Merge 5579440640 into 410084384e

.github/workflows/deploy.yml

@@ -0,0 +1,208 @@
+on:
+  push:
+    tags:
+    - 'v*'
+
+name: Create release and build artifacts
+
+jobs:
+  build_win:
+    name: Build Windows artifacts
+    runs-on: windows-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@master
+      - name: Checkout deps
+        run:  git clone https://github.com/xmrig/xmrig-deps.git
+      - name: Build project on Windows
+        run: |
+          cmake . -G "MinGW Makefiles" -DXMRIG_DEPS=xmrig-deps\gcc\x64
+          make -j2
+          copy src\config.json .
+          copy bin\WinRing0\WinRing0x64.sys .
+          7z a -tzip -mx windows_build.zip xmrig.exe config.json WinRing0x64.sys
+      - name: Upload Windows build artifacts
+        uses: actions/upload-artifact@v1
+        with:
+          name: windows_build
+          path: windows_build.zip
+
+  build_lin:
+    name: Build Ubuntu artifacts
+    runs-on: ubuntu-20.04
+    steps:
+      - name: Prepare Ubuntu tools
+        run: |
+          sudo apt update
+          sudo apt install -y git build-essential cmake libuv1-dev libssl-dev libhwloc-dev
+      - name: Checkout code
+        uses: actions/checkout@master
+      - name: Build project on Ubuntu
+        run: |
+          cmake .
+          make -j$(nproc)
+          cp src/config.json .
+          tar cfz ubuntu_build.tar.gz xmrig config.json
+      - name: Upload Ubuntu build artifacts
+        uses: actions/upload-artifact@v1
+        with:
+          name: ubuntu_build
+          path: ubuntu_build.tar.gz
+
+  build_macos:
+    name: Build MacOS artifacts
+    runs-on: macos-latest
+    steps:
+      - name: Prepare MacOS tools
+        run: |
+          brew install cmake libuv openssl hwloc
+      - name: Checkout code
+        uses: actions/checkout@master
+      - name: Build hwloc on MacOS
+        run: |
+          curl -O https://download.open-mpi.org/release/hwloc/v2.1/hwloc-2.1.0.tar.bz2
+          tar xjf hwloc-2.1.0.tar.bz2
+          cd hwloc-2.1.0
+          ./configure --disable-shared --enable-static --disable-io --disable-libxml2
+          make -j$(sysctl -n hw.logicalcpu)
+          cd ..
+      - name: Build project on MacOS
+        run: |
+          cmake . -DOPENSSL_ROOT_DIR=/usr/local/opt/openssl -DHWLOC_INCLUDE_DIR=hwloc-2.1.0/include -DHWLOC_LIBRARY=hwloc-2.1.0/hwloc/.libs/libhwloc.a
+          make -j$(sysctl -n hw.logicalcpu)
+          cp src/config.json .
+          tar cfz macos_build.tar.gz xmrig config.json
+      - name: Upload MacOS build artifacts
+        uses: actions/upload-artifact@v1
+        with:
+          name: macos_build
+          path: macos_build.tar.gz
+
+  build_lin_rh7:
+    name: Build CentOS 7 artifacts
+    runs-on: ubuntu-latest
+    container: centos:7
+    steps:
+      - name: Prepare CentOS 7 tools
+        run: |
+          yum install -y centos-release-scl epel-release
+          yum install -y devtoolset-9
+          yum install -y wget git cmake3 automake libtool autoconf libstdc++-static glibc-static
+      - name: Checkout code
+        run: |
+          git clone https://github.com/MoneroOcean/xmrig.git .
+          git checkout ${GITHUB_REF:10}
+      - name: Build project on CentOS 7
+        run: |
+          source /opt/rh/devtoolset-9/enable
+          cd scripts
+          ./build_deps.sh
+          cd ..
+          cmake3 . -DXMRIG_DEPS=scripts/deps
+          make -j$(nproc)
+          cp src/config.json .
+          tar cfz centos7_build.tar.gz xmrig config.json
+      - name: Upload CentOS 7 build artifacts
+        uses: actions/upload-artifact@v1
+        with:
+          name: centos7_build
+          path: centos7_build.tar.gz
+
+  deploy:
+    needs: [build_win, build_lin, build_macos, build_lin_rh7]
+    name: Create release and upload artifacts
+    runs-on: ubuntu-latest
+    steps:
+      - name: Create Release
+        id: create_release
+        uses: actions/create-release@v1.0.0
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          tag_name: ${{ github.ref }}
+          release_name: Release ${{ github.ref }}
+          draft: false
+          prerelease: false
+      - name: Set version
+        id: version
+        run: echo ::set-output name=VERSION::${GITHUB_REF:10}
+      - name: Download Windows build artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: windows_build
+      - name: Download Ubuntu build artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: ubuntu_build
+      - name: Download MacOS build artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: macos_build
+      - name: Download CentOS 7 build artifacts
+        uses: actions/download-artifact@v1
+        with:
+          name: centos7_build
+      - name: Upload Windows build release asset
+        uses: actions/upload-release-asset@v1.0.1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: windows_build/windows_build.zip
+          asset_name: xmrig-${{steps.version.outputs.VERSION}}-win64.zip
+          asset_content_type: application/zip
+      - name: Upload Ubuntu build release asset
+        uses: actions/upload-release-asset@v1.0.1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: ubuntu_build/ubuntu_build.tar.gz
+          asset_name: xmrig-${{steps.version.outputs.VERSION}}-lin64.tar.gz
+          asset_content_type: application/zip
+      - name: Upload MacOS build release asset
+        uses: actions/upload-release-asset@v1.0.1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: macos_build/macos_build.tar.gz
+          asset_name: xmrig-${{steps.version.outputs.VERSION}}-mac64.tar.gz
+          asset_content_type: application/zip
+      - name: Upload CentOS 7 build release asset
+        uses: actions/upload-release-asset@v1.0.1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ steps.create_release.outputs.upload_url }}
+          asset_path: centos7_build/centos7_build.tar.gz
+          asset_name: xmrig-${{steps.version.outputs.VERSION}}-lin64-compat.tar.gz
+          asset_content_type: application/zip
+      - name: Update xmrig_setup repo
+        run: |
+          git clone https://$GITHUB_ACTOR:${{secrets.xmrig_setup_key}}@github.com/MoneroOcean/xmrig_setup.git
+          cd xmrig_setup
+          git config user.name  MoneroOcean
+          git config user.email support@moneroocean.stream
+          cp ../centos7_build/centos7_build.tar.gz xmrig.tar.gz
+          cp ../windows_build/windows_build.zip xmrig.zip
+          unzip xmrig.zip
+          zip -u offline_miner_setup.zip xmrig.exe config.json WinRing0x64.sys
+          git commit -m "xmrig "${GITHUB_REF:10}" based release" xmrig.tar.gz xmrig.zip offline_miner_setup.zip
+          git push
+          cd ..
+      - name: Update hiveos repo
+        run: |
+          git clone https://$GITHUB_ACTOR:${{secrets.xmrig_setup_key}}@github.com/MoneroOcean/hiveos.git
+          cd hiveos
+          git config user.name  MoneroOcean
+          git config user.email support@moneroocean.stream
+          tar xf ../centos7_build/centos7_build.tar.gz
+          mv xmrig mo_xmrig/xmrig
+          mv config.json mo_xmrig/config_global.json
+          export VER=${GITHUB_REF:10}
+          export VER=${VER//-/_}
+          tar -zcvf mo_xmrig-$VER.tar.gz mo_xmrig
+          git add mo_xmrig-$VER.tar.gz mo_xmrig/xmrig mo_xmrig/config_global.json
+          git commit -m "xmrig "${GITHUB_REF:10}" based release" mo_xmrig-$VER.tar.gz
+          git push

.github/workflows/test.yml

@@ -0,0 +1,32 @@
+on: push
+
+name: Test builds
+
+jobs:
+  build_win:
+    name: Windows build
+    runs-on: windows-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@master
+      - name: Checkout deps
+        run:  git clone https://github.com/xmrig/xmrig-deps.git
+      - name: Build project on Windows
+        run: |
+          cmake . -G "MinGW Makefiles" -DXMRIG_DEPS=xmrig-deps\gcc\x64
+          make -j2
+
+  build_lin:
+    name: Ubuntu build
+    runs-on: ubuntu-latest
+    steps:
+      - name: Prepare Ubuntu tools
+        run: |
+          sudo apt update
+          sudo apt install -y git build-essential cmake libuv1-dev libssl-dev libhwloc-dev
+      - name: Checkout code
+        uses: actions/checkout@master
+      - name: Build project on Ubuntu
+        run: |
+          cmake .
+          make -j$(nproc)

CMakeLists.txt

@@ -5,6 +5,7 @@ option(WITH_HWLOC           "Enable hwloc support" ON)
 option(WITH_CN_LITE         "Enable CryptoNight-Lite algorithms family" ON)
 option(WITH_CN_HEAVY        "Enable CryptoNight-Heavy algorithms family" ON)
 option(WITH_CN_PICO         "Enable CryptoNight-Pico algorithm" ON)
+option(WITH_CN_GPU          "Enable CryptoNight-GPU algorithm" ON)
 option(WITH_CN_FEMTO        "Enable CryptoNight-UPX2 algorithm" ON)
 option(WITH_RANDOMX         "Enable RandomX algorithms family" ON)
 option(WITH_ARGON2          "Enable Argon2 algorithms family" ON)
@@ -23,6 +24,7 @@ option(WITH_NVML            "Enable NVML (NVIDIA Management Library) support (on
 option(WITH_ADL             "Enable ADL (AMD Display Library) or sysfs support (only if OpenCL backend enabled)" ON)
 option(WITH_STRICT_CACHE    "Enable strict checks for OpenCL cache" ON)
 option(WITH_INTERLEAVE_DEBUG_LOG "Enable debug log for threads interleave" OFF)
+option(WITH_MO_BENCHMARK    "Enable Benchmark module and algo-perf feature (for MoneroOcean)" ON)
 option(WITH_PROFILING       "Enable profiling for developers" OFF)
 option(WITH_SSE4_1          "Enable SSE 4.1 for Blake2" ON)
 option(WITH_BENCHMARK       "Enable builtin RandomX benchmark and stress test" ON)
@@ -124,6 +126,13 @@ set(SOURCES_CRYPTO
     src/crypto/common/VirtualMemory.cpp
    )
 
+if (WITH_MO_BENCHMARK)
+    list(APPEND SOURCES
+        src/core/MoBenchmark.cpp
+        )
+    add_definitions(/DXMRIG_FEATURE_MO_BENCHMARK)
+endif()
+
 if (WITH_HWLOC)
     list(APPEND HEADERS_CRYPTO
         src/crypto/common/NUMAMemoryPool.h
@@ -184,6 +193,7 @@ include(cmake/astrobwt.cmake)
 include(cmake/kawpow.cmake)
 include(cmake/OpenSSL.cmake)
 include(cmake/asm.cmake)
+include(cmake/cn-gpu.cmake)
 
 if (WITH_CN_LITE)
     add_definitions(/DXMRIG_ALGO_CN_LITE)

README.md

@@ -1,21 +1,21 @@
 # XMRig
 
-[![Github All Releases](https://img.shields.io/github/downloads/xmrig/xmrig/total.svg)](https://github.com/xmrig/xmrig/releases)
+[![Github All Releases](https://img.shields.io/github/downloads/xmrig/xmrig/total.svg)](https://github.com/MoneroOcean/xmrig/releases)
-[![GitHub release](https://img.shields.io/github/release/xmrig/xmrig/all.svg)](https://github.com/xmrig/xmrig/releases)
+[![GitHub release](https://img.shields.io/github/release/xmrig/xmrig/all.svg)](https://github.com/MoneroOcean/xmrig/releases)
-[![GitHub Release Date](https://img.shields.io/github/release-date/xmrig/xmrig.svg)](https://github.com/xmrig/xmrig/releases)
+[![GitHub Release Date](https://img.shields.io/github/release-date/xmrig/xmrig.svg)](https://github.com/MoneroOcean/xmrig/releases)
-[![GitHub license](https://img.shields.io/github/license/xmrig/xmrig.svg)](https://github.com/xmrig/xmrig/blob/master/LICENSE)
+[![GitHub license](https://img.shields.io/github/license/xmrig/xmrig.svg)](https://github.com/MoneroOcean/xmrig/blob/master/LICENSE)
-[![GitHub stars](https://img.shields.io/github/stars/xmrig/xmrig.svg)](https://github.com/xmrig/xmrig/stargazers)
+[![GitHub stars](https://img.shields.io/github/stars/xmrig/xmrig.svg)](https://github.com/MoneroOcean/xmrig/stargazers)
-[![GitHub forks](https://img.shields.io/github/forks/xmrig/xmrig.svg)](https://github.com/xmrig/xmrig/network)
+[![GitHub forks](https://img.shields.io/github/forks/xmrig/xmrig.svg)](https://github.com/MoneroOcean/xmrig/network)
 
 XMRig is a high performance, open source, cross platform RandomX, KawPow, CryptoNight and AstroBWT unified CPU/GPU miner and [RandomX benchmark](https://xmrig.com/benchmark). Official binaries are available for Windows, Linux, macOS and FreeBSD.
 
 ## Mining backends
 - **CPU** (x64/ARMv8)
 - **OpenCL** for AMD GPUs.
-- **CUDA** for NVIDIA GPUs via external [CUDA plugin](https://github.com/xmrig/xmrig-cuda).
+- **CUDA** for NVIDIA GPUs via external [CUDA plugin](https://github.com/MoneroOcean/xmrig-cuda).
 
 ## Download
-* **[Binary releases](https://github.com/xmrig/xmrig/releases)**
+* **[Binary releases](https://github.com/MoneroOcean/xmrig/releases)**
 * **[Build from source](https://xmrig.com/docs/miner/build)**
 
 ## Usage

cmake/cn-gpu.cmake

@@ -0,0 +1,31 @@
+if (WITH_CN_GPU AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+    add_definitions(/DXMRIG_ALGO_CN_GPU)
+
+    if (XMRIG_ARM)
+        list(APPEND SOURCES_CRYPTO
+            src/crypto/cn/gpu/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/cn_gpu_arm.cpp PROPERTIES COMPILE_FLAGS "-O3")
+        endif()
+    else()
+        list(APPEND SOURCES_CRYPTO
+            src/crypto/cn/gpu/cn_gpu_avx.cpp
+            src/crypto/cn/gpu/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/cn_gpu_avx.cpp PROPERTIES COMPILE_FLAGS "-O3 -mavx2")
+            set_source_files_properties(src/crypto/cn/gpu/cn_gpu_ssse3.cpp PROPERTIES COMPILE_FLAGS "-O3")
+        elseif (CMAKE_CXX_COMPILER_ID MATCHES Intel)
+            set_source_files_properties(src/crypto/cn/gpu/cn_gpu_avx.cpp PROPERTIES COMPILE_FLAGS "-O3 -mavx2")
+            set_source_files_properties(src/crypto/cn/gpu/cn_gpu_ssse3.cpp PROPERTIES COMPILE_FLAGS "-O1")
+        elseif (CMAKE_CXX_COMPILER_ID MATCHES MSVC)
+            set_source_files_properties(src/crypto/cn/gpu/cn_gpu_avx.cpp PROPERTIES COMPILE_FLAGS "/arch:AVX")
+        endif()
+    endif()
+
+else()
+    remove_definitions(/DXMRIG_ALGO_CN_GPU)
+endif()

cmake/randomx.cmake

@@ -40,6 +40,13 @@ if (WITH_RANDOMX)
         src/crypto/rx/RxDataset.cpp
         src/crypto/rx/RxQueue.cpp
         src/crypto/rx/RxVm.cpp
+		
+		### Removed useless includes		
+		src/crypto/randomx/panthera/sha256.c
+		src/crypto/randomx/panthera/KangarooTwelve.c
+		src/crypto/randomx/panthera/KeccakP-1600-reference.c
+		src/crypto/randomx/panthera/KeccakSpongeWidth1600.c
+		src/crypto/randomx/panthera/yespower-opt.c
     )
 
     if (WITH_ASM AND CMAKE_C_COMPILER_ID MATCHES MSVC)

doc/ALGORITHMS.md

@@ -28,6 +28,7 @@ Option `coin` useful for pools without [algorithm negotiation](https://xmrig.com
 | `cn/zls` | 2 MB | 2.14.0+ | CryptoNight variant 2 with 3/4 iterations. |  |
 | `cn/double` | 2 MB | 2.14.0+ | CryptoNight variant 2 with double iterations. |  |
 | `cn/r` | 2 MB | 2.13.0+ | CryptoNightR (Monero's variant 4). |  |
+| `cn/gpu` | 2 MB | 2.11.0+ | CryptoNight-GPU. |  |
 | `cn-pico` | 256 KB | 2.10.0+ | CryptoNight-Pico. |  |
 | `cn/half` | 2 MB | 2.9.0+ | CryptoNight variant 2 with half iterations. |  |
 | `cn/2` | 2 MB | 2.8.0+ | CryptoNight variant 2. |  |

doc/build/CMAKE_OPTIONS.md

@@ -6,6 +6,7 @@
 * **`-DWITH_CN_LITE=OFF`** disable all CryptoNight-Lite algorithms (`cn-lite/0`, `cn-lite/1`).
 * **`-DWITH_CN_HEAVY=OFF`** disable all CryptoNight-Heavy algorithms (`cn-heavy/0`, `cn-heavy/xhv`, `cn-heavy/tube`).
 * **`-DWITH_CN_PICO=OFF`** disable CryptoNight-Pico algorithm (`cn-pico`).
+* **`-DWITH_CN_GPU=OFF`** disable CryptoNight-GPU algorithm (`cn/gpu`).
 * **`-DWITH_RANDOMX=OFF`** disable RandomX algorithms (`rx/loki`, `rx/wow`).
 * **`-DWITH_ARGON2=OFF`** disable Argon2 algorithms (`argon2/chukwa`, `argon2/wrkz`).
 

scripts/generate_cl.js

@@ -86,10 +86,20 @@ function kawpow()
 }
 
 
+function cn_gpu()
+{
+    const cn_gpu = opencl_minify(addIncludes('cryptonight_gpu.cl', [ 'wolf-aes.cl', 'keccak.cl' ]));
+
+    // fs.writeFileSync('cryptonight_gpu_gen.cl', cn_gpu);
+    fs.writeFileSync('cryptonight_gpu_cl.h', text2h(cn_gpu, 'xmrig', 'cryptonight_gpu_cl'));
+}
+
+
 process.chdir(path.resolve('src/backend/opencl/cl/cn'));
 
 cn();
 cn_r();
+cn_gpu();
 
 process.chdir(cwd);
 process.chdir(path.resolve('src/backend/opencl/cl/rx'));

src/App.cpp

@@ -85,7 +85,21 @@ int xmrig::App::exec()
         return 0;
     }
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    m_controller->pre_start();
+    m_controller->config()->benchmark().set_controller(m_controller);
+
+    if (m_controller->config()->benchmark().isNewBenchRun() || m_controller->config()->isRebenchAlgo()) {
+        if (m_controller->config()->isShouldSave()) {
+            m_controller->config()->save();
+        }
+        m_controller->config()->benchmark().start();
+    } else {
+        m_controller->start();
+    }
+#   else
     m_controller->start();
+#   endif
 
     rc = uv_run(uv_default_loop(), UV_RUN_DEFAULT);
     uv_loop_close(uv_default_loop());

src/backend/common/Tags.h

@@ -46,6 +46,10 @@ const char *ocl_tag();
 const char *cuda_tag();
 #endif
 
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+const char *bm_tag();
+#endif
+
 
 } // namespace xmrig
 

src/backend/common/Threads.cpp

@@ -23,6 +23,8 @@
  */
 
 
+#include <type_traits>
+
 #include "backend/common/Threads.h"
 #include "3rdparty/rapidjson/document.h"
 #include "backend/cpu/CpuThreads.h"
@@ -136,6 +138,8 @@ xmrig::String xmrig::Threads<T>::profileName(const Algorithm &algorithm, bool st
         }
     }
 
+    if (std::is_same<T, CpuThreads>::value && (name == "defyx" || name == "panthera") && has("rx")) return "rx";
+
     if (has(kAsterisk)) {
         return kAsterisk;
     }

src/backend/cpu/CpuBackend.cpp

@@ -266,7 +266,7 @@ bool xmrig::CpuBackend::isEnabled() const
 
 bool xmrig::CpuBackend::isEnabled(const Algorithm &algorithm) const
 {
-    return !d_ptr->controller->config()->cpu().threads().get(algorithm).isEmpty();
+    return algorithm.isValid() && !d_ptr->controller->config()->cpu().threads().get(algorithm).isEmpty();
 }
 
 

src/backend/cpu/CpuConfig_gen.h

@@ -60,6 +60,10 @@ size_t inline generate<Algorithm::CN>(Threads<CpuThreads> &threads, uint32_t lim
         ++count;
     }
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    count += generate("cn/gpu", threads, Algorithm::CN_GPU, limit);
+#   endif
+
     return count;
 }
 
@@ -143,6 +147,10 @@ size_t inline generate<Algorithm::RANDOM_X>(Threads<CpuThreads> &threads, uint32
         count += threads.move("rx/wow", std::move(wow));
     }
 
+    if (!threads.isExist(Algorithm::RX_XLA)) {
+        count += generate("panthera", threads, Algorithm::RX_XLA, limit);
+    }
+
     count += generate("rx", threads, Algorithm::RX_0, limit);
 
     return count;

src/backend/cpu/CpuWorker.cpp

@@ -166,9 +166,20 @@ bool xmrig::CpuWorker<N>::selfTest()
                         verify(Algorithm::CN_RWZ,    test_output_rwz)  &&
                         verify(Algorithm::CN_ZLS,    test_output_zls)  &&
                         verify(Algorithm::CN_CCX,    test_output_ccx)  &&
+#                       ifdef XMRIG_ALGO_CN_GPU
+                        verify(Algorithm::CN_GPU,    test_output_gpu)  &&
+#                       endif
                         verify(Algorithm::CN_DOUBLE, test_output_double);
 
+#       ifdef XMRIG_ALGO_CN_GPU
+        if (!rc || N > 1) {
+            return rc;
+        }
+
+        return verify(Algorithm::CN_GPU, test_output_gpu);
+#       else
         return rc;
+#       endif
     }
 
 #   ifdef XMRIG_ALGO_CN_LITE
@@ -279,12 +290,13 @@ void xmrig::CpuWorker<N>::start()
 
 #           ifdef XMRIG_ALGO_RANDOMX
             if (job.algorithm().family() == Algorithm::RANDOM_X) {
+
                 if (first) {
                     first = false;
                     if (job.hasMinerSignature()) {
                         job.generateMinerSignature(m_job.blob(), job.size(), miner_signature_ptr);
                     }
-                    randomx_calculate_hash_first(m_vm, tempHash, m_job.blob(), job.size());
+                    randomx_calculate_hash_first(m_vm, tempHash, m_job.blob(), job.size(), job.algorithm());
                 }
 
                 if (!nextRound()) {
@@ -295,7 +307,7 @@ void xmrig::CpuWorker<N>::start()
                     memcpy(miner_signature_saved, miner_signature_ptr, sizeof(miner_signature_saved));
                     job.generateMinerSignature(m_job.blob(), job.size(), miner_signature_ptr);
                 }
-                randomx_calculate_hash_next(m_vm, tempHash, m_job.blob(), job.size(), m_hash);
+                randomx_calculate_hash_next(m_vm, tempHash, m_job.blob(), job.size(), m_hash, job.algorithm());
             }
             else
 #           endif

src/backend/cpu/platform/BasicCpuInfo.cpp

@@ -301,9 +301,12 @@ const char *xmrig::BasicCpuInfo::backend() const
 }
 
 
-xmrig::CpuThreads xmrig::BasicCpuInfo::threads(const Algorithm &algorithm, uint32_t) const
+xmrig::CpuThreads xmrig::BasicCpuInfo::threads(const Algorithm &algorithm, uint32_t limit) const
 {
-    const size_t count = std::thread::hardware_concurrency();
+    const uint32_t count = std::thread::hardware_concurrency();
+    const uint32_t count_limit  = std::max(static_cast<uint32_t>(count * (limit / 100.0f)), 1U);
+    const uint32_t count_limit2 = std::max(static_cast<uint32_t>(count / 2), count_limit);
+    const uint32_t count_limit4 = std::max(static_cast<uint32_t>(count / 4), count_limit);
 
     if (count == 1) {
         return 1;
@@ -313,13 +316,13 @@ xmrig::CpuThreads xmrig::BasicCpuInfo::threads(const Algorithm &algorithm, uint3
 
 #   ifdef XMRIG_ALGO_CN_LITE
     if (f == Algorithm::CN_LITE) {
-        return CpuThreads(count, 1);
+        return CpuThreads(count_limit, 1);
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_CN_PICO
     if (f == Algorithm::CN_PICO) {
-        return CpuThreads(count, 2);
+        return CpuThreads(count_limit, 2);
     }
 #   endif
 
@@ -331,37 +334,51 @@ xmrig::CpuThreads xmrig::BasicCpuInfo::threads(const Algorithm &algorithm, uint3
 
 #   ifdef XMRIG_ALGO_CN_HEAVY
     if (f == Algorithm::CN_HEAVY) {
-        return CpuThreads(std::max<size_t>(count / 4, 1), 1);
+        return CpuThreads(count_limit4, 1);
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_RANDOMX
     if (f == Algorithm::RANDOM_X) {
         if (algorithm == Algorithm::RX_WOW) {
-            return count;
+            return count_limit;
         }
 
-        return std::max<size_t>(count / 2, 1);
+        if (algorithm == Algorithm::RX_XLA) {
+            CpuThreads threads;
+            for (size_t i = 0; i < count_limit2; ++i) {
+                threads.add(i, 0);
+            }
+            return threads;
+        }
+
+        return count_limit2;
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_ARGON2
     if (f == Algorithm::ARGON2) {
-        return count;
+        return count_limit;
     }
 #   endif
 
 #   ifdef XMRIG_ALGO_ASTROBWT
     if (f == Algorithm::ASTROBWT) {
         CpuThreads threads;
-        for (size_t i = 0; i < count; ++i) {
+        for (size_t i = 0; i < count_limit; ++i) {
             threads.add(i, 0);
         }
         return threads;
     }
 #   endif
 
-    return CpuThreads(std::max<size_t>(count / 2, 1), 1);
+#   ifdef XMRIG_ALGO_CN_GPU
+    if (algorithm == Algorithm::CN_GPU) {
+        return count_limit;
+    }
+#   endif
+
+    return CpuThreads(count_limit2, 1);
 }
 
 

src/backend/cpu/platform/HwlocCpuInfo.cpp

@@ -341,10 +341,18 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
         intensity = 2;
     }
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    if (algorithm == Algorithm::CN_GPU) {
+        cacheHashes = PUs;
+    }
+#   endif
+
 #   ifdef XMRIG_ALGO_RANDOMX
     if (extra == 0 && algorithm.l2() > 0) {
         cacheHashes = std::min<size_t>(std::max<size_t>(L2 / algorithm.l2(), cores.size()), cacheHashes);
     }
+    if (algorithm == Algorithm::RX_XLA) cacheHashes = cores.size();
+
 #   endif
 
     if (limit > 0) {

src/backend/cuda/CudaConfig_gen.h

@@ -64,6 +64,10 @@ size_t inline generate<Algorithm::CN>(Threads<CudaThreads> &threads, const std::
         count++;
     }
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    count += generate("cn/gpu", threads, Algorithm::CN_GPU, devices);
+#   endif
+
     return count;
 }
 

src/backend/opencl/OclConfig_gen.h

@@ -63,6 +63,10 @@ size_t inline generate<Algorithm::CN>(Threads<OclThreads> &threads, const std::v
         count++;
     }
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    count += generate("cn/gpu", threads, Algorithm::CN_GPU, devices);
+#   endif
+
     return count;
 }
 

src/backend/opencl/OclThread.h

@@ -95,6 +95,20 @@ public:
     }
 #   endif
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    OclThread(uint32_t index, uint32_t intensity, uint32_t worksize, uint32_t threads, uint32_t unrollFactor) :
+        m_fields(0),
+        m_threads(threads, -1),
+        m_index(index),
+        m_memChunk(0),
+        m_stridedIndex(0),
+        m_unrollFactor(unrollFactor),
+        m_worksize(worksize)
+    {
+        setIntensity(intensity);
+    }
+#   endif
+
     OclThread(const rapidjson::Value &value);
 
     inline bool isAsm() const                               { return m_gcnAsm; }

src/backend/opencl/OclWorker.cpp

@@ -42,6 +42,10 @@
 #   include "backend/opencl/runners/OclKawPowRunner.h"
 #endif
 
+#ifdef XMRIG_ALGO_CN_GPU
+#   include "backend/opencl/runners/OclRyoRunner.h"
+#endif
+
 #include <cassert>
 #include <thread>
 
@@ -102,6 +106,12 @@ xmrig::OclWorker::OclWorker(size_t id, const OclLaunchData &data) :
         break;
 
     default:
+#       ifdef XMRIG_ALGO_CN_GPU
+        if (m_algorithm == Algorithm::CN_GPU) {
+            m_runner = new OclRyoRunner(id, data);
+        }
+        else
+#       endif
         m_runner = new OclCnRunner(id, data);
         break;
     }

src/backend/opencl/cl/OclSource.cpp

@@ -41,6 +41,10 @@
 #   include "backend/opencl/cl/kawpow/kawpow_dag_cl.h"
 #endif
 
+#ifdef XMRIG_ALGO_CN_GPU
+#   include "backend/opencl/cl/cn/cryptonight_gpu_cl.h"
+#endif
+
 
 const char *xmrig::OclSource::get(const Algorithm &algorithm)
 {
@@ -62,5 +66,11 @@ const char *xmrig::OclSource::get(const Algorithm &algorithm)
     }
 #   endif
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    if (algorithm == Algorithm::CN_GPU) {
+        return cryptonight_gpu_cl;
+    }
+#   endif
+
     return cryptonight_cl;
 }

src/backend/opencl/cl/cn/algorithm.cl

@@ -17,17 +17,19 @@
 #define ALGO_CN_PICO_0      16
 #define ALGO_CN_PICO_TLO    17
 #define ALGO_CN_CCX         18
-#define ALGO_CN_UPX2        19
+#define ALGO_CN_GPU         19
-#define ALGO_RX_0           20
+#define ALGO_CN_UPX2        20
-#define ALGO_RX_WOW         21
+#define ALGO_RX_0           21
-#define ALGO_RX_ARQMA       22
+#define ALGO_RX_WOW         22
-#define ALGO_RX_SFX         23
+#define ALGO_RX_ARQ         23
-#define ALGO_RX_KEVA        24
+#define ALGO_RX_SFX         24
-#define ALGO_AR2_CHUKWA     25
+#define ALGO_RX_KEVA        25
-#define ALGO_AR2_CHUKWA_V2  26
+#define ALGO_AR2_CHUKWA     26
-#define ALGO_AR2_WRKZ       27
+#define ALGO_AR2_CHUKWA_V2  27
-#define ALGO_ASTROBWT_DERO  28
+#define ALGO_AR2_WRKZ       28
-#define ALGO_KAWPOW_RVN     29
+#define ALGO_ASTROBWT_DERO  29
+#define ALGO_KAWPOW_RVN     30
+#define ALGO_RX_XLA         31
 
 #define FAMILY_UNKNOWN      0
 #define FAMILY_CN           1

src/backend/opencl/cl/cn/cryptonight_gpu.cl

@@ -0,0 +1,520 @@
+#include "wolf-aes.cl"
+#include "keccak.cl"
+
+
+inline uint getIdx()
+{
+    return get_global_id(0) - get_global_offset(0);
+}
+
+
+#define IDX(x) (x)
+
+
+inline float4 _mm_add_ps(float4 a, float4 b)
+{
+    return a + b;
+}
+
+
+inline float4 _mm_sub_ps(float4 a, float4 b)
+{
+    return a - b;
+}
+
+
+inline float4 _mm_mul_ps(float4 a, float4 b)
+{
+    return a * b;
+}
+
+
+inline float4 _mm_div_ps(float4 a, float4 b)
+{
+    return a / b;
+}
+
+
+inline float4 _mm_and_ps(float4 a, int b)
+{
+    return as_float4(as_int4(a) & (int4)(b));
+}
+
+
+inline float4 _mm_or_ps(float4 a, int b)
+{
+    return as_float4(as_int4(a) | (int4)(b));
+}
+
+
+inline float4 _mm_fmod_ps(float4 v, float dc)
+{
+    float4 d = (float4)(dc);
+    float4 c = _mm_div_ps(v, d);
+    c = trunc(c);
+    c = _mm_mul_ps(c, d);
+    return _mm_sub_ps(v, c);
+}
+
+
+inline int4 _mm_xor_si128(int4 a, int4 b)
+{
+    return a ^ b;
+}
+
+
+inline float4 _mm_xor_ps(float4 a, int b)
+{
+    return as_float4(as_int4(a) ^ (int4)(b));
+}
+
+
+inline int4 _mm_alignr_epi8(int4 a, const uint rot)
+{
+    const uint right = 8 * rot;
+    const uint left = (32 - 8 * rot);
+    return (int4)(
+        ((uint)a.x >> right) | ( a.y << left ),
+        ((uint)a.y >> right) | ( a.z << left ),
+        ((uint)a.z >> right) | ( a.w << left ),
+        ((uint)a.w >> right) | ( a.x << left )
+    );
+}
+
+
+inline global int4* scratchpad_ptr(uint idx, uint n, __global int *lpad) { return (__global int4*)((__global char*)lpad + (idx & MASK) + n * 16); }
+
+
+inline float4 fma_break(float4 x)
+{
+    // Break the dependency chain by setting the exp to ?????01
+    x = _mm_and_ps(x, 0xFEFFFFFF);
+    return _mm_or_ps(x, 0x00800000);
+}
+
+
+inline void sub_round(float4 n0, float4 n1, float4 n2, float4 n3, float4 rnd_c, float4* n, float4* d, float4* c)
+{
+    n1 = _mm_add_ps(n1, *c);
+    float4 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);
+    float4 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, (float4)(0.734375f));
+    float4 r = _mm_add_ps(nn, dd);
+    r = _mm_and_ps(r, 0x807FFFFF);
+    r = _mm_or_ps(r, 0x40000000);
+    *c = _mm_add_ps(*c, r);
+
+}
+
+
+// 9*8 + 2 = 74
+inline void round_compute(float4 n0, float4 n1, float4 n2, float4 n3, float4 rnd_c, float4* c, float4* r)
+{
+    float4 n = (float4)(0.0f);
+    float4 d = (float4)(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
+    d = _mm_and_ps(d, 0xFF7FFFFF);
+    d = _mm_or_ps(d, 0x40000000);
+    *r =_mm_add_ps(*r, _mm_div_ps(n,d));
+}
+
+
+inline int4 single_compute(float4 n0, float4 n1, float4 n2, float4 n3, float cnt, float4 rnd_c, __local float4* sum)
+{
+    float4 c= (float4)(cnt);
+    // 35 maths calls follow (140 FLOPS)
+    float4 r = (float4)(0.0f);
+
+    for (int i = 0; i < 4; ++i) {
+        round_compute(n0, n1, n2, n3, rnd_c, &c, &r);
+    }
+
+    // do a quick fmod by setting exp to 2
+    r = _mm_and_ps(r, 0x807FFFFF);
+    r = _mm_or_ps(r, 0x40000000);
+    *sum = r; // 34
+    float4 x = (float4)(536870880.0f);
+    r = _mm_mul_ps(r, x); // 35
+    return convert_int4_rte(r);
+}
+
+
+inline void single_compute_wrap(const uint rot, int4 v0, int4 v1, int4 v2, int4 v3, float cnt, float4 rnd_c, __local float4* sum, __local int4* out)
+{
+    float4 n0 = convert_float4_rte(v0);
+    float4 n1 = convert_float4_rte(v1);
+    float4 n2 = convert_float4_rte(v2);
+    float4 n3 = convert_float4_rte(v3);
+
+    int4 r = single_compute(n0, n1, n2, n3, cnt, rnd_c, sum);
+    *out = rot == 0 ? r : _mm_alignr_epi8(r, rot);
+}
+
+
+static const __constant uint look[16][4] = {
+    {0, 1, 2, 3},
+    {0, 2, 3, 1},
+    {0, 3, 1, 2},
+    {0, 3, 2, 1},
+
+    {1, 0, 2, 3},
+    {1, 2, 3, 0},
+    {1, 3, 0, 2},
+    {1, 3, 2, 0},
+
+    {2, 1, 0, 3},
+    {2, 0, 3, 1},
+    {2, 3, 1, 0},
+    {2, 3, 0, 1},
+
+    {3, 1, 2, 0},
+    {3, 2, 0, 1},
+    {3, 0, 1, 2},
+    {3, 0, 2, 1}
+};
+
+
+static const __constant float ccnt[16] = {
+    1.34375f,
+    1.28125f,
+    1.359375f,
+    1.3671875f,
+
+    1.4296875f,
+    1.3984375f,
+    1.3828125f,
+    1.3046875f,
+
+    1.4140625f,
+    1.2734375f,
+    1.2578125f,
+    1.2890625f,
+
+    1.3203125f,
+    1.3515625f,
+    1.3359375f,
+    1.4609375f
+};
+
+
+struct SharedMemChunk
+{
+    int4 out[16];
+    float4 va[16];
+};
+
+
+__attribute__((reqd_work_group_size(WORKSIZE * 16, 1, 1)))
+__kernel void cn1(__global int *lpad_in, __global int *spad, uint numThreads)
+{
+    const uint gIdx = getIdx();
+    uint chunk = get_local_id(0) / 16;
+
+    __global int* lpad = (__global int*)((__global char*)lpad_in + MEMORY * (gIdx/16));
+
+    __local struct SharedMemChunk smem_in[WORKSIZE];
+    __local struct SharedMemChunk* smem = smem_in + chunk;
+
+    uint tid = get_local_id(0) % 16;
+
+    uint idxHash = gIdx/16;
+    uint s = ((__global uint*)spad)[idxHash * 50] >> 8;
+    float4 vs = (float4)(0);
+
+    // tid divided
+    const uint tidd = tid / 4;
+    // tid modulo
+    const uint tidm = tid % 4;
+    const uint block = tidd * 16 + tidm;
+
+    #pragma unroll CN_UNROLL
+    for (uint i = 0; i < ITERATIONS; i++)   {
+        mem_fence(CLK_LOCAL_MEM_FENCE);
+        int tmp = ((__global int*)scratchpad_ptr(s, tidd, lpad))[tidm];
+        ((__local int*)(smem->out))[tid] = tmp;
+        mem_fence(CLK_LOCAL_MEM_FENCE);
+
+        {
+            single_compute_wrap(
+                tidm,
+                *(smem->out + look[tid][0]),
+                *(smem->out + look[tid][1]),
+                *(smem->out + look[tid][2]),
+                *(smem->out + look[tid][3]),
+                ccnt[tid], vs, smem->va + tid,
+                smem->out + tid
+            );
+        }
+        mem_fence(CLK_LOCAL_MEM_FENCE);
+
+        int outXor = ((__local int*)smem->out)[block];
+        for (uint dd = block + 4; dd < (tidd + 1) * 16; dd += 4) {
+            outXor ^= ((__local int*)smem->out)[dd];
+        }
+
+        ((__global int*)scratchpad_ptr(s, tidd, lpad))[tidm] = outXor ^ tmp;
+        ((__local int*)smem->out)[tid] = outXor;
+
+        float va_tmp1 = ((__local float*)smem->va)[block] + ((__local float*)smem->va)[block + 4];
+        float va_tmp2 = ((__local float*)smem->va)[block+ 8] + ((__local float*)smem->va)[block + 12];
+        ((__local float*)smem->va)[tid] = va_tmp1 + va_tmp2;
+
+        mem_fence(CLK_LOCAL_MEM_FENCE);
+
+        int out2 = ((__local int*)smem->out)[tid] ^ ((__local int*)smem->out)[tid + 4 ] ^ ((__local int*)smem->out)[tid + 8] ^ ((__local int*)smem->out)[tid + 12];
+        va_tmp1 = ((__local float*)smem->va)[block] + ((__local float*)smem->va)[block + 4];
+        va_tmp2 = ((__local float*)smem->va)[block + 8] + ((__local float*)smem->va)[block + 12];
+        va_tmp1 = va_tmp1 + va_tmp2;
+        va_tmp1 = fabs(va_tmp1);
+
+        float xx = va_tmp1 * 16777216.0f;
+        int xx_int = (int)xx;
+        ((__local int*)smem->out)[tid] = out2 ^ xx_int;
+        ((__local float*)smem->va)[tid] = va_tmp1 / 64.0f;
+
+        mem_fence(CLK_LOCAL_MEM_FENCE);
+
+        vs = smem->va[0];
+        s = smem->out[0].x ^ smem->out[0].y ^ smem->out[0].z ^ smem->out[0].w;
+    }
+}
+
+
+static const __constant uint skip[3] = {
+    20,22,22
+};
+
+
+inline void generate_512(uint idx, __local ulong* in, __global ulong* out)
+{
+    ulong hash[25];
+
+    hash[0] = in[0] ^ idx;
+    for (int i = 1; i < 25; ++i) {
+        hash[i] = in[i];
+    }
+
+    for (int a = 0; a < 3; ++a) {
+        keccakf1600_1(hash);
+        for (int i = 0; i < skip[a]; ++i) {
+            out[i] = hash[i];
+        }
+
+        out += skip[a];
+    }
+}
+
+
+__attribute__((reqd_work_group_size(8, 8, 1)))
+__kernel void cn0(__global ulong *input, int inlen, __global int *Scratchpad, __global ulong *states, uint Threads)
+{
+    const uint gIdx = getIdx();
+    __local ulong State_buf[8 * 25];
+    __local ulong* State = State_buf + get_local_id(0) * 25;
+
+    {
+        states += 25 * gIdx;
+
+        Scratchpad = (__global int*)((__global char*)Scratchpad + MEMORY * gIdx);
+
+        if (get_local_id(1) == 0) {
+
+#           ifdef __NV_CL_C_VERSION
+            for(uint i = 0; i < 8; ++i)
+                State[i] = input[i];
+#           else
+            ((__local ulong8 *)State)[0] = vload8(0, input);
+#           endif
+
+            State[8]  = input[8];
+            State[9]  = input[9];
+            State[10] = input[10];
+
+            ((__local uint *)State)[9]  &= 0x00FFFFFFU;
+            ((__local uint *)State)[9]  |= (((uint)get_global_id(0)) & 0xFF) << 24;
+            ((__local uint *)State)[10] &= 0xFF000000U;
+            /* explicit cast to `uint` is required because some OpenCL implementations (e.g. NVIDIA)
+             * handle get_global_id and get_global_offset as signed long long int and add
+             * 0xFFFFFFFF... to `get_global_id` if we set on host side a 32bit offset where the first bit is `1`
+             * (even if it is correct casted to unsigned on the host)
+             */
+            ((__local uint *)State)[10] |= (((uint)get_global_id(0) >> 8));
+
+            for (int i = 11; i < 25; ++i) {
+                State[i] = 0x00UL;
+            }
+
+            // Last bit of padding
+            State[16] = 0x8000000000000000UL;
+
+            keccakf1600_2(State);
+
+            #pragma unroll
+            for (int i = 0; i < 25; ++i) {
+                states[i] = State[i];
+            }
+        }
+    }
+}
+
+
+__attribute__((reqd_work_group_size(64, 1, 1)))
+__kernel void cn00(__global int *Scratchpad, __global ulong *states)
+{
+    const uint gIdx = getIdx() / 64;
+    __local ulong State[25];
+
+    states += 25 * gIdx;
+
+    Scratchpad = (__global int*)((__global char*)Scratchpad + MEMORY * gIdx);
+
+    for (int i = get_local_id(0); i < 25; i += get_local_size(0)) {
+        State[i] = states[i];
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    for (uint i = get_local_id(0); i < MEMORY / 512; i += get_local_size(0)) {
+        generate_512(i, State, (__global ulong*)((__global uchar*)Scratchpad + i * 512));
+    }
+}
+
+
+__attribute__((reqd_work_group_size(8, 8, 1)))
+__kernel void cn2(__global uint4 *Scratchpad, __global ulong *states, __global uint *output, ulong Target, uint Threads)
+{
+    __local uint AES0[256], AES1[256], AES2[256], AES3[256];
+    uint ExpandedKey2[40];
+    uint4 text;
+
+    const uint gIdx = getIdx();
+
+    for (int i = get_local_id(1) * 8 + get_local_id(0); i < 256; i += 8 * 8) {
+        const uint tmp = AES0_C[i];
+        AES0[i] = tmp;
+        AES1[i] = rotate(tmp, 8U);
+        AES2[i] = rotate(tmp, 16U);
+        AES3[i] = rotate(tmp, 24U);
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local uint4 xin1[8][8];
+    __local uint4 xin2[8][8];
+
+    {
+        states += 25 * gIdx;
+        Scratchpad += gIdx * (MEMORY >> 4);
+
+        #if defined(__Tahiti__) || defined(__Pitcairn__)
+        for(int i = 0; i < 4; ++i) ((ulong *)ExpandedKey2)[i] = states[i + 4];
+        text = vload4(get_local_id(1) + 4, (__global uint *)states);
+        #else
+        text = vload4(get_local_id(1) + 4, (__global uint *)states);
+        ((uint8 *)ExpandedKey2)[0] = vload8(1, (__global uint *)states);
+        #endif
+
+        AESExpandKey256(ExpandedKey2);
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local uint4* xin1_store = &xin1[get_local_id(1)][get_local_id(0)];
+    __local uint4* xin1_load = &xin1[(get_local_id(1) + 1) % 8][get_local_id(0)];
+    __local uint4* xin2_store = &xin2[get_local_id(1)][get_local_id(0)];
+    __local uint4* xin2_load = &xin2[(get_local_id(1) + 1) % 8][get_local_id(0)];
+    *xin2_store = (uint4)(0, 0, 0, 0);
+
+    {
+
+        #pragma unroll 2
+        for (int i = 0, i1 = get_local_id(1); i < (MEMORY >> 7); ++i, i1 = (i1 + 16) % (MEMORY >> 4)) {
+            text ^= Scratchpad[(uint)i1];
+            barrier(CLK_LOCAL_MEM_FENCE);
+            text ^= *xin2_load;
+
+            #pragma unroll 10
+            for(int j = 0; j < 10; ++j)
+                text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
+
+            *xin1_store = text;
+
+            text ^= Scratchpad[(uint)i1 + 8u];
+            barrier(CLK_LOCAL_MEM_FENCE);
+            text ^= *xin1_load;
+
+            #pragma unroll 10
+            for(int j = 0; j < 10; ++j)
+                text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
+
+            *xin2_store = text;
+        }
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+        text ^= *xin2_load;
+    }
+
+    /* Also left over threads performe this loop.
+     * The left over thread results will be ignored
+     */
+    #pragma unroll 16
+    for(size_t i = 0; i < 16; i++)
+    {
+        #pragma unroll 10
+        for (int j = 0; j < 10; ++j) {
+            text = AES_Round(AES0, AES1, AES2, AES3, text, ((uint4 *)ExpandedKey2)[j]);
+        }
+
+        barrier(CLK_LOCAL_MEM_FENCE);
+        *xin1_store = text;
+        barrier(CLK_LOCAL_MEM_FENCE);
+        text ^= *xin1_load;
+    }
+
+    __local ulong State_buf[8 * 25];
+    {
+        vstore2(as_ulong2(text), get_local_id(1) + 4, states);
+    }
+
+    barrier(CLK_GLOBAL_MEM_FENCE);
+
+    {
+        if(!get_local_id(1))
+        {
+            __local ulong* State = State_buf + get_local_id(0) * 25;
+
+            for(int i = 0; i < 25; ++i) State[i] = states[i];
+
+            keccakf1600_2(State);
+
+            if(State[3] <= Target)
+            {
+                ulong outIdx = atomic_inc(output + 0xFF);
+                if(outIdx < 0xFF)
+                    output[outIdx] = get_global_id(0);
+            }
+        }
+    }
+    mem_fence(CLK_GLOBAL_MEM_FENCE);
+}

src/backend/opencl/cl/cn/cryptonight_gpu_cl.h

@@ -0,0 +1,639 @@
+#pragma once
+
+namespace xmrig {
+
+static const char cryptonight_gpu_cl[20176] = {
+    0x23,0x69,0x66,0x6e,0x64,0x65,0x66,0x20,0x57,0x4f,0x4c,0x46,0x5f,0x41,0x45,0x53,0x5f,0x43,0x4c,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x57,0x4f,0x4c,0x46,
+    0x5f,0x41,0x45,0x53,0x5f,0x43,0x4c,0x0a,0x23,0x69,0x66,0x64,0x65,0x66,0x20,0x63,0x6c,0x5f,0x61,0x6d,0x64,0x5f,0x6d,0x65,0x64,0x69,0x61,0x5f,0x6f,0x70,0x73,0x32,
+    0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x4f,0x50,0x45,0x4e,0x43,0x4c,0x20,0x45,0x58,0x54,0x45,0x4e,0x53,0x49,0x4f,0x4e,0x20,0x63,0x6c,0x5f,0x61,0x6d,0x64,
+    0x5f,0x6d,0x65,0x64,0x69,0x61,0x5f,0x6f,0x70,0x73,0x32,0x20,0x3a,0x20,0x65,0x6e,0x61,0x62,0x6c,0x65,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x78,0x6d,0x72,
+    0x69,0x67,0x5f,0x61,0x6d,0x64,0x5f,0x62,0x66,0x65,0x28,0x73,0x72,0x63,0x30,0x2c,0x20,0x73,0x72,0x63,0x31,0x2c,0x20,0x73,0x72,0x63,0x32,0x29,0x20,0x61,0x6d,0x64,
+    0x5f,0x62,0x66,0x65,0x28,0x73,0x72,0x63,0x30,0x2c,0x20,0x73,0x72,0x63,0x31,0x2c,0x20,0x73,0x72,0x63,0x32,0x29,0x0a,0x23,0x65,0x6c,0x73,0x65,0x0a,0x69,0x6e,0x6c,
+    0x69,0x6e,0x65,0x20,0x69,0x6e,0x74,0x20,0x78,0x6d,0x72,0x69,0x67,0x5f,0x61,0x6d,0x64,0x5f,0x62,0x66,0x65,0x28,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,
+    0x20,0x73,0x72,0x63,0x30,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x6f,0x66,0x66,0x73,0x65,0x74,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,
+    0x6e,0x74,0x20,0x77,0x69,0x64,0x74,0x68,0x29,0x0a,0x7b,0x0a,0x69,0x66,0x28,0x28,0x6f,0x66,0x66,0x73,0x65,0x74,0x2b,0x77,0x69,0x64,0x74,0x68,0x29,0x3c,0x33,0x32,
+    0x75,0x29,0x20,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x28,0x73,0x72,0x63,0x30,0x3c,0x3c,0x28,0x33,0x32,0x75,0x2d,0x6f,0x66,0x66,0x73,0x65,0x74,0x2d,0x77,
+    0x69,0x64,0x74,0x68,0x29,0x29,0x3e,0x3e,0x28,0x33,0x32,0x75,0x2d,0x77,0x69,0x64,0x74,0x68,0x29,0x3b,0x0a,0x7d,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x73,0x72,
+    0x63,0x30,0x3e,0x3e,0x6f,0x66,0x66,0x73,0x65,0x74,0x3b,0x0a,0x7d,0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,
+    0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x41,0x45,0x53,0x30,0x5f,0x43,0x5b,0x32,0x35,0x36,0x5d,0x20,0x3d,0x0a,
+    0x7b,0x0a,0x30,0x78,0x41,0x35,0x36,0x33,0x36,0x33,0x43,0x36,0x55,0x2c,0x30,0x78,0x38,0x34,0x37,0x43,0x37,0x43,0x46,0x38,0x55,0x2c,0x30,0x78,0x39,0x39,0x37,0x37,
+    0x37,0x37,0x45,0x45,0x55,0x2c,0x30,0x78,0x38,0x44,0x37,0x42,0x37,0x42,0x46,0x36,0x55,0x2c,0x0a,0x30,0x78,0x30,0x44,0x46,0x32,0x46,0x32,0x46,0x46,0x55,0x2c,0x30,
+    0x78,0x42,0x44,0x36,0x42,0x36,0x42,0x44,0x36,0x55,0x2c,0x30,0x78,0x42,0x31,0x36,0x46,0x36,0x46,0x44,0x45,0x55,0x2c,0x30,0x78,0x35,0x34,0x43,0x35,0x43,0x35,0x39,
+    0x31,0x55,0x2c,0x0a,0x30,0x78,0x35,0x30,0x33,0x30,0x33,0x30,0x36,0x30,0x55,0x2c,0x30,0x78,0x30,0x33,0x30,0x31,0x30,0x31,0x30,0x32,0x55,0x2c,0x30,0x78,0x41,0x39,
+    0x36,0x37,0x36,0x37,0x43,0x45,0x55,0x2c,0x30,0x78,0x37,0x44,0x32,0x42,0x32,0x42,0x35,0x36,0x55,0x2c,0x0a,0x30,0x78,0x31,0x39,0x46,0x45,0x46,0x45,0x45,0x37,0x55,
+    0x2c,0x30,0x78,0x36,0x32,0x44,0x37,0x44,0x37,0x42,0x35,0x55,0x2c,0x30,0x78,0x45,0x36,0x41,0x42,0x41,0x42,0x34,0x44,0x55,0x2c,0x30,0x78,0x39,0x41,0x37,0x36,0x37,
+    0x36,0x45,0x43,0x55,0x2c,0x0a,0x30,0x78,0x34,0x35,0x43,0x41,0x43,0x41,0x38,0x46,0x55,0x2c,0x30,0x78,0x39,0x44,0x38,0x32,0x38,0x32,0x31,0x46,0x55,0x2c,0x30,0x78,
+    0x34,0x30,0x43,0x39,0x43,0x39,0x38,0x39,0x55,0x2c,0x30,0x78,0x38,0x37,0x37,0x44,0x37,0x44,0x46,0x41,0x55,0x2c,0x0a,0x30,0x78,0x31,0x35,0x46,0x41,0x46,0x41,0x45,
+    0x46,0x55,0x2c,0x30,0x78,0x45,0x42,0x35,0x39,0x35,0x39,0x42,0x32,0x55,0x2c,0x30,0x78,0x43,0x39,0x34,0x37,0x34,0x37,0x38,0x45,0x55,0x2c,0x30,0x78,0x30,0x42,0x46,
+    0x30,0x46,0x30,0x46,0x42,0x55,0x2c,0x0a,0x30,0x78,0x45,0x43,0x41,0x44,0x41,0x44,0x34,0x31,0x55,0x2c,0x30,0x78,0x36,0x37,0x44,0x34,0x44,0x34,0x42,0x33,0x55,0x2c,
+    0x30,0x78,0x46,0x44,0x41,0x32,0x41,0x32,0x35,0x46,0x55,0x2c,0x30,0x78,0x45,0x41,0x41,0x46,0x41,0x46,0x34,0x35,0x55,0x2c,0x0a,0x30,0x78,0x42,0x46,0x39,0x43,0x39,
+    0x43,0x32,0x33,0x55,0x2c,0x30,0x78,0x46,0x37,0x41,0x34,0x41,0x34,0x35,0x33,0x55,0x2c,0x30,0x78,0x39,0x36,0x37,0x32,0x37,0x32,0x45,0x34,0x55,0x2c,0x30,0x78,0x35,
+    0x42,0x43,0x30,0x43,0x30,0x39,0x42,0x55,0x2c,0x0a,0x30,0x78,0x43,0x32,0x42,0x37,0x42,0x37,0x37,0x35,0x55,0x2c,0x30,0x78,0x31,0x43,0x46,0x44,0x46,0x44,0x45,0x31,
+    0x55,0x2c,0x30,0x78,0x41,0x45,0x39,0x33,0x39,0x33,0x33,0x44,0x55,0x2c,0x30,0x78,0x36,0x41,0x32,0x36,0x32,0x36,0x34,0x43,0x55,0x2c,0x0a,0x30,0x78,0x35,0x41,0x33,
+    0x36,0x33,0x36,0x36,0x43,0x55,0x2c,0x30,0x78,0x34,0x31,0x33,0x46,0x33,0x46,0x37,0x45,0x55,0x2c,0x30,0x78,0x30,0x32,0x46,0x37,0x46,0x37,0x46,0x35,0x55,0x2c,0x30,
+    0x78,0x34,0x46,0x43,0x43,0x43,0x43,0x38,0x33,0x55,0x2c,0x0a,0x30,0x78,0x35,0x43,0x33,0x34,0x33,0x34,0x36,0x38,0x55,0x2c,0x30,0x78,0x46,0x34,0x41,0x35,0x41,0x35,
+    0x35,0x31,0x55,0x2c,0x30,0x78,0x33,0x34,0x45,0x35,0x45,0x35,0x44,0x31,0x55,0x2c,0x30,0x78,0x30,0x38,0x46,0x31,0x46,0x31,0x46,0x39,0x55,0x2c,0x0a,0x30,0x78,0x39,
+    0x33,0x37,0x31,0x37,0x31,0x45,0x32,0x55,0x2c,0x30,0x78,0x37,0x33,0x44,0x38,0x44,0x38,0x41,0x42,0x55,0x2c,0x30,0x78,0x35,0x33,0x33,0x31,0x33,0x31,0x36,0x32,0x55,
+    0x2c,0x30,0x78,0x33,0x46,0x31,0x35,0x31,0x35,0x32,0x41,0x55,0x2c,0x0a,0x30,0x78,0x30,0x43,0x30,0x34,0x30,0x34,0x30,0x38,0x55,0x2c,0x30,0x78,0x35,0x32,0x43,0x37,
+    0x43,0x37,0x39,0x35,0x55,0x2c,0x30,0x78,0x36,0x35,0x32,0x33,0x32,0x33,0x34,0x36,0x55,0x2c,0x30,0x78,0x35,0x45,0x43,0x33,0x43,0x33,0x39,0x44,0x55,0x2c,0x0a,0x30,
+    0x78,0x32,0x38,0x31,0x38,0x31,0x38,0x33,0x30,0x55,0x2c,0x30,0x78,0x41,0x31,0x39,0x36,0x39,0x36,0x33,0x37,0x55,0x2c,0x30,0x78,0x30,0x46,0x30,0x35,0x30,0x35,0x30,
+    0x41,0x55,0x2c,0x30,0x78,0x42,0x35,0x39,0x41,0x39,0x41,0x32,0x46,0x55,0x2c,0x0a,0x30,0x78,0x30,0x39,0x30,0x37,0x30,0x37,0x30,0x45,0x55,0x2c,0x30,0x78,0x33,0x36,
+    0x31,0x32,0x31,0x32,0x32,0x34,0x55,0x2c,0x30,0x78,0x39,0x42,0x38,0x30,0x38,0x30,0x31,0x42,0x55,0x2c,0x30,0x78,0x33,0x44,0x45,0x32,0x45,0x32,0x44,0x46,0x55,0x2c,
+    0x0a,0x30,0x78,0x32,0x36,0x45,0x42,0x45,0x42,0x43,0x44,0x55,0x2c,0x30,0x78,0x36,0x39,0x32,0x37,0x32,0x37,0x34,0x45,0x55,0x2c,0x30,0x78,0x43,0x44,0x42,0x32,0x42,
+    0x32,0x37,0x46,0x55,0x2c,0x30,0x78,0x39,0x46,0x37,0x35,0x37,0x35,0x45,0x41,0x55,0x2c,0x0a,0x30,0x78,0x31,0x42,0x30,0x39,0x30,0x39,0x31,0x32,0x55,0x2c,0x30,0x78,
+    0x39,0x45,0x38,0x33,0x38,0x33,0x31,0x44,0x55,0x2c,0x30,0x78,0x37,0x34,0x32,0x43,0x32,0x43,0x35,0x38,0x55,0x2c,0x30,0x78,0x32,0x45,0x31,0x41,0x31,0x41,0x33,0x34,
+    0x55,0x2c,0x0a,0x30,0x78,0x32,0x44,0x31,0x42,0x31,0x42,0x33,0x36,0x55,0x2c,0x30,0x78,0x42,0x32,0x36,0x45,0x36,0x45,0x44,0x43,0x55,0x2c,0x30,0x78,0x45,0x45,0x35,
+    0x41,0x35,0x41,0x42,0x34,0x55,0x2c,0x30,0x78,0x46,0x42,0x41,0x30,0x41,0x30,0x35,0x42,0x55,0x2c,0x0a,0x30,0x78,0x46,0x36,0x35,0x32,0x35,0x32,0x41,0x34,0x55,0x2c,
+    0x30,0x78,0x34,0x44,0x33,0x42,0x33,0x42,0x37,0x36,0x55,0x2c,0x30,0x78,0x36,0x31,0x44,0x36,0x44,0x36,0x42,0x37,0x55,0x2c,0x30,0x78,0x43,0x45,0x42,0x33,0x42,0x33,
+    0x37,0x44,0x55,0x2c,0x0a,0x30,0x78,0x37,0x42,0x32,0x39,0x32,0x39,0x35,0x32,0x55,0x2c,0x30,0x78,0x33,0x45,0x45,0x33,0x45,0x33,0x44,0x44,0x55,0x2c,0x30,0x78,0x37,
+    0x31,0x32,0x46,0x32,0x46,0x35,0x45,0x55,0x2c,0x30,0x78,0x39,0x37,0x38,0x34,0x38,0x34,0x31,0x33,0x55,0x2c,0x0a,0x30,0x78,0x46,0x35,0x35,0x33,0x35,0x33,0x41,0x36,
+    0x55,0x2c,0x30,0x78,0x36,0x38,0x44,0x31,0x44,0x31,0x42,0x39,0x55,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x55,0x2c,0x30,0x78,0x32,0x43,0x45,0x44,
+    0x45,0x44,0x43,0x31,0x55,0x2c,0x0a,0x30,0x78,0x36,0x30,0x32,0x30,0x32,0x30,0x34,0x30,0x55,0x2c,0x30,0x78,0x31,0x46,0x46,0x43,0x46,0x43,0x45,0x33,0x55,0x2c,0x30,
+    0x78,0x43,0x38,0x42,0x31,0x42,0x31,0x37,0x39,0x55,0x2c,0x30,0x78,0x45,0x44,0x35,0x42,0x35,0x42,0x42,0x36,0x55,0x2c,0x0a,0x30,0x78,0x42,0x45,0x36,0x41,0x36,0x41,
+    0x44,0x34,0x55,0x2c,0x30,0x78,0x34,0x36,0x43,0x42,0x43,0x42,0x38,0x44,0x55,0x2c,0x30,0x78,0x44,0x39,0x42,0x45,0x42,0x45,0x36,0x37,0x55,0x2c,0x30,0x78,0x34,0x42,
+    0x33,0x39,0x33,0x39,0x37,0x32,0x55,0x2c,0x0a,0x30,0x78,0x44,0x45,0x34,0x41,0x34,0x41,0x39,0x34,0x55,0x2c,0x30,0x78,0x44,0x34,0x34,0x43,0x34,0x43,0x39,0x38,0x55,
+    0x2c,0x30,0x78,0x45,0x38,0x35,0x38,0x35,0x38,0x42,0x30,0x55,0x2c,0x30,0x78,0x34,0x41,0x43,0x46,0x43,0x46,0x38,0x35,0x55,0x2c,0x0a,0x30,0x78,0x36,0x42,0x44,0x30,
+    0x44,0x30,0x42,0x42,0x55,0x2c,0x30,0x78,0x32,0x41,0x45,0x46,0x45,0x46,0x43,0x35,0x55,0x2c,0x30,0x78,0x45,0x35,0x41,0x41,0x41,0x41,0x34,0x46,0x55,0x2c,0x30,0x78,
+    0x31,0x36,0x46,0x42,0x46,0x42,0x45,0x44,0x55,0x2c,0x0a,0x30,0x78,0x43,0x35,0x34,0x33,0x34,0x33,0x38,0x36,0x55,0x2c,0x30,0x78,0x44,0x37,0x34,0x44,0x34,0x44,0x39,
+    0x41,0x55,0x2c,0x30,0x78,0x35,0x35,0x33,0x33,0x33,0x33,0x36,0x36,0x55,0x2c,0x30,0x78,0x39,0x34,0x38,0x35,0x38,0x35,0x31,0x31,0x55,0x2c,0x0a,0x30,0x78,0x43,0x46,
+    0x34,0x35,0x34,0x35,0x38,0x41,0x55,0x2c,0x30,0x78,0x31,0x30,0x46,0x39,0x46,0x39,0x45,0x39,0x55,0x2c,0x30,0x78,0x30,0x36,0x30,0x32,0x30,0x32,0x30,0x34,0x55,0x2c,
+    0x30,0x78,0x38,0x31,0x37,0x46,0x37,0x46,0x46,0x45,0x55,0x2c,0x0a,0x30,0x78,0x46,0x30,0x35,0x30,0x35,0x30,0x41,0x30,0x55,0x2c,0x30,0x78,0x34,0x34,0x33,0x43,0x33,
+    0x43,0x37,0x38,0x55,0x2c,0x30,0x78,0x42,0x41,0x39,0x46,0x39,0x46,0x32,0x35,0x55,0x2c,0x30,0x78,0x45,0x33,0x41,0x38,0x41,0x38,0x34,0x42,0x55,0x2c,0x0a,0x30,0x78,
+    0x46,0x33,0x35,0x31,0x35,0x31,0x41,0x32,0x55,0x2c,0x30,0x78,0x46,0x45,0x41,0x33,0x41,0x33,0x35,0x44,0x55,0x2c,0x30,0x78,0x43,0x30,0x34,0x30,0x34,0x30,0x38,0x30,
+    0x55,0x2c,0x30,0x78,0x38,0x41,0x38,0x46,0x38,0x46,0x30,0x35,0x55,0x2c,0x0a,0x30,0x78,0x41,0x44,0x39,0x32,0x39,0x32,0x33,0x46,0x55,0x2c,0x30,0x78,0x42,0x43,0x39,
+    0x44,0x39,0x44,0x32,0x31,0x55,0x2c,0x30,0x78,0x34,0x38,0x33,0x38,0x33,0x38,0x37,0x30,0x55,0x2c,0x30,0x78,0x30,0x34,0x46,0x35,0x46,0x35,0x46,0x31,0x55,0x2c,0x0a,
+    0x30,0x78,0x44,0x46,0x42,0x43,0x42,0x43,0x36,0x33,0x55,0x2c,0x30,0x78,0x43,0x31,0x42,0x36,0x42,0x36,0x37,0x37,0x55,0x2c,0x30,0x78,0x37,0x35,0x44,0x41,0x44,0x41,
+    0x41,0x46,0x55,0x2c,0x30,0x78,0x36,0x33,0x32,0x31,0x32,0x31,0x34,0x32,0x55,0x2c,0x0a,0x30,0x78,0x33,0x30,0x31,0x30,0x31,0x30,0x32,0x30,0x55,0x2c,0x30,0x78,0x31,
+    0x41,0x46,0x46,0x46,0x46,0x45,0x35,0x55,0x2c,0x30,0x78,0x30,0x45,0x46,0x33,0x46,0x33,0x46,0x44,0x55,0x2c,0x30,0x78,0x36,0x44,0x44,0x32,0x44,0x32,0x42,0x46,0x55,
+    0x2c,0x0a,0x30,0x78,0x34,0x43,0x43,0x44,0x43,0x44,0x38,0x31,0x55,0x2c,0x30,0x78,0x31,0x34,0x30,0x43,0x30,0x43,0x31,0x38,0x55,0x2c,0x30,0x78,0x33,0x35,0x31,0x33,
+    0x31,0x33,0x32,0x36,0x55,0x2c,0x30,0x78,0x32,0x46,0x45,0x43,0x45,0x43,0x43,0x33,0x55,0x2c,0x0a,0x30,0x78,0x45,0x31,0x35,0x46,0x35,0x46,0x42,0x45,0x55,0x2c,0x30,
+    0x78,0x41,0x32,0x39,0x37,0x39,0x37,0x33,0x35,0x55,0x2c,0x30,0x78,0x43,0x43,0x34,0x34,0x34,0x34,0x38,0x38,0x55,0x2c,0x30,0x78,0x33,0x39,0x31,0x37,0x31,0x37,0x32,
+    0x45,0x55,0x2c,0x0a,0x30,0x78,0x35,0x37,0x43,0x34,0x43,0x34,0x39,0x33,0x55,0x2c,0x30,0x78,0x46,0x32,0x41,0x37,0x41,0x37,0x35,0x35,0x55,0x2c,0x30,0x78,0x38,0x32,
+    0x37,0x45,0x37,0x45,0x46,0x43,0x55,0x2c,0x30,0x78,0x34,0x37,0x33,0x44,0x33,0x44,0x37,0x41,0x55,0x2c,0x0a,0x30,0x78,0x41,0x43,0x36,0x34,0x36,0x34,0x43,0x38,0x55,
+    0x2c,0x30,0x78,0x45,0x37,0x35,0x44,0x35,0x44,0x42,0x41,0x55,0x2c,0x30,0x78,0x32,0x42,0x31,0x39,0x31,0x39,0x33,0x32,0x55,0x2c,0x30,0x78,0x39,0x35,0x37,0x33,0x37,
+    0x33,0x45,0x36,0x55,0x2c,0x0a,0x30,0x78,0x41,0x30,0x36,0x30,0x36,0x30,0x43,0x30,0x55,0x2c,0x30,0x78,0x39,0x38,0x38,0x31,0x38,0x31,0x31,0x39,0x55,0x2c,0x30,0x78,
+    0x44,0x31,0x34,0x46,0x34,0x46,0x39,0x45,0x55,0x2c,0x30,0x78,0x37,0x46,0x44,0x43,0x44,0x43,0x41,0x33,0x55,0x2c,0x0a,0x30,0x78,0x36,0x36,0x32,0x32,0x32,0x32,0x34,
+    0x34,0x55,0x2c,0x30,0x78,0x37,0x45,0x32,0x41,0x32,0x41,0x35,0x34,0x55,0x2c,0x30,0x78,0x41,0x42,0x39,0x30,0x39,0x30,0x33,0x42,0x55,0x2c,0x30,0x78,0x38,0x33,0x38,
+    0x38,0x38,0x38,0x30,0x42,0x55,0x2c,0x0a,0x30,0x78,0x43,0x41,0x34,0x36,0x34,0x36,0x38,0x43,0x55,0x2c,0x30,0x78,0x32,0x39,0x45,0x45,0x45,0x45,0x43,0x37,0x55,0x2c,
+    0x30,0x78,0x44,0x33,0x42,0x38,0x42,0x38,0x36,0x42,0x55,0x2c,0x30,0x78,0x33,0x43,0x31,0x34,0x31,0x34,0x32,0x38,0x55,0x2c,0x0a,0x30,0x78,0x37,0x39,0x44,0x45,0x44,
+    0x45,0x41,0x37,0x55,0x2c,0x30,0x78,0x45,0x32,0x35,0x45,0x35,0x45,0x42,0x43,0x55,0x2c,0x30,0x78,0x31,0x44,0x30,0x42,0x30,0x42,0x31,0x36,0x55,0x2c,0x30,0x78,0x37,
+    0x36,0x44,0x42,0x44,0x42,0x41,0x44,0x55,0x2c,0x0a,0x30,0x78,0x33,0x42,0x45,0x30,0x45,0x30,0x44,0x42,0x55,0x2c,0x30,0x78,0x35,0x36,0x33,0x32,0x33,0x32,0x36,0x34,
+    0x55,0x2c,0x30,0x78,0x34,0x45,0x33,0x41,0x33,0x41,0x37,0x34,0x55,0x2c,0x30,0x78,0x31,0x45,0x30,0x41,0x30,0x41,0x31,0x34,0x55,0x2c,0x0a,0x30,0x78,0x44,0x42,0x34,
+    0x39,0x34,0x39,0x39,0x32,0x55,0x2c,0x30,0x78,0x30,0x41,0x30,0x36,0x30,0x36,0x30,0x43,0x55,0x2c,0x30,0x78,0x36,0x43,0x32,0x34,0x32,0x34,0x34,0x38,0x55,0x2c,0x30,
+    0x78,0x45,0x34,0x35,0x43,0x35,0x43,0x42,0x38,0x55,0x2c,0x0a,0x30,0x78,0x35,0x44,0x43,0x32,0x43,0x32,0x39,0x46,0x55,0x2c,0x30,0x78,0x36,0x45,0x44,0x33,0x44,0x33,
+    0x42,0x44,0x55,0x2c,0x30,0x78,0x45,0x46,0x41,0x43,0x41,0x43,0x34,0x33,0x55,0x2c,0x30,0x78,0x41,0x36,0x36,0x32,0x36,0x32,0x43,0x34,0x55,0x2c,0x0a,0x30,0x78,0x41,
+    0x38,0x39,0x31,0x39,0x31,0x33,0x39,0x55,0x2c,0x30,0x78,0x41,0x34,0x39,0x35,0x39,0x35,0x33,0x31,0x55,0x2c,0x30,0x78,0x33,0x37,0x45,0x34,0x45,0x34,0x44,0x33,0x55,
+    0x2c,0x30,0x78,0x38,0x42,0x37,0x39,0x37,0x39,0x46,0x32,0x55,0x2c,0x0a,0x30,0x78,0x33,0x32,0x45,0x37,0x45,0x37,0x44,0x35,0x55,0x2c,0x30,0x78,0x34,0x33,0x43,0x38,
+    0x43,0x38,0x38,0x42,0x55,0x2c,0x30,0x78,0x35,0x39,0x33,0x37,0x33,0x37,0x36,0x45,0x55,0x2c,0x30,0x78,0x42,0x37,0x36,0x44,0x36,0x44,0x44,0x41,0x55,0x2c,0x0a,0x30,
+    0x78,0x38,0x43,0x38,0x44,0x38,0x44,0x30,0x31,0x55,0x2c,0x30,0x78,0x36,0x34,0x44,0x35,0x44,0x35,0x42,0x31,0x55,0x2c,0x30,0x78,0x44,0x32,0x34,0x45,0x34,0x45,0x39,
+    0x43,0x55,0x2c,0x30,0x78,0x45,0x30,0x41,0x39,0x41,0x39,0x34,0x39,0x55,0x2c,0x0a,0x30,0x78,0x42,0x34,0x36,0x43,0x36,0x43,0x44,0x38,0x55,0x2c,0x30,0x78,0x46,0x41,
+    0x35,0x36,0x35,0x36,0x41,0x43,0x55,0x2c,0x30,0x78,0x30,0x37,0x46,0x34,0x46,0x34,0x46,0x33,0x55,0x2c,0x30,0x78,0x32,0x35,0x45,0x41,0x45,0x41,0x43,0x46,0x55,0x2c,
+    0x0a,0x30,0x78,0x41,0x46,0x36,0x35,0x36,0x35,0x43,0x41,0x55,0x2c,0x30,0x78,0x38,0x45,0x37,0x41,0x37,0x41,0x46,0x34,0x55,0x2c,0x30,0x78,0x45,0x39,0x41,0x45,0x41,
+    0x45,0x34,0x37,0x55,0x2c,0x30,0x78,0x31,0x38,0x30,0x38,0x30,0x38,0x31,0x30,0x55,0x2c,0x0a,0x30,0x78,0x44,0x35,0x42,0x41,0x42,0x41,0x36,0x46,0x55,0x2c,0x30,0x78,
+    0x38,0x38,0x37,0x38,0x37,0x38,0x46,0x30,0x55,0x2c,0x30,0x78,0x36,0x46,0x32,0x35,0x32,0x35,0x34,0x41,0x55,0x2c,0x30,0x78,0x37,0x32,0x32,0x45,0x32,0x45,0x35,0x43,
+    0x55,0x2c,0x0a,0x30,0x78,0x32,0x34,0x31,0x43,0x31,0x43,0x33,0x38,0x55,0x2c,0x30,0x78,0x46,0x31,0x41,0x36,0x41,0x36,0x35,0x37,0x55,0x2c,0x30,0x78,0x43,0x37,0x42,
+    0x34,0x42,0x34,0x37,0x33,0x55,0x2c,0x30,0x78,0x35,0x31,0x43,0x36,0x43,0x36,0x39,0x37,0x55,0x2c,0x0a,0x30,0x78,0x32,0x33,0x45,0x38,0x45,0x38,0x43,0x42,0x55,0x2c,
+    0x30,0x78,0x37,0x43,0x44,0x44,0x44,0x44,0x41,0x31,0x55,0x2c,0x30,0x78,0x39,0x43,0x37,0x34,0x37,0x34,0x45,0x38,0x55,0x2c,0x30,0x78,0x32,0x31,0x31,0x46,0x31,0x46,
+    0x33,0x45,0x55,0x2c,0x0a,0x30,0x78,0x44,0x44,0x34,0x42,0x34,0x42,0x39,0x36,0x55,0x2c,0x30,0x78,0x44,0x43,0x42,0x44,0x42,0x44,0x36,0x31,0x55,0x2c,0x30,0x78,0x38,
+    0x36,0x38,0x42,0x38,0x42,0x30,0x44,0x55,0x2c,0x30,0x78,0x38,0x35,0x38,0x41,0x38,0x41,0x30,0x46,0x55,0x2c,0x0a,0x30,0x78,0x39,0x30,0x37,0x30,0x37,0x30,0x45,0x30,
+    0x55,0x2c,0x30,0x78,0x34,0x32,0x33,0x45,0x33,0x45,0x37,0x43,0x55,0x2c,0x30,0x78,0x43,0x34,0x42,0x35,0x42,0x35,0x37,0x31,0x55,0x2c,0x30,0x78,0x41,0x41,0x36,0x36,
+    0x36,0x36,0x43,0x43,0x55,0x2c,0x0a,0x30,0x78,0x44,0x38,0x34,0x38,0x34,0x38,0x39,0x30,0x55,0x2c,0x30,0x78,0x30,0x35,0x30,0x33,0x30,0x33,0x30,0x36,0x55,0x2c,0x30,
+    0x78,0x30,0x31,0x46,0x36,0x46,0x36,0x46,0x37,0x55,0x2c,0x30,0x78,0x31,0x32,0x30,0x45,0x30,0x45,0x31,0x43,0x55,0x2c,0x0a,0x30,0x78,0x41,0x33,0x36,0x31,0x36,0x31,
+    0x43,0x32,0x55,0x2c,0x30,0x78,0x35,0x46,0x33,0x35,0x33,0x35,0x36,0x41,0x55,0x2c,0x30,0x78,0x46,0x39,0x35,0x37,0x35,0x37,0x41,0x45,0x55,0x2c,0x30,0x78,0x44,0x30,
+    0x42,0x39,0x42,0x39,0x36,0x39,0x55,0x2c,0x0a,0x30,0x78,0x39,0x31,0x38,0x36,0x38,0x36,0x31,0x37,0x55,0x2c,0x30,0x78,0x35,0x38,0x43,0x31,0x43,0x31,0x39,0x39,0x55,
+    0x2c,0x30,0x78,0x32,0x37,0x31,0x44,0x31,0x44,0x33,0x41,0x55,0x2c,0x30,0x78,0x42,0x39,0x39,0x45,0x39,0x45,0x32,0x37,0x55,0x2c,0x0a,0x30,0x78,0x33,0x38,0x45,0x31,
+    0x45,0x31,0x44,0x39,0x55,0x2c,0x30,0x78,0x31,0x33,0x46,0x38,0x46,0x38,0x45,0x42,0x55,0x2c,0x30,0x78,0x42,0x33,0x39,0x38,0x39,0x38,0x32,0x42,0x55,0x2c,0x30,0x78,
+    0x33,0x33,0x31,0x31,0x31,0x31,0x32,0x32,0x55,0x2c,0x0a,0x30,0x78,0x42,0x42,0x36,0x39,0x36,0x39,0x44,0x32,0x55,0x2c,0x30,0x78,0x37,0x30,0x44,0x39,0x44,0x39,0x41,
+    0x39,0x55,0x2c,0x30,0x78,0x38,0x39,0x38,0x45,0x38,0x45,0x30,0x37,0x55,0x2c,0x30,0x78,0x41,0x37,0x39,0x34,0x39,0x34,0x33,0x33,0x55,0x2c,0x0a,0x30,0x78,0x42,0x36,
+    0x39,0x42,0x39,0x42,0x32,0x44,0x55,0x2c,0x30,0x78,0x32,0x32,0x31,0x45,0x31,0x45,0x33,0x43,0x55,0x2c,0x30,0x78,0x39,0x32,0x38,0x37,0x38,0x37,0x31,0x35,0x55,0x2c,
+    0x30,0x78,0x32,0x30,0x45,0x39,0x45,0x39,0x43,0x39,0x55,0x2c,0x0a,0x30,0x78,0x34,0x39,0x43,0x45,0x43,0x45,0x38,0x37,0x55,0x2c,0x30,0x78,0x46,0x46,0x35,0x35,0x35,
+    0x35,0x41,0x41,0x55,0x2c,0x30,0x78,0x37,0x38,0x32,0x38,0x32,0x38,0x35,0x30,0x55,0x2c,0x30,0x78,0x37,0x41,0x44,0x46,0x44,0x46,0x41,0x35,0x55,0x2c,0x0a,0x30,0x78,
+    0x38,0x46,0x38,0x43,0x38,0x43,0x30,0x33,0x55,0x2c,0x30,0x78,0x46,0x38,0x41,0x31,0x41,0x31,0x35,0x39,0x55,0x2c,0x30,0x78,0x38,0x30,0x38,0x39,0x38,0x39,0x30,0x39,
+    0x55,0x2c,0x30,0x78,0x31,0x37,0x30,0x44,0x30,0x44,0x31,0x41,0x55,0x2c,0x0a,0x30,0x78,0x44,0x41,0x42,0x46,0x42,0x46,0x36,0x35,0x55,0x2c,0x30,0x78,0x33,0x31,0x45,
+    0x36,0x45,0x36,0x44,0x37,0x55,0x2c,0x30,0x78,0x43,0x36,0x34,0x32,0x34,0x32,0x38,0x34,0x55,0x2c,0x30,0x78,0x42,0x38,0x36,0x38,0x36,0x38,0x44,0x30,0x55,0x2c,0x0a,
+    0x30,0x78,0x43,0x33,0x34,0x31,0x34,0x31,0x38,0x32,0x55,0x2c,0x30,0x78,0x42,0x30,0x39,0x39,0x39,0x39,0x32,0x39,0x55,0x2c,0x30,0x78,0x37,0x37,0x32,0x44,0x32,0x44,
+    0x35,0x41,0x55,0x2c,0x30,0x78,0x31,0x31,0x30,0x46,0x30,0x46,0x31,0x45,0x55,0x2c,0x0a,0x30,0x78,0x43,0x42,0x42,0x30,0x42,0x30,0x37,0x42,0x55,0x2c,0x30,0x78,0x46,
+    0x43,0x35,0x34,0x35,0x34,0x41,0x38,0x55,0x2c,0x30,0x78,0x44,0x36,0x42,0x42,0x42,0x42,0x36,0x44,0x55,0x2c,0x30,0x78,0x33,0x41,0x31,0x36,0x31,0x36,0x32,0x43,0x55,
+    0x0a,0x7d,0x3b,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x42,0x59,0x54,0x45,0x28,0x78,0x2c,0x20,0x79,0x29,0x20,0x28,0x78,0x6d,0x72,0x69,0x67,0x5f,0x61,0x6d,
+    0x64,0x5f,0x62,0x66,0x65,0x28,0x28,0x78,0x29,0x2c,0x20,0x28,0x79,0x29,0x20,0x3c,0x3c,0x20,0x33,0x55,0x2c,0x20,0x38,0x55,0x29,0x29,0x0a,0x23,0x69,0x66,0x20,0x28,
+    0x41,0x4c,0x47,0x4f,0x20,0x3d,0x3d,0x20,0x41,0x4c,0x47,0x4f,0x5f,0x43,0x4e,0x5f,0x48,0x45,0x41,0x56,0x59,0x5f,0x54,0x55,0x42,0x45,0x29,0x0a,0x69,0x6e,0x6c,0x69,
+    0x6e,0x65,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x5f,0x62,0x69,0x74,0x74,0x75,0x62,0x65,0x32,0x28,0x63,0x6f,0x6e,0x73,
+    0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x30,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,
+    0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x31,0x2c,0x75,0x69,0x6e,0x74,0x34,0x20,0x78,0x2c,0x75,0x69,0x6e,0x74,0x34,0x20,0x6b,0x29,0x0a,0x7b,
+    0x0a,0x78,0x3d,0x7e,0x78,0x3b,0x0a,0x6b,0x2e,0x73,0x30,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x30,0x2c,0x30,0x29,
+    0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x31,0x2c,0x31,0x29,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,
+    0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x32,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x33,0x2c,0x33,0x29,
+    0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x78,0x2e,0x73,0x30,0x20,0x5e,0x3d,0x20,0x6b,0x2e,0x73,0x30,0x3b,0x0a,0x6b,0x2e,0x73,0x31,0x20,0x5e,0x3d,0x20,0x41,0x45,
+    0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x31,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x32,0x2c,
+    0x31,0x29,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x33,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,
+    0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x30,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x78,0x2e,0x73,0x31,0x20,0x5e,0x3d,0x20,0x6b,
+    0x2e,0x73,0x31,0x3b,0x0a,0x6b,0x2e,0x73,0x32,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x32,0x2c,0x30,0x29,0x5d,0x5e,
+    0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x33,0x2c,0x31,0x29,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,
+    0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x30,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x31,0x2c,0x33,0x29,0x5d,0x2c,
+    0x31,0x36,0x55,0x29,0x3b,0x0a,0x78,0x2e,0x73,0x32,0x20,0x5e,0x3d,0x20,0x6b,0x2e,0x73,0x32,0x3b,0x0a,0x6b,0x2e,0x73,0x33,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,
+    0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x33,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x30,0x2c,0x31,0x29,
+    0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x31,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,
+    0x5b,0x42,0x59,0x54,0x45,0x28,0x78,0x2e,0x73,0x32,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x6b,0x3b,0x0a,0x7d,
+    0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x75,0x69,0x6e,0x74,0x34,0x20,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x28,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,
+    0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x30,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,
+    0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x31,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,
+    0x53,0x32,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x33,0x2c,0x63,0x6f,0x6e,0x73,
+    0x74,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x58,0x2c,0x75,0x69,0x6e,0x74,0x34,0x20,0x6b,0x65,0x79,0x29,0x0a,0x7b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x30,0x20,0x5e,0x3d,
+    0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x30,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,
+    0x73,0x31,0x2c,0x31,0x29,0x5d,0x5e,0x41,0x45,0x53,0x32,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x33,0x5b,0x42,
+    0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x33,0x29,0x5d,0x3b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x31,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,
+    0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x31,0x29,0x5d,0x5e,0x41,0x45,
+    0x53,0x32,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x33,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x30,0x2c,
+    0x33,0x29,0x5d,0x3b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x32,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x30,0x29,
+    0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x31,0x29,0x5d,0x5e,0x41,0x45,0x53,0x32,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,
+    0x2e,0x73,0x30,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x33,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x33,0x29,0x5d,0x3b,0x0a,0x6b,0x65,0x79,0x2e,
+    0x73,0x33,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,
+    0x54,0x45,0x28,0x58,0x2e,0x73,0x30,0x2c,0x31,0x29,0x5d,0x5e,0x41,0x45,0x53,0x32,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x32,0x29,0x5d,0x5e,0x41,
+    0x45,0x53,0x33,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x33,0x29,0x5d,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x6b,0x65,0x79,0x3b,0x0a,0x7d,
+    0x0a,0x75,0x69,0x6e,0x74,0x34,0x20,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x5f,0x54,0x77,0x6f,0x5f,0x54,0x61,0x62,0x6c,0x65,0x73,0x28,0x63,0x6f,0x6e,0x73,
+    0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x30,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x6c,0x6f,0x63,
+    0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x41,0x45,0x53,0x31,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x58,0x2c,0x75,0x69,0x6e,0x74,
+    0x34,0x20,0x6b,0x65,0x79,0x29,0x0a,0x7b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x30,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,
+    0x30,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x31,0x29,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,
+    0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,
+    0x33,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x31,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,
+    0x58,0x2e,0x73,0x31,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x31,0x29,0x5d,0x5e,0x72,0x6f,0x74,0x61,
+    0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,
+    0x58,0x2e,0x73,0x30,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x32,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,
+    0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x31,0x29,0x5d,0x5e,0x72,
+    0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x30,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,
+    0x54,0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x6b,0x65,0x79,0x2e,0x73,0x33,0x20,0x5e,0x3d,0x20,0x41,0x45,0x53,0x30,
+    0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x33,0x2c,0x30,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x30,0x2c,0x31,0x29,
+    0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x41,0x45,0x53,0x30,0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x31,0x2c,0x32,0x29,0x5d,0x5e,0x41,0x45,0x53,0x31,
+    0x5b,0x42,0x59,0x54,0x45,0x28,0x58,0x2e,0x73,0x32,0x2c,0x33,0x29,0x5d,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x6b,0x65,0x79,0x3b,
+    0x0a,0x7d,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x63,0x68,0x61,0x72,
+    0x20,0x72,0x63,0x6f,0x6e,0x5b,0x38,0x5d,0x3d,0x7b,0x20,0x30,0x78,0x38,0x64,0x2c,0x30,0x78,0x30,0x31,0x2c,0x30,0x78,0x30,0x32,0x2c,0x30,0x78,0x30,0x34,0x2c,0x30,
+    0x78,0x30,0x38,0x2c,0x30,0x78,0x31,0x30,0x2c,0x30,0x78,0x32,0x30,0x2c,0x30,0x78,0x34,0x30,0x20,0x7d,0x3b,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,
+    0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x63,0x68,0x61,0x72,0x20,0x73,0x62,0x6f,0x78,0x5b,0x32,0x35,0x36,0x5d,0x20,0x3d,0x0a,
+    0x7b,0x0a,0x30,0x78,0x36,0x33,0x2c,0x30,0x78,0x37,0x43,0x2c,0x30,0x78,0x37,0x37,0x2c,0x30,0x78,0x37,0x42,0x2c,0x30,0x78,0x46,0x32,0x2c,0x30,0x78,0x36,0x42,0x2c,
+    0x30,0x78,0x36,0x46,0x2c,0x30,0x78,0x43,0x35,0x2c,0x30,0x78,0x33,0x30,0x2c,0x30,0x78,0x30,0x31,0x2c,0x30,0x78,0x36,0x37,0x2c,0x30,0x78,0x32,0x42,0x2c,0x30,0x78,
+    0x46,0x45,0x2c,0x30,0x78,0x44,0x37,0x2c,0x30,0x78,0x41,0x42,0x2c,0x30,0x78,0x37,0x36,0x2c,0x0a,0x30,0x78,0x43,0x41,0x2c,0x30,0x78,0x38,0x32,0x2c,0x30,0x78,0x43,
+    0x39,0x2c,0x30,0x78,0x37,0x44,0x2c,0x30,0x78,0x46,0x41,0x2c,0x30,0x78,0x35,0x39,0x2c,0x30,0x78,0x34,0x37,0x2c,0x30,0x78,0x46,0x30,0x2c,0x30,0x78,0x41,0x44,0x2c,
+    0x30,0x78,0x44,0x34,0x2c,0x30,0x78,0x41,0x32,0x2c,0x30,0x78,0x41,0x46,0x2c,0x30,0x78,0x39,0x43,0x2c,0x30,0x78,0x41,0x34,0x2c,0x30,0x78,0x37,0x32,0x2c,0x30,0x78,
+    0x43,0x30,0x2c,0x0a,0x30,0x78,0x42,0x37,0x2c,0x30,0x78,0x46,0x44,0x2c,0x30,0x78,0x39,0x33,0x2c,0x30,0x78,0x32,0x36,0x2c,0x30,0x78,0x33,0x36,0x2c,0x30,0x78,0x33,
+    0x46,0x2c,0x30,0x78,0x46,0x37,0x2c,0x30,0x78,0x43,0x43,0x2c,0x30,0x78,0x33,0x34,0x2c,0x30,0x78,0x41,0x35,0x2c,0x30,0x78,0x45,0x35,0x2c,0x30,0x78,0x46,0x31,0x2c,
+    0x30,0x78,0x37,0x31,0x2c,0x30,0x78,0x44,0x38,0x2c,0x30,0x78,0x33,0x31,0x2c,0x30,0x78,0x31,0x35,0x2c,0x0a,0x30,0x78,0x30,0x34,0x2c,0x30,0x78,0x43,0x37,0x2c,0x30,
+    0x78,0x32,0x33,0x2c,0x30,0x78,0x43,0x33,0x2c,0x30,0x78,0x31,0x38,0x2c,0x30,0x78,0x39,0x36,0x2c,0x30,0x78,0x30,0x35,0x2c,0x30,0x78,0x39,0x41,0x2c,0x30,0x78,0x30,
+    0x37,0x2c,0x30,0x78,0x31,0x32,0x2c,0x30,0x78,0x38,0x30,0x2c,0x30,0x78,0x45,0x32,0x2c,0x30,0x78,0x45,0x42,0x2c,0x30,0x78,0x32,0x37,0x2c,0x30,0x78,0x42,0x32,0x2c,
+    0x30,0x78,0x37,0x35,0x2c,0x0a,0x30,0x78,0x30,0x39,0x2c,0x30,0x78,0x38,0x33,0x2c,0x30,0x78,0x32,0x43,0x2c,0x30,0x78,0x31,0x41,0x2c,0x30,0x78,0x31,0x42,0x2c,0x30,
+    0x78,0x36,0x45,0x2c,0x30,0x78,0x35,0x41,0x2c,0x30,0x78,0x41,0x30,0x2c,0x30,0x78,0x35,0x32,0x2c,0x30,0x78,0x33,0x42,0x2c,0x30,0x78,0x44,0x36,0x2c,0x30,0x78,0x42,
+    0x33,0x2c,0x30,0x78,0x32,0x39,0x2c,0x30,0x78,0x45,0x33,0x2c,0x30,0x78,0x32,0x46,0x2c,0x30,0x78,0x38,0x34,0x2c,0x0a,0x30,0x78,0x35,0x33,0x2c,0x30,0x78,0x44,0x31,
+    0x2c,0x30,0x78,0x30,0x30,0x2c,0x30,0x78,0x45,0x44,0x2c,0x30,0x78,0x32,0x30,0x2c,0x30,0x78,0x46,0x43,0x2c,0x30,0x78,0x42,0x31,0x2c,0x30,0x78,0x35,0x42,0x2c,0x30,
+    0x78,0x36,0x41,0x2c,0x30,0x78,0x43,0x42,0x2c,0x30,0x78,0x42,0x45,0x2c,0x30,0x78,0x33,0x39,0x2c,0x30,0x78,0x34,0x41,0x2c,0x30,0x78,0x34,0x43,0x2c,0x30,0x78,0x35,
+    0x38,0x2c,0x30,0x78,0x43,0x46,0x2c,0x0a,0x30,0x78,0x44,0x30,0x2c,0x30,0x78,0x45,0x46,0x2c,0x30,0x78,0x41,0x41,0x2c,0x30,0x78,0x46,0x42,0x2c,0x30,0x78,0x34,0x33,
+    0x2c,0x30,0x78,0x34,0x44,0x2c,0x30,0x78,0x33,0x33,0x2c,0x30,0x78,0x38,0x35,0x2c,0x30,0x78,0x34,0x35,0x2c,0x30,0x78,0x46,0x39,0x2c,0x30,0x78,0x30,0x32,0x2c,0x30,
+    0x78,0x37,0x46,0x2c,0x30,0x78,0x35,0x30,0x2c,0x30,0x78,0x33,0x43,0x2c,0x30,0x78,0x39,0x46,0x2c,0x30,0x78,0x41,0x38,0x2c,0x0a,0x30,0x78,0x35,0x31,0x2c,0x30,0x78,
+    0x41,0x33,0x2c,0x30,0x78,0x34,0x30,0x2c,0x30,0x78,0x38,0x46,0x2c,0x30,0x78,0x39,0x32,0x2c,0x30,0x78,0x39,0x44,0x2c,0x30,0x78,0x33,0x38,0x2c,0x30,0x78,0x46,0x35,
+    0x2c,0x30,0x78,0x42,0x43,0x2c,0x30,0x78,0x42,0x36,0x2c,0x30,0x78,0x44,0x41,0x2c,0x30,0x78,0x32,0x31,0x2c,0x30,0x78,0x31,0x30,0x2c,0x30,0x78,0x46,0x46,0x2c,0x30,
+    0x78,0x46,0x33,0x2c,0x30,0x78,0x44,0x32,0x2c,0x0a,0x30,0x78,0x43,0x44,0x2c,0x30,0x78,0x30,0x43,0x2c,0x30,0x78,0x31,0x33,0x2c,0x30,0x78,0x45,0x43,0x2c,0x30,0x78,
+    0x35,0x46,0x2c,0x30,0x78,0x39,0x37,0x2c,0x30,0x78,0x34,0x34,0x2c,0x30,0x78,0x31,0x37,0x2c,0x30,0x78,0x43,0x34,0x2c,0x30,0x78,0x41,0x37,0x2c,0x30,0x78,0x37,0x45,
+    0x2c,0x30,0x78,0x33,0x44,0x2c,0x30,0x78,0x36,0x34,0x2c,0x30,0x78,0x35,0x44,0x2c,0x30,0x78,0x31,0x39,0x2c,0x30,0x78,0x37,0x33,0x2c,0x0a,0x30,0x78,0x36,0x30,0x2c,
+    0x30,0x78,0x38,0x31,0x2c,0x30,0x78,0x34,0x46,0x2c,0x30,0x78,0x44,0x43,0x2c,0x30,0x78,0x32,0x32,0x2c,0x30,0x78,0x32,0x41,0x2c,0x30,0x78,0x39,0x30,0x2c,0x30,0x78,
+    0x38,0x38,0x2c,0x30,0x78,0x34,0x36,0x2c,0x30,0x78,0x45,0x45,0x2c,0x30,0x78,0x42,0x38,0x2c,0x30,0x78,0x31,0x34,0x2c,0x30,0x78,0x44,0x45,0x2c,0x30,0x78,0x35,0x45,
+    0x2c,0x30,0x78,0x30,0x42,0x2c,0x30,0x78,0x44,0x42,0x2c,0x0a,0x30,0x78,0x45,0x30,0x2c,0x30,0x78,0x33,0x32,0x2c,0x30,0x78,0x33,0x41,0x2c,0x30,0x78,0x30,0x41,0x2c,
+    0x30,0x78,0x34,0x39,0x2c,0x30,0x78,0x30,0x36,0x2c,0x30,0x78,0x32,0x34,0x2c,0x30,0x78,0x35,0x43,0x2c,0x30,0x78,0x43,0x32,0x2c,0x30,0x78,0x44,0x33,0x2c,0x30,0x78,
+    0x41,0x43,0x2c,0x30,0x78,0x36,0x32,0x2c,0x30,0x78,0x39,0x31,0x2c,0x30,0x78,0x39,0x35,0x2c,0x30,0x78,0x45,0x34,0x2c,0x30,0x78,0x37,0x39,0x2c,0x0a,0x30,0x78,0x45,
+    0x37,0x2c,0x30,0x78,0x43,0x38,0x2c,0x30,0x78,0x33,0x37,0x2c,0x30,0x78,0x36,0x44,0x2c,0x30,0x78,0x38,0x44,0x2c,0x30,0x78,0x44,0x35,0x2c,0x30,0x78,0x34,0x45,0x2c,
+    0x30,0x78,0x41,0x39,0x2c,0x30,0x78,0x36,0x43,0x2c,0x30,0x78,0x35,0x36,0x2c,0x30,0x78,0x46,0x34,0x2c,0x30,0x78,0x45,0x41,0x2c,0x30,0x78,0x36,0x35,0x2c,0x30,0x78,
+    0x37,0x41,0x2c,0x30,0x78,0x41,0x45,0x2c,0x30,0x78,0x30,0x38,0x2c,0x0a,0x30,0x78,0x42,0x41,0x2c,0x30,0x78,0x37,0x38,0x2c,0x30,0x78,0x32,0x35,0x2c,0x30,0x78,0x32,
+    0x45,0x2c,0x30,0x78,0x31,0x43,0x2c,0x30,0x78,0x41,0x36,0x2c,0x30,0x78,0x42,0x34,0x2c,0x30,0x78,0x43,0x36,0x2c,0x30,0x78,0x45,0x38,0x2c,0x30,0x78,0x44,0x44,0x2c,
+    0x30,0x78,0x37,0x34,0x2c,0x30,0x78,0x31,0x46,0x2c,0x30,0x78,0x34,0x42,0x2c,0x30,0x78,0x42,0x44,0x2c,0x30,0x78,0x38,0x42,0x2c,0x30,0x78,0x38,0x41,0x2c,0x0a,0x30,
+    0x78,0x37,0x30,0x2c,0x30,0x78,0x33,0x45,0x2c,0x30,0x78,0x42,0x35,0x2c,0x30,0x78,0x36,0x36,0x2c,0x30,0x78,0x34,0x38,0x2c,0x30,0x78,0x30,0x33,0x2c,0x30,0x78,0x46,
+    0x36,0x2c,0x30,0x78,0x30,0x45,0x2c,0x30,0x78,0x36,0x31,0x2c,0x30,0x78,0x33,0x35,0x2c,0x30,0x78,0x35,0x37,0x2c,0x30,0x78,0x42,0x39,0x2c,0x30,0x78,0x38,0x36,0x2c,
+    0x30,0x78,0x43,0x31,0x2c,0x30,0x78,0x31,0x44,0x2c,0x30,0x78,0x39,0x45,0x2c,0x0a,0x30,0x78,0x45,0x31,0x2c,0x30,0x78,0x46,0x38,0x2c,0x30,0x78,0x39,0x38,0x2c,0x30,
+    0x78,0x31,0x31,0x2c,0x30,0x78,0x36,0x39,0x2c,0x30,0x78,0x44,0x39,0x2c,0x30,0x78,0x38,0x45,0x2c,0x30,0x78,0x39,0x34,0x2c,0x30,0x78,0x39,0x42,0x2c,0x30,0x78,0x31,
+    0x45,0x2c,0x30,0x78,0x38,0x37,0x2c,0x30,0x78,0x45,0x39,0x2c,0x30,0x78,0x43,0x45,0x2c,0x30,0x78,0x35,0x35,0x2c,0x30,0x78,0x32,0x38,0x2c,0x30,0x78,0x44,0x46,0x2c,
+    0x0a,0x30,0x78,0x38,0x43,0x2c,0x30,0x78,0x41,0x31,0x2c,0x30,0x78,0x38,0x39,0x2c,0x30,0x78,0x30,0x44,0x2c,0x30,0x78,0x42,0x46,0x2c,0x30,0x78,0x45,0x36,0x2c,0x30,
+    0x78,0x34,0x32,0x2c,0x30,0x78,0x36,0x38,0x2c,0x30,0x78,0x34,0x31,0x2c,0x30,0x78,0x39,0x39,0x2c,0x30,0x78,0x32,0x44,0x2c,0x30,0x78,0x30,0x46,0x2c,0x30,0x78,0x42,
+    0x30,0x2c,0x30,0x78,0x35,0x34,0x2c,0x30,0x78,0x42,0x42,0x2c,0x30,0x78,0x31,0x36,0x0a,0x7d,0x3b,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x53,0x75,0x62,0x57,
+    0x6f,0x72,0x64,0x28,0x69,0x6e,0x77,0x29,0x20,0x28,0x28,0x73,0x62,0x6f,0x78,0x5b,0x42,0x59,0x54,0x45,0x28,0x69,0x6e,0x77,0x2c,0x20,0x33,0x29,0x5d,0x20,0x3c,0x3c,
+    0x20,0x32,0x34,0x29,0x20,0x7c,0x20,0x28,0x73,0x62,0x6f,0x78,0x5b,0x42,0x59,0x54,0x45,0x28,0x69,0x6e,0x77,0x2c,0x20,0x32,0x29,0x5d,0x20,0x3c,0x3c,0x20,0x31,0x36,
+    0x29,0x20,0x7c,0x20,0x28,0x73,0x62,0x6f,0x78,0x5b,0x42,0x59,0x54,0x45,0x28,0x69,0x6e,0x77,0x2c,0x20,0x31,0x29,0x5d,0x20,0x3c,0x3c,0x20,0x38,0x29,0x20,0x7c,0x20,
+    0x73,0x62,0x6f,0x78,0x5b,0x42,0x59,0x54,0x45,0x28,0x69,0x6e,0x77,0x2c,0x20,0x30,0x29,0x5d,0x29,0x0a,0x76,0x6f,0x69,0x64,0x20,0x41,0x45,0x53,0x45,0x78,0x70,0x61,
+    0x6e,0x64,0x4b,0x65,0x79,0x32,0x35,0x36,0x28,0x75,0x69,0x6e,0x74,0x20,0x2a,0x6b,0x65,0x79,0x62,0x75,0x66,0x29,0x0a,0x7b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x75,0x69,
+    0x6e,0x74,0x20,0x63,0x3d,0x38,0x2c,0x69,0x3d,0x31,0x3b,0x20,0x63,0x3c,0x34,0x30,0x3b,0x20,0x2b,0x2b,0x63,0x29,0x20,0x7b,0x0a,0x75,0x69,0x6e,0x74,0x20,0x74,0x3d,
+    0x28,0x28,0x21,0x28,0x63,0x26,0x37,0x29,0x29,0x7c,0x7c,0x28,0x28,0x63,0x26,0x37,0x29,0x3d,0x3d,0x34,0x29,0x29,0x3f,0x53,0x75,0x62,0x57,0x6f,0x72,0x64,0x28,0x6b,
+    0x65,0x79,0x62,0x75,0x66,0x5b,0x63,0x2d,0x31,0x5d,0x29,0x3a,0x6b,0x65,0x79,0x62,0x75,0x66,0x5b,0x63,0x2d,0x31,0x5d,0x3b,0x0a,0x6b,0x65,0x79,0x62,0x75,0x66,0x5b,
+    0x63,0x5d,0x3d,0x6b,0x65,0x79,0x62,0x75,0x66,0x5b,0x63,0x2d,0x38,0x5d,0x5e,0x28,0x28,0x21,0x28,0x63,0x26,0x37,0x29,0x29,0x3f,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,
+    0x74,0x2c,0x32,0x34,0x55,0x29,0x5e,0x61,0x73,0x5f,0x75,0x69,0x6e,0x74,0x28,0x28,0x75,0x63,0x68,0x61,0x72,0x34,0x29,0x28,0x72,0x63,0x6f,0x6e,0x5b,0x69,0x2b,0x2b,
+    0x5d,0x2c,0x30,0x55,0x2c,0x30,0x55,0x2c,0x30,0x55,0x29,0x29,0x3a,0x74,0x29,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x23,0x69,0x66,0x6e,
+    0x64,0x65,0x66,0x20,0x58,0x4d,0x52,0x49,0x47,0x5f,0x4b,0x45,0x43,0x43,0x41,0x4b,0x5f,0x43,0x4c,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x58,0x4d,0x52,0x49,
+    0x47,0x5f,0x4b,0x45,0x43,0x43,0x41,0x4b,0x5f,0x43,0x4c,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,
+    0x61,0x6e,0x74,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,0x6e,0x64,0x63,0x5b,0x32,0x34,0x5d,0x20,0x3d,0x0a,0x7b,0x0a,0x30,
+    0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x31,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,
+    0x38,0x30,0x38,0x32,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x38,0x61,0x2c,0x0a,0x30,0x78,0x38,0x30,0x30,0x30,0x30,
+    0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x38,0x62,0x2c,0x30,
+    0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x31,0x2c,0x0a,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,
+    0x30,0x38,0x30,0x38,0x31,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x39,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x61,0x2c,0x0a,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x38,0x2c,
+    0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x38,0x30,0x30,0x39,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x61,0x2c,0x0a,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x38,0x30,0x38,0x62,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x62,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x38,0x39,0x2c,
+    0x0a,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x33,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,
+    0x30,0x30,0x38,0x30,0x30,0x32,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x2c,0x0a,0x30,0x78,0x30,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x61,0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x61,
+    0x2c,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x38,0x30,0x38,0x31,0x2c,0x0a,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,
+    0x30,0x30,0x30,0x38,0x30,0x38,0x30,0x2c,0x30,0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x31,0x2c,0x30,0x78,0x38,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x30,0x38,0x30,0x30,0x30,0x38,0x30,0x30,0x38,0x0a,0x7d,0x3b,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,
+    0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,0x6f,0x74,0x63,0x5b,0x32,0x34,0x5d,0x20,0x3d,
+    0x0a,0x7b,0x0a,0x31,0x2c,0x33,0x2c,0x36,0x2c,0x31,0x30,0x2c,0x31,0x35,0x2c,0x32,0x31,0x2c,0x32,0x38,0x2c,0x33,0x36,0x2c,0x34,0x35,0x2c,0x35,0x35,0x2c,0x32,0x2c,
+    0x31,0x34,0x2c,0x0a,0x32,0x37,0x2c,0x34,0x31,0x2c,0x35,0x36,0x2c,0x38,0x2c,0x32,0x35,0x2c,0x34,0x33,0x2c,0x36,0x32,0x2c,0x31,0x38,0x2c,0x33,0x39,0x2c,0x36,0x31,
+    0x2c,0x32,0x30,0x2c,0x34,0x34,0x0a,0x7d,0x3b,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,
+    0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x70,0x69,0x6c,0x6e,0x5b,0x32,0x34,0x5d,0x20,0x3d,0x0a,0x7b,0x0a,0x31,0x30,0x2c,0x37,
+    0x2c,0x31,0x31,0x2c,0x31,0x37,0x2c,0x31,0x38,0x2c,0x33,0x2c,0x35,0x2c,0x31,0x36,0x2c,0x38,0x2c,0x32,0x31,0x2c,0x32,0x34,0x2c,0x34,0x2c,0x0a,0x31,0x35,0x2c,0x32,
+    0x33,0x2c,0x31,0x39,0x2c,0x31,0x33,0x2c,0x31,0x32,0x2c,0x32,0x2c,0x32,0x30,0x2c,0x31,0x34,0x2c,0x32,0x32,0x2c,0x39,0x2c,0x36,0x2c,0x31,0x0a,0x7d,0x3b,0x0a,0x76,
+    0x6f,0x69,0x64,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x31,0x36,0x30,0x30,0x5f,0x31,0x28,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x73,0x74,0x29,0x0a,0x7b,0x0a,0x69,
+    0x6e,0x74,0x20,0x69,0x2c,0x72,0x6f,0x75,0x6e,0x64,0x3b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x74,0x2c,0x62,0x63,0x5b,0x35,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,
+    0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x72,0x6f,0x75,0x6e,0x64,0x3d,0x30,0x3b,0x20,0x72,0x6f,0x75,0x6e,0x64,0x3c,
+    0x32,0x34,0x3b,0x20,0x2b,0x2b,0x72,0x6f,0x75,0x6e,0x64,0x29,0x20,0x7b,0x0a,0x62,0x63,0x5b,0x30,0x5d,0x3d,0x73,0x74,0x5b,0x30,0x5d,0x5e,0x73,0x74,0x5b,0x35,0x5d,
+    0x5e,0x73,0x74,0x5b,0x31,0x30,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x35,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x30,0x5d,0x3b,0x0a,0x62,0x63,0x5b,0x31,0x5d,0x3d,0x73,0x74,0x5b,
+    0x31,0x5d,0x5e,0x73,0x74,0x5b,0x36,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x31,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x36,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x31,0x5d,0x3b,0x0a,0x62,
+    0x63,0x5b,0x32,0x5d,0x3d,0x73,0x74,0x5b,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x37,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x37,0x5d,0x5e,0x73,
+    0x74,0x5b,0x32,0x32,0x5d,0x3b,0x0a,0x62,0x63,0x5b,0x33,0x5d,0x3d,0x73,0x74,0x5b,0x33,0x5d,0x5e,0x73,0x74,0x5b,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x33,0x5d,0x5e,
+    0x73,0x74,0x5b,0x31,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x33,0x5d,0x3b,0x0a,0x62,0x63,0x5b,0x34,0x5d,0x3d,0x73,0x74,0x5b,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x39,0x5d,
+    0x5e,0x73,0x74,0x5b,0x31,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x39,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x34,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,
+    0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x35,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,0x0a,0x74,0x3d,
+    0x62,0x63,0x5b,0x28,0x69,0x2b,0x34,0x29,0x20,0x25,0x20,0x35,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x62,0x63,0x5b,0x28,0x69,0x2b,0x31,0x29,0x20,0x25,0x20,
+    0x35,0x5d,0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x20,0x5d,0x20,0x5e,0x3d,0x20,0x74,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x35,0x5d,0x20,0x5e,0x3d,
+    0x20,0x74,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x31,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x74,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x31,0x35,0x5d,0x20,0x5e,0x3d,0x20,0x74,
+    0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x32,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x74,0x3b,0x0a,0x7d,0x0a,0x74,0x3d,0x73,0x74,0x5b,0x31,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,
+    0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x32,0x34,0x3b,0x20,0x2b,0x2b,0x69,
+    0x29,0x20,0x7b,0x0a,0x62,0x63,0x5b,0x30,0x5d,0x3d,0x73,0x74,0x5b,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x70,0x69,0x6c,0x6e,0x5b,0x69,0x5d,0x5d,0x3b,0x0a,0x73,
+    0x74,0x5b,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x70,0x69,0x6c,0x6e,0x5b,0x69,0x5d,0x5d,0x3d,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x74,0x2c,0x28,0x75,0x6c,0x6f,
+    0x6e,0x67,0x29,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,0x6f,0x74,0x63,0x5b,0x69,0x5d,0x29,0x3b,0x0a,0x74,0x3d,0x62,0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x7d,0x0a,
+    0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,
+    0x3c,0x32,0x35,0x3b,0x20,0x69,0x2b,0x3d,0x35,0x29,0x20,0x7b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x74,0x6d,0x70,0x5b,0x35,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,
+    0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x78,0x3d,0x30,0x3b,0x20,0x78,0x3c,0x35,0x3b,0x20,0x2b,
+    0x2b,0x78,0x29,0x20,0x7b,0x0a,0x74,0x6d,0x70,0x5b,0x78,0x5d,0x3d,0x62,0x69,0x74,0x73,0x65,0x6c,0x65,0x63,0x74,0x28,0x73,0x74,0x5b,0x69,0x2b,0x78,0x5d,0x5e,0x73,
+    0x74,0x5b,0x69,0x2b,0x28,0x28,0x78,0x2b,0x32,0x29,0x20,0x25,0x20,0x35,0x29,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x78,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x28,0x28,
+    0x78,0x2b,0x31,0x29,0x20,0x25,0x20,0x35,0x29,0x5d,0x29,0x3b,0x0a,0x7d,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,
+    0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x78,0x3d,0x30,0x3b,0x20,0x78,0x3c,0x35,0x3b,0x20,0x2b,0x2b,0x78,0x29,0x20,0x7b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x78,
+    0x5d,0x3d,0x74,0x6d,0x70,0x5b,0x78,0x5d,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x73,0x74,0x5b,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,
+    0x6e,0x64,0x63,0x5b,0x72,0x6f,0x75,0x6e,0x64,0x5d,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x76,0x6f,0x69,0x64,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x31,0x36,0x30,0x30,
+    0x5f,0x32,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x73,0x74,0x29,0x0a,0x7b,0x0a,0x69,0x6e,0x74,0x20,0x69,0x2c,0x72,0x6f,
+    0x75,0x6e,0x64,0x3b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x74,0x2c,0x62,0x63,0x5b,0x35,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,
+    0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x72,0x6f,0x75,0x6e,0x64,0x3d,0x30,0x3b,0x20,0x72,0x6f,0x75,0x6e,0x64,0x3c,0x32,0x34,0x3b,0x20,0x2b,0x2b,0x72,
+    0x6f,0x75,0x6e,0x64,0x29,0x20,0x7b,0x0a,0x62,0x63,0x5b,0x30,0x5d,0x3d,0x73,0x74,0x5b,0x30,0x5d,0x5e,0x73,0x74,0x5b,0x35,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x30,0x5d,
+    0x5e,0x73,0x74,0x5b,0x31,0x35,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x30,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x73,0x74,0x5b,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x37,
+    0x5d,0x5e,0x73,0x74,0x5b,0x31,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x37,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x32,0x5d,0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x62,0x63,0x5b,
+    0x31,0x5d,0x3d,0x73,0x74,0x5b,0x31,0x5d,0x5e,0x73,0x74,0x5b,0x36,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x31,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x36,0x5d,0x5e,0x73,0x74,0x5b,
+    0x32,0x31,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x73,0x74,0x5b,0x33,0x5d,0x5e,0x73,0x74,0x5b,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x33,0x5d,0x5e,0x73,0x74,
+    0x5b,0x31,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x33,0x5d,0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x62,0x63,0x5b,0x32,0x5d,0x3d,0x73,0x74,0x5b,0x32,0x5d,0x5e,0x73,0x74,
+    0x5b,0x37,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x37,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x32,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,
+    0x73,0x74,0x5b,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x39,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x39,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x34,0x5d,
+    0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x62,0x63,0x5b,0x33,0x5d,0x3d,0x73,0x74,0x5b,0x33,0x5d,0x5e,0x73,0x74,0x5b,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x33,0x5d,0x5e,
+    0x73,0x74,0x5b,0x31,0x38,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x33,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x73,0x74,0x5b,0x30,0x5d,0x5e,0x73,0x74,0x5b,0x35,0x5d,
+    0x5e,0x73,0x74,0x5b,0x31,0x30,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x35,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x30,0x5d,0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x62,0x63,0x5b,0x34,
+    0x5d,0x3d,0x73,0x74,0x5b,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x39,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x34,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x39,0x5d,0x5e,0x73,0x74,0x5b,0x32,
+    0x34,0x5d,0x5e,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x73,0x74,0x5b,0x31,0x5d,0x5e,0x73,0x74,0x5b,0x36,0x5d,0x5e,0x73,0x74,0x5b,0x31,0x31,0x5d,0x5e,0x73,0x74,0x5b,
+    0x31,0x36,0x5d,0x5e,0x73,0x74,0x5b,0x32,0x31,0x5d,0x2c,0x31,0x55,0x4c,0x29,0x3b,0x0a,0x73,0x74,0x5b,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x34,0x5d,0x3b,
+    0x0a,0x73,0x74,0x5b,0x35,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x34,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x34,0x5d,
+    0x3b,0x0a,0x73,0x74,0x5b,0x31,0x35,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x34,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x32,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,
+    0x34,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x36,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,
+    0x30,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x31,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x36,0x5d,0x20,0x5e,0x3d,0x20,0x62,
+    0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x32,0x31,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x32,0x5d,0x20,0x5e,0x3d,0x20,
+    0x62,0x63,0x5b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x37,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x32,0x5d,0x20,0x5e,0x3d,
+    0x20,0x62,0x63,0x5b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x37,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x32,0x32,0x5d,0x20,
+    0x5e,0x3d,0x20,0x62,0x63,0x5b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x33,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x32,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x38,0x5d,0x20,
+    0x5e,0x3d,0x20,0x62,0x63,0x5b,0x32,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x33,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x32,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x38,
+    0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x32,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x32,0x33,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x32,0x5d,0x3b,0x0a,0x73,0x74,0x5b,
+    0x34,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x33,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x39,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x33,0x5d,0x3b,0x0a,0x73,0x74,0x5b,
+    0x31,0x34,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x33,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x31,0x39,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x33,0x5d,0x3b,0x0a,0x73,
+    0x74,0x5b,0x32,0x34,0x5d,0x20,0x5e,0x3d,0x20,0x62,0x63,0x5b,0x33,0x5d,0x3b,0x0a,0x74,0x3d,0x73,0x74,0x5b,0x31,0x5d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,
+    0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x32,0x34,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,
+    0x0a,0x62,0x63,0x5b,0x30,0x5d,0x3d,0x73,0x74,0x5b,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x70,0x69,0x6c,0x6e,0x5b,0x69,0x5d,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x6b,
+    0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x70,0x69,0x6c,0x6e,0x5b,0x69,0x5d,0x5d,0x3d,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x74,0x2c,0x28,0x75,0x6c,0x6f,0x6e,0x67,0x29,
+    0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,0x6f,0x74,0x63,0x5b,0x69,0x5d,0x29,0x3b,0x0a,0x74,0x3d,0x62,0x63,0x5b,0x30,0x5d,0x3b,0x0a,0x7d,0x0a,0x23,0x70,0x72,
+    0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x32,0x35,
+    0x3b,0x20,0x69,0x2b,0x3d,0x35,0x29,0x20,0x7b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x74,0x6d,0x70,0x31,0x3d,0x73,0x74,0x5b,0x69,0x5d,0x2c,0x74,0x6d,0x70,0x32,0x3d,
+    0x73,0x74,0x5b,0x69,0x2b,0x31,0x5d,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x5d,0x3d,0x62,0x69,0x74,0x73,0x65,0x6c,0x65,0x63,0x74,0x28,0x73,0x74,0x5b,0x69,0x5d,0x5e,0x73,
+    0x74,0x5b,0x69,0x2b,0x32,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x31,0x5d,0x29,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x31,0x5d,0x3d,0x62,
+    0x69,0x74,0x73,0x65,0x6c,0x65,0x63,0x74,0x28,0x73,0x74,0x5b,0x69,0x2b,0x31,0x5d,0x5e,0x73,0x74,0x5b,0x69,0x2b,0x33,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x31,0x5d,
+    0x2c,0x73,0x74,0x5b,0x69,0x2b,0x32,0x5d,0x29,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x32,0x5d,0x3d,0x62,0x69,0x74,0x73,0x65,0x6c,0x65,0x63,0x74,0x28,0x73,0x74,0x5b,
+    0x69,0x2b,0x32,0x5d,0x5e,0x73,0x74,0x5b,0x69,0x2b,0x34,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x32,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x33,0x5d,0x29,0x3b,0x0a,0x73,
+    0x74,0x5b,0x69,0x2b,0x33,0x5d,0x3d,0x62,0x69,0x74,0x73,0x65,0x6c,0x65,0x63,0x74,0x28,0x73,0x74,0x5b,0x69,0x2b,0x33,0x5d,0x5e,0x74,0x6d,0x70,0x31,0x2c,0x73,0x74,
+    0x5b,0x69,0x2b,0x33,0x5d,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x34,0x5d,0x29,0x3b,0x0a,0x73,0x74,0x5b,0x69,0x2b,0x34,0x5d,0x3d,0x62,0x69,0x74,0x73,0x65,0x6c,0x65,0x63,
+    0x74,0x28,0x73,0x74,0x5b,0x69,0x2b,0x34,0x5d,0x5e,0x74,0x6d,0x70,0x32,0x2c,0x73,0x74,0x5b,0x69,0x2b,0x34,0x5d,0x2c,0x74,0x6d,0x70,0x31,0x29,0x3b,0x0a,0x7d,0x0a,
+    0x73,0x74,0x5b,0x30,0x5d,0x20,0x5e,0x3d,0x20,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x5f,0x72,0x6e,0x64,0x63,0x5b,0x72,0x6f,0x75,0x6e,0x64,0x5d,0x3b,0x0a,0x7d,0x0a,
+    0x7d,0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x75,0x69,0x6e,0x74,0x20,0x67,0x65,0x74,0x49,0x64,0x78,0x28,0x29,0x0a,0x7b,0x0a,
+    0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x67,0x65,0x74,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x2d,0x67,0x65,0x74,0x5f,0x67,0x6c,0x6f,0x62,
+    0x61,0x6c,0x5f,0x6f,0x66,0x66,0x73,0x65,0x74,0x28,0x30,0x29,0x3b,0x0a,0x7d,0x0a,0x23,0x64,0x65,0x66,0x69,0x6e,0x65,0x20,0x49,0x44,0x58,0x28,0x78,0x29,0x20,0x28,
+    0x78,0x29,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,
+    0x74,0x34,0x20,0x61,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x2b,0x62,0x3b,0x0a,0x7d,0x0a,0x69,
+    0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x73,0x75,0x62,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,
+    0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x2d,0x62,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,
+    0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,0x2c,0x66,0x6c,0x6f,
+    0x61,0x74,0x34,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x2a,0x62,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,
+    0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x64,0x69,0x76,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,
+    0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x2f,0x62,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,
+    0x20,0x5f,0x6d,0x6d,0x5f,0x61,0x6e,0x64,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,0x2c,0x69,0x6e,0x74,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,
+    0x74,0x75,0x72,0x6e,0x20,0x61,0x73,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x28,0x61,0x73,0x5f,0x69,0x6e,0x74,0x34,0x28,0x61,0x29,0x26,0x28,0x69,0x6e,0x74,0x34,0x29,
+    0x28,0x62,0x29,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x6f,0x72,0x5f,0x70,0x73,0x28,
+    0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,0x2c,0x69,0x6e,0x74,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x73,0x5f,0x66,0x6c,0x6f,0x61,
+    0x74,0x34,0x28,0x61,0x73,0x5f,0x69,0x6e,0x74,0x34,0x28,0x61,0x29,0x7c,0x28,0x69,0x6e,0x74,0x34,0x29,0x28,0x62,0x29,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,
+    0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x66,0x6d,0x6f,0x64,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x76,0x2c,0x66,
+    0x6c,0x6f,0x61,0x74,0x20,0x64,0x63,0x29,0x0a,0x7b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x64,0x3d,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x64,0x63,0x29,
+    0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x63,0x3d,0x5f,0x6d,0x6d,0x5f,0x64,0x69,0x76,0x5f,0x70,0x73,0x28,0x76,0x2c,0x64,0x29,0x3b,0x0a,0x63,0x3d,0x74,0x72,
+    0x75,0x6e,0x63,0x28,0x63,0x29,0x3b,0x0a,0x63,0x3d,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x63,0x2c,0x64,0x29,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,
+    0x6e,0x20,0x5f,0x6d,0x6d,0x5f,0x73,0x75,0x62,0x5f,0x70,0x73,0x28,0x76,0x2c,0x63,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x69,0x6e,0x74,0x34,
+    0x20,0x5f,0x6d,0x6d,0x5f,0x78,0x6f,0x72,0x5f,0x73,0x69,0x31,0x32,0x38,0x28,0x69,0x6e,0x74,0x34,0x20,0x61,0x2c,0x69,0x6e,0x74,0x34,0x20,0x62,0x29,0x0a,0x7b,0x0a,
+    0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x61,0x5e,0x62,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,
+    0x78,0x6f,0x72,0x5f,0x70,0x73,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x61,0x2c,0x69,0x6e,0x74,0x20,0x62,0x29,0x0a,0x7b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,
+    0x61,0x73,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x28,0x61,0x73,0x5f,0x69,0x6e,0x74,0x34,0x28,0x61,0x29,0x5e,0x28,0x69,0x6e,0x74,0x34,0x29,0x28,0x62,0x29,0x29,0x3b,
+    0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x69,0x6e,0x74,0x34,0x20,0x5f,0x6d,0x6d,0x5f,0x61,0x6c,0x69,0x67,0x6e,0x72,0x5f,0x65,0x70,0x69,0x38,0x28,0x69,
+    0x6e,0x74,0x34,0x20,0x61,0x2c,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x72,0x6f,0x74,0x29,0x0a,0x7b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,
+    0x6e,0x74,0x20,0x72,0x69,0x67,0x68,0x74,0x3d,0x38,0x2a,0x72,0x6f,0x74,0x3b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x6c,0x65,0x66,0x74,0x3d,
+    0x28,0x33,0x32,0x2d,0x38,0x2a,0x72,0x6f,0x74,0x29,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x28,0x69,0x6e,0x74,0x34,0x29,0x28,0x0a,0x28,0x28,0x75,0x69,0x6e,
+    0x74,0x29,0x61,0x2e,0x78,0x3e,0x3e,0x72,0x69,0x67,0x68,0x74,0x29,0x7c,0x28,0x20,0x61,0x2e,0x79,0x3c,0x3c,0x6c,0x65,0x66,0x74,0x20,0x29,0x2c,0x0a,0x28,0x28,0x75,
+    0x69,0x6e,0x74,0x29,0x61,0x2e,0x79,0x3e,0x3e,0x72,0x69,0x67,0x68,0x74,0x29,0x7c,0x28,0x20,0x61,0x2e,0x7a,0x3c,0x3c,0x6c,0x65,0x66,0x74,0x20,0x29,0x2c,0x0a,0x28,
+    0x28,0x75,0x69,0x6e,0x74,0x29,0x61,0x2e,0x7a,0x3e,0x3e,0x72,0x69,0x67,0x68,0x74,0x29,0x7c,0x28,0x20,0x61,0x2e,0x77,0x3c,0x3c,0x6c,0x65,0x66,0x74,0x20,0x29,0x2c,
+    0x0a,0x28,0x28,0x75,0x69,0x6e,0x74,0x29,0x61,0x2e,0x77,0x3e,0x3e,0x72,0x69,0x67,0x68,0x74,0x29,0x7c,0x28,0x20,0x61,0x2e,0x78,0x3c,0x3c,0x6c,0x65,0x66,0x74,0x20,
+    0x29,0x0a,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x34,0x2a,0x20,0x73,0x63,0x72,0x61,0x74,
+    0x63,0x68,0x70,0x61,0x64,0x5f,0x70,0x74,0x72,0x28,0x75,0x69,0x6e,0x74,0x20,0x69,0x64,0x78,0x2c,0x75,0x69,0x6e,0x74,0x20,0x6e,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,
+    0x61,0x6c,0x20,0x69,0x6e,0x74,0x20,0x2a,0x6c,0x70,0x61,0x64,0x29,0x20,0x7b,0x20,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,
+    0x20,0x69,0x6e,0x74,0x34,0x2a,0x29,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x63,0x68,0x61,0x72,0x2a,0x29,0x6c,0x70,0x61,0x64,0x2b,0x28,0x69,0x64,
+    0x78,0x26,0x4d,0x41,0x53,0x4b,0x29,0x2b,0x6e,0x2a,0x31,0x36,0x29,0x3b,0x20,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x66,
+    0x6d,0x61,0x5f,0x62,0x72,0x65,0x61,0x6b,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x78,0x29,0x0a,0x7b,0x0a,0x78,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x6e,0x64,0x5f,0x70,
+    0x73,0x28,0x78,0x2c,0x30,0x78,0x46,0x45,0x46,0x46,0x46,0x46,0x46,0x46,0x29,0x3b,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x5f,0x6d,0x6d,0x5f,0x6f,0x72,0x5f,0x70,
+    0x73,0x28,0x78,0x2c,0x30,0x78,0x30,0x30,0x38,0x30,0x30,0x30,0x30,0x30,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x76,0x6f,0x69,0x64,0x20,0x73,
+    0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x30,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x31,0x2c,0x66,0x6c,0x6f,
+    0x61,0x74,0x34,0x20,0x6e,0x32,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x33,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x66,0x6c,
+    0x6f,0x61,0x74,0x34,0x2a,0x20,0x6e,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,0x64,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,0x63,0x29,0x0a,0x7b,0x0a,0x6e,
+    0x31,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x6e,0x31,0x2c,0x2a,0x63,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x6e,0x3d,0x5f,
+    0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x6e,0x30,0x2c,0x2a,0x63,0x29,0x3b,0x0a,0x6e,0x6e,0x3d,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,
+    0x6e,0x31,0x2c,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x6e,0x6e,0x2c,0x6e,0x6e,0x29,0x29,0x3b,0x0a,0x6e,0x6e,0x3d,0x66,0x6d,0x61,0x5f,0x62,0x72,
+    0x65,0x61,0x6b,0x28,0x6e,0x6e,0x29,0x3b,0x0a,0x2a,0x6e,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x2a,0x6e,0x2c,0x6e,0x6e,0x29,0x3b,0x0a,0x6e,
+    0x33,0x3d,0x5f,0x6d,0x6d,0x5f,0x73,0x75,0x62,0x5f,0x70,0x73,0x28,0x6e,0x33,0x2c,0x2a,0x63,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x64,0x64,0x3d,0x5f,
+    0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x6e,0x32,0x2c,0x2a,0x63,0x29,0x3b,0x0a,0x64,0x64,0x3d,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,
+    0x6e,0x33,0x2c,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,0x70,0x73,0x28,0x64,0x64,0x2c,0x64,0x64,0x29,0x29,0x3b,0x0a,0x64,0x64,0x3d,0x66,0x6d,0x61,0x5f,0x62,0x72,
+    0x65,0x61,0x6b,0x28,0x64,0x64,0x29,0x3b,0x0a,0x2a,0x64,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x2a,0x64,0x2c,0x64,0x64,0x29,0x3b,0x0a,0x2a,
+    0x63,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x2a,0x63,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x29,0x3b,0x0a,0x2a,0x63,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,
+    0x64,0x64,0x5f,0x70,0x73,0x28,0x2a,0x63,0x2c,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x30,0x2e,0x37,0x33,0x34,0x33,0x37,0x35,0x66,0x29,0x29,0x3b,0x0a,0x66,
+    0x6c,0x6f,0x61,0x74,0x34,0x20,0x72,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x6e,0x6e,0x2c,0x64,0x64,0x29,0x3b,0x0a,0x72,0x3d,0x5f,0x6d,0x6d,
+    0x5f,0x61,0x6e,0x64,0x5f,0x70,0x73,0x28,0x72,0x2c,0x30,0x78,0x38,0x30,0x37,0x46,0x46,0x46,0x46,0x46,0x29,0x3b,0x0a,0x72,0x3d,0x5f,0x6d,0x6d,0x5f,0x6f,0x72,0x5f,
+    0x70,0x73,0x28,0x72,0x2c,0x30,0x78,0x34,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x29,0x3b,0x0a,0x2a,0x63,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,
+    0x2a,0x63,0x2c,0x72,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x76,0x6f,0x69,0x64,0x20,0x72,0x6f,0x75,0x6e,0x64,0x5f,0x63,0x6f,0x6d,0x70,0x75,
+    0x74,0x65,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x30,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x31,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x32,
+    0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x33,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,
+    0x63,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,0x72,0x29,0x0a,0x7b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x3d,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,
+    0x28,0x30,0x2e,0x30,0x66,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x64,0x3d,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x30,0x2e,0x30,0x66,0x29,0x3b,
+    0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x30,0x2c,0x6e,0x31,0x2c,0x6e,0x32,0x2c,0x6e,0x33,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,
+    0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x31,0x2c,0x6e,0x32,0x2c,0x6e,0x33,0x2c,0x6e,0x30,0x2c,0x72,0x6e,0x64,
+    0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x32,0x2c,0x6e,0x33,0x2c,0x6e,0x30,0x2c,
+    0x6e,0x31,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x33,0x2c,
+    0x6e,0x30,0x2c,0x6e,0x31,0x2c,0x6e,0x32,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,
+    0x6e,0x64,0x28,0x6e,0x33,0x2c,0x6e,0x32,0x2c,0x6e,0x31,0x2c,0x6e,0x30,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,
+    0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x32,0x2c,0x6e,0x31,0x2c,0x6e,0x30,0x2c,0x6e,0x33,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,
+    0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x31,0x2c,0x6e,0x30,0x2c,0x6e,0x33,0x2c,0x6e,0x32,0x2c,0x72,0x6e,0x64,0x5f,0x63,
+    0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x73,0x75,0x62,0x5f,0x72,0x6f,0x75,0x6e,0x64,0x28,0x6e,0x30,0x2c,0x6e,0x33,0x2c,0x6e,0x32,0x2c,0x6e,0x31,
+    0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x6e,0x2c,0x26,0x64,0x2c,0x63,0x29,0x3b,0x0a,0x64,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x6e,0x64,0x5f,0x70,0x73,0x28,0x64,0x2c,
+    0x30,0x78,0x46,0x46,0x37,0x46,0x46,0x46,0x46,0x46,0x29,0x3b,0x0a,0x64,0x3d,0x5f,0x6d,0x6d,0x5f,0x6f,0x72,0x5f,0x70,0x73,0x28,0x64,0x2c,0x30,0x78,0x34,0x30,0x30,
+    0x30,0x30,0x30,0x30,0x30,0x29,0x3b,0x0a,0x2a,0x72,0x20,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x64,0x64,0x5f,0x70,0x73,0x28,0x2a,0x72,0x2c,0x5f,0x6d,0x6d,0x5f,0x64,0x69,
+    0x76,0x5f,0x70,0x73,0x28,0x6e,0x2c,0x64,0x29,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x69,0x6e,0x74,0x34,0x20,0x73,0x69,0x6e,0x67,0x6c,0x65,
+    0x5f,0x63,0x6f,0x6d,0x70,0x75,0x74,0x65,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x30,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x31,0x2c,0x66,0x6c,0x6f,
+    0x61,0x74,0x34,0x20,0x6e,0x32,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x33,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x20,0x63,0x6e,0x74,0x2c,0x66,0x6c,0x6f,0x61,0x74,
+    0x34,0x20,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,0x73,0x75,0x6d,0x29,0x0a,0x7b,0x0a,0x66,
+    0x6c,0x6f,0x61,0x74,0x34,0x20,0x63,0x3d,0x20,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x63,0x6e,0x74,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x72,
+    0x3d,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x30,0x2e,0x30,0x66,0x29,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,
+    0x3c,0x34,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,0x0a,0x72,0x6f,0x75,0x6e,0x64,0x5f,0x63,0x6f,0x6d,0x70,0x75,0x74,0x65,0x28,0x6e,0x30,0x2c,0x6e,0x31,0x2c,0x6e,
+    0x32,0x2c,0x6e,0x33,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x26,0x63,0x2c,0x26,0x72,0x29,0x3b,0x0a,0x7d,0x0a,0x72,0x3d,0x5f,0x6d,0x6d,0x5f,0x61,0x6e,0x64,0x5f,0x70,
+    0x73,0x28,0x72,0x2c,0x30,0x78,0x38,0x30,0x37,0x46,0x46,0x46,0x46,0x46,0x29,0x3b,0x0a,0x72,0x3d,0x5f,0x6d,0x6d,0x5f,0x6f,0x72,0x5f,0x70,0x73,0x28,0x72,0x2c,0x30,
+    0x78,0x34,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x29,0x3b,0x0a,0x2a,0x73,0x75,0x6d,0x3d,0x72,0x3b,0x20,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x78,0x3d,0x28,0x66,
+    0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x35,0x33,0x36,0x38,0x37,0x30,0x38,0x38,0x30,0x2e,0x30,0x66,0x29,0x3b,0x0a,0x72,0x3d,0x5f,0x6d,0x6d,0x5f,0x6d,0x75,0x6c,0x5f,
+    0x70,0x73,0x28,0x72,0x2c,0x78,0x29,0x3b,0x20,0x0a,0x72,0x65,0x74,0x75,0x72,0x6e,0x20,0x63,0x6f,0x6e,0x76,0x65,0x72,0x74,0x5f,0x69,0x6e,0x74,0x34,0x5f,0x72,0x74,
+    0x65,0x28,0x72,0x29,0x3b,0x0a,0x7d,0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x76,0x6f,0x69,0x64,0x20,0x73,0x69,0x6e,0x67,0x6c,0x65,0x5f,0x63,0x6f,0x6d,0x70,0x75,
+    0x74,0x65,0x5f,0x77,0x72,0x61,0x70,0x28,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x72,0x6f,0x74,0x2c,0x69,0x6e,0x74,0x34,0x20,0x76,0x30,0x2c,0x69,
+    0x6e,0x74,0x34,0x20,0x76,0x31,0x2c,0x69,0x6e,0x74,0x34,0x20,0x76,0x32,0x2c,0x69,0x6e,0x74,0x34,0x20,0x76,0x33,0x2c,0x66,0x6c,0x6f,0x61,0x74,0x20,0x63,0x6e,0x74,
+    0x2c,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x34,0x2a,0x20,0x73,0x75,
+    0x6d,0x2c,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x34,0x2a,0x20,0x6f,0x75,0x74,0x29,0x0a,0x7b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x6e,0x30,
+    0x3d,0x63,0x6f,0x6e,0x76,0x65,0x72,0x74,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x5f,0x72,0x74,0x65,0x28,0x76,0x30,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,
+    0x6e,0x31,0x3d,0x63,0x6f,0x6e,0x76,0x65,0x72,0x74,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x5f,0x72,0x74,0x65,0x28,0x76,0x31,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,
+    0x34,0x20,0x6e,0x32,0x3d,0x63,0x6f,0x6e,0x76,0x65,0x72,0x74,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x5f,0x72,0x74,0x65,0x28,0x76,0x32,0x29,0x3b,0x0a,0x66,0x6c,0x6f,
+    0x61,0x74,0x34,0x20,0x6e,0x33,0x3d,0x63,0x6f,0x6e,0x76,0x65,0x72,0x74,0x5f,0x66,0x6c,0x6f,0x61,0x74,0x34,0x5f,0x72,0x74,0x65,0x28,0x76,0x33,0x29,0x3b,0x0a,0x69,
+    0x6e,0x74,0x34,0x20,0x72,0x3d,0x73,0x69,0x6e,0x67,0x6c,0x65,0x5f,0x63,0x6f,0x6d,0x70,0x75,0x74,0x65,0x28,0x6e,0x30,0x2c,0x6e,0x31,0x2c,0x6e,0x32,0x2c,0x6e,0x33,
+    0x2c,0x63,0x6e,0x74,0x2c,0x72,0x6e,0x64,0x5f,0x63,0x2c,0x73,0x75,0x6d,0x29,0x3b,0x0a,0x2a,0x6f,0x75,0x74,0x3d,0x72,0x6f,0x74,0x3d,0x3d,0x30,0x3f,0x72,0x3a,0x5f,
+    0x6d,0x6d,0x5f,0x61,0x6c,0x69,0x67,0x6e,0x72,0x5f,0x65,0x70,0x69,0x38,0x28,0x72,0x2c,0x72,0x6f,0x74,0x29,0x3b,0x0a,0x7d,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,
+    0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x6c,0x6f,0x6f,0x6b,0x5b,0x31,0x36,0x5d,0x5b,0x34,
+    0x5d,0x3d,0x7b,0x0a,0x7b,0x30,0x2c,0x31,0x2c,0x32,0x2c,0x33,0x7d,0x2c,0x0a,0x7b,0x30,0x2c,0x32,0x2c,0x33,0x2c,0x31,0x7d,0x2c,0x0a,0x7b,0x30,0x2c,0x33,0x2c,0x31,
+    0x2c,0x32,0x7d,0x2c,0x0a,0x7b,0x30,0x2c,0x33,0x2c,0x32,0x2c,0x31,0x7d,0x2c,0x0a,0x7b,0x31,0x2c,0x30,0x2c,0x32,0x2c,0x33,0x7d,0x2c,0x0a,0x7b,0x31,0x2c,0x32,0x2c,
+    0x33,0x2c,0x30,0x7d,0x2c,0x0a,0x7b,0x31,0x2c,0x33,0x2c,0x30,0x2c,0x32,0x7d,0x2c,0x0a,0x7b,0x31,0x2c,0x33,0x2c,0x32,0x2c,0x30,0x7d,0x2c,0x0a,0x7b,0x32,0x2c,0x31,
+    0x2c,0x30,0x2c,0x33,0x7d,0x2c,0x0a,0x7b,0x32,0x2c,0x30,0x2c,0x33,0x2c,0x31,0x7d,0x2c,0x0a,0x7b,0x32,0x2c,0x33,0x2c,0x31,0x2c,0x30,0x7d,0x2c,0x0a,0x7b,0x32,0x2c,
+    0x33,0x2c,0x30,0x2c,0x31,0x7d,0x2c,0x0a,0x7b,0x33,0x2c,0x31,0x2c,0x32,0x2c,0x30,0x7d,0x2c,0x0a,0x7b,0x33,0x2c,0x32,0x2c,0x30,0x2c,0x31,0x7d,0x2c,0x0a,0x7b,0x33,
+    0x2c,0x30,0x2c,0x31,0x2c,0x32,0x7d,0x2c,0x0a,0x7b,0x33,0x2c,0x30,0x2c,0x32,0x2c,0x31,0x7d,0x0a,0x7d,0x3b,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,
+    0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,0x73,0x74,0x61,0x6e,0x74,0x20,0x66,0x6c,0x6f,0x61,0x74,0x20,0x63,0x63,0x6e,0x74,0x5b,0x31,0x36,0x5d,0x3d,0x7b,0x0a,0x31,
+    0x2e,0x33,0x34,0x33,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x32,0x38,0x31,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x33,0x35,0x39,0x33,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,
+    0x33,0x36,0x37,0x31,0x38,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x34,0x32,0x39,0x36,0x38,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x33,0x39,0x38,0x34,0x33,0x37,0x35,0x66,
+    0x2c,0x0a,0x31,0x2e,0x33,0x38,0x32,0x38,0x31,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x33,0x30,0x34,0x36,0x38,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x34,0x31,0x34,0x30,
+    0x36,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x32,0x37,0x33,0x34,0x33,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x32,0x35,0x37,0x38,0x31,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,
+    0x32,0x38,0x39,0x30,0x36,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x33,0x32,0x30,0x33,0x31,0x32,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x33,0x35,0x31,0x35,0x36,0x32,0x35,0x66,
+    0x2c,0x0a,0x31,0x2e,0x33,0x33,0x35,0x39,0x33,0x37,0x35,0x66,0x2c,0x0a,0x31,0x2e,0x34,0x36,0x30,0x39,0x33,0x37,0x35,0x66,0x0a,0x7d,0x3b,0x0a,0x73,0x74,0x72,0x75,
+    0x63,0x74,0x20,0x53,0x68,0x61,0x72,0x65,0x64,0x4d,0x65,0x6d,0x43,0x68,0x75,0x6e,0x6b,0x0a,0x7b,0x0a,0x69,0x6e,0x74,0x34,0x20,0x6f,0x75,0x74,0x5b,0x31,0x36,0x5d,
+    0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x76,0x61,0x5b,0x31,0x36,0x5d,0x3b,0x0a,0x7d,0x3b,0x0a,0x5f,0x5f,0x61,0x74,0x74,0x72,0x69,0x62,0x75,0x74,0x65,0x5f,
+    0x5f,0x28,0x28,0x72,0x65,0x71,0x64,0x5f,0x77,0x6f,0x72,0x6b,0x5f,0x67,0x72,0x6f,0x75,0x70,0x5f,0x73,0x69,0x7a,0x65,0x28,0x57,0x4f,0x52,0x4b,0x53,0x49,0x5a,0x45,
+    0x2a,0x31,0x36,0x2c,0x31,0x2c,0x31,0x29,0x29,0x29,0x0a,0x5f,0x5f,0x6b,0x65,0x72,0x6e,0x65,0x6c,0x20,0x76,0x6f,0x69,0x64,0x20,0x63,0x6e,0x31,0x28,0x5f,0x5f,0x67,
+    0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x20,0x2a,0x6c,0x70,0x61,0x64,0x5f,0x69,0x6e,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x20,
+    0x2a,0x73,0x70,0x61,0x64,0x2c,0x75,0x69,0x6e,0x74,0x20,0x6e,0x75,0x6d,0x54,0x68,0x72,0x65,0x61,0x64,0x73,0x29,0x0a,0x7b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,
+    0x69,0x6e,0x74,0x20,0x67,0x49,0x64,0x78,0x3d,0x67,0x65,0x74,0x49,0x64,0x78,0x28,0x29,0x3b,0x0a,0x75,0x69,0x6e,0x74,0x20,0x63,0x68,0x75,0x6e,0x6b,0x3d,0x67,0x65,
+    0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x2f,0x31,0x36,0x3b,0x0a,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x20,
+    0x6c,0x70,0x61,0x64,0x3d,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x63,
+    0x68,0x61,0x72,0x2a,0x29,0x6c,0x70,0x61,0x64,0x5f,0x69,0x6e,0x2b,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x2a,0x28,0x67,0x49,0x64,0x78,0x2f,0x31,0x36,0x29,0x29,0x3b,0x0a,
+    0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x73,0x74,0x72,0x75,0x63,0x74,0x20,0x53,0x68,0x61,0x72,0x65,0x64,0x4d,0x65,0x6d,0x43,0x68,0x75,0x6e,0x6b,0x20,0x73,0x6d,
+    0x65,0x6d,0x5f,0x69,0x6e,0x5b,0x57,0x4f,0x52,0x4b,0x53,0x49,0x5a,0x45,0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x73,0x74,0x72,0x75,0x63,0x74,0x20,
+    0x53,0x68,0x61,0x72,0x65,0x64,0x4d,0x65,0x6d,0x43,0x68,0x75,0x6e,0x6b,0x2a,0x20,0x73,0x6d,0x65,0x6d,0x3d,0x73,0x6d,0x65,0x6d,0x5f,0x69,0x6e,0x2b,0x63,0x68,0x75,
+    0x6e,0x6b,0x3b,0x0a,0x75,0x69,0x6e,0x74,0x20,0x74,0x69,0x64,0x3d,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x20,0x25,0x20,0x31,
+    0x36,0x3b,0x0a,0x75,0x69,0x6e,0x74,0x20,0x69,0x64,0x78,0x48,0x61,0x73,0x68,0x3d,0x67,0x49,0x64,0x78,0x2f,0x31,0x36,0x3b,0x0a,0x75,0x69,0x6e,0x74,0x20,0x73,0x3d,
+    0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x2a,0x29,0x73,0x70,0x61,0x64,0x29,0x5b,0x69,0x64,0x78,0x48,0x61,0x73,0x68,0x2a,0x35,
+    0x30,0x5d,0x3e,0x3e,0x38,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x34,0x20,0x76,0x73,0x3d,0x28,0x66,0x6c,0x6f,0x61,0x74,0x34,0x29,0x28,0x30,0x29,0x3b,0x0a,0x63,0x6f,
+    0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x74,0x69,0x64,0x64,0x3d,0x74,0x69,0x64,0x2f,0x34,0x3b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,
+    0x74,0x69,0x64,0x6d,0x3d,0x74,0x69,0x64,0x20,0x25,0x20,0x34,0x3b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x62,0x6c,0x6f,0x63,0x6b,0x3d,0x74,
+    0x69,0x64,0x64,0x2a,0x31,0x36,0x2b,0x74,0x69,0x64,0x6d,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x43,0x4e,0x5f,0x55,
+    0x4e,0x52,0x4f,0x4c,0x4c,0x0a,0x66,0x6f,0x72,0x20,0x28,0x75,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x49,0x54,0x45,0x52,0x41,0x54,0x49,0x4f,0x4e,
+    0x53,0x3b,0x20,0x69,0x2b,0x2b,0x29,0x20,0x7b,0x0a,0x6d,0x65,0x6d,0x5f,0x66,0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,
+    0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x69,0x6e,0x74,0x20,0x74,0x6d,0x70,0x3d,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,
+    0x2a,0x29,0x73,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x5f,0x70,0x74,0x72,0x28,0x73,0x2c,0x74,0x69,0x64,0x64,0x2c,0x6c,0x70,0x61,0x64,0x29,0x29,0x5b,0x74,
+    0x69,0x64,0x6d,0x5d,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x28,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,
+    0x29,0x5b,0x74,0x69,0x64,0x5d,0x3d,0x74,0x6d,0x70,0x3b,0x0a,0x6d,0x65,0x6d,0x5f,0x66,0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,
+    0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x7b,0x0a,0x73,0x69,0x6e,0x67,0x6c,0x65,0x5f,0x63,0x6f,0x6d,0x70,0x75,0x74,0x65,0x5f,0x77,0x72,0x61,
+    0x70,0x28,0x0a,0x74,0x69,0x64,0x6d,0x2c,0x0a,0x2a,0x28,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x2b,0x6c,0x6f,0x6f,0x6b,0x5b,0x74,0x69,0x64,0x5d,0x5b,0x30,
+    0x5d,0x29,0x2c,0x0a,0x2a,0x28,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x2b,0x6c,0x6f,0x6f,0x6b,0x5b,0x74,0x69,0x64,0x5d,0x5b,0x31,0x5d,0x29,0x2c,0x0a,0x2a,
+    0x28,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x2b,0x6c,0x6f,0x6f,0x6b,0x5b,0x74,0x69,0x64,0x5d,0x5b,0x32,0x5d,0x29,0x2c,0x0a,0x2a,0x28,0x73,0x6d,0x65,0x6d,
+    0x2d,0x3e,0x6f,0x75,0x74,0x2b,0x6c,0x6f,0x6f,0x6b,0x5b,0x74,0x69,0x64,0x5d,0x5b,0x33,0x5d,0x29,0x2c,0x0a,0x63,0x63,0x6e,0x74,0x5b,0x74,0x69,0x64,0x5d,0x2c,0x76,
+    0x73,0x2c,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x2b,0x74,0x69,0x64,0x2c,0x0a,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x2b,0x74,0x69,0x64,0x0a,0x29,0x3b,
+    0x0a,0x7d,0x0a,0x6d,0x65,0x6d,0x5f,0x66,0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,
+    0x29,0x3b,0x0a,0x69,0x6e,0x74,0x20,0x6f,0x75,0x74,0x58,0x6f,0x72,0x3d,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,
+    0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x5d,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x75,0x69,0x6e,0x74,0x20,0x64,0x64,0x3d,0x62,0x6c,0x6f,
+    0x63,0x6b,0x2b,0x34,0x3b,0x20,0x64,0x64,0x3c,0x28,0x74,0x69,0x64,0x64,0x2b,0x31,0x29,0x2a,0x31,0x36,0x3b,0x20,0x64,0x64,0x2b,0x3d,0x34,0x29,0x20,0x7b,0x0a,0x6f,
+    0x75,0x74,0x58,0x6f,0x72,0x20,0x5e,0x3d,0x20,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,
+    0x74,0x29,0x5b,0x64,0x64,0x5d,0x3b,0x0a,0x7d,0x0a,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x63,0x72,0x61,0x74,0x63,
+    0x68,0x70,0x61,0x64,0x5f,0x70,0x74,0x72,0x28,0x73,0x2c,0x74,0x69,0x64,0x64,0x2c,0x6c,0x70,0x61,0x64,0x29,0x29,0x5b,0x74,0x69,0x64,0x6d,0x5d,0x3d,0x6f,0x75,0x74,
+    0x58,0x6f,0x72,0x5e,0x74,0x6d,0x70,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,
+    0x74,0x29,0x5b,0x74,0x69,0x64,0x5d,0x3d,0x6f,0x75,0x74,0x58,0x6f,0x72,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x20,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x3d,0x28,0x28,
+    0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x5d,0x2b,
+    0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,
+    0x2b,0x34,0x5d,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x20,0x76,0x61,0x5f,0x74,0x6d,0x70,0x32,0x3d,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,
+    0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x2b,0x20,0x38,0x5d,0x2b,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,
+    0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x2b,0x31,0x32,0x5d,0x3b,0x0a,0x28,0x28,
+    0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x74,0x69,0x64,0x5d,0x3d,0x76,0x61,
+    0x5f,0x74,0x6d,0x70,0x31,0x2b,0x76,0x61,0x5f,0x74,0x6d,0x70,0x32,0x3b,0x0a,0x6d,0x65,0x6d,0x5f,0x66,0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,
+    0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x69,0x6e,0x74,0x20,0x6f,0x75,0x74,0x32,0x3d,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,
+    0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x74,0x69,0x64,0x5d,0x5e,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,
+    0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x74,0x69,0x64,0x2b,0x34,0x20,0x5d,0x5e,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,
+    0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x74,0x69,0x64,0x2b,0x38,0x5d,0x5e,0x28,0x28,0x5f,0x5f,0x6c,0x6f,
+    0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x74,0x69,0x64,0x2b,0x31,0x32,0x5d,0x3b,0x0a,0x76,0x61,0x5f,
+    0x74,0x6d,0x70,0x31,0x3d,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,
+    0x62,0x6c,0x6f,0x63,0x6b,0x5d,0x2b,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,
+    0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x2b,0x34,0x5d,0x3b,0x0a,0x76,0x61,0x5f,0x74,0x6d,0x70,0x32,0x3d,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,
+    0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x2b,0x38,0x5d,0x2b,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,
+    0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x62,0x6c,0x6f,0x63,0x6b,0x2b,0x31,0x32,0x5d,0x3b,0x0a,0x76,0x61,
+    0x5f,0x74,0x6d,0x70,0x31,0x3d,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x2b,0x76,0x61,0x5f,0x74,0x6d,0x70,0x32,0x3b,0x0a,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x3d,0x66,
+    0x61,0x62,0x73,0x28,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x29,0x3b,0x0a,0x66,0x6c,0x6f,0x61,0x74,0x20,0x78,0x78,0x3d,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x2a,0x31,
+    0x36,0x37,0x37,0x37,0x32,0x31,0x36,0x2e,0x30,0x66,0x3b,0x0a,0x69,0x6e,0x74,0x20,0x78,0x78,0x5f,0x69,0x6e,0x74,0x3d,0x28,0x69,0x6e,0x74,0x29,0x78,0x78,0x3b,0x0a,
+    0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x29,0x5b,0x74,0x69,0x64,0x5d,0x3d,0x6f,
+    0x75,0x74,0x32,0x5e,0x78,0x78,0x5f,0x69,0x6e,0x74,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x66,0x6c,0x6f,0x61,0x74,0x2a,0x29,0x73,0x6d,0x65,
+    0x6d,0x2d,0x3e,0x76,0x61,0x29,0x5b,0x74,0x69,0x64,0x5d,0x3d,0x76,0x61,0x5f,0x74,0x6d,0x70,0x31,0x2f,0x36,0x34,0x2e,0x30,0x66,0x3b,0x0a,0x6d,0x65,0x6d,0x5f,0x66,
+    0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x76,0x73,0x3d,0x73,0x6d,
+    0x65,0x6d,0x2d,0x3e,0x76,0x61,0x5b,0x30,0x5d,0x3b,0x0a,0x73,0x3d,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x5b,0x30,0x5d,0x2e,0x78,0x5e,0x73,0x6d,0x65,0x6d,
+    0x2d,0x3e,0x6f,0x75,0x74,0x5b,0x30,0x5d,0x2e,0x79,0x5e,0x73,0x6d,0x65,0x6d,0x2d,0x3e,0x6f,0x75,0x74,0x5b,0x30,0x5d,0x2e,0x7a,0x5e,0x73,0x6d,0x65,0x6d,0x2d,0x3e,
+    0x6f,0x75,0x74,0x5b,0x30,0x5d,0x2e,0x77,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x73,0x74,0x61,0x74,0x69,0x63,0x20,0x63,0x6f,0x6e,0x73,0x74,0x20,0x5f,0x5f,0x63,0x6f,0x6e,
+    0x73,0x74,0x61,0x6e,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x73,0x6b,0x69,0x70,0x5b,0x33,0x5d,0x3d,0x7b,0x0a,0x32,0x30,0x2c,0x32,0x32,0x2c,0x32,0x32,0x0a,0x7d,0x3b,
+    0x0a,0x69,0x6e,0x6c,0x69,0x6e,0x65,0x20,0x76,0x6f,0x69,0x64,0x20,0x67,0x65,0x6e,0x65,0x72,0x61,0x74,0x65,0x5f,0x35,0x31,0x32,0x28,0x75,0x69,0x6e,0x74,0x20,0x69,
+    0x64,0x78,0x2c,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x2a,0x20,0x69,0x6e,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,
+    0x6f,0x6e,0x67,0x2a,0x20,0x6f,0x75,0x74,0x29,0x0a,0x7b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x68,0x61,0x73,0x68,0x5b,0x32,0x35,0x5d,0x3b,0x0a,0x68,0x61,0x73,0x68,
+    0x5b,0x30,0x5d,0x3d,0x69,0x6e,0x5b,0x30,0x5d,0x5e,0x69,0x64,0x78,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x31,0x3b,0x20,0x69,0x3c,0x32,
+    0x35,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,0x0a,0x68,0x61,0x73,0x68,0x5b,0x69,0x5d,0x3d,0x69,0x6e,0x5b,0x69,0x5d,0x3b,0x0a,0x7d,0x0a,0x66,0x6f,0x72,0x20,0x28,
+    0x69,0x6e,0x74,0x20,0x61,0x3d,0x30,0x3b,0x20,0x61,0x3c,0x33,0x3b,0x20,0x2b,0x2b,0x61,0x29,0x20,0x7b,0x0a,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x31,0x36,0x30,0x30,
+    0x5f,0x31,0x28,0x68,0x61,0x73,0x68,0x29,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x73,0x6b,0x69,0x70,0x5b,0x61,
+    0x5d,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,0x0a,0x6f,0x75,0x74,0x5b,0x69,0x5d,0x3d,0x68,0x61,0x73,0x68,0x5b,0x69,0x5d,0x3b,0x0a,0x7d,0x0a,0x6f,0x75,0x74,0x2b,
+    0x3d,0x73,0x6b,0x69,0x70,0x5b,0x61,0x5d,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x5f,0x5f,0x61,0x74,0x74,0x72,0x69,0x62,0x75,0x74,0x65,0x5f,0x5f,0x28,0x28,0x72,0x65,0x71,
+    0x64,0x5f,0x77,0x6f,0x72,0x6b,0x5f,0x67,0x72,0x6f,0x75,0x70,0x5f,0x73,0x69,0x7a,0x65,0x28,0x38,0x2c,0x38,0x2c,0x31,0x29,0x29,0x29,0x0a,0x5f,0x5f,0x6b,0x65,0x72,
+    0x6e,0x65,0x6c,0x20,0x76,0x6f,0x69,0x64,0x20,0x63,0x6e,0x30,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x69,0x6e,0x70,
+    0x75,0x74,0x2c,0x69,0x6e,0x74,0x20,0x69,0x6e,0x6c,0x65,0x6e,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x20,0x2a,0x53,0x63,0x72,0x61,0x74,
+    0x63,0x68,0x70,0x61,0x64,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x73,0x74,0x61,0x74,0x65,0x73,0x2c,0x75,0x69,0x6e,
+    0x74,0x20,0x54,0x68,0x72,0x65,0x61,0x64,0x73,0x29,0x0a,0x7b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x67,0x49,0x64,0x78,0x3d,0x67,0x65,0x74,
+    0x49,0x64,0x78,0x28,0x29,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x53,0x74,0x61,0x74,0x65,0x5f,0x62,0x75,0x66,0x5b,0x38,
+    0x2a,0x32,0x35,0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x2a,0x20,0x53,0x74,0x61,0x74,0x65,0x3d,0x53,0x74,0x61,0x74,0x65,
+    0x5f,0x62,0x75,0x66,0x2b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x2a,0x32,0x35,0x3b,0x0a,0x7b,0x0a,0x73,0x74,0x61,0x74,0x65,
+    0x73,0x2b,0x3d,0x32,0x35,0x2a,0x67,0x49,0x64,0x78,0x3b,0x0a,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x3d,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,
+    0x20,0x69,0x6e,0x74,0x2a,0x29,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x63,0x68,0x61,0x72,0x2a,0x29,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,
+    0x64,0x2b,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x2a,0x67,0x49,0x64,0x78,0x29,0x3b,0x0a,0x69,0x66,0x28,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,
+    0x31,0x29,0x3d,0x3d,0x30,0x29,0x20,0x7b,0x0a,0x23,0x69,0x66,0x64,0x65,0x66,0x20,0x5f,0x5f,0x4e,0x56,0x5f,0x43,0x4c,0x5f,0x43,0x5f,0x56,0x45,0x52,0x53,0x49,0x4f,
+    0x4e,0x0a,0x66,0x6f,0x72,0x28,0x75,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x38,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x0a,0x53,0x74,0x61,0x74,0x65,0x5b,
+    0x69,0x5d,0x3d,0x69,0x6e,0x70,0x75,0x74,0x5b,0x69,0x5d,0x3b,0x0a,0x23,0x65,0x6c,0x73,0x65,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,
+    0x6e,0x67,0x38,0x20,0x2a,0x29,0x53,0x74,0x61,0x74,0x65,0x29,0x5b,0x30,0x5d,0x3d,0x76,0x6c,0x6f,0x61,0x64,0x38,0x28,0x30,0x2c,0x69,0x6e,0x70,0x75,0x74,0x29,0x3b,
+    0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x53,0x74,0x61,0x74,0x65,0x5b,0x38,0x5d,0x3d,0x69,0x6e,0x70,0x75,0x74,0x5b,0x38,0x5d,0x3b,0x0a,0x53,0x74,0x61,0x74,0x65,
+    0x5b,0x39,0x5d,0x3d,0x69,0x6e,0x70,0x75,0x74,0x5b,0x39,0x5d,0x3b,0x0a,0x53,0x74,0x61,0x74,0x65,0x5b,0x31,0x30,0x5d,0x3d,0x69,0x6e,0x70,0x75,0x74,0x5b,0x31,0x30,
+    0x5d,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x53,0x74,0x61,0x74,0x65,0x29,0x5b,0x39,0x5d,0x20,0x26,0x3d,
+    0x20,0x30,0x78,0x30,0x30,0x46,0x46,0x46,0x46,0x46,0x46,0x55,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x53,
+    0x74,0x61,0x74,0x65,0x29,0x5b,0x39,0x5d,0x7c,0x3d,0x28,0x28,0x28,0x75,0x69,0x6e,0x74,0x29,0x67,0x65,0x74,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x5f,0x69,0x64,0x28,
+    0x30,0x29,0x29,0x26,0x30,0x78,0x46,0x46,0x29,0x3c,0x3c,0x32,0x34,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,
+    0x53,0x74,0x61,0x74,0x65,0x29,0x5b,0x31,0x30,0x5d,0x20,0x26,0x3d,0x20,0x30,0x78,0x46,0x46,0x30,0x30,0x30,0x30,0x30,0x30,0x55,0x3b,0x0a,0x28,0x28,0x5f,0x5f,0x6c,
+    0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x53,0x74,0x61,0x74,0x65,0x29,0x5b,0x31,0x30,0x5d,0x7c,0x3d,0x28,0x28,0x28,0x75,0x69,0x6e,0x74,0x29,
+    0x67,0x65,0x74,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x3e,0x3e,0x38,0x29,0x29,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,
+    0x69,0x3d,0x31,0x31,0x3b,0x20,0x69,0x3c,0x32,0x35,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x7b,0x0a,0x53,0x74,0x61,0x74,0x65,0x5b,0x69,0x5d,0x3d,0x30,0x78,0x30,0x30,
+    0x55,0x4c,0x3b,0x0a,0x7d,0x0a,0x53,0x74,0x61,0x74,0x65,0x5b,0x31,0x36,0x5d,0x3d,0x30,0x78,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,
+    0x30,0x30,0x55,0x4c,0x3b,0x0a,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x31,0x36,0x30,0x30,0x5f,0x32,0x28,0x53,0x74,0x61,0x74,0x65,0x29,0x3b,0x0a,0x23,0x70,0x72,0x61,
+    0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x32,0x35,0x3b,0x20,0x2b,
+    0x2b,0x69,0x29,0x20,0x7b,0x0a,0x73,0x74,0x61,0x74,0x65,0x73,0x5b,0x69,0x5d,0x3d,0x53,0x74,0x61,0x74,0x65,0x5b,0x69,0x5d,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x7d,0x0a,
+    0x7d,0x0a,0x5f,0x5f,0x61,0x74,0x74,0x72,0x69,0x62,0x75,0x74,0x65,0x5f,0x5f,0x28,0x28,0x72,0x65,0x71,0x64,0x5f,0x77,0x6f,0x72,0x6b,0x5f,0x67,0x72,0x6f,0x75,0x70,
+    0x5f,0x73,0x69,0x7a,0x65,0x28,0x36,0x34,0x2c,0x31,0x2c,0x31,0x29,0x29,0x29,0x0a,0x5f,0x5f,0x6b,0x65,0x72,0x6e,0x65,0x6c,0x20,0x76,0x6f,0x69,0x64,0x20,0x63,0x6e,
+    0x30,0x30,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x20,0x2a,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x2c,0x5f,0x5f,0x67,0x6c,
+    0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x73,0x74,0x61,0x74,0x65,0x73,0x29,0x0a,0x7b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,
+    0x20,0x67,0x49,0x64,0x78,0x3d,0x67,0x65,0x74,0x49,0x64,0x78,0x28,0x29,0x2f,0x36,0x34,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,
+    0x20,0x53,0x74,0x61,0x74,0x65,0x5b,0x32,0x35,0x5d,0x3b,0x0a,0x73,0x74,0x61,0x74,0x65,0x73,0x2b,0x3d,0x32,0x35,0x2a,0x67,0x49,0x64,0x78,0x3b,0x0a,0x53,0x63,0x72,
+    0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x3d,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x69,0x6e,0x74,0x2a,0x29,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,
+    0x6c,0x20,0x63,0x68,0x61,0x72,0x2a,0x29,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x2b,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x2a,0x67,0x49,0x64,0x78,0x29,0x3b,
+    0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x3b,0x20,0x69,0x3c,0x32,
+    0x35,0x3b,0x20,0x69,0x2b,0x3d,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x73,0x69,0x7a,0x65,0x28,0x30,0x29,0x29,0x20,0x7b,0x0a,0x53,0x74,0x61,0x74,0x65,
+    0x5b,0x69,0x5d,0x3d,0x73,0x74,0x61,0x74,0x65,0x73,0x5b,0x69,0x5d,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,
+    0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x75,0x69,0x6e,0x74,0x20,0x69,0x3d,0x67,0x65,0x74,0x5f,0x6c,
+    0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x3b,0x20,0x69,0x3c,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x2f,0x35,0x31,0x32,0x3b,0x20,0x69,0x2b,0x3d,0x67,0x65,0x74,
+    0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x73,0x69,0x7a,0x65,0x28,0x30,0x29,0x29,0x20,0x7b,0x0a,0x67,0x65,0x6e,0x65,0x72,0x61,0x74,0x65,0x5f,0x35,0x31,0x32,0x28,0x69,
+    0x2c,0x53,0x74,0x61,0x74,0x65,0x2c,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x2a,0x29,0x28,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,
+    0x61,0x6c,0x20,0x75,0x63,0x68,0x61,0x72,0x2a,0x29,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x2b,0x69,0x2a,0x35,0x31,0x32,0x29,0x29,0x3b,0x0a,0x7d,0x0a,
+    0x7d,0x0a,0x5f,0x5f,0x61,0x74,0x74,0x72,0x69,0x62,0x75,0x74,0x65,0x5f,0x5f,0x28,0x28,0x72,0x65,0x71,0x64,0x5f,0x77,0x6f,0x72,0x6b,0x5f,0x67,0x72,0x6f,0x75,0x70,
+    0x5f,0x73,0x69,0x7a,0x65,0x28,0x38,0x2c,0x38,0x2c,0x31,0x29,0x29,0x29,0x0a,0x5f,0x5f,0x6b,0x65,0x72,0x6e,0x65,0x6c,0x20,0x76,0x6f,0x69,0x64,0x20,0x63,0x6e,0x32,
+    0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x2a,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x2c,0x5f,0x5f,0x67,0x6c,
+    0x6f,0x62,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x73,0x74,0x61,0x74,0x65,0x73,0x2c,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,
+    0x20,0x2a,0x6f,0x75,0x74,0x70,0x75,0x74,0x2c,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x54,0x61,0x72,0x67,0x65,0x74,0x2c,0x75,0x69,0x6e,0x74,0x20,0x54,0x68,0x72,0x65,0x61,
+    0x64,0x73,0x29,0x0a,0x7b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x41,0x45,0x53,0x30,0x5b,0x32,0x35,0x36,0x5d,0x2c,0x41,0x45,0x53,
+    0x31,0x5b,0x32,0x35,0x36,0x5d,0x2c,0x41,0x45,0x53,0x32,0x5b,0x32,0x35,0x36,0x5d,0x2c,0x41,0x45,0x53,0x33,0x5b,0x32,0x35,0x36,0x5d,0x3b,0x0a,0x75,0x69,0x6e,0x74,
+    0x20,0x45,0x78,0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x5b,0x34,0x30,0x5d,0x3b,0x0a,0x75,0x69,0x6e,0x74,0x34,0x20,0x74,0x65,0x78,0x74,0x3b,0x0a,0x63,
+    0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,0x6e,0x74,0x20,0x67,0x49,0x64,0x78,0x3d,0x67,0x65,0x74,0x49,0x64,0x78,0x28,0x29,0x3b,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,
+    0x74,0x20,0x69,0x3d,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2a,0x38,0x2b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,
+    0x69,0x64,0x28,0x30,0x29,0x3b,0x20,0x69,0x3c,0x32,0x35,0x36,0x3b,0x20,0x69,0x2b,0x3d,0x38,0x2a,0x38,0x29,0x20,0x7b,0x0a,0x63,0x6f,0x6e,0x73,0x74,0x20,0x75,0x69,
+    0x6e,0x74,0x20,0x74,0x6d,0x70,0x3d,0x41,0x45,0x53,0x30,0x5f,0x43,0x5b,0x69,0x5d,0x3b,0x0a,0x41,0x45,0x53,0x30,0x5b,0x69,0x5d,0x3d,0x74,0x6d,0x70,0x3b,0x0a,0x41,
+    0x45,0x53,0x31,0x5b,0x69,0x5d,0x3d,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x74,0x6d,0x70,0x2c,0x38,0x55,0x29,0x3b,0x0a,0x41,0x45,0x53,0x32,0x5b,0x69,0x5d,0x3d,0x72,
+    0x6f,0x74,0x61,0x74,0x65,0x28,0x74,0x6d,0x70,0x2c,0x31,0x36,0x55,0x29,0x3b,0x0a,0x41,0x45,0x53,0x33,0x5b,0x69,0x5d,0x3d,0x72,0x6f,0x74,0x61,0x74,0x65,0x28,0x74,
+    0x6d,0x70,0x2c,0x32,0x34,0x55,0x29,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,
+    0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x78,0x69,0x6e,0x31,0x5b,0x38,0x5d,0x5b,0x38,
+    0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x20,0x78,0x69,0x6e,0x32,0x5b,0x38,0x5d,0x5b,0x38,0x5d,0x3b,0x0a,0x7b,0x0a,0x73,
+    0x74,0x61,0x74,0x65,0x73,0x2b,0x3d,0x32,0x35,0x2a,0x67,0x49,0x64,0x78,0x3b,0x0a,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x2b,0x3d,0x67,0x49,0x64,0x78,
+    0x2a,0x28,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x3e,0x3e,0x34,0x29,0x3b,0x0a,0x23,0x69,0x66,0x20,0x64,0x65,0x66,0x69,0x6e,0x65,0x64,0x28,0x5f,0x5f,0x54,0x61,0x68,0x69,
+    0x74,0x69,0x5f,0x5f,0x29,0x20,0x7c,0x7c,0x20,0x64,0x65,0x66,0x69,0x6e,0x65,0x64,0x28,0x5f,0x5f,0x50,0x69,0x74,0x63,0x61,0x69,0x72,0x6e,0x5f,0x5f,0x29,0x0a,0x66,
+    0x6f,0x72,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x34,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x28,0x28,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x2a,0x29,
+    0x45,0x78,0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,0x5b,0x69,0x5d,0x3d,0x73,0x74,0x61,0x74,0x65,0x73,0x5b,0x69,0x2b,0x34,0x5d,0x3b,0x0a,0x74,0x65,
+    0x78,0x74,0x3d,0x76,0x6c,0x6f,0x61,0x64,0x34,0x28,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2b,0x34,0x2c,0x28,0x5f,0x5f,0x67,
+    0x6c,0x6f,0x62,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x73,0x74,0x61,0x74,0x65,0x73,0x29,0x3b,0x0a,0x23,0x65,0x6c,0x73,0x65,0x0a,0x74,0x65,0x78,0x74,
+    0x3d,0x76,0x6c,0x6f,0x61,0x64,0x34,0x28,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2b,0x34,0x2c,0x28,0x5f,0x5f,0x67,0x6c,0x6f,
+    0x62,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x73,0x74,0x61,0x74,0x65,0x73,0x29,0x3b,0x0a,0x28,0x28,0x75,0x69,0x6e,0x74,0x38,0x20,0x2a,0x29,0x45,0x78,
+    0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,0x5b,0x30,0x5d,0x3d,0x76,0x6c,0x6f,0x61,0x64,0x38,0x28,0x31,0x2c,0x28,0x5f,0x5f,0x67,0x6c,0x6f,0x62,0x61,
+    0x6c,0x20,0x75,0x69,0x6e,0x74,0x20,0x2a,0x29,0x73,0x74,0x61,0x74,0x65,0x73,0x29,0x3b,0x0a,0x23,0x65,0x6e,0x64,0x69,0x66,0x0a,0x41,0x45,0x53,0x45,0x78,0x70,0x61,
+    0x6e,0x64,0x4b,0x65,0x79,0x32,0x35,0x36,0x28,0x45,0x78,0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,
+    0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,
+    0x75,0x69,0x6e,0x74,0x34,0x2a,0x20,0x78,0x69,0x6e,0x31,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x26,0x78,0x69,0x6e,0x31,0x5b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,
+    0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x5d,0x5b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,
+    0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x2a,0x20,0x78,0x69,0x6e,0x31,0x5f,0x6c,0x6f,0x61,0x64,0x3d,0x26,0x78,0x69,0x6e,0x31,0x5b,0x28,0x67,0x65,0x74,0x5f,0x6c,
+    0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2b,0x31,0x29,0x20,0x25,0x20,0x38,0x5d,0x5b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,
+    0x30,0x29,0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x2a,0x20,0x78,0x69,0x6e,0x32,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x26,
+    0x78,0x69,0x6e,0x32,0x5b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x5d,0x5b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,
+    0x69,0x64,0x28,0x30,0x29,0x5d,0x3b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x69,0x6e,0x74,0x34,0x2a,0x20,0x78,0x69,0x6e,0x32,0x5f,0x6c,0x6f,0x61,0x64,
+    0x3d,0x26,0x78,0x69,0x6e,0x32,0x5b,0x28,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2b,0x31,0x29,0x20,0x25,0x20,0x38,0x5d,0x5b,
+    0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x5d,0x3b,0x0a,0x2a,0x78,0x69,0x6e,0x32,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x28,0x75,
+    0x69,0x6e,0x74,0x34,0x29,0x28,0x30,0x2c,0x30,0x2c,0x30,0x2c,0x30,0x29,0x3b,0x0a,0x7b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,
+    0x20,0x32,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,0x30,0x2c,0x69,0x31,0x3d,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,
+    0x31,0x29,0x3b,0x20,0x69,0x3c,0x28,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x3e,0x3e,0x37,0x29,0x3b,0x20,0x2b,0x2b,0x69,0x2c,0x69,0x31,0x3d,0x28,0x69,0x31,0x2b,0x31,0x36,
+    0x29,0x20,0x25,0x20,0x28,0x4d,0x45,0x4d,0x4f,0x52,0x59,0x3e,0x3e,0x34,0x29,0x29,0x20,0x7b,0x0a,0x74,0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,0x53,0x63,0x72,0x61,0x74,
+    0x63,0x68,0x70,0x61,0x64,0x5b,0x28,0x75,0x69,0x6e,0x74,0x29,0x69,0x31,0x5d,0x3b,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,
+    0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x74,0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,0x2a,0x78,0x69,0x6e,0x32,0x5f,0x6c,0x6f,0x61,
+    0x64,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x30,0x0a,0x66,0x6f,0x72,0x28,0x69,0x6e,0x74,0x20,0x6a,0x3d,0x30,
+    0x3b,0x20,0x6a,0x3c,0x31,0x30,0x3b,0x20,0x2b,0x2b,0x6a,0x29,0x0a,0x74,0x65,0x78,0x74,0x3d,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x28,0x41,0x45,0x53,0x30,
+    0x2c,0x41,0x45,0x53,0x31,0x2c,0x41,0x45,0x53,0x32,0x2c,0x41,0x45,0x53,0x33,0x2c,0x74,0x65,0x78,0x74,0x2c,0x28,0x28,0x75,0x69,0x6e,0x74,0x34,0x20,0x2a,0x29,0x45,
+    0x78,0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,0x5b,0x6a,0x5d,0x29,0x3b,0x0a,0x2a,0x78,0x69,0x6e,0x31,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x74,0x65,
+    0x78,0x74,0x3b,0x0a,0x74,0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,0x53,0x63,0x72,0x61,0x74,0x63,0x68,0x70,0x61,0x64,0x5b,0x28,0x75,0x69,0x6e,0x74,0x29,0x69,0x31,0x2b,
+    0x38,0x75,0x5d,0x3b,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,
+    0x29,0x3b,0x0a,0x74,0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,0x2a,0x78,0x69,0x6e,0x31,0x5f,0x6c,0x6f,0x61,0x64,0x3b,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,
+    0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x30,0x0a,0x66,0x6f,0x72,0x28,0x69,0x6e,0x74,0x20,0x6a,0x3d,0x30,0x3b,0x20,0x6a,0x3c,0x31,0x30,0x3b,0x20,0x2b,0x2b,0x6a,0x29,
+    0x0a,0x74,0x65,0x78,0x74,0x3d,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x28,0x41,0x45,0x53,0x30,0x2c,0x41,0x45,0x53,0x31,0x2c,0x41,0x45,0x53,0x32,0x2c,0x41,
+    0x45,0x53,0x33,0x2c,0x74,0x65,0x78,0x74,0x2c,0x28,0x28,0x75,0x69,0x6e,0x74,0x34,0x20,0x2a,0x29,0x45,0x78,0x70,0x61,0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,
+    0x5b,0x6a,0x5d,0x29,0x3b,0x0a,0x2a,0x78,0x69,0x6e,0x32,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x74,0x65,0x78,0x74,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,
+    0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x74,0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,
+    0x2a,0x78,0x69,0x6e,0x32,0x5f,0x6c,0x6f,0x61,0x64,0x3b,0x0a,0x7d,0x0a,0x23,0x70,0x72,0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x36,0x0a,
+    0x66,0x6f,0x72,0x28,0x73,0x69,0x7a,0x65,0x5f,0x74,0x20,0x69,0x3d,0x30,0x3b,0x20,0x69,0x3c,0x31,0x36,0x3b,0x20,0x69,0x2b,0x2b,0x29,0x0a,0x7b,0x0a,0x23,0x70,0x72,
+    0x61,0x67,0x6d,0x61,0x20,0x75,0x6e,0x72,0x6f,0x6c,0x6c,0x20,0x31,0x30,0x0a,0x66,0x6f,0x72,0x20,0x28,0x69,0x6e,0x74,0x20,0x6a,0x3d,0x30,0x3b,0x20,0x6a,0x3c,0x31,
+    0x30,0x3b,0x20,0x2b,0x2b,0x6a,0x29,0x20,0x7b,0x0a,0x74,0x65,0x78,0x74,0x3d,0x41,0x45,0x53,0x5f,0x52,0x6f,0x75,0x6e,0x64,0x28,0x41,0x45,0x53,0x30,0x2c,0x41,0x45,
+    0x53,0x31,0x2c,0x41,0x45,0x53,0x32,0x2c,0x41,0x45,0x53,0x33,0x2c,0x74,0x65,0x78,0x74,0x2c,0x28,0x28,0x75,0x69,0x6e,0x74,0x34,0x20,0x2a,0x29,0x45,0x78,0x70,0x61,
+    0x6e,0x64,0x65,0x64,0x4b,0x65,0x79,0x32,0x29,0x5b,0x6a,0x5d,0x29,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,
+    0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x2a,0x78,0x69,0x6e,0x31,0x5f,0x73,0x74,0x6f,0x72,0x65,0x3d,0x74,0x65,0x78,0x74,0x3b,
+    0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x4c,0x4f,0x43,0x41,0x4c,0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x74,
+    0x65,0x78,0x74,0x20,0x5e,0x3d,0x20,0x2a,0x78,0x69,0x6e,0x31,0x5f,0x6c,0x6f,0x61,0x64,0x3b,0x0a,0x7d,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,
+    0x6e,0x67,0x20,0x53,0x74,0x61,0x74,0x65,0x5f,0x62,0x75,0x66,0x5b,0x38,0x2a,0x32,0x35,0x5d,0x3b,0x0a,0x7b,0x0a,0x76,0x73,0x74,0x6f,0x72,0x65,0x32,0x28,0x61,0x73,
+    0x5f,0x75,0x6c,0x6f,0x6e,0x67,0x32,0x28,0x74,0x65,0x78,0x74,0x29,0x2c,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x2b,0x34,0x2c,
+    0x73,0x74,0x61,0x74,0x65,0x73,0x29,0x3b,0x0a,0x7d,0x0a,0x62,0x61,0x72,0x72,0x69,0x65,0x72,0x28,0x43,0x4c,0x4b,0x5f,0x47,0x4c,0x4f,0x42,0x41,0x4c,0x5f,0x4d,0x45,
+    0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x7b,0x0a,0x69,0x66,0x28,0x21,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x31,0x29,0x29,
+    0x0a,0x7b,0x0a,0x5f,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x20,0x75,0x6c,0x6f,0x6e,0x67,0x2a,0x20,0x53,0x74,0x61,0x74,0x65,0x3d,0x53,0x74,0x61,0x74,0x65,0x5f,0x62,0x75,
+    0x66,0x2b,0x67,0x65,0x74,0x5f,0x6c,0x6f,0x63,0x61,0x6c,0x5f,0x69,0x64,0x28,0x30,0x29,0x2a,0x32,0x35,0x3b,0x0a,0x66,0x6f,0x72,0x28,0x69,0x6e,0x74,0x20,0x69,0x3d,
+    0x30,0x3b,0x20,0x69,0x3c,0x32,0x35,0x3b,0x20,0x2b,0x2b,0x69,0x29,0x20,0x53,0x74,0x61,0x74,0x65,0x5b,0x69,0x5d,0x3d,0x73,0x74,0x61,0x74,0x65,0x73,0x5b,0x69,0x5d,
+    0x3b,0x0a,0x6b,0x65,0x63,0x63,0x61,0x6b,0x66,0x31,0x36,0x30,0x30,0x5f,0x32,0x28,0x53,0x74,0x61,0x74,0x65,0x29,0x3b,0x0a,0x69,0x66,0x28,0x53,0x74,0x61,0x74,0x65,
+    0x5b,0x33,0x5d,0x3c,0x3d,0x54,0x61,0x72,0x67,0x65,0x74,0x29,0x0a,0x7b,0x0a,0x75,0x6c,0x6f,0x6e,0x67,0x20,0x6f,0x75,0x74,0x49,0x64,0x78,0x3d,0x61,0x74,0x6f,0x6d,
+    0x69,0x63,0x5f,0x69,0x6e,0x63,0x28,0x6f,0x75,0x74,0x70,0x75,0x74,0x2b,0x30,0x78,0x46,0x46,0x29,0x3b,0x0a,0x69,0x66,0x28,0x6f,0x75,0x74,0x49,0x64,0x78,0x3c,0x30,
+    0x78,0x46,0x46,0x29,0x0a,0x6f,0x75,0x74,0x70,0x75,0x74,0x5b,0x6f,0x75,0x74,0x49,0x64,0x78,0x5d,0x3d,0x67,0x65,0x74,0x5f,0x67,0x6c,0x6f,0x62,0x61,0x6c,0x5f,0x69,
+    0x64,0x28,0x30,0x29,0x3b,0x0a,0x7d,0x0a,0x7d,0x0a,0x7d,0x0a,0x6d,0x65,0x6d,0x5f,0x66,0x65,0x6e,0x63,0x65,0x28,0x43,0x4c,0x4b,0x5f,0x47,0x4c,0x4f,0x42,0x41,0x4c,
+    0x5f,0x4d,0x45,0x4d,0x5f,0x46,0x45,0x4e,0x43,0x45,0x29,0x3b,0x0a,0x7d,0x0a,0x00
+};
+
+} // namespace xmrig

src/backend/opencl/cl/rx/randomx.cl

@@ -4,7 +4,7 @@
 #include "randomx_constants_monero.h"
 #elif (ALGO == ALGO_RX_WOW)
 #include "randomx_constants_wow.h"
-#elif (ALGO == ALGO_RX_ARQMA)
+#elif (ALGO == ALGO_RX_ARQ)
 #include "randomx_constants_arqma.h"
 #elif (ALGO == ALGO_RX_KEVA)
 #include "randomx_constants_keva.h"

src/backend/opencl/generators/ocl_generic_cn_gpu_generator.cpp

@@ -0,0 +1,87 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018      Lee Clagett <https://github.com/vtnerd>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ *
+ *   This program is free software: you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, either version 3 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+
+#include "backend/opencl/OclThreads.h"
+#include "backend/opencl/wrappers/OclDevice.h"
+#include "base/crypto/Algorithm.h"
+
+
+#include <algorithm>
+
+
+namespace xmrig {
+
+
+constexpr const size_t oneMiB = 1024u * 1024u;
+
+
+
+bool ocl_generic_cn_gpu_generator(const OclDevice &device, const Algorithm &algorithm, OclThreads &threads)
+{
+    if (algorithm != Algorithm::CN_GPU) {
+        return false;
+    }
+
+    uint32_t worksize   = 8;
+    uint32_t numThreads = 1u;
+    size_t minFreeMem   = 128u * oneMiB;
+
+    if (device.type() == OclDevice::Vega_10 || device.type() == OclDevice::Vega_20) {
+        minFreeMem  = oneMiB;
+        worksize    = 16;
+    }
+    else if (device.type() == OclDevice::Navi_10) {
+        numThreads = 2u;
+    }
+    else if (device.name() == "Fiji") {
+        worksize = 16;
+    }
+
+    size_t maxThreads = device.computeUnits() * 6 * 8;
+
+    const size_t maxAvailableFreeMem = device.freeMemSize() - minFreeMem;
+    const size_t memPerThread        = std::min(device.maxMemAllocSize(), maxAvailableFreeMem);
+
+    size_t memPerHash           = algorithm.l3() + 240u;
+    size_t maxIntensity         = memPerThread / memPerHash;
+    size_t possibleIntensity    = std::min(maxThreads, maxIntensity);
+    size_t intensity            = 0;
+    size_t cuUtilization        = ((possibleIntensity * 100)  / (worksize * device.computeUnits())) % 100;
+
+    if (cuUtilization >= 75) {
+        intensity = (possibleIntensity / worksize) * worksize;
+    }
+    else {
+        intensity = (possibleIntensity / (worksize * device.computeUnits())) * device.computeUnits() * worksize;
+    }
+
+    threads.add(OclThread(device.index(), intensity, worksize, numThreads, 1));
+
+    return true;
+}
+
+
+} // namespace xmrig

src/backend/opencl/kernels/Cn00RyoKernel.cpp

@@ -0,0 +1,44 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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 "backend/opencl/kernels/Cn00RyoKernel.h"
+#include "backend/opencl/wrappers/OclLib.h"
+
+
+void xmrig::Cn00RyoKernel::enqueue(cl_command_queue queue, size_t threads)
+{
+    const size_t gthreads = threads * 64;
+    const size_t lthreads = 64;
+
+    enqueueNDRange(queue, 1, nullptr, &gthreads, &lthreads);
+}
+
+
+// __kernel void cn00(__global int *Scratchpad, __global ulong *states)
+void xmrig::Cn00RyoKernel::setArgs(cl_mem scratchpads, cl_mem states)
+{
+    setArg(0, sizeof(cl_mem), &scratchpads);
+    setArg(1, sizeof(cl_mem), &states);
+}

src/backend/opencl/kernels/Cn00RyoKernel.h

@@ -0,0 +1,48 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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/>.
+ */
+
+#ifndef XMRIG_CN00RYOKERNEL_H
+#define XMRIG_CN00RYOKERNEL_H
+
+
+#include "backend/opencl/wrappers/OclKernel.h"
+
+
+namespace xmrig {
+
+
+class Cn00RyoKernel : public OclKernel
+{
+public:
+    inline Cn00RyoKernel(cl_program program) : OclKernel(program, "cn00") {}
+
+    void enqueue(cl_command_queue queue, size_t threads);
+    void setArgs(cl_mem scratchpads, cl_mem states);
+};
+
+
+} // namespace xmrig
+
+
+#endif /* XMRIG_CN00RYOKERNEL_H */

src/backend/opencl/kernels/Cn1RyoKernel.cpp

@@ -0,0 +1,48 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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 <string>
+
+
+#include "backend/opencl/kernels/Cn1RyoKernel.h"
+#include "backend/opencl/wrappers/OclLib.h"
+
+
+void xmrig::Cn1RyoKernel::enqueue(cl_command_queue queue, size_t threads, size_t worksize)
+{
+    const size_t gthreads = threads * 16;
+    const size_t lthreads = worksize * 16;
+
+    enqueueNDRange(queue, 1, nullptr, &gthreads, &lthreads);
+}
+
+
+// __kernel void cn1(__global int *lpad_in, __global int *spad, uint numThreads)
+void xmrig::Cn1RyoKernel::setArgs(cl_mem scratchpads, cl_mem states, uint32_t threads)
+{
+    setArg(0, sizeof(cl_mem), &scratchpads);
+    setArg(1, sizeof(cl_mem), &states);
+    setArg(2, sizeof(uint32_t), &threads);
+}

src/backend/opencl/kernels/Cn1RyoKernel.h

@@ -0,0 +1,48 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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/>.
+ */
+
+#ifndef XMRIG_CN1RYOKERNEL_H
+#define XMRIG_CN1RYOKERNEL_H
+
+
+#include "backend/opencl/wrappers/OclKernel.h"
+
+
+namespace xmrig {
+
+
+class Cn1RyoKernel : public OclKernel
+{
+public:
+    inline Cn1RyoKernel(cl_program program) : OclKernel(program, "cn1") {}
+
+    void enqueue(cl_command_queue queue, size_t threads, size_t worksize);
+    void setArgs(cl_mem scratchpads, cl_mem states, uint32_t threads);
+};
+
+
+} // namespace xmrig
+
+
+#endif /* XMRIG_CN1RYOKERNEL_H */

src/backend/opencl/kernels/Cn2RyoKernel.cpp

@@ -0,0 +1,53 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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 "backend/opencl/kernels/Cn2RyoKernel.h"
+#include "backend/opencl/wrappers/OclLib.h"
+
+
+void xmrig::Cn2RyoKernel::enqueue(cl_command_queue queue, uint32_t nonce, size_t threads)
+{
+    const size_t offset[2]          = { nonce, 1 };
+    const size_t gthreads[2]        = { threads, 8 };
+    static const size_t lthreads[2] = { 8, 8 };
+
+    enqueueNDRange(queue, 2, offset, gthreads, lthreads);
+}
+
+
+// __kernel void cn2(__global uint4 *Scratchpad, __global ulong *states, __global uint *output, ulong Target, uint Threads)
+void xmrig::Cn2RyoKernel::setArgs(cl_mem scratchpads, cl_mem states, cl_mem output, uint32_t threads)
+{
+    setArg(0, sizeof(cl_mem), &scratchpads);
+    setArg(1, sizeof(cl_mem), &states);
+    setArg(2, sizeof(cl_mem), &output);
+    setArg(4, sizeof(uint32_t), &threads);
+}
+
+
+void xmrig::Cn2RyoKernel::setTarget(uint64_t target)
+{
+    setArg(3, sizeof(cl_ulong), &target);
+}

src/backend/opencl/kernels/Cn2RyoKernel.h

@@ -0,0 +1,49 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * 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
+ *   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/>.
+ */
+
+#ifndef XMRIG_CN2RYOKERNEL_H
+#define XMRIG_CN2RYOKERNEL_H
+
+
+#include "backend/opencl/wrappers/OclKernel.h"
+
+
+namespace xmrig {
+
+
+class Cn2RyoKernel : public OclKernel
+{
+public:
+    inline Cn2RyoKernel(cl_program program) : OclKernel(program, "cn2") {}
+
+    void enqueue(cl_command_queue queue, uint32_t nonce, size_t threads);
+    void setArgs(cl_mem scratchpads, cl_mem states, cl_mem output, uint32_t threads);
+    void setTarget(uint64_t target);
+};
+
+
+} // namespace xmrig
+
+
+#endif /* XMRIG_CN2RYOKERNEL_H */

src/backend/opencl/opencl.cmake

@@ -146,6 +146,23 @@ if (WITH_OPENCL)
              )
     endif()
 
+    if (WITH_CN_GPU AND CMAKE_SIZEOF_VOID_P EQUAL 8)
+        list(APPEND HEADERS_BACKEND_OPENCL
+             src/backend/opencl/kernels/Cn00RyoKernel.h
+             src/backend/opencl/kernels/Cn1RyoKernel.h
+             src/backend/opencl/kernels/Cn2RyoKernel.h
+             src/backend/opencl/runners/OclRyoRunner.h
+             )
+
+        list(APPEND SOURCES_BACKEND_OPENCL
+             src/backend/opencl/generators/ocl_generic_cn_gpu_generator.cpp
+             src/backend/opencl/kernels/Cn00RyoKernel.cpp
+             src/backend/opencl/kernels/Cn1RyoKernel.cpp
+             src/backend/opencl/kernels/Cn2RyoKernel.cpp
+             src/backend/opencl/runners/OclRyoRunner.cpp
+             )
+    endif()
+
     if (WITH_STRICT_CACHE)
         add_definitions(/DXMRIG_STRICT_OPENCL_CACHE)
     else()

src/backend/opencl/runners/OclRyoRunner.cpp

@@ -0,0 +1,128 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ *
+ *   This program is free software: you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, either version 3 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+
+#include "backend/opencl/runners/OclRyoRunner.h"
+#include "backend/opencl/kernels/Cn00RyoKernel.h"
+#include "backend/opencl/kernels/Cn0Kernel.h"
+#include "backend/opencl/kernels/Cn1RyoKernel.h"
+#include "backend/opencl/kernels/Cn2RyoKernel.h"
+#include "backend/opencl/kernels/CnBranchKernel.h"
+#include "backend/opencl/OclLaunchData.h"
+#include "backend/opencl/wrappers/OclLib.h"
+#include "base/io/log/Log.h"
+#include "base/net/stratum/Job.h"
+#include "crypto/cn/CnAlgo.h"
+
+
+xmrig::OclRyoRunner::OclRyoRunner(size_t index, const OclLaunchData &data) : OclBaseRunner(index, data)
+{
+    m_options += " -DITERATIONS="   + std::to_string(CnAlgo<>::iterations(m_algorithm)) + "U";
+    m_options += " -DMASK="         + std::to_string(CnAlgo<>::mask(m_algorithm)) + "U";
+    m_options += " -DWORKSIZE="     + std::to_string(data.thread.worksize()) + "U";
+    m_options += " -DMEMORY="       + std::to_string(m_algorithm.l3()) + "LU";
+    m_options += " -DCN_UNROLL="    + std::to_string(data.thread.unrollFactor());
+
+    m_options += " -cl-fp32-correctly-rounded-divide-sqrt";
+}
+
+
+xmrig::OclRyoRunner::~OclRyoRunner()
+{
+    delete m_cn00;
+    delete m_cn0;
+    delete m_cn1;
+    delete m_cn2;
+
+    OclLib::release(m_scratchpads);
+    OclLib::release(m_states);
+}
+
+
+size_t xmrig::OclRyoRunner::bufferSize() const
+{
+    return OclBaseRunner::bufferSize() + align(data().algorithm.l3() * m_intensity) + align(200 * m_intensity);
+}
+
+
+void xmrig::OclRyoRunner::run(uint32_t nonce, uint32_t *hashOutput)
+{
+    static const cl_uint zero = 0;
+
+    const size_t w_size = data().thread.worksize();
+    const size_t g_thd  = ((m_intensity + w_size - 1u) / w_size) * w_size;
+
+    assert(g_thd % w_size == 0);
+
+    enqueueWriteBuffer(m_output, CL_FALSE, sizeof(cl_uint) * 0xFF, sizeof(cl_uint), &zero);
+
+    m_cn0->enqueue(m_queue, nonce, g_thd);
+    m_cn00->enqueue(m_queue, g_thd);
+    m_cn1->enqueue(m_queue, g_thd, w_size);
+    m_cn2->enqueue(m_queue, nonce, g_thd);
+
+    finalize(hashOutput);
+}
+
+
+void xmrig::OclRyoRunner::set(const Job &job, uint8_t *blob)
+{
+    if (job.size() > (Job::kMaxBlobSize - 4)) {
+        throw std::length_error("job size too big");
+    }
+
+    blob[job.size()] = 0x01;
+    memset(blob + job.size() + 1, 0, Job::kMaxBlobSize - job.size() - 1);
+
+    enqueueWriteBuffer(m_input, CL_TRUE, 0, Job::kMaxBlobSize, blob);
+
+    m_cn2->setTarget(job.target());
+}
+
+
+void xmrig::OclRyoRunner::build()
+{
+    OclBaseRunner::build();
+
+    m_cn00 = new Cn00RyoKernel(m_program);
+    m_cn00->setArgs(m_scratchpads, m_states);
+
+    m_cn0 = new Cn0Kernel(m_program);
+    m_cn0->setArgs(m_input, 0, m_scratchpads, m_states, m_intensity);
+
+    m_cn1 = new Cn1RyoKernel(m_program);
+    m_cn1->setArgs(m_scratchpads, m_states, m_intensity);
+
+    m_cn2 = new Cn2RyoKernel(m_program);
+    m_cn2->setArgs(m_scratchpads, m_states, m_output, m_intensity);
+}
+
+
+void xmrig::OclRyoRunner::init()
+{
+    OclBaseRunner::init();
+
+    m_scratchpads = createSubBuffer(CL_MEM_READ_WRITE, data().algorithm.l3() * m_intensity);
+    m_states      = createSubBuffer(CL_MEM_READ_WRITE, 200 * m_intensity);
+}

src/backend/opencl/runners/OclRyoRunner.h

@@ -0,0 +1,70 @@
+/* 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 2017-2018 XMR-Stak    <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
+ * Copyright 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 XMRig       <https://github.com/xmrig>, <support@xmrig.com>
+ *
+ *   This program is free software: you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, either version 3 of the License, or
+ *   (at your option) any later version.
+ *
+ *   This program is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef XMRIG_OCLRYORUNNER_H
+#define XMRIG_OCLRYORUNNER_H
+
+
+#include "backend/opencl/runners/OclBaseRunner.h"
+
+
+namespace xmrig {
+
+
+class Cn00RyoKernel;
+class Cn0Kernel;
+class Cn1RyoKernel;
+class Cn2RyoKernel;
+
+
+class OclRyoRunner : public OclBaseRunner
+{
+public:
+    XMRIG_DISABLE_COPY_MOVE_DEFAULT(OclRyoRunner)
+
+    OclRyoRunner(size_t index, const OclLaunchData &data);
+
+    ~OclRyoRunner() override;
+
+protected:
+    size_t bufferSize() const override;
+    void run(uint32_t nonce, uint32_t *hashOutput) override;
+    void set(const Job &job, uint8_t *blob) override;
+    void build() override;
+    void init() override;
+
+private:
+    cl_mem m_scratchpads    = nullptr;
+    cl_mem m_states         = nullptr;
+    Cn00RyoKernel *m_cn00   = nullptr;
+    Cn0Kernel *m_cn0        = nullptr;
+    Cn1RyoKernel *m_cn1     = nullptr;
+    Cn2RyoKernel *m_cn2     = nullptr;
+};
+
+
+} /* namespace xmrig */
+
+
+#endif // XMRIG_OCLRYORUNNER_H

src/backend/opencl/wrappers/OclDevice.cpp

@@ -64,6 +64,10 @@ extern bool ocl_generic_kawpow_generator(const OclDevice& device, const Algorith
 extern bool ocl_vega_cn_generator(const OclDevice &device, const Algorithm &algorithm, OclThreads &threads);
 extern bool ocl_generic_cn_generator(const OclDevice &device, const Algorithm &algorithm, OclThreads &threads);
 
+#ifdef XMRIG_ALGO_CN_GPU
+extern bool ocl_generic_cn_gpu_generator(const OclDevice &device, const Algorithm &algorithm, OclThreads &threads);
+#endif
+
 
 static ocl_gen_config_fun generators[] = {
 #   ifdef XMRIG_ALGO_RANDOMX
@@ -76,7 +80,10 @@ static ocl_gen_config_fun generators[] = {
     ocl_generic_kawpow_generator,
 #   endif
     ocl_vega_cn_generator,
-    ocl_generic_cn_generator
+    ocl_generic_cn_generator,
+#   ifdef XMRIG_ALGO_CN_GPU
+    ocl_generic_cn_gpu_generator,
+#   endif
 };
 
 

src/base/crypto/Algorithm.cpp

@@ -111,6 +111,8 @@ static AlgoName const algorithm_names[] = {
     { "RandomSFX",                 nullptr,            Algorithm::RX_SFX          },
     { "randomx/keva",              "rx/keva",          Algorithm::RX_KEVA         },
     { "RandomKEVA",                nullptr,            Algorithm::RX_KEVA         },
+    { "panthera",                  "panthera",         Algorithm::RX_XLA        },
+    { "Panthera",                  "panthera",         Algorithm::RX_XLA        },
 #   endif
 #   ifdef XMRIG_ALGO_ARGON2
     { "argon2/chukwa",             nullptr,            Algorithm::AR2_CHUKWA      },
@@ -130,6 +132,10 @@ static AlgoName const algorithm_names[] = {
 #   endif
     { "cryptonight/ccx",           "cn/ccx",           Algorithm::CN_CCX          },
     { "cryptonight/conceal",       "cn/conceal",       Algorithm::CN_CCX          },
+#   ifdef XMRIG_ALGO_CN_GPU
+    { "cryptonight/gpu",           "cn/gpu",           Algorithm::CN_GPU          },
+    { "cryptonight_gpu",           nullptr,            Algorithm::CN_GPU          },
+#   endif
 #   ifdef XMRIG_ALGO_CN_FEMTO
     { "cryptonight/upx2",          "cn/upx2",          Algorithm::CN_UPX2         },
     { "cn-extremelite/upx2",       nullptr,            Algorithm::CN_UPX2         },
@@ -171,6 +177,7 @@ size_t xmrig::Algorithm::l2() const
 
     case RX_WOW:
     case RX_KEVA:
+    case RX_XLA:
         return 0x20000;
 
     case RX_ARQ:
@@ -224,6 +231,7 @@ size_t xmrig::Algorithm::l3() const
             return oneMiB;
 
         case RX_ARQ:
+        case RX_XLA:
             return oneMiB / 4;
 
         default:
@@ -298,6 +306,12 @@ uint32_t xmrig::Algorithm::maxIntensity() const
     }
 #   endif
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    if (m_id == CN_GPU) {
+        return 1;
+    }
+#   endif
+
     return 5;
 }
 
@@ -317,6 +331,9 @@ xmrig::Algorithm::Family xmrig::Algorithm::family(Id id)
     case CN_ZLS:
     case CN_DOUBLE:
     case CN_CCX:
+#   ifdef XMRIG_ALGO_CN_GPU
+    case CN_GPU:
+#   endif
         return CN;
 
 #   ifdef XMRIG_ALGO_CN_LITE
@@ -349,6 +366,7 @@ xmrig::Algorithm::Family xmrig::Algorithm::family(Id id)
     case RX_ARQ:
     case RX_SFX:
     case RX_KEVA:
+    case RX_XLA:
         return RANDOM_X;
 #   endif
 

src/base/crypto/Algorithm.h

@@ -44,7 +44,6 @@ public:
     // src/backend/opencl/cl/cn/algorithm.cl
     //
     enum Id : int {
-        INVALID = -1,
         CN_0,          // "cn/0"             CryptoNight (original).
         CN_1,          // "cn/1"             CryptoNight variant 1 also known as Monero7 and CryptoNightV7.
         CN_2,          // "cn/2"             CryptoNight variant 2.
@@ -64,7 +63,11 @@ public:
         CN_PICO_0,     // "cn-pico"          CryptoNight-Pico
         CN_PICO_TLO,   // "cn-pico/tlo"      CryptoNight-Pico (TLO)
         CN_CCX,        // "cn/ccx"           Conceal (CCX)
+        CN_GPU,        // "cn/gpu"           CryptoNight-GPU (Ryo).
         CN_UPX2,       // "cn/upx2"          Uplexa (UPX2)
+        // CryptoNight variants must be above this line
+        // (index of RX_0 is used in loops as "end of all CN families" marker)
+        // next line MUST be RX_0
         RX_0,          // "rx/0"             RandomX (reference configuration).
         RX_WOW,        // "rx/wow"           RandomWOW (Wownero).
         RX_ARQ,        // "rx/arq"           RandomARQ (Arqma).
@@ -75,7 +78,10 @@ public:
         AR2_WRKZ,      // "argon2/wrkz"      Argon2id (WRKZ)
         ASTROBWT_DERO, // "astrobwt"         AstroBWT (Dero)
         KAWPOW_RVN,    // "kawpow/rvn"       KawPow (RVN)
-        MAX
+        RX_XLA,        // "panthera"         Panthera (Scala2).
+        MAX,
+        MIN = 0,
+        INVALID = -1,
     };
 
     enum Family : int {

src/base/crypto/Coin.cpp

@@ -49,8 +49,8 @@ struct CoinName
 static CoinName const coin_names[] = {
     { "monero",     Coin::MONERO  },
     { "xmr",        Coin::MONERO  },
-    { "arqma",      Coin::ARQMA   },
+    { "arqma",      Coin::ARQ     },
-    { "arq",        Coin::ARQMA   },
+    { "arq",        Coin::ARQ     },
     { "dero",       Coin::DERO    },
     { "keva",       Coin::KEVA    },
     { "ravencoin",  Coin::RAVEN   },
@@ -71,7 +71,7 @@ xmrig::Algorithm::Id xmrig::Coin::algorithm(uint8_t blobVersion) const
     case MONERO:
         return (blobVersion >= 12) ? Algorithm::RX_0 : Algorithm::CN_R;
 
-    case ARQMA:
+    case ARQ:
         return (blobVersion >= 15) ? Algorithm::RX_ARQ : Algorithm::CN_PICO_0;
 
     case DERO:

src/base/crypto/Coin.h

@@ -40,7 +40,7 @@ public:
     enum Id : int {
         INVALID = -1,
         MONERO,
-        ARQMA,
+        ARQ,
         DERO,
         KEVA,
         RAVEN,

src/base/io/log/Log.h

@@ -97,6 +97,7 @@ private:
 #define WHITE_S             CSI "0;37m" // another name for LT.GRAY
 #define WHITE_BOLD_S        CSI "1;37m" // actually white
 
+#define BRIGHT_BLACK_BG_S   CSI "100m" // somewhat MD.GRAY
 #define RED_BG_BOLD_S       CSI "41;1m"
 #define GREEN_BG_BOLD_S     CSI "42;1m"
 #define YELLOW_BG_BOLD_S    CSI "43;1m"
@@ -125,6 +126,7 @@ private:
 #define WHITE(x)            WHITE_S x CLEAR
 #define WHITE_BOLD(x)       WHITE_BOLD_S x CLEAR
 
+#define BRIGHT_BLACK_BG(x)  BRIGHT_BLACK_BG_S x CLEAR
 #define RED_BG_BOLD(x)      RED_BG_BOLD_S x CLEAR
 #define GREEN_BG_BOLD(x)    GREEN_BG_BOLD_S x CLEAR
 #define YELLOW_BG_BOLD(x)   YELLOW_BG_BOLD_S x CLEAR

src/base/io/log/Tags.cpp

@@ -121,6 +121,15 @@ const char *xmrig::Tags::opencl()
 #endif
 
 
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+const char *xmrig::Tags::benchmark()
+{
+    static const char *tag = BRIGHT_BLACK_BG(CYAN_BOLD_S " benchmk ");
+
+    return tag;
+}
+#endif
+
 #ifdef XMRIG_FEATURE_PROFILING
 const char* xmrig::Tags::profiler()
 {

src/base/io/log/Tags.h

@@ -58,6 +58,10 @@ public:
     static const char *opencl();
 #   endif
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    static const char *benchmark();
+#   endif
+
 #   ifdef XMRIG_FEATURE_PROFILING
     static const char* profiler();
 #   endif

src/base/kernel/config/BaseConfig.cpp

@@ -47,16 +47,25 @@
 namespace xmrig {
 
 
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+const char *BaseConfig::kAlgoPerf       = "algo-perf";
+#endif
 const char *BaseConfig::kApi            = "api";
 const char *BaseConfig::kApiId          = "id";
 const char *BaseConfig::kApiWorkerId    = "worker-id";
 const char *BaseConfig::kAutosave       = "autosave";
 const char *BaseConfig::kBackground     = "background";
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+const char *BaseConfig::kBenchAlgoTime  = "bench-algo-time";
+#endif
 const char *BaseConfig::kColors         = "colors";
 const char *BaseConfig::kDryRun         = "dry-run";
 const char *BaseConfig::kHttp           = "http";
 const char *BaseConfig::kLogFile        = "log-file";
 const char *BaseConfig::kPrintTime      = "print-time";
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+const char *BaseConfig::kRebenchAlgo    = "rebench-algo";
+#endif
 const char *BaseConfig::kSyslog         = "syslog";
 const char *BaseConfig::kTitle          = "title";
 const char *BaseConfig::kUserAgent      = "user-agent";
@@ -83,6 +92,9 @@ bool xmrig::BaseConfig::read(const IJsonReader &reader, const char *fileName)
     m_autoSave          = reader.getBool(kAutosave, m_autoSave);
     m_background        = reader.getBool(kBackground, m_background);
     m_dryRun            = reader.getBool(kDryRun, m_dryRun);
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    m_rebenchAlgo  = reader.getBool(kRebenchAlgo, m_rebenchAlgo);
+#   endif
     m_syslog            = reader.getBool(kSyslog, m_syslog);
     m_watch             = reader.getBool(kWatch, m_watch);
     m_logFile           = reader.getString(kLogFile);
@@ -95,6 +107,9 @@ bool xmrig::BaseConfig::read(const IJsonReader &reader, const char *fileName)
 #   endif
 
     Log::setColors(reader.getBool(kColors, Log::isColors()));
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    m_benchAlgoTime = reader.getInt(kBenchAlgoTime, m_benchAlgoTime);
+#   endif
     setVerbose(reader.getValue(kVerbose));
 
     const auto &api = reader.getObject(kApi);

src/base/kernel/config/BaseConfig.h

@@ -40,16 +40,25 @@ class IJsonReader;
 class BaseConfig : public IConfig
 {
 public:
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    static const char *kAlgoPerf;
+#   endif
     static const char *kApi;
     static const char *kApiId;
     static const char *kApiWorkerId;
     static const char *kAutosave;
     static const char *kBackground;
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    static const char *kBenchAlgoTime;
+#   endif
     static const char *kColors;
     static const char *kDryRun;
     static const char *kHttp;
     static const char *kLogFile;
     static const char *kPrintTime;
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    static const char *kRebenchAlgo;
+#   endif
     static const char *kSyslog;
     static const char *kTitle;
     static const char *kUserAgent;
@@ -75,6 +84,11 @@ public:
     inline const Title &title() const                       { return m_title; }
     inline uint32_t printTime() const                       { return m_printTime; }
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    inline bool isRebenchAlgo() const                       { return m_rebenchAlgo; }
+    inline int  benchAlgoTime() const                       { return m_benchAlgoTime; }
+#   endif
+
 #   ifdef XMRIG_FEATURE_TLS
     inline const TlsConfig &tls() const                     { return m_tls; }
 #   endif
@@ -105,6 +119,11 @@ protected:
     Title m_title;
     uint32_t m_printTime    = 60;
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    bool m_rebenchAlgo   = false;
+    int  m_benchAlgoTime = 10;
+#   endif
+
 #   ifdef XMRIG_FEATURE_TLS
     TlsConfig m_tls;
 #   endif

src/base/kernel/config/BaseTransform.cpp

@@ -243,6 +243,9 @@ void xmrig::BaseTransform::transform(rapidjson::Document &doc, int key, const ch
 #   endif
 
     case IConfig::RetriesKey:       /* --retries */
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    case IConfig::BenchAlgoTimeKey: /* --bench-algo-time */
+#   endif
     case IConfig::RetryPauseKey:    /* --retry-pause */
     case IConfig::PrintTimeKey:     /* --print-time */
     case IConfig::HttpPort:         /* --http-port */
@@ -256,10 +259,18 @@ void xmrig::BaseTransform::transform(rapidjson::Document &doc, int key, const ch
     case IConfig::SyslogKey:      /* --syslog */
     case IConfig::KeepAliveKey:   /* --keepalive */
     case IConfig::NicehashKey:    /* --nicehash */
+#   ifdef XMRIG_FEATURE_TLS
     case IConfig::TlsKey:         /* --tls */
+#   endif
     case IConfig::DryRunKey:      /* --dry-run */
+#   ifdef XMRIG_FEATURE_HTTP
     case IConfig::HttpEnabledKey: /* --http-enabled */
     case IConfig::DaemonKey:      /* --daemon */
+#   endif
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    case IConfig::RebenchAlgoKey: /* --rebench-algo */
+#   endif
+    case IConfig::PauseOnBatteryKey: /* --pause-on-battery */
     case IConfig::SubmitToOriginKey: /* --submit-to-origin */
     case IConfig::VerboseKey:     /* --verbose */
     case IConfig::DnsIPv6Key:     /* --dns-ipv6 */
@@ -323,6 +334,11 @@ void xmrig::BaseTransform::transformBoolean(rapidjson::Document &doc, int key, b
     case IConfig::NoTitleKey: /* --no-title */
         return set(doc, BaseConfig::kTitle, enable);
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    case IConfig::RebenchAlgoKey: /* --rebench-algo */
+        return set(doc, BaseConfig::kRebenchAlgo, enable);
+#   endif
+
     case IConfig::DnsIPv6Key: /* --dns-ipv6 */
         return set(doc, DnsConfig::kField, DnsConfig::kIPv6, enable);
 
@@ -365,6 +381,11 @@ void xmrig::BaseTransform::transformUint64(rapidjson::Document &doc, int key, ui
         return add(doc, Pools::kPools, Pool::kDaemonZMQPort, arg);
 #   endif
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    case IConfig::BenchAlgoTimeKey: /* --bench-algo-time */
+        return set(doc, BaseConfig::kBenchAlgoTime, arg);
+#   endif
+
     default:
         break;
     }

src/base/kernel/interfaces/IConfig.h

@@ -93,6 +93,11 @@ public:
         NicehashKey          = 1006,
         PrintTimeKey         = 1007,
 
+#       ifdef XMRIG_FEATURE_MO_BENCHMARK
+        RebenchAlgoKey       = 10001,
+        BenchAlgoTimeKey     = 10002,
+#       endif
+
         // xmrig cpu
         CPUKey               = 1024,
         AVKey                = 'v',

src/base/net/stratum/AutoClient.cpp

@@ -20,6 +20,7 @@
 #include "base/net/stratum/AutoClient.h"
 #include "3rdparty/rapidjson/document.h"
 #include "base/io/json/Json.h"
+#include "net/JobResult.h"
 
 
 xmrig::AutoClient::AutoClient(int id, const char *agent, IClientListener *listener) :
@@ -72,7 +73,7 @@ bool xmrig::AutoClient::parseLogin(const rapidjson::Value &result, int *code)
 
 int64_t xmrig::AutoClient::submit(const JobResult &result)
 {
-    if (m_mode == DEFAULT_MODE) {
+    if (result.algorithm.family() != Algorithm::KAWPOW) {
         return Client::submit(result);
     }
 
@@ -82,9 +83,11 @@ int64_t xmrig::AutoClient::submit(const JobResult &result)
 
 void xmrig::AutoClient::parseNotification(const char *method, const rapidjson::Value &params, const rapidjson::Value &error)
 {
-    if (m_mode == DEFAULT_MODE) {
+    if (strcmp(method, "job") == 0) {
+        m_mode = DEFAULT_MODE;
         return Client::parseNotification(method, params, error);
     }
 
+    m_mode = ETH_MODE;
     return EthStratumClient::parseNotification(method, params, error);
 }

src/base/net/stratum/Client.cpp

@@ -182,6 +182,8 @@ int64_t xmrig::Client::send(const rapidjson::Value &obj)
 
 int64_t xmrig::Client::submit(const JobResult &result)
 {
+    if (m_rpcId.isNull()) return 0; // ignore leftout benchmark jobs
+
 #   ifndef XMRIG_PROXY_PROJECT
     if (result.clientId != m_rpcId || m_rpcId.isNull() || m_state != ConnectedState) {
         return -1;

src/base/net/stratum/EthStratumClient.cpp

@@ -177,7 +177,7 @@ void xmrig::EthStratumClient::parseNotification(const char *method, const rapidj
         Job job;
         job.setId(arr[0].GetString());
 
-        auto algo = m_pool.algorithm();
+        auto algo = Algorithm(Algorithm::KAWPOW_RVN); //m_pool.algorithm();
         if (!algo.isValid()) {
             algo = m_pool.coin().algorithm();
         }

src/base/net/stratum/Pool.h

@@ -46,7 +46,9 @@ public:
         MODE_POOL,
         MODE_DAEMON,
         MODE_SELF_SELECT,
+#       ifdef XMRIG_ALGO_KAWPOW
         MODE_AUTO_ETH,
+#       endif
 #       ifdef XMRIG_FEATURE_BENCHMARK
         MODE_BENCHMARK,
 #       endif
@@ -146,7 +148,11 @@ private:
     bool m_submitToOrigin           = false;
     Coin m_coin;
     int m_keepAlive                 = 0;
+#   ifdef XMRIG_ALGO_KAWPOW
+    Mode m_mode                     = MODE_AUTO_ETH;
+#   else
     Mode m_mode                     = MODE_POOL;
+#   endif
     ProxyUrl m_proxy;
     std::bitset<FLAG_MAX> m_flags   = 0;
     String m_fingerprint;

src/base/net/stratum/Pools.cpp

@@ -147,6 +147,7 @@ void xmrig::Pools::load(const IJsonReader &reader)
         return;
     }
 
+    bool mo = false;
     for (const rapidjson::Value &value : pools.GetArray()) {
         if (!value.IsObject()) {
             continue;
@@ -154,10 +155,12 @@ void xmrig::Pools::load(const IJsonReader &reader)
 
         Pool pool(value);
         if (pool.isValid()) {
+            if (m_data.empty() && strstr(pool.host(), "moneroocean.stream")) mo = true;
             m_data.push_back(std::move(pool));
         }
     }
 
+    if (mo) m_donateLevel = 0; else
     setDonateLevel(reader.getInt(kDonateLevel, kDefaultDonateLevel));
     setProxyDonate(reader.getInt(kDonateOverProxy, PROXY_DONATE_AUTO));
     setRetries(reader.getInt(kRetries));

src/config.json

@@ -16,7 +16,7 @@
     "title": true,
     "randomx": {
         "init": -1,
-        "init-avx2": -1,
+        "init-avx2": 0,
         "mode": "auto",
         "1gb-pages": false,
         "rdmsr": true,
@@ -31,7 +31,7 @@
         "huge-pages-jit": false,
         "hw-aes": null,
         "priority": null,
-        "memory-pool": false,
+        "memory-pool": true,
         "yield": true,
         "max-threads-hint": 100,
         "asm": true,
@@ -48,14 +48,17 @@
         "platform": "AMD",
         "adl": true,
         "cn/0": false,
-        "cn-lite/0": false
+        "cn-lite/0": false,
+        "panthera": false
     },
     "cuda": {
         "enabled": false,
         "loader": null,
         "nvml": true,
         "cn/0": false,
-        "cn-lite/0": false
+        "cn-lite/0": false,
+        "panthera": false,
+        "astrobwt": false
     },
     "donate-level": 1,
     "donate-over-proxy": 1,
@@ -64,12 +67,12 @@
         {
             "algo": null,
             "coin": null,
-            "url": "donate.v2.xmrig.com:3333",
+            "url": "gulf.moneroocean.stream:10128",
             "user": "YOUR_WALLET_ADDRESS",
             "pass": "x",
             "rig-id": null,
             "nicehash": false,
-            "keepalive": false,
+            "keepalive": true,
             "enabled": true,
             "tls": false,
             "tls-fingerprint": null,
@@ -97,6 +100,8 @@
     "user-agent": null,
     "verbose": 0,
     "watch": true,
+    "rebench-algo": false,
+    "bench-algo-time": 20,
     "pause-on-battery": false,
     "pause-on-active": false
 }

src/core/Controller.cpp

@@ -62,12 +62,19 @@ int xmrig::Controller::init()
     return 0;
 }
 
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+void xmrig::Controller::pre_start()
+{
+    m_miner = std::make_shared<Miner>(this);
+}
+#endif
+
 
 void xmrig::Controller::start()
 {
     Base::start();
 
-    m_miner = std::make_shared<Miner>(this);
+    if (m_miner == nullptr) m_miner = std::make_shared<Miner>(this);
 
     network()->connect();
 }

src/core/Controller.h

@@ -44,6 +44,9 @@ public:
     ~Controller() override;
 
     int init() override;
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    void pre_start();
+#   endif
     void start() override;
     void stop() override;
 

src/core/MoBenchmark.cpp

@@ -0,0 +1,237 @@
+/* XMRig
+ * Copyright 2018-2020 MoneroOcean <https://github.com/MoneroOcean>, <support@moneroocean.stream>
+ *
+ *   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 "core/MoBenchmark.h"
+#include "3rdparty/rapidjson/document.h"
+#include "backend/common/Hashrate.h"
+#include "backend/common/interfaces/IBackend.h"
+#include "backend/common/Tags.h"
+#include "base/io/log/Log.h"
+#include "base/io/log/Tags.h"
+#include "base/net/stratum/Job.h"
+#include "core/config/Config.h"
+#include "core/Controller.h"
+#include "core/Miner.h"
+#include "net/JobResult.h"
+#include "net/JobResults.h"
+#include "net/Network.h"
+
+#include <chrono>
+
+namespace xmrig {
+
+MoBenchmark::MoBenchmark() : m_controller(nullptr), m_isNewBenchRun(true) {
+  for (BenchAlgo bench_algo = BenchAlgo::MIN; bench_algo != BenchAlgo::MAX; bench_algo = static_cast<BenchAlgo>(bench_algo + 1)) {
+    m_bench_job[bench_algo] = new Job(false, Algorithm(ba2a[bench_algo]), "benchmark");
+  }
+}
+
+MoBenchmark::~MoBenchmark() {
+  for (BenchAlgo bench_algo = BenchAlgo::MIN; bench_algo != BenchAlgo::MAX; bench_algo = static_cast<BenchAlgo>(bench_algo + 1)) {
+    delete m_bench_job[bench_algo];
+  }
+}
+
+// start performance measurements from the first bench_algo
+void MoBenchmark::start() {
+    JobResults::setListener(this, m_controller->config()->cpu().isHwAES()); // register benchmark as job result listener to compute hashrates there
+    // write text before first benchmark round
+    LOG_INFO("%s " BRIGHT_BLACK_BG(CYAN_BOLD_S " STARTING ALGO PERFORMANCE CALIBRATION (with " MAGENTA_BOLD_S "%i" CYAN_BOLD_S " seconds round) "), Tags::benchmark(), m_controller->config()->benchAlgoTime());
+    // start benchmarking from first PerfAlgo in the list
+    start(BenchAlgo::MIN);
+    m_isNewBenchRun = true;
+}
+
+// end of benchmarks, switch to jobs from the pool (network), fill algo_perf
+void MoBenchmark::finish() {
+    for (Algorithm::Id algo = Algorithm::MIN; algo != Algorithm::MAX; algo = static_cast<Algorithm::Id>(algo + 1)) {
+        algo_perf[algo] = get_algo_perf(algo);
+    }
+    m_bench_algo = BenchAlgo::INVALID;
+    LOG_INFO("%s " BRIGHT_BLACK_BG(CYAN_BOLD_S " ALGO PERFORMANCE CALIBRATION COMPLETE "), Tags::benchmark());
+    m_controller->miner()->pause(); // do not compute anything before job from the pool
+    JobResults::stop();
+    JobResults::setListener(m_controller->network(), m_controller->config()->cpu().isHwAES());
+    m_controller->start();
+}
+
+rapidjson::Value MoBenchmark::toJSON(rapidjson::Document &doc) const
+{
+    using namespace rapidjson;
+    auto &allocator = doc.GetAllocator();
+
+    Value obj(kObjectType);
+
+    for (const auto &a : m_controller->miner()->algorithms()) {
+        if (algo_perf[a.id()] == 0.0f) continue;
+        obj.AddMember(StringRef(a.shortName()), algo_perf[a.id()], allocator);
+    }
+
+    return obj;
+}
+
+void MoBenchmark::read(const rapidjson::Value &value)
+{
+    for (Algorithm::Id algo = Algorithm::MIN; algo != Algorithm::MAX; algo = static_cast<Algorithm::Id>(algo + 1)) {
+        algo_perf[algo] = 0.0f;
+    }
+    if (value.IsObject()) {
+        for (auto &member : value.GetObject()) {
+            const Algorithm algo(member.name.GetString());
+            if (!algo.isValid()) {
+                LOG_INFO("%s " BRIGHT_BLACK_BG(MAGENTA_BOLD_S " Ignoring wrong name for algo-perf[%s] "), Tags::benchmark(), member.name.GetString());
+                continue;
+            }
+            if (member.value.IsFloat()) {
+                algo_perf[algo.id()] = member.value.GetFloat();
+                m_isNewBenchRun = false;
+                continue;
+            }
+            if (member.value.IsInt()) {
+                algo_perf[algo.id()] = member.value.GetInt();
+                m_isNewBenchRun = false;
+                continue;
+            }
+            LOG_INFO("%s " BRIGHT_BLACK_BG(MAGENTA_BOLD_S " Ignoring wrong value for algo-perf[%s] "), Tags::benchmark(), member.name.GetString());
+        }
+    }
+}
+
+double MoBenchmark::get_algo_perf(Algorithm::Id algo) const {
+    switch (algo) {
+        case Algorithm::CN_CCX:        return m_bench_algo_perf[BenchAlgo::CN_CCX];
+        case Algorithm::CN_0:          return m_bench_algo_perf[BenchAlgo::CN_CCX] / 2;
+        case Algorithm::CN_1:          return m_bench_algo_perf[BenchAlgo::CN_R];
+        case Algorithm::CN_2:          return m_bench_algo_perf[BenchAlgo::CN_R];
+        case Algorithm::CN_R:          return m_bench_algo_perf[BenchAlgo::CN_R];
+        case Algorithm::CN_RTO:        return m_bench_algo_perf[BenchAlgo::CN_R];
+        case Algorithm::CN_XAO:        return m_bench_algo_perf[BenchAlgo::CN_R];
+        case Algorithm::CN_FAST:       return m_bench_algo_perf[BenchAlgo::CN_R] * 2;
+        case Algorithm::CN_HALF:       return m_bench_algo_perf[BenchAlgo::CN_R] * 2;
+        case Algorithm::CN_RWZ:        return m_bench_algo_perf[BenchAlgo::CN_R] / 3 * 4;
+        case Algorithm::CN_ZLS:        return m_bench_algo_perf[BenchAlgo::CN_R] / 3 * 4;
+        case Algorithm::CN_DOUBLE:     return m_bench_algo_perf[BenchAlgo::CN_R] / 2;
+        case Algorithm::CN_LITE_0:     return m_bench_algo_perf[BenchAlgo::CN_LITE_1];
+        case Algorithm::CN_LITE_1:     return m_bench_algo_perf[BenchAlgo::CN_LITE_1];
+        case Algorithm::CN_HEAVY_XHV:  return m_bench_algo_perf[BenchAlgo::CN_HEAVY_XHV];
+        case Algorithm::CN_PICO_0:     return m_bench_algo_perf[BenchAlgo::CN_PICO_0];
+        case Algorithm::CN_PICO_TLO:   return m_bench_algo_perf[BenchAlgo::CN_PICO_0];
+        case Algorithm::CN_GPU:        return m_bench_algo_perf[BenchAlgo::CN_GPU];
+        case Algorithm::AR2_CHUKWA_V2: return m_bench_algo_perf[BenchAlgo::AR2_CHUKWA_V2];
+        case Algorithm::ASTROBWT_DERO: return m_bench_algo_perf[BenchAlgo::ASTROBWT_DERO];
+        case Algorithm::KAWPOW_RVN:    return m_bench_algo_perf[BenchAlgo::KAWPOW_RVN];
+        case Algorithm::RX_0:          return m_bench_algo_perf[BenchAlgo::RX_0];
+        case Algorithm::RX_SFX:        return m_bench_algo_perf[BenchAlgo::RX_0];
+        case Algorithm::RX_WOW:        return m_bench_algo_perf[BenchAlgo::RX_WOW];
+        case Algorithm::RX_ARQ:        return m_bench_algo_perf[BenchAlgo::RX_ARQ];
+        case Algorithm::RX_XLA:        return m_bench_algo_perf[BenchAlgo::RX_XLA];
+        default: return 0.0f;
+    }
+}
+
+// start performance measurements for specified perf bench_algo
+void MoBenchmark::start(const BenchAlgo bench_algo) {
+    // calculate number of active miner backends in m_enabled_backend_count
+    m_enabled_backend_count = 0;
+    const Algorithm algo(ba2a[bench_algo]);
+    for (auto backend : m_controller->miner()->backends()) if (backend->isEnabled() && backend->isEnabled(algo)) ++ m_enabled_backend_count;
+    if (m_enabled_backend_count == 0) {
+        run_next_bench_algo(bench_algo);
+        return;
+    }
+    LOG_INFO("%s " BRIGHT_BLACK_BG(WHITE_BOLD_S " Algo " MAGENTA_BOLD_S "%s" WHITE_BOLD_S " Preparation "), Tags::benchmark(), algo.shortName());
+    // prepare test job for benchmark runs ("benchmark" client id is to make sure we can detect benchmark jobs)
+    Job& job = *m_bench_job[bench_algo];
+    job.setId(algo.shortName()); // need to set different id so that workers will see job change
+    if (bench_algo == BenchAlgo::KAWPOW_RVN) {
+      job.setBlob("4c38e8a5f7b2944d1e4274635d828519b97bc64a1f1c7896ecdbb139988aa0e80000000000000000000000000000000000000000000000000000000000000000000000000000000000000000");
+      job.setDiff(Job::toDiff(strtoull("000000639c000000", nullptr, 16)));
+      job.setHeight(1500000);
+    } else {
+      // 99 here to trigger all future bench_algo versions for auto veriant detection based on block version
+      job.setBlob("9905A0DBD6BF05CF16E503F3A66F78007CBF34144332ECBFC22ED95C8700383B309ACE1923A0964B00000008BA939A62724C0D7581FCE5761E9D8A0E6A1C3F924FDD8493D1115649C05EB601");
+      job.setTarget("FFFFFFFFFFFFFF20"); // set difficulty to 8 cause onJobResult after every 8-th computed hash
+      job.setHeight(1000);
+      job.setSeedHash("0000000000000000000000000000000000000000000000000000000000000001");
+    }
+    m_bench_algo  = bench_algo; // current perf bench_algo
+    m_hash_count  = 0;          // number of hashes calculated for current perf bench_algo
+    m_time_start  = 0;          // init time of the first result (in ms) during the first onJobResult
+    m_bench_start = 0;          // init time of measurements start (in ms) during the first onJobResult
+    m_backends_started.clear();
+    m_controller->miner()->setJob(job, false); // set job for workers to compute
+}
+
+// run next bench algo or finish benchmark for the last one
+void MoBenchmark::run_next_bench_algo(const BenchAlgo bench_algo) {
+    const BenchAlgo next_bench_algo = static_cast<BenchAlgo>(bench_algo + 1); // compute next perf bench_algo to benchmark
+    if (next_bench_algo != BenchAlgo::MAX) {
+        start(next_bench_algo);
+    } else {
+        finish();
+    }
+}
+
+void MoBenchmark::onJobResult(const JobResult& result) {
+    if (result.clientId != String("benchmark")) { // switch to network pool jobs
+        JobResults::setListener(m_controller->network(), m_controller->config()->cpu().isHwAES());
+        static_cast<IJobResultListener*>(m_controller->network())->onJobResult(result);
+        return;
+    }
+    // ignore benchmark results for other perf bench_algo
+    if (m_bench_algo == BenchAlgo::INVALID || result.jobId != String(Algorithm(ba2a[m_bench_algo]).shortName())) return;
+    const uint64_t now = get_now();
+    if (!m_time_start) m_time_start = now; // time of the first result (in ms)
+    m_backends_started.insert(result.backend);
+    // waiting for all backends to start
+    if (m_backends_started.size() < m_enabled_backend_count && (now - m_time_start < static_cast<unsigned>(3*60*1000))) return;
+    ++ m_hash_count;
+    if (!m_bench_start) {
+       LOG_INFO("%s " BRIGHT_BLACK_BG(WHITE_BOLD_S " Algo " MAGENTA_BOLD_S "%s" WHITE_BOLD_S " Starting test "), Tags::benchmark(), Algorithm(ba2a[m_bench_algo]).shortName());
+       m_bench_start = now; // time of measurements start (in ms)
+    } else if (now - m_bench_start > static_cast<unsigned>(m_controller->config()->benchAlgoTime()*1000)) { // end of benchmark round for m_bench_algo
+        double t[3] = { 0.0 };
+        for (auto backend : m_controller->miner()->backends()) {
+            const Hashrate *hr = backend->hashrate();
+            if (!hr) continue;
+            t[0] += hr->calc(Hashrate::ShortInterval);
+            t[1] += hr->calc(Hashrate::MediumInterval);
+            t[2] += hr->calc(Hashrate::LargeInterval);
+        }
+        double hashrate = 0.0f;
+        if (!(hashrate = t[2]))
+            if (!(hashrate = t[1]))
+                if (!(hashrate = t[0]))
+                    hashrate = static_cast<double>(m_hash_count) * result.diff / (now - m_bench_start) * 1000.0f;
+        if (m_bench_algo == KAWPOW_RVN) hashrate /= ((double)0xFFFFFFFFFFFFFFFF) / 0xFF000000;
+        m_bench_algo_perf[m_bench_algo] = hashrate; // store hashrate result
+        LOG_INFO("%s " BRIGHT_BLACK_BG(WHITE_BOLD_S " Algo " MAGENTA_BOLD_S "%s" WHITE_BOLD_S " hashrate: " CYAN_BOLD_S "%f "), Tags::benchmark(), Algorithm(ba2a[m_bench_algo]).shortName(), hashrate);
+        run_next_bench_algo(m_bench_algo);
+    }
+}
+
+uint64_t MoBenchmark::get_now() const { // get current time in ms
+    using namespace std::chrono;
+    return time_point_cast<milliseconds>(high_resolution_clock::now()).time_since_epoch().count();
+}
+
+} // namespace xmrig
+
+const char *xmrig::bm_tag()
+{
+    return Tags::benchmark();
+}

src/core/MoBenchmark.h

@@ -0,0 +1,104 @@
+/* XMRig
+ * Copyright 2018-2020 MoneroOcean <https://github.com/MoneroOcean>, <support@moneroocean.stream>
+ *
+ *   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/>.
+ */
+
+#pragma once
+
+#include <set>
+#include "net/interfaces/IJobResultListener.h"
+#include "base/crypto/Algorithm.h"
+#include "rapidjson/fwd.h"
+
+#include <memory>
+
+namespace xmrig {
+
+class Controller;
+class Miner;
+class Job;
+
+class MoBenchmark : public IJobResultListener {
+
+        enum BenchAlgo : int {
+            CN_R,          // "cn/r"             CryptoNightR (Monero's variant 4).
+            CN_LITE_1,     // "cn-lite/1"        CryptoNight-Lite variant 1.
+            CN_HEAVY_XHV,  // "cn-heavy/xhv"     CryptoNight-Heavy (modified, Haven Protocol only).
+            CN_PICO_0,     // "cn-pico"          CryptoNight-Pico.
+            CN_CCX,        // "cn/ccx"           Conceal (CCX).
+            CN_GPU,        // "cn/gpu"           CryptoNight-GPU (Ryo).
+            AR2_CHUKWA_V2, // "argon2/chukwav2"  Argon2id (Chukwa v2).
+            KAWPOW_RVN,    // "kawpow/rvn"       KawPow (RVN)
+            ASTROBWT_DERO, // "astrobwt"         AstroBWT (Dero).
+            RX_0,          // "rx/0"             RandomX (Monero).
+            RX_WOW,        // "rx/wow"           RandomWOW (Wownero).
+            RX_ARQ,        // "rx/arq"           RandomARQ (Arqma).
+            RX_XLA,        // "panthera"         Panthera (Scala2).
+            MAX,
+            MIN = 0,
+            INVALID = -1,
+        };
+
+        const Algorithm::Id ba2a[BenchAlgo::MAX] = {
+            Algorithm::CN_R,
+            Algorithm::CN_LITE_1,
+            Algorithm::CN_HEAVY_XHV,
+            Algorithm::CN_PICO_0,
+            Algorithm::CN_CCX,
+            Algorithm::CN_GPU,
+            Algorithm::AR2_CHUKWA_V2,
+            Algorithm::KAWPOW_RVN,
+            Algorithm::ASTROBWT_DERO,
+            Algorithm::RX_0,
+            Algorithm::RX_WOW,
+            Algorithm::RX_ARQ,
+            Algorithm::RX_XLA,
+        };
+
+        Job* m_bench_job[BenchAlgo::MAX];
+        double m_bench_algo_perf[BenchAlgo::MAX];
+
+        std::shared_ptr<Controller> m_controller;          // to get access to config and network
+        bool m_isNewBenchRun;              // true if benchmark is need to be executed or was executed
+        MoBenchmark::BenchAlgo m_bench_algo; // current perf algo we benchmark
+        uint64_t m_hash_count;             // number of hashes calculated for current perf algo
+        uint64_t m_time_start;             // time of the first resultt for current perf algo (in ms)
+        uint64_t m_bench_start;            // time of measurements start for current perf algo (in ms) after all backends are started
+        unsigned m_enabled_backend_count;  // number of active miner backends
+        std::set<uint32_t> m_backends_started; // id of backend started for benchmark
+
+        uint64_t get_now() const;                      // get current time in ms
+        double get_algo_perf(Algorithm::Id algo) const; // get algo perf based on m_bench_algo_perf
+        void start(const MoBenchmark::BenchAlgo);        // start benchmark for specified perf algo
+        void finish();                                 // end of benchmarks, switch to jobs from the pool (network), fill algo_perf
+        void onJobResult(const JobResult&) override;   // onJobResult is called after each computed benchmark hash
+        void run_next_bench_algo(BenchAlgo);           // run next bench algo or finish benchmark for the last one
+
+    public:
+        MoBenchmark();
+        virtual ~MoBenchmark();
+
+        void set_controller(std::shared_ptr<Controller> controller) { m_controller = controller; }
+
+        void start(); // start benchmarks
+
+        bool isNewBenchRun() const { return m_isNewBenchRun; }
+        double algo_perf[Algorithm::MAX];
+
+        rapidjson::Value toJSON(rapidjson::Document &doc) const;
+        void read(const rapidjson::Value &value);
+};
+
+} // namespace xmrig

src/core/config/Config.cpp

@@ -203,6 +203,12 @@ bool xmrig::Config::isShouldSave() const
     }
 #   endif
 
+#   ifdef XMRIG_FEATURE_BENCHMARK
+    if (m_benchmark.isNewBenchRun()) {
+        return true;
+    }
+#   endif
+
     return (m_upgrade || cpu().isShouldSave());
 }
 
@@ -240,6 +246,10 @@ bool xmrig::Config::read(const IJsonReader &reader, const char *fileName)
     d_ptr->healthPrintTime = reader.getUint(kHealthPrintTime, d_ptr->healthPrintTime);
 #   endif
 
+#   ifdef XMRIG_FEATURE_BENCHMARK
+    m_benchmark.read(reader.getValue(kAlgoPerf));
+#   endif
+
 #   ifdef XMRIG_FEATURE_DMI
     d_ptr->dmi = reader.getBool(kDMI, d_ptr->dmi);
 #   endif
@@ -304,6 +314,13 @@ void xmrig::Config::getJSON(rapidjson::Document &doc) const
     doc.AddMember(StringRef(kUserAgent),                m_userAgent.toJSON(), allocator);
     doc.AddMember(StringRef(kVerbose),                  Log::verbose(), allocator);
     doc.AddMember(StringRef(kWatch),                    m_watch, allocator);
+
+#   ifdef XMRIG_FEATURE_BENCHMARK
+    doc.AddMember(StringRef(kRebenchAlgo),              isRebenchAlgo(), allocator);
+    doc.AddMember(StringRef(kBenchAlgoTime),            benchAlgoTime(), allocator);
+    doc.AddMember(StringRef(kAlgoPerf),                 m_benchmark.toJSON(doc), allocator);
+#   endif
+
     doc.AddMember(StringRef(kPauseOnBattery),           isPauseOnBattery(), allocator);
     doc.AddMember(StringRef(kPauseOnActive),            (d_ptr->idleTime == 0U || d_ptr->idleTime == kIdleTime) ? Value(isPauseOnActive()) : Value(d_ptr->idleTime), allocator);
 }

src/core/config/Config.h

@@ -27,6 +27,9 @@
 #include "backend/cpu/CpuConfig.h"
 #include "base/kernel/config/BaseConfig.h"
 #include "base/tools/Object.h"
+#ifdef XMRIG_FEATURE_MO_BENCHMARK
+#include "core/MoBenchmark.h"
+#endif
 
 
 namespace xmrig {
@@ -100,8 +103,15 @@ public:
     bool read(const IJsonReader &reader, const char *fileName) override;
     void getJSON(rapidjson::Document &doc) const override;
 
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    inline MoBenchmark &benchmark()           { return m_benchmark; }
+#   endif
+
 private:
     ConfigPrivate *d_ptr;
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    MoBenchmark m_benchmark;
+#   endif
 };
 
 

src/core/config/ConfigTransform.cpp

@@ -44,6 +44,9 @@ static const char *kAsterisk    = "*";
 static const char *kEnabled     = "enabled";
 static const char *kIntensity   = "intensity";
 static const char *kThreads     = "threads";
+#ifdef XMRIG_ALGO_KAWPOW
+static const char *kKawPow      = "kawpow";
+#endif
 
 
 static inline uint64_t intensity(uint64_t av)
@@ -103,6 +106,9 @@ void xmrig::ConfigTransform::finalize(rapidjson::Document &doc)
         profile.AddMember(StringRef(kThreads),   m_threads, allocator);
         profile.AddMember(StringRef(kAffinity),  m_affinity, allocator);
 
+#       ifdef XMRIG_ALGO_KAWPOW
+        doc[CpuConfig::kField].AddMember(StringRef(kKawPow), false, doc.GetAllocator());
+#       endif
         doc[CpuConfig::kField].AddMember(StringRef(kAsterisk), profile, doc.GetAllocator());
     }
 

src/core/config/Config_platform.h

@@ -65,6 +65,10 @@ static const option options[] = {
     { "keepalive",             0, nullptr, IConfig::KeepAliveKey          },
     { "log-file",              1, nullptr, IConfig::LogFileKey            },
     { "nicehash",              0, nullptr, IConfig::NicehashKey           },
+#   ifdef XMRIG_FEATURE_MO_BENCHMARK
+    { "rebench-algo",          0, nullptr, IConfig::RebenchAlgoKey        },
+    { "bench-algo-time",       1, nullptr, IConfig::BenchAlgoTimeKey      },
+#   endif
     { "no-color",              0, nullptr, IConfig::ColorKey              },
     { "no-huge-pages",         0, nullptr, IConfig::HugePagesKey          },
     { "no-hugepages",          0, nullptr, IConfig::HugePagesKey          },

src/crypto/cn/CnAlgo.h

@@ -96,6 +96,11 @@ public:
             return CN_ITER / 8;
 #       endif
 
+#       ifdef XMRIG_ALGO_CN_GPU
+        case Algorithm::CN_GPU:
+            return 0xC000;
+#       endif
+
 #       ifdef XMRIG_ALGO_CN_FEMTO
         case Algorithm::CN_UPX2:
             return CN_ITER / 32;
@@ -116,6 +121,12 @@ public:
         }
 #       endif
 
+#       ifdef XMRIG_ALGO_CN_GPU
+        if (algo == Algorithm::CN_GPU) {
+            return 0x1FFFC0;
+	}
+#       endif
+
 #       ifdef XMRIG_ALGO_CN_FEMTO
         if (algo == Algorithm::CN_UPX2) {
             return 0x1FFF0;
@@ -166,6 +177,11 @@ public:
 #       endif
             return Algorithm::CN_2;
 
+#       ifdef XMRIG_ALGO_CN_GPU
+        case Algorithm::CN_GPU:
+            return Algorithm::CN_GPU;
+#       endif
+
         default:
             break;
         }
@@ -207,6 +223,7 @@ template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_ZLS>::iterations() con
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_PICO_0>::iterations() const       { return CN_ITER / 8; }
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_PICO_TLO>::iterations() const     { return CN_ITER / 8; }
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_CCX>::iterations() const          { return CN_ITER / 2; }
+template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_GPU>::iterations() const          { return 0xC000; }
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_UPX2>::iterations() const         { return CN_ITER / 32; }
 
 
@@ -221,6 +238,7 @@ template<> constexpr inline size_t CnAlgo<Algorithm::CN_UPX2>::memory() const
 
 
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_PICO_0>::mask() const             { return 0x1FFF0; }
+template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_GPU>::mask() const                { return 0x1FFFC0; }
 template<> constexpr inline uint32_t CnAlgo<Algorithm::CN_UPX2>::mask() const               { return 0x1FFF0; }
 
 

src/crypto/cn/CnHash.cpp

@@ -311,6 +311,11 @@ xmrig::CnHash::CnHash()
     m_map[Algorithm::ASTROBWT_DERO][AV_SINGLE_SOFT][Assembly::NONE] = astrobwt::single_hash<Algorithm::ASTROBWT_DERO>;
 #   endif
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    m_map[Algorithm::CN_GPU][AV_SINGLE][Assembly::NONE]      = cryptonight_single_hash_gpu<Algorithm::CN_GPU, false>;
+    m_map[Algorithm::CN_GPU][AV_SINGLE_SOFT][Assembly::NONE] = cryptonight_single_hash_gpu<Algorithm::CN_GPU, true>;
+#   endif
+
 #   ifdef XMRIG_FEATURE_ASM
     patchAsmVariants();
 #   endif

src/crypto/cn/CryptoNight.h

@@ -30,7 +30,7 @@
 #include <stddef.h>
 #include <stdint.h>
 
-#if defined _MSC_VER || defined XMRIG_ARM
+#if defined _MSC_VER || defined XMRIG_ARM || defined __INTEL_COMPILER
 #   define ABI_ATTRIBUTE
 #else
 #   define ABI_ATTRIBUTE __attribute__((ms_abi))

src/crypto/cn/CryptoNight_arm.h

@@ -230,7 +230,11 @@ static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
 {
     constexpr CnAlgo<ALGO> props;
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    constexpr bool IS_HEAVY = props.isHeavy() || ALGO == Algorithm::CN_GPU;
+#   else
     constexpr bool IS_HEAVY = props.isHeavy();
+#   endif
 
     __m128i xout0, xout1, xout2, xout3, xout4, xout5, xout6, xout7;
     __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
@@ -545,6 +549,66 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 }
 
 
+} /* namespace xmrig */
+
+
+#ifdef XMRIG_ALGO_CN_GPU
+template<size_t ITER, uint32_t MASK>
+void cn_gpu_inner_arm(const uint8_t *spad, uint8_t *lpad);
+
+
+namespace xmrig {
+
+
+template<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;
+    }
+}
+
+
+template<xmrig::Algorithm::Id ALGO, bool SOFT_AES>
+inline void cryptonight_single_hash_gpu(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx, uint64_t height)
+{
+    constexpr CnAlgo<ALGO> props;
+
+    keccak(input, size, ctx[0]->state);
+    cn_explode_scratchpad_gpu<props.memory()>(ctx[0]->state, ctx[0]->memory);
+
+    fesetround(FE_TONEAREST);
+
+    cn_gpu_inner_arm<props.iterations(), props.mask()>(ctx[0]->state, ctx[0]->memory);
+
+    cn_implode_scratchpad<ALGO, SOFT_AES>(reinterpret_cast<const __m128i *>(ctx[0]->memory), reinterpret_cast<__m128i *>(ctx[0]->state));
+    keccakf(reinterpret_cast<uint64_t*>(ctx[0]->state), 24);
+    memcpy(output, ctx[0]->state, 32);
+}
+
+} /* namespace xmrig */
+#endif
+
+
+namespace xmrig {
+
+
 template<Algorithm::Id ALGO, bool SOFT_AES>
 inline void cryptonight_double_hash(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, struct cryptonight_ctx **__restrict__ ctx, uint64_t height)
 {

src/crypto/cn/CryptoNight_test.h

@@ -370,6 +370,22 @@ const static uint8_t test_output_pico_tlo[160] = {
 #endif
 
 
+#ifdef XMRIG_ALGO_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
+
 #ifdef XMRIG_ALGO_CN_FEMTO
 // "cn/upx2"
 const static uint8_t test_output_femto_upx2[160] = {

src/crypto/cn/CryptoNight_x86.h

@@ -359,7 +359,11 @@ static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
 {
     constexpr CnAlgo<ALGO> props;
 
+#   ifdef XMRIG_ALGO_CN_GPU
+    constexpr bool IS_HEAVY = props.isHeavy() || ALGO == Algorithm::CN_GPU;
+#   else
     constexpr bool IS_HEAVY = props.isHeavy();
+#   endif
 
     __m128i xout0, xout1, xout2, xout3, xout4, xout5, xout6, xout7;
     __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
@@ -510,8 +514,10 @@ static inline __m128i int_sqrt_v2(const uint64_t n0)
     r >>= 19;
 
     uint64_t x2 = (s - (1022ULL << 32)) * (r - s - (1022ULL << 32) + 1);
-#   if (defined(_MSC_VER) || __GNUC__ > 7 || (__GNUC__ == 7 && __GNUC_MINOR__ > 1)) && (defined(__x86_64__) || defined(_M_AMD64))
+#   if (defined(_MSC_VER) || __GNUC__ > 7 || (__GNUC__ == 7 && __GNUC_MINOR__ > 1)) && (defined(__x86_64__) || defined(_M_AMD64)) && !defined(__INTEL_COMPILER)
     _addcarry_u64(_subborrow_u64(0, x2, n0, (unsigned long long int*)&x2), r, 0, (unsigned long long int*)&r);
+#   elif defined(__INTEL_COMPILER)
+    _addcarry_u64(_subborrow_u64(0, x2, n0, (unsigned long int*)&x2), r, 0, (unsigned long int*)&r);
 #   else
     if (x2 < n0) ++r;
 #   endif
@@ -751,6 +757,73 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 } /* namespace xmrig */
 
 
+#ifdef XMRIG_ALGO_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);
+
+
+namespace xmrig {
+
+
+template<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;
+    }
+}
+
+
+template<Algorithm::Id ALGO, bool SOFT_AES>
+inline void cryptonight_single_hash_gpu(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx, uint64_t)
+{
+    constexpr CnAlgo<ALGO> props;
+
+    keccak(input, size, ctx[0]->state);
+    cn_explode_scratchpad_gpu<props.memory()>(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<props.iterations(), props.mask()>(ctx[0]->state, ctx[0]->memory);
+    } else {
+        cn_gpu_inner_ssse3<props.iterations(), props.mask()>(ctx[0]->state, ctx[0]->memory);
+    }
+
+    cn_implode_scratchpad<ALGO, SOFT_AES, 0>(reinterpret_cast<const __m128i *>(ctx[0]->memory), reinterpret_cast<__m128i *>(ctx[0]->state));
+    keccakf(reinterpret_cast<uint64_t*>(ctx[0]->state), 24);
+    memcpy(output, ctx[0]->state, 32);
+}
+
+
+} /* namespace xmrig */
+#endif
+
+
 #ifdef XMRIG_FEATURE_ASM
 extern "C" void cnv2_mainloop_ivybridge_asm(cryptonight_ctx **ctx);
 extern "C" void cnv2_mainloop_ryzen_asm(cryptonight_ctx **ctx);

src/crypto/cn/gpu/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/cn/CnAlgo.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::CnAlgo<xmrig::Algorithm::CN_GPU>().iterations(), xmrig::CnAlgo<xmrig::Algorithm::CN_GPU>().mask()>(const uint8_t* spad, uint8_t* lpad);

src/crypto/cn/gpu/cn_gpu_avx.cpp

@@ -0,0 +1,211 @@
+/* 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 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 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/cn/CnAlgo.h"
+
+
+#ifdef __GNUC__
+#   include <x86intrin.h>
+#else
+#   include <intrin.h>
+#   define __restrict__ __restrict
+#endif
+#ifndef _mm256_bslli_epi128
+	#define _mm256_bslli_epi128(a, count) _mm256_slli_si256((a), (count))
+#endif
+#ifndef _mm256_bsrli_epi128
+	#define _mm256_bsrli_epi128(a, count) _mm256_srli_si256((a), (count))
+#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 setting 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::CnAlgo<xmrig::Algorithm::CN_GPU>().iterations(), xmrig::CnAlgo<xmrig::Algorithm::CN_GPU>().mask()>(const uint8_t* spad, uint8_t* lpad);

src/crypto/cn/gpu/cn_gpu_ssse3.cpp

@@ -0,0 +1,212 @@
+/* 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 2018-2020 SChernykh   <https://github.com/SChernykh>
+ * Copyright 2016-2020 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/cn/CnAlgo.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 setting 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::CnAlgo<xmrig::Algorithm::CN_GPU>().iterations(), xmrig::CnAlgo<xmrig::Algorithm::CN_GPU>().mask()>(const uint8_t* spad, uint8_t* lpad);

src/crypto/common/MemoryPool.cpp

@@ -47,11 +47,11 @@ xmrig::MemoryPool::MemoryPool(size_t size, bool hugePages, uint32_t node)
         return;
     }
 
-    constexpr size_t alignment = 1 << 24;
+    constexpr size_t alignment = 0; //1 << 24;
 
     m_memory = new VirtualMemory(size * pageSize + alignment, hugePages, false, false, node);
 
-    m_alignOffset = (alignment - (((size_t)m_memory->scratchpad()) % alignment)) % alignment;
+    //m_alignOffset = (alignment - (((size_t)m_memory->scratchpad()) % alignment)) % alignment;
 }
 
 

src/crypto/randomx/asm/program_prologue_linux.inc

@@ -22,7 +22,6 @@
 	mov rsi, rdx                ;# uint8_t* scratchpad
 
 	mov rax, rbp
-	ror rbp, 32
 
 	;# zero integer registers
 	xor r8, r8

src/crypto/randomx/asm/program_prologue_win64.inc

@@ -35,7 +35,6 @@
 	mov rbx, r9                 ;# loop counter
 
 	mov rax, rbp
-	ror rbp, 32
 
 	;# zero integer registers
 	xor r8, r8

src/crypto/randomx/asm/program_read_dataset.inc

@@ -1,16 +1,17 @@
-	mov ecx, ebp                       ;# ecx = ma
-	and ecx, RANDOMX_DATASET_BASE_MASK
-	xor r8, qword ptr [rdi+rcx]
-	ror rbp, 32                        ;# swap "ma" and "mx"
 	xor rbp, rax                       ;# modify "mx"
 	mov edx, ebp                       ;# edx = mx
 	and edx, RANDOMX_DATASET_BASE_MASK
 	prefetchnta byte ptr [rdi+rdx]
-	xor r9,  qword ptr [rdi+rcx+8]
+	ror rbp, 32                        ;# swap "ma" and "mx"
-	xor r10, qword ptr [rdi+rcx+16]
+	mov edx, ebp                       ;# edx = ma
-	xor r11, qword ptr [rdi+rcx+24]
+	and edx, RANDOMX_DATASET_BASE_MASK
-	xor r12, qword ptr [rdi+rcx+32]
+	lea rcx, [rdi+rdx]                 ;# dataset cache line
-	xor r13, qword ptr [rdi+rcx+40]
+	xor r8,  qword ptr [rcx+0]
-	xor r14, qword ptr [rdi+rcx+48]
+	xor r9,  qword ptr [rcx+8]
-	xor r15, qword ptr [rdi+rcx+56]
+	xor r10, qword ptr [rcx+16]
+	xor r11, qword ptr [rcx+24]
+	xor r12, qword ptr [rcx+32]
+	xor r13, qword ptr [rcx+40]
+	xor r14, qword ptr [rcx+48]
+	xor r15, qword ptr [rcx+56]
 	

src/crypto/randomx/asm/program_read_dataset_ryzen.inc

@@ -0,0 +1,17 @@
+	mov rcx, rbp                       ;# ecx = ma
+	shr rcx, 32
+	and ecx, RANDOMX_DATASET_BASE_MASK
+	xor r8, qword ptr [rdi+rcx]
+	xor rbp, rax                       ;# modify "mx"
+	mov edx, ebp                       ;# edx = mx
+	and edx, RANDOMX_DATASET_BASE_MASK
+	prefetchnta byte ptr [rdi+rdx]
+	ror rbp, 32                        ;# swap "ma" and "mx"
+	xor r9,  qword ptr [rdi+rcx+8]
+	xor r10, qword ptr [rdi+rcx+16]
+	xor r11, qword ptr [rdi+rcx+24]
+	xor r12, qword ptr [rdi+rcx+32]
+	xor r13, qword ptr [rdi+rcx+40]
+	xor r14, qword ptr [rdi+rcx+48]
+	xor r15, qword ptr [rdi+rcx+56]
+	

src/crypto/randomx/common.hpp

@@ -74,7 +74,7 @@ namespace randomx {
 	constexpr int SuperscalarMaxSize = 3 * RANDOMX_SUPERSCALAR_MAX_LATENCY + 2;
 	constexpr size_t CacheLineSize = RANDOMX_DATASET_ITEM_SIZE;
 	#define ScratchpadSize RandomX_CurrentConfig.ScratchpadL3_Size
-	#define CacheLineAlignMask RandomX_ConfigurationBase::CacheLineAlignMask_Calculated
+        #define CacheLineAlignMask RandomX_CurrentConfig.CacheLineAlignMask_Calculated
 	#define DatasetExtraItems RandomX_ConfigurationBase::DatasetExtraItems_Calculated
 	constexpr int StoreL3Condition = 14;
 

src/crypto/randomx/jit_compiler_a64.cpp

@@ -90,7 +90,7 @@ static size_t CalcDatasetItemSize()
 	// Prologue
 	((uint8_t*)randomx_calc_dataset_item_aarch64_prefetch - (uint8_t*)randomx_calc_dataset_item_aarch64) +
 	// Main loop
-	RandomX_ConfigurationBase::CacheAccesses * (
+	RandomX_CurrentConfig.CacheAccesses * (
 		// Main loop prologue
 		((uint8_t*)randomx_calc_dataset_item_aarch64_mix - ((uint8_t*)randomx_calc_dataset_item_aarch64_prefetch)) + 4 +
 		// Inner main loop (instructions)
@@ -263,7 +263,7 @@ void JitCompilerA64::generateSuperscalarHash(SuperscalarProgram(&programs)[N])
 	num32bitLiterals = 64;
 	constexpr uint32_t tmp_reg = 12;
 
-	for (size_t i = 0; i < RandomX_ConfigurationBase::CacheAccesses; ++i)
+	for (size_t i = 0; i < RandomX_CurrentConfig.CacheAccesses; ++i)
 	{
 		// and x11, x10, CacheSize / CacheLineSize - 1
 		emit32(0x92400000 | 11 | (10 << 5) | ((RandomX_CurrentConfig.Log2_CacheSize - 1) << 10), code, codePos);

src/crypto/randomx/jit_compiler_x86.cpp

@@ -115,7 +115,6 @@ namespace randomx {
 	#define codeLoopLoad ADDR(randomx_program_loop_load)
 	#define codeLoopLoadXOP ADDR(randomx_program_loop_load_xop)
 	#define codeProgamStart ADDR(randomx_program_start)
-	#define codeReadDataset ADDR(randomx_program_read_dataset)
 	#define codeReadDatasetLightSshInit ADDR(randomx_program_read_dataset_sshash_init)
 	#define codeReadDatasetLightSshFin ADDR(randomx_program_read_dataset_sshash_fin)
 	#define codeDatasetInit ADDR(randomx_dataset_init)
@@ -136,7 +135,6 @@ namespace randomx {
 	#define prologueSize (codeLoopBegin - codePrologue)
 	#define loopLoadSize (codeLoopLoadXOP - codeLoopLoad)
 	#define loopLoadXOPSize (codeProgamStart - codeLoopLoadXOP)
-	#define readDatasetSize (codeReadDatasetLightSshInit - codeReadDataset)
 	#define readDatasetLightInitSize (codeReadDatasetLightSshFin - codeReadDatasetLightSshInit)
 	#define readDatasetLightFinSize (codeLoopStore - codeReadDatasetLightSshFin)
 	#define loopStoreSize (codeLoopEnd - codeLoopStore)
@@ -320,7 +318,20 @@ namespace randomx {
 		vm_flags = flags;
 
 		generateProgramPrologue(prog, pcfg);
-		emit(codeReadDataset, readDatasetSize, code, codePos);
+
+		uint8_t* p;
+		uint32_t n;
+		if (flags & RANDOMX_FLAG_AMD) {
+			p = RandomX_CurrentConfig.codeReadDatasetRyzenTweaked;
+			n = RandomX_CurrentConfig.codeReadDatasetRyzenTweakedSize;
+		}
+		else {
+			p = RandomX_CurrentConfig.codeReadDatasetTweaked;
+			n = RandomX_CurrentConfig.codeReadDatasetTweakedSize;
+		}
+		memcpy(code + codePos, p, n);
+		codePos += n;
+
 		generateProgramEpilogue(prog, pcfg);
 	}
 

src/crypto/randomx/jit_compiler_x86_static.S

@@ -48,6 +48,7 @@
 .global DECL(randomx_program_loop_load_xop)
 .global DECL(randomx_program_start)
 .global DECL(randomx_program_read_dataset)
+.global DECL(randomx_program_read_dataset_ryzen)
 .global DECL(randomx_program_read_dataset_sshash_init)
 .global DECL(randomx_program_read_dataset_sshash_fin)
 .global DECL(randomx_program_loop_store)
@@ -139,6 +140,9 @@ DECL(randomx_program_start):
 DECL(randomx_program_read_dataset):
 	#include "asm/program_read_dataset.inc"
 
+DECL(randomx_program_read_dataset_ryzen):
+	#include "asm/program_read_dataset_ryzen.inc"
+
 DECL(randomx_program_read_dataset_sshash_init):
 	#include "asm/program_read_dataset_sshash_init.inc"
 

src/crypto/randomx/jit_compiler_x86_static.asm

@@ -39,6 +39,7 @@ PUBLIC randomx_program_loop_load
 PUBLIC randomx_program_loop_load_xop
 PUBLIC randomx_program_start
 PUBLIC randomx_program_read_dataset
+PUBLIC randomx_program_read_dataset_ryzen
 PUBLIC randomx_program_read_dataset_sshash_init
 PUBLIC randomx_program_read_dataset_sshash_fin
 PUBLIC randomx_dataset_init
@@ -135,6 +136,10 @@ randomx_program_read_dataset PROC
 	include asm/program_read_dataset.inc
 randomx_program_read_dataset ENDP
 
+randomx_program_read_dataset_ryzen PROC
+	include asm/program_read_dataset_ryzen.inc
+randomx_program_read_dataset_ryzen ENDP
+
 randomx_program_read_dataset_sshash_init PROC
 	include asm/program_read_dataset_sshash_init.inc
 randomx_program_read_dataset_sshash_init ENDP

src/crypto/randomx/jit_compiler_x86_static.hpp

@@ -40,6 +40,7 @@ extern "C" {
 	void randomx_program_loop_load_xop();
 	void randomx_program_start();
 	void randomx_program_read_dataset();
+	void randomx_program_read_dataset_ryzen();
 	void randomx_program_read_dataset_sshash_init();
 	void randomx_program_read_dataset_sshash_fin();
 	void randomx_program_loop_store();

src/crypto/randomx/panthera/KangarooTwelve.c

@@ -0,0 +1,271 @@
+/*
+Implementation by Ronny Van Keer, hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#include <string.h>
+#include "KangarooTwelve.h"
+#ifndef KeccakP1600timesN_excluded
+    // #include "KeccakP-1600-times2-SnP.h"
+    // #include "KeccakP-1600-times4-SnP.h"
+    // #include "KeccakP-1600-times8-SnP.h"
+#endif
+
+#define chunkSize       8192
+#define laneSize        8
+#define suffixLeaