n0, n3, rnd_c, n, d, c);
+    sub_round(n1, n0, n3, n2, rnd_c, n, d, c);
+    sub_round(n0, n3, n2, n1, rnd_c, n, d, c);
+
+    // Make sure abs(d) > 2.0 - this prevents division by zero and accidental overflows by division by < 1.0
+    vandq_f32(d, 0xFF7FFFFF);
+    vorq_f32(d, 0x40000000);
+    r = vaddq_f32(r, vdivq_f32(n, d));
+}
+
+
+// 112×4 = 448
+template <bool add>
+inline int32x4_t single_compute(const float32x4_t &n0, const float32x4_t &n1, const float32x4_t &n2, const float32x4_t &n3, float cnt, const float32x4_t &rnd_c, float32x4_t &sum)
+{
+    float32x4_t c = vdupq_n_f32(cnt);
+    float32x4_t r = vdupq_n_f32(0.0f);
+
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+
+    // do a quick fmod by setting exp to 2
+    vandq_f32(r, 0x807FFFFF);
+    vorq_f32(r, 0x40000000);
+
+    if (add) {
+        sum = vaddq_f32(sum, r);
+    } else {
+        sum = r;
+    }
+
+    const float32x4_t cc2 = vdupq_n_f32(536870880.0f);
+    r = vmulq_f32(r, cc2); // 35
+    return vcvtq_s32_f32(r);
+}
+
+
+template<size_t rot>
+inline void single_compute_wrap(const float32x4_t &n0, const float32x4_t &n1, const float32x4_t &n2, const float32x4_t &n3, float cnt, const float32x4_t &rnd_c, float32x4_t &sum, int32x4_t &out)
+{
+    int32x4_t r = single_compute<rot % 2 != 0>(n0, n1, n2, n3, cnt, rnd_c, sum);
+    vrot_si32<rot>(r);
+    out = veorq_s32(out, r);
+}
+
+
+template<uint32_t MASK>
+inline int32_t *scratchpad_ptr(uint8_t* lpad, uint32_t idx, size_t n) { return reinterpret_cast<int32_t *>(lpad + (idx & MASK) + n * 16); }
+
+
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_arm(const uint8_t *spad, uint8_t *lpad)
+{
+    uint32_t s = reinterpret_cast<const uint32_t*>(spad)[0] >> 8;
+    int32_t *idx0 = scratchpad_ptr<MASK>(lpad, s, 0);
+    int32_t *idx1 = scratchpad_ptr<MASK>(lpad, s, 1);
+    int32_t *idx2 = scratchpad_ptr<MASK>(lpad, s, 2);
+    int32_t *idx3 = scratchpad_ptr<MASK>(lpad, s, 3);
+    float32x4_t sum0 = vdupq_n_f32(0.0f);
+
+    for (size_t i = 0; i < ITER; i++) {
+        float32x4_t n0, n1, n2, n3;
+        int32x4_t v0, v1, v2, v3;
+        float32x4_t suma, sumb, sum1, sum2, sum3;
+
+        prep_dv(idx0, v0, n0);
+        prep_dv(idx1, v1, n1);
+        prep_dv(idx2, v2, n2);
+        prep_dv(idx3, v3, n3);
+        float32x4_t rc = sum0;
+
+        int32x4_t out, out2;
+        out = vdupq_n_s32(0);
+        single_compute_wrap<0>(n0, n1, n2, n3, 1.3437500f, rc, suma, out);
+        single_compute_wrap<1>(n0, n2, n3, n1, 1.2812500f, rc, suma, out);
+        single_compute_wrap<2>(n0, n3, n1, n2, 1.3593750f, rc, sumb, out);
+        single_compute_wrap<3>(n0, n3, n2, n1, 1.3671875f, rc, sumb, out);
+        sum0 = vaddq_f32(suma, sumb);
+        vst1q_s32(idx0, veorq_s32(v0, out));
+        out2 = out;
+
+        out = vdupq_n_s32(0);
+        single_compute_wrap<0>(n1, n0, n2, n3, 1.4296875f, rc, suma, out);
+        single_compute_wrap<1>(n1, n2, n3, n0, 1.3984375f, rc, suma, out);
+        single_compute_wrap<2>(n1, n3, n0, n2, 1.3828125f, rc, sumb, out);
+        single_compute_wrap<3>(n1, n3, n2, n0, 1.3046875f, rc, sumb, out);
+        sum1 = vaddq_f32(suma, sumb);
+        vst1q_s32(idx1, veorq_s32(v1, out));
+        out2 = veorq_s32(out2, out);
+
+        out = vdupq_n_s32(0);
+        single_compute_wrap<0>(n2, n1, n0, n3, 1.4140625f, rc, suma, out);
+        single_compute_wrap<1>(n2, n0, n3, n1, 1.2734375f, rc, suma, out);
+        single_compute_wrap<2>(n2, n3, n1, n0, 1.2578125f, rc, sumb, out);
+        single_compute_wrap<3>(n2, n3, n0, n1, 1.2890625f, rc, sumb, out);
+        sum2 = vaddq_f32(suma, sumb);
+        vst1q_s32(idx2, veorq_s32(v2, out));
+        out2 = veorq_s32(out2, out);
+
+        out = vdupq_n_s32(0);
+        single_compute_wrap<0>(n3, n1, n2, n0, 1.3203125f, rc, suma, out);
+        single_compute_wrap<1>(n3, n2, n0, n1, 1.3515625f, rc, suma, out);
+        single_compute_wrap<2>(n3, n0, n1, n2, 1.3359375f, rc, sumb, out);
+        single_compute_wrap<3>(n3, n0, n2, n1, 1.4609375f, rc, sumb, out);
+        sum3 = vaddq_f32(suma, sumb);
+        vst1q_s32(idx3, veorq_s32(v3, out));
+        out2 = veorq_s32(out2, out);
+
+        sum0 = vaddq_f32(sum0, sum1);
+        sum2 = vaddq_f32(sum2, sum3);
+        sum0 = vaddq_f32(sum0, sum2);
+
+        const float32x4_t cc1 = vdupq_n_f32(16777216.0f);
+        const float32x4_t cc2 = vdupq_n_f32(64.0f);
+        vandq_f32(sum0, 0x7fffffff); // take abs(va) by masking the float sign bit
+        // vs range 0 - 64
+        n0 = vmulq_f32(sum0, cc1);
+        v0 = vcvtq_s32_f32(n0);
+        v0 = veorq_s32(v0, out2);
+        uint32_t n = vheor_s32(v0);
+
+        // vs is now between 0 and 1
+        sum0 = vdivq_f32(sum0, cc2);
+        idx0 = scratchpad_ptr<MASK>(lpad, n, 0);
+        idx1 = scratchpad_ptr<MASK>(lpad, n, 1);
+        idx2 = scratchpad_ptr<MASK>(lpad, n, 2);
+        idx3 = scratchpad_ptr<MASK>(lpad, n, 3);
+    }
+}
+
+template void cn_gpu_inner_arm<xmrig::CRYPTONIGHT_GPU_ITER, xmrig::CRYPTONIGHT_GPU_MASK>(const uint8_t* spad, uint8_t* lpad);

src/crypto/cn_gpu_avx.cpp

@@ -0,0 +1,203 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
+ * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
+ * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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 "crypto/CryptoNight_constants.h"
+
+#ifdef __GNUC__
+#   include <x86intrin.h>
+#else
+#   include <intrin.h>
+#   define __restrict__ __restrict
+#endif
+
+inline void prep_dv_avx(__m256i* idx, __m256i& v, __m256& n01)
+{
+    v = _mm256_load_si256(idx);
+    n01 = _mm256_cvtepi32_ps(v);
+}
+
+inline __m256 fma_break(const __m256& x) 
+{ 
+    // Break the dependency chain by setitng the exp to ?????01 
+    __m256 xx = _mm256_and_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0xFEFFFFFF)), x); 
+    return _mm256_or_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x00800000)), xx); 
+}
+
+// 14
+inline void sub_round(const __m256& n0, const __m256& n1, const __m256& n2, const __m256& n3, const __m256& rnd_c, __m256& n, __m256& d, __m256& c)
+{
+    __m256 nn = _mm256_mul_ps(n0, c);
+    nn = _mm256_mul_ps(_mm256_add_ps(n1, c), _mm256_mul_ps(nn, nn));
+    nn = fma_break(nn);
+    n = _mm256_add_ps(n, nn);
+
+    __m256 dd = _mm256_mul_ps(n2, c);
+    dd = _mm256_mul_ps(_mm256_sub_ps(n3, c), _mm256_mul_ps(dd, dd));
+    dd = fma_break(dd);
+    d = _mm256_add_ps(d, dd);
+
+    //Constant feedback
+    c = _mm256_add_ps(c, rnd_c);
+    c = _mm256_add_ps(c, _mm256_set1_ps(0.734375f));
+    __m256 r = _mm256_add_ps(nn, dd);
+    r = _mm256_and_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x807FFFFF)), r);
+    r = _mm256_or_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x40000000)), r);
+    c = _mm256_add_ps(c, r);
+}
+
+// 14*8 + 2 = 112
+inline void round_compute(const __m256& n0, const __m256& n1, const __m256& n2, const __m256& n3, const __m256& rnd_c, __m256& c, __m256& r)
+{
+    __m256 n = _mm256_setzero_ps(), d = _mm256_setzero_ps();
+
+    sub_round(n0, n1, n2, n3, rnd_c, n, d, c);
+    sub_round(n1, n2, n3, n0, rnd_c, n, d, c);
+    sub_round(n2, n3, n0, n1, rnd_c, n, d, c);
+    sub_round(n3, n0, n1, n2, rnd_c, n, d, c);
+    sub_round(n3, n2, n1, n0, rnd_c, n, d, c);
+    sub_round(n2, n1, n0, n3, rnd_c, n, d, c);
+    sub_round(n1, n0, n3, n2, rnd_c, n, d, c);
+    sub_round(n0, n3, n2, n1, rnd_c, n, d, c);
+
+    // Make sure abs(d) > 2.0 - this prevents division by zero and accidental overflows by division by < 1.0
+    d = _mm256_and_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0xFF7FFFFF)), d);
+    d = _mm256_or_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x40000000)), d);
+    r = _mm256_add_ps(r, _mm256_div_ps(n, d));
+}
+
+// 112×4 = 448
+template <bool add>
+inline __m256i double_compute(const __m256& n0, const __m256& n1, const __m256& n2, const __m256& n3,
+                              float lcnt, float hcnt, const __m256& rnd_c, __m256& sum)
+{
+    __m256 c = _mm256_insertf128_ps(_mm256_castps128_ps256(_mm_set1_ps(lcnt)), _mm_set1_ps(hcnt), 1);
+    __m256 r = _mm256_setzero_ps();
+
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+
+    // do a quick fmod by setting exp to 2
+    r = _mm256_and_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x807FFFFF)), r);
+    r = _mm256_or_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x40000000)), r);
+
+    if(add)
+        sum = _mm256_add_ps(sum, r);
+    else
+        sum = r;
+
+    r = _mm256_mul_ps(r, _mm256_set1_ps(536870880.0f)); // 35
+    return _mm256_cvttps_epi32(r);
+}
+
+template <size_t rot>
+inline void double_compute_wrap(const __m256& n0, const __m256& n1, const __m256& n2, const __m256& n3,
+                                float lcnt, float hcnt, const __m256& rnd_c, __m256& sum, __m256i& out)
+{
+    __m256i r = double_compute<rot % 2 != 0>(n0, n1, n2, n3, lcnt, hcnt, rnd_c, sum);
+    if(rot != 0)
+        r = _mm256_or_si256(_mm256_bslli_epi128(r, 16 - rot), _mm256_bsrli_epi128(r, rot));
+
+    out = _mm256_xor_si256(out, r);
+}
+
+template<uint32_t MASK>
+inline __m256i* scratchpad_ptr(uint8_t* lpad, uint32_t idx, size_t n) { return reinterpret_cast<__m256i*>(lpad + (idx & MASK) + n*16); }
+
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_avx(const uint8_t* spad, uint8_t* lpad)
+{
+    uint32_t s = reinterpret_cast<const uint32_t*>(spad)[0] >> 8;
+    __m256i* idx0 = scratchpad_ptr<MASK>(lpad, s, 0);
+    __m256i* idx2 = scratchpad_ptr<MASK>(lpad, s, 2);
+    __m256 sum0 = _mm256_setzero_ps();
+
+    for(size_t i = 0; i < ITER; i++)
+    {
+        __m256i v01, v23;
+        __m256 suma, sumb, sum1;
+        __m256 rc = sum0;
+
+        __m256 n01, n23;
+        prep_dv_avx(idx0, v01, n01);
+        prep_dv_avx(idx2, v23, n23);
+        
+        __m256i out, out2;
+        __m256 n10, n22, n33;
+        n10 = _mm256_permute2f128_ps(n01, n01, 0x01);
+        n22 = _mm256_permute2f128_ps(n23, n23, 0x00);
+        n33 = _mm256_permute2f128_ps(n23, n23, 0x11);
+        
+        out = _mm256_setzero_si256();
+        double_compute_wrap<0>(n01, n10, n22, n33, 1.3437500f, 1.4296875f, rc, suma, out);
+        double_compute_wrap<1>(n01, n22, n33, n10, 1.2812500f, 1.3984375f, rc, suma, out);
+        double_compute_wrap<2>(n01, n33, n10, n22, 1.3593750f, 1.3828125f, rc, sumb, out);
+        double_compute_wrap<3>(n01, n33, n22, n10, 1.3671875f, 1.3046875f, rc, sumb, out);
+        _mm256_store_si256(idx0, _mm256_xor_si256(v01, out));
+        sum0 = _mm256_add_ps(suma, sumb);
+        out2 = out;
+        
+        __m256 n11, n02, n30;
+        n11 = _mm256_permute2f128_ps(n01, n01, 0x11);
+        n02 = _mm256_permute2f128_ps(n01, n23, 0x20);
+        n30 = _mm256_permute2f128_ps(n01, n23, 0x03);
+
+        out = _mm256_setzero_si256();
+        double_compute_wrap<0>(n23, n11, n02, n30, 1.4140625f, 1.3203125f, rc, suma, out);
+        double_compute_wrap<1>(n23, n02, n30, n11, 1.2734375f, 1.3515625f, rc, suma, out);
+        double_compute_wrap<2>(n23, n30, n11, n02, 1.2578125f, 1.3359375f, rc, sumb, out);
+        double_compute_wrap<3>(n23, n30, n02, n11, 1.2890625f, 1.4609375f, rc, sumb, out);
+        _mm256_store_si256(idx2, _mm256_xor_si256(v23, out));
+        sum1 = _mm256_add_ps(suma, sumb);
+
+        out2 = _mm256_xor_si256(out2, out);
+        out2 = _mm256_xor_si256(_mm256_permute2x128_si256(out2,out2,0x41), out2);
+        suma = _mm256_permute2f128_ps(sum0, sum1, 0x30);
+        sumb = _mm256_permute2f128_ps(sum0, sum1, 0x21);
+        sum0 = _mm256_add_ps(suma, sumb);
+        sum0 = _mm256_add_ps(sum0, _mm256_permute2f128_ps(sum0, sum0, 0x41));
+
+        // Clear the high 128 bits
+        __m128 sum = _mm256_castps256_ps128(sum0);
+
+        sum = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0x7fffffff)), sum); // take abs(va) by masking the float sign bit
+        // vs range 0 - 64 
+        __m128i v0 = _mm_cvttps_epi32(_mm_mul_ps(sum, _mm_set1_ps(16777216.0f)));
+        v0 = _mm_xor_si128(v0, _mm256_castsi256_si128(out2));
+        __m128i v1 = _mm_shuffle_epi32(v0, _MM_SHUFFLE(0, 1, 2, 3));
+        v0 = _mm_xor_si128(v0, v1);
+        v1 = _mm_shuffle_epi32(v0, _MM_SHUFFLE(0, 1, 0, 1));
+        v0 = _mm_xor_si128(v0, v1);
+
+        // vs is now between 0 and 1
+        sum = _mm_div_ps(sum, _mm_set1_ps(64.0f));
+        sum0 = _mm256_insertf128_ps(_mm256_castps128_ps256(sum), sum, 1);
+        uint32_t n = _mm_cvtsi128_si32(v0);
+        idx0 = scratchpad_ptr<MASK>(lpad, n, 0);
+        idx2 = scratchpad_ptr<MASK>(lpad, n, 2);
+    }
+}
+
+template void cn_gpu_inner_avx<xmrig::CRYPTONIGHT_GPU_ITER, xmrig::CRYPTONIGHT_GPU_MASK>(const uint8_t* spad, uint8_t* lpad);

src/crypto/cn_gpu_ssse3.cpp

@@ -0,0 +1,210 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * Copyright 2014      Lucas Jones <https://github.com/lucasjones>
+ * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
+ * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
+ * Copyright 2017-2019 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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 "crypto/CryptoNight_constants.h"
+
+#ifdef __GNUC__
+#   include <x86intrin.h>
+#else
+#   include <intrin.h>
+#   define __restrict__ __restrict
+#endif
+
+inline void prep_dv(__m128i* idx, __m128i& v, __m128& n)
+{
+    v = _mm_load_si128(idx);
+    n = _mm_cvtepi32_ps(v);
+}
+
+inline __m128 fma_break(__m128 x) 
+{ 
+    // Break the dependency chain by setitng the exp to ?????01 
+    x = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0xFEFFFFFF)), x); 
+    return _mm_or_ps(_mm_castsi128_ps(_mm_set1_epi32(0x00800000)), x); 
+}
+
+// 14
+inline void sub_round(__m128 n0, __m128 n1, __m128 n2, __m128 n3, __m128 rnd_c, __m128& n, __m128& d, __m128& c)
+{
+    n1 = _mm_add_ps(n1, c);
+    __m128 nn = _mm_mul_ps(n0, c);
+    nn = _mm_mul_ps(n1, _mm_mul_ps(nn,nn));
+    nn = fma_break(nn);
+    n = _mm_add_ps(n, nn);
+
+    n3 = _mm_sub_ps(n3, c);
+    __m128 dd = _mm_mul_ps(n2, c);
+    dd = _mm_mul_ps(n3, _mm_mul_ps(dd,dd));
+    dd = fma_break(dd);
+    d = _mm_add_ps(d, dd);
+
+    //Constant feedback
+    c = _mm_add_ps(c, rnd_c);
+    c = _mm_add_ps(c, _mm_set1_ps(0.734375f));
+    __m128 r = _mm_add_ps(nn, dd);
+    r = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0x807FFFFF)), r);
+    r = _mm_or_ps(_mm_castsi128_ps(_mm_set1_epi32(0x40000000)), r);
+    c = _mm_add_ps(c, r);
+}
+
+// 14*8 + 2 = 112
+inline void round_compute(__m128 n0, __m128 n1, __m128 n2, __m128 n3, __m128 rnd_c, __m128& c, __m128& r)
+{
+    __m128 n = _mm_setzero_ps(), d = _mm_setzero_ps();
+
+    sub_round(n0, n1, n2, n3, rnd_c, n, d, c);
+    sub_round(n1, n2, n3, n0, rnd_c, n, d, c);
+    sub_round(n2, n3, n0, n1, rnd_c, n, d, c);
+    sub_round(n3, n0, n1, n2, rnd_c, n, d, c);
+    sub_round(n3, n2, n1, n0, rnd_c, n, d, c);
+    sub_round(n2, n1, n0, n3, rnd_c, n, d, c);
+    sub_round(n1, n0, n3, n2, rnd_c, n, d, c);
+    sub_round(n0, n3, n2, n1, rnd_c, n, d, c);
+
+    // Make sure abs(d) > 2.0 - this prevents division by zero and accidental overflows by division by < 1.0
+    d = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0xFF7FFFFF)), d);
+    d = _mm_or_ps(_mm_castsi128_ps(_mm_set1_epi32(0x40000000)), d);
+    r =_mm_add_ps(r, _mm_div_ps(n,d));
+}
+
+// 112×4 = 448
+template<bool add>
+inline __m128i single_compute(__m128 n0, __m128 n1,  __m128 n2,  __m128 n3, float cnt, __m128 rnd_c, __m128& sum)
+{
+    __m128 c = _mm_set1_ps(cnt);
+    __m128 r = _mm_setzero_ps();
+
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+    round_compute(n0, n1, n2, n3, rnd_c, c, r);
+
+    // do a quick fmod by setting exp to 2
+    r = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0x807FFFFF)), r);
+    r = _mm_or_ps(_mm_castsi128_ps(_mm_set1_epi32(0x40000000)), r);
+
+    if(add)
+        sum = _mm_add_ps(sum, r);
+    else
+        sum = r;
+
+    r = _mm_mul_ps(r, _mm_set1_ps(536870880.0f)); // 35
+    return _mm_cvttps_epi32(r);
+}
+
+template<size_t rot>
+inline void single_compute_wrap(__m128 n0, __m128 n1, __m128 n2,  __m128 n3, float cnt, __m128 rnd_c, __m128& sum, __m128i& out)
+{
+    __m128i r = single_compute<rot % 2 != 0>(n0, n1, n2, n3, cnt, rnd_c, sum);
+    if(rot != 0)
+        r = _mm_or_si128(_mm_slli_si128(r, 16 - rot), _mm_srli_si128(r, rot));
+    out = _mm_xor_si128(out, r);
+}
+
+template<uint32_t MASK>
+inline __m128i* scratchpad_ptr(uint8_t* lpad, uint32_t idx, size_t n) { return reinterpret_cast<__m128i*>(lpad + (idx & MASK) + n*16); }
+
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_ssse3(const uint8_t* spad, uint8_t* lpad)
+{
+    uint32_t s = reinterpret_cast<const uint32_t*>(spad)[0] >> 8;
+    __m128i* idx0 = scratchpad_ptr<MASK>(lpad, s, 0);
+    __m128i* idx1 = scratchpad_ptr<MASK>(lpad, s, 1);
+    __m128i* idx2 = scratchpad_ptr<MASK>(lpad, s, 2);
+    __m128i* idx3 = scratchpad_ptr<MASK>(lpad, s, 3);
+    __m128 sum0 = _mm_setzero_ps();
+    
+    for(size_t i = 0; i < ITER; i++)
+    {
+        __m128 n0, n1, n2, n3;
+        __m128i v0, v1, v2, v3;
+        __m128 suma, sumb, sum1, sum2, sum3;
+        
+        prep_dv(idx0, v0, n0);
+        prep_dv(idx1, v1, n1);
+        prep_dv(idx2, v2, n2);
+        prep_dv(idx3, v3, n3);
+        __m128 rc = sum0;
+
+        __m128i out, out2;
+        out = _mm_setzero_si128();
+        single_compute_wrap<0>(n0, n1, n2, n3, 1.3437500f, rc, suma, out);
+        single_compute_wrap<1>(n0, n2, n3, n1, 1.2812500f, rc, suma, out);
+        single_compute_wrap<2>(n0, n3, n1, n2, 1.3593750f, rc, sumb, out);
+        single_compute_wrap<3>(n0, n3, n2, n1, 1.3671875f, rc, sumb, out);
+        sum0 = _mm_add_ps(suma, sumb);
+        _mm_store_si128(idx0, _mm_xor_si128(v0, out));
+        out2 = out;
+    
+        out = _mm_setzero_si128();
+        single_compute_wrap<0>(n1, n0, n2, n3, 1.4296875f, rc, suma, out);
+        single_compute_wrap<1>(n1, n2, n3, n0, 1.3984375f, rc, suma, out);
+        single_compute_wrap<2>(n1, n3, n0, n2, 1.3828125f, rc, sumb, out);
+        single_compute_wrap<3>(n1, n3, n2, n0, 1.3046875f, rc, sumb, out);
+        sum1 = _mm_add_ps(suma, sumb);
+        _mm_store_si128(idx1, _mm_xor_si128(v1, out));
+        out2 = _mm_xor_si128(out2, out);
+
+        out = _mm_setzero_si128();
+        single_compute_wrap<0>(n2, n1, n0, n3, 1.4140625f, rc, suma, out);
+        single_compute_wrap<1>(n2, n0, n3, n1, 1.2734375f, rc, suma, out);
+        single_compute_wrap<2>(n2, n3, n1, n0, 1.2578125f, rc, sumb, out);
+        single_compute_wrap<3>(n2, n3, n0, n1, 1.2890625f, rc, sumb, out);
+        sum2 = _mm_add_ps(suma, sumb);
+        _mm_store_si128(idx2, _mm_xor_si128(v2, out));
+        out2 = _mm_xor_si128(out2, out);
+
+        out = _mm_setzero_si128();
+        single_compute_wrap<0>(n3, n1, n2, n0, 1.3203125f, rc, suma, out);
+        single_compute_wrap<1>(n3, n2, n0, n1, 1.3515625f, rc, suma, out);
+        single_compute_wrap<2>(n3, n0, n1, n2, 1.3359375f, rc, sumb, out);
+        single_compute_wrap<3>(n3, n0, n2, n1, 1.4609375f, rc, sumb, out);
+        sum3 = _mm_add_ps(suma, sumb);
+        _mm_store_si128(idx3, _mm_xor_si128(v3, out));
+        out2 = _mm_xor_si128(out2, out);
+        sum0 = _mm_add_ps(sum0, sum1);
+        sum2 = _mm_add_ps(sum2, sum3);
+        sum0 = _mm_add_ps(sum0, sum2);
+
+        sum0 = _mm_and_ps(_mm_castsi128_ps(_mm_set1_epi32(0x7fffffff)), sum0); // take abs(va) by masking the float sign bit
+        // vs range 0 - 64 
+        n0 = _mm_mul_ps(sum0, _mm_set1_ps(16777216.0f));
+        v0 = _mm_cvttps_epi32(n0);
+        v0 = _mm_xor_si128(v0, out2);
+        v1 = _mm_shuffle_epi32(v0, _MM_SHUFFLE(0, 1, 2, 3));
+        v0 = _mm_xor_si128(v0, v1);
+        v1 = _mm_shuffle_epi32(v0, _MM_SHUFFLE(0, 1, 0, 1));
+        v0 = _mm_xor_si128(v0, v1);
+
+        // vs is now between 0 and 1
+        sum0 = _mm_div_ps(sum0, _mm_set1_ps(64.0f));
+        uint32_t n = _mm_cvtsi128_si32(v0);
+        idx0 = scratchpad_ptr<MASK>(lpad, n, 0);
+        idx1 = scratchpad_ptr<MASK>(lpad, n, 1);
+        idx2 = scratchpad_ptr<MASK>(lpad, n, 2);
+        idx3 = scratchpad_ptr<MASK>(lpad, n, 3);
+    }
+}
+
+template void cn_gpu_inner_ssse3<xmrig::CRYPTONIGHT_GPU_ITER, xmrig::CRYPTONIGHT_GPU_MASK>(const uint8_t* spad, uint8_t* lpad);

src/net/Network.cpp

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -154,12 +155,12 @@ void Network::onResultAccepted(IStrategy *strategy, Client *client, const Submit
     m_state.add(result, error);
 
     if (error) {
-        LOG_INFO(isColors() ? "\x1B[01;31mrejected\x1B[0m (%" PRId64 "/%" PRId64 ") diff \x1B[01;37m%u\x1B[0m \x1B[31m\"%s\"\x1B[0m \x1B[01;30m(%" PRIu64 " ms)"
+        LOG_INFO(isColors() ? "\x1B[1;31mrejected\x1B[0m (%" PRId64 "/%" PRId64 ") diff \x1B[1;37m%u\x1B[0m \x1B[31m\"%s\"\x1B[0m \x1B[1;30m(%" PRIu64 " ms)"
                             : "rejected (%" PRId64 "/%" PRId64 ") diff %u \"%s\" (%" PRIu64 " ms)",
                  m_state.accepted, m_state.rejected, result.diff, error, result.elapsed);
     }
     else {
-        LOG_INFO(isColors() ? "\x1B[01;32maccepted\x1B[0m (%" PRId64 "/%" PRId64 ") diff \x1B[01;37m%u\x1B[0m \x1B[01;30m(%" PRIu64 " ms)"
+        LOG_INFO(isColors() ? "\x1B[1;32maccepted\x1B[0m (%" PRId64 "/%" PRId64 ") diff \x1B[1;37m%u\x1B[0m \x1B[1;30m(%" PRIu64 " ms)"
                             : "accepted (%" PRId64 "/%" PRId64 ") diff %u (%" PRIu64 " ms)",
                  m_state.accepted, m_state.rejected, result.diff, result.elapsed);
     }
@@ -178,6 +179,10 @@ void Network::setJob(Client *client, const Job &job, bool donate)
                         : "new job from %s:%d diff %d algo %s",
              client->host(), client->port(), job.diff(), job.algorithm().shortName());
 
+    if (!donate && m_donate) {
+        m_donate->setAlgo(job.algorithm());
+    }
+
     m_state.diff = job.diff();
     Workers::setJob(job, donate);
 }

src/net/Network.h

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -21,8 +22,8 @@
  *   along with this program. If not, see <http://www.gnu.org/licenses/>.
  */
 
-#ifndef __NETWORK_H__
+#ifndef XMRIG_NETWORK_H
-#define __NETWORK_H__
+#define XMRIG_NETWORK_H
 
 
 #include <vector>
@@ -47,7 +48,7 @@ class Network : public IJobResultListener, public IStrategyListener
 {
 public:
   Network(xmrig::Controller *controller);
-  ~Network();
+  ~Network() override;
 
   void connect();
   void stop();
@@ -76,4 +77,4 @@ private:
 };
 
 
-#endif /* __NETWORK_H__ */
+#endif /* XMRIG_NETWORK_H */

src/net/strategies/DonateStrategy.cpp

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -43,7 +44,9 @@ DonateStrategy::DonateStrategy(int level, const char *user, xmrig::Algo algo, IS
     m_donateTime(level * 60 * 1000),
     m_idleTime((100 - level) * 60 * 1000),
     m_strategy(nullptr),
-    m_listener(listener)
+    m_listener(listener),
+    m_now(0),
+    m_stop(0)
 {
     uint8_t hash[200];
     char userId[65] = { 0 };
@@ -93,6 +96,12 @@ void DonateStrategy::connect()
 }
 
 
+void DonateStrategy::setAlgo(const xmrig::Algorithm &algo)
+{
+    m_strategy->setAlgo(algo);
+}
+
+
 void DonateStrategy::stop()
 {
     uv_timer_stop(&m_timer);
@@ -102,7 +111,14 @@ void DonateStrategy::stop()
 
 void DonateStrategy::tick(uint64_t now)
 {
+    m_now = now;
+
     m_strategy->tick(now);
+
+    if (m_stop && now > m_stop) {
+        m_strategy->stop();
+        m_stop = 0;
+    }
 }
 
 
@@ -119,7 +135,9 @@ void DonateStrategy::onActive(IStrategy *strategy, Client *client)
 
 void DonateStrategy::onJob(IStrategy *strategy, Client *client, const Job &job)
 {
-    m_listener->onJob(this, client, job);
+    if (isActive()) {
+        m_listener->onJob(this, client, job);
+    }
 }
 
 
@@ -142,7 +160,11 @@ void DonateStrategy::idle(uint64_t timeout)
 
 void DonateStrategy::suspend()
 {
-    m_strategy->stop();
+#   if defined(XMRIG_AMD_PROJECT) || defined(XMRIG_NVIDIA_PROJECT)
+    m_stop = m_now + 5000;
+#   else
+    m_stop = m_now + 500;
+#   endif
 
     m_active = false;
     m_listener->onPause(this);

src/net/strategies/DonateStrategy.h

@@ -5,7 +5,8 @@
  * Copyright 2014-2016 Wolf9466    <https://github.com/OhGodAPet>
  * Copyright 2016      Jay D Dee   <jayddee246@gmail.com>
  * Copyright 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
- * Copyright 2016-2018 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ * Copyright 2018-2019 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2019 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
@@ -44,7 +45,7 @@ class DonateStrategy : public IStrategy, public IStrategyListener
 {
 public:
     DonateStrategy(int level, const char *user, xmrig::Algo algo, IStrategyListener *listener);
-    ~DonateStrategy();
+    ~DonateStrategy() override;
 
 public:
     inline bool isActive() const override  { return m_active; }
@@ -52,6 +53,7 @@ public:
 
     int64_t submit(const JobResult &result) override;
     void connect() override;
+    void setAlgo(const xmrig::Algorithm &algo) override;
     void stop() override;
     void tick(uint64_t now) override;
 
@@ -68,11 +70,13 @@ private:
     static void onTimer(uv_timer_t *handle);
 
     bool m_active;
-    const int m_donateTime;
+    const uint64_t m_donateTime;
-    const int m_idleTime;
+    const uint64_t m_idleTime;
     IStrategy *m_strategy;
     IStrategyListener *m_listener;
     std::vector<Pool> m_pools;
+    uint64_t m_now;
+    uint64_t m_stop;
     uv_timer_t m_timer;
 };
 

src/version.h

@@ -28,15 +28,15 @@
 #define APP_ID        "xmrig"
 #define APP_NAME      "XMRig"
 #define APP_DESC      "XMRig CPU miner"
-#define APP_VERSION   "2.9.4"
+#define APP_VERSION   "2.11.0"
 #define APP_DOMAIN    "xmrig.com"
 #define APP_SITE      "www.xmrig.com"
 #define APP_COPYRIGHT "Copyright (C) 2016-2019 xmrig.com"
 #define APP_KIND      "cpu"
 
 #define APP_VER_MAJOR  2
-#define APP_VER_MINOR  9
+#define APP_VER_MINOR  11
-#define APP_VER_PATCH  4
+#define APP_VER_PATCH  0
 
 #ifdef _MSC_VER
 #   if (_MSC_VER >= 1910)

src/workers/CpuThread.cpp

@@ -97,11 +97,16 @@ extern "C" void cnv2_mainloop_bulldozer_asm(cryptonight_ctx *ctx);
 extern "C" void cnv2_double_mainloop_sandybridge_asm(cryptonight_ctx *ctx0, cryptonight_ctx *ctx1);
 
 
-xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_ivybridge_asm = nullptr;
+xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_ivybridge_asm          = nullptr;
-xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_ryzen_asm = nullptr;
+xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_ryzen_asm              = nullptr;
-xmrig::CpuThread::cn_mainloop_fun cn_half_mainloop_bulldozer_asm = nullptr;
+xmrig::CpuThread::cn_mainloop_fun        cn_half_mainloop_bulldozer_asm          = nullptr;
 xmrig::CpuThread::cn_mainloop_double_fun cn_half_double_mainloop_sandybridge_asm = nullptr;
 
+xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_ivybridge_asm          = nullptr;
+xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_ryzen_asm              = nullptr;
+xmrig::CpuThread::cn_mainloop_fun        cn_trtl_mainloop_bulldozer_asm          = nullptr;
+xmrig::CpuThread::cn_mainloop_double_fun cn_trtl_double_mainloop_sandybridge_asm = nullptr;
+
 
 void xmrig::CpuThread::patchAsmVariants()
 {
@@ -113,11 +118,21 @@ void xmrig::CpuThread::patchAsmVariants()
     cn_half_mainloop_bulldozer_asm              = reinterpret_cast<cn_mainloop_fun>         (base + 0x2000);
     cn_half_double_mainloop_sandybridge_asm     = reinterpret_cast<cn_mainloop_double_fun>  (base + 0x3000);
 
+    cn_trtl_mainloop_ivybridge_asm              = reinterpret_cast<cn_mainloop_fun>         (base + 0x4000);
+    cn_trtl_mainloop_ryzen_asm                  = reinterpret_cast<cn_mainloop_fun>         (base + 0x5000);
+    cn_trtl_mainloop_bulldozer_asm              = reinterpret_cast<cn_mainloop_fun>         (base + 0x6000);
+    cn_trtl_double_mainloop_sandybridge_asm     = reinterpret_cast<cn_mainloop_double_fun>  (base + 0x7000);
+
     patchCode(cn_half_mainloop_ivybridge_asm,           cnv2_mainloop_ivybridge_asm,            xmrig::CRYPTONIGHT_HALF_ITER,   xmrig::CRYPTONIGHT_MASK);
     patchCode(cn_half_mainloop_ryzen_asm,               cnv2_mainloop_ryzen_asm,                xmrig::CRYPTONIGHT_HALF_ITER,   xmrig::CRYPTONIGHT_MASK);
     patchCode(cn_half_mainloop_bulldozer_asm,           cnv2_mainloop_bulldozer_asm,            xmrig::CRYPTONIGHT_HALF_ITER,   xmrig::CRYPTONIGHT_MASK);
     patchCode(cn_half_double_mainloop_sandybridge_asm,  cnv2_double_mainloop_sandybridge_asm,   xmrig::CRYPTONIGHT_HALF_ITER,   xmrig::CRYPTONIGHT_MASK);
 
+    patchCode(cn_trtl_mainloop_ivybridge_asm,           cnv2_mainloop_ivybridge_asm,            xmrig::CRYPTONIGHT_TRTL_ITER,   xmrig::CRYPTONIGHT_PICO_MASK);
+    patchCode(cn_trtl_mainloop_ryzen_asm,               cnv2_mainloop_ryzen_asm,                xmrig::CRYPTONIGHT_TRTL_ITER,   xmrig::CRYPTONIGHT_PICO_MASK);
+    patchCode(cn_trtl_mainloop_bulldozer_asm,           cnv2_mainloop_bulldozer_asm,            xmrig::CRYPTONIGHT_TRTL_ITER,   xmrig::CRYPTONIGHT_PICO_MASK);
+    patchCode(cn_trtl_double_mainloop_sandybridge_asm,  cnv2_double_mainloop_sandybridge_asm,   xmrig::CRYPTONIGHT_TRTL_ITER,   xmrig::CRYPTONIGHT_PICO_MASK);
+
     Mem::protectExecutableMemory(base, allocation_size);
     Mem::flushInstructionCache(base, allocation_size);
 }
@@ -135,12 +150,12 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
     assert(variant >= VARIANT_0 && variant < VARIANT_MAX);
 
 #   ifndef XMRIG_NO_ASM
-    constexpr const size_t count = VARIANT_MAX * 10 * 3 + 8;
+    constexpr const size_t count = VARIANT_MAX * 10 * ALGO_MAX + 12;
 #   else
-    constexpr const size_t count = VARIANT_MAX * 10 * 3;
+    constexpr const size_t count = VARIANT_MAX * 10 * ALGO_MAX;
 #   endif
 
-    static const cn_hash_fun func_table[count] = {
+    static const cn_hash_fun func_table[] = {
         cryptonight_single_hash<CRYPTONIGHT, false, VARIANT_0>,
         cryptonight_double_hash<CRYPTONIGHT, false, VARIANT_0>,
         cryptonight_single_hash<CRYPTONIGHT, true,  VARIANT_0>,
@@ -233,6 +248,23 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
         cryptonight_quad_hash<CRYPTONIGHT,   true,  VARIANT_HALF>,
         cryptonight_penta_hash<CRYPTONIGHT,  true,  VARIANT_HALF>,
 
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+
+#       ifndef XMRIG_NO_CN_GPU
+        cryptonight_single_hash_gpu<CRYPTONIGHT, false, VARIANT_GPU>,
+        nullptr,
+        cryptonight_single_hash_gpu<CRYPTONIGHT, true,  VARIANT_GPU>,
+        nullptr,
+        nullptr,
+        nullptr,
+        nullptr,
+        nullptr,
+        nullptr,
+        nullptr,
+#       else
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
+#       endif
+
 #       ifndef XMRIG_NO_AEON
         cryptonight_single_hash<CRYPTONIGHT_LITE, false, VARIANT_0>,
         cryptonight_double_hash<CRYPTONIGHT_LITE, false, VARIANT_0>,
@@ -263,18 +295,22 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL2
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
 #       else
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TUBE
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_MSR
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XHV
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
 #       endif
 
 #       ifndef XMRIG_NO_SUMO
@@ -319,18 +355,60 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
         nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL2
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
 #       else
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TUBE
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_MSR
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XHV
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
-        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
+#       endif
+#       ifndef XMRIG_NO_CN_PICO
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TUBE
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_MSR
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XHV
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+
+        cryptonight_single_hash<CRYPTONIGHT_PICO, false, VARIANT_TRTL>,
+        cryptonight_double_hash<CRYPTONIGHT_PICO, false, VARIANT_TRTL>,
+        cryptonight_single_hash<CRYPTONIGHT_PICO, true,  VARIANT_TRTL>,
+        cryptonight_double_hash<CRYPTONIGHT_PICO, true,  VARIANT_TRTL>,
+        cryptonight_triple_hash<CRYPTONIGHT_PICO, false, VARIANT_TRTL>,
+        cryptonight_quad_hash<CRYPTONIGHT_PICO,   false, VARIANT_TRTL>,
+        cryptonight_penta_hash<CRYPTONIGHT_PICO,  false, VARIANT_TRTL>,
+        cryptonight_triple_hash<CRYPTONIGHT_PICO, true,  VARIANT_TRTL>,
+        cryptonight_quad_hash<CRYPTONIGHT_PICO,   true,  VARIANT_TRTL>,
+        cryptonight_penta_hash<CRYPTONIGHT_PICO,  true,  VARIANT_TRTL>,
+
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
+#       else
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_0
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_1
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TUBE
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_MSR
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XHV
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_XAO
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_RTO
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_2
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_HALF
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_TRTL
+        nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, // VARIANT_GPU
 #       endif
 #       ifndef XMRIG_NO_ASM
         cryptonight_single_hash_asm<CRYPTONIGHT, VARIANT_2, ASM_INTEL>,
@@ -341,10 +419,18 @@ xmrig::CpuThread::cn_hash_fun xmrig::CpuThread::fn(Algo algorithm, AlgoVariant a
         cryptonight_single_hash_asm<CRYPTONIGHT, VARIANT_HALF, ASM_INTEL>,
         cryptonight_single_hash_asm<CRYPTONIGHT, VARIANT_HALF, ASM_RYZEN>,
         cryptonight_single_hash_asm<CRYPTONIGHT, VARIANT_HALF, ASM_BULLDOZER>,
-        cryptonight_double_hash_asm<CRYPTONIGHT, VARIANT_HALF, ASM_INTEL>
+        cryptonight_double_hash_asm<CRYPTONIGHT, VARIANT_HALF, ASM_INTEL>,
+
+        cryptonight_single_hash_asm<CRYPTONIGHT_PICO, VARIANT_TRTL, ASM_INTEL>,
+        cryptonight_single_hash_asm<CRYPTONIGHT_PICO, VARIANT_TRTL, ASM_RYZEN>,
+        cryptonight_single_hash_asm<CRYPTONIGHT_PICO, VARIANT_TRTL, ASM_BULLDOZER>,
+        cryptonight_double_hash_asm<CRYPTONIGHT_PICO, VARIANT_TRTL, ASM_INTEL>
 #       endif
     };
 
+
+    static_assert(count == sizeof(func_table) / sizeof(func_table[0]), "func_table size mismatch");
+
 #   ifndef NDEBUG
     const size_t index = fnIndex(algorithm, av, variant, assembly);
     cn_hash_fun func = func_table[index];
@@ -531,7 +617,7 @@ size_t xmrig::CpuThread::fnIndex(Algo algorithm, AlgoVariant av, Variant variant
         return index;
     }
 
-    constexpr const size_t offset = VARIANT_MAX * 10 * 3;
+    constexpr const size_t offset = VARIANT_MAX * 10 * ALGO_MAX;
     size_t extra_offset           = 0;
 
     if (algorithm == CRYPTONIGHT && (variant == VARIANT_2 || variant == VARIANT_HALF)) {
@@ -543,6 +629,17 @@ size_t xmrig::CpuThread::fnIndex(Algo algorithm, AlgoVariant av, Variant variant
             return offset + extra_offset + assembly - 2;
         }
 
+        if (av == AV_DOUBLE) {
+            return offset + 3 + extra_offset;
+        }
+    }
+    else if (algorithm == CRYPTONIGHT_PICO && variant == VARIANT_TRTL) {
+        extra_offset = 8;
+
+        if (av == AV_SINGLE) {
+            return offset + extra_offset + assembly - 2;
+        }
+
         if (av == AV_DOUBLE) {
             return offset + 3 + extra_offset;
         }

src/workers/MultiWorker.cpp

@@ -54,14 +54,24 @@ bool MultiWorker<N>::selfTest()
     using namespace xmrig;
 
     if (m_thread->algorithm() == CRYPTONIGHT) {
-        return verify(VARIANT_0,    test_output_v0)  &&
+        const bool rc = verify(VARIANT_0,    test_output_v0)  &&
-               verify(VARIANT_1,    test_output_v1)  &&
+                        verify(VARIANT_1,    test_output_v1)  &&
-               verify(VARIANT_2,    test_output_v2)  &&
+                        verify(VARIANT_2,    test_output_v2)  &&
-               verify(VARIANT_XTL,  test_output_xtl) &&
+                        verify(VARIANT_XTL,  test_output_xtl) &&
-               verify(VARIANT_MSR,  test_output_msr) &&
+                        verify(VARIANT_MSR,  test_output_msr) &&
-               verify(VARIANT_XAO,  test_output_xao) &&
+                        verify(VARIANT_XAO,  test_output_xao) &&
-               verify(VARIANT_RTO,  test_output_rto) &&
+                        verify(VARIANT_RTO,  test_output_rto) &&
-               verify(VARIANT_HALF, test_output_half);
+                        verify(VARIANT_HALF, test_output_half);
+
+#       ifndef XMRIG_NO_CN_GPU
+        if (!rc || N > 1) {
+            return rc;
+        }
+
+        return verify(VARIANT_GPU, test_output_gpu);
+#       else
+        return rc;
+#       endif
     }
 
 #   ifndef XMRIG_NO_AEON
@@ -79,6 +89,12 @@ bool MultiWorker<N>::selfTest()
     }
 #   endif
 
+#   ifndef XMRIG_NO_CN_PICO
+    if (m_thread->algorithm() == CRYPTONIGHT_PICO) {
+        return verify(VARIANT_TRTL, test_output_pico_trtl);
+    }
+#   endif
+
     return false;
 }
 

src/workers/Workers.cpp

@@ -346,16 +346,16 @@ void Workers::start(IWorker *worker)
 
     if (m_status.started == m_status.threads) {
         const double percent = (double) m_status.hugePages / m_status.pages * 100.0;
-        const size_t memory  = m_status.ways * xmrig::cn_select_memory(m_status.algo) / 1048576;
+        const size_t memory  = m_status.ways * xmrig::cn_select_memory(m_status.algo) / 1024;
 
         if (m_status.colors) {
-            LOG_INFO(GREEN_BOLD("READY (CPU)") " threads " CYAN_BOLD("%zu(%zu)") " huge pages %s%zu/%zu %1.0f%%\x1B[0m memory " CYAN_BOLD("%zu.0 MB") "",
+            LOG_INFO(GREEN_BOLD("READY (CPU)") " threads " CYAN_BOLD("%zu(%zu)") " huge pages %s%zu/%zu %1.0f%%\x1B[0m memory " CYAN_BOLD("%zu KB") "",
                      m_status.threads, m_status.ways,
                      (m_status.hugePages == m_status.pages ? "\x1B[1;32m" : (m_status.hugePages == 0 ? "\x1B[1;31m" : "\x1B[1;33m")),
                      m_status.hugePages, m_status.pages, percent, memory);
         }
         else {
-            LOG_INFO("READY (CPU) threads %zu(%zu) huge pages %zu/%zu %1.0f%% memory %zu.0 MB",
+            LOG_INFO("READY (CPU) threads %zu(%zu) huge pages %zu/%zu %1.0f%% memory %zu KB",
                      m_status.threads, m_status.ways, m_status.hugePages, m_status.pages, percent, memory);
         }
     }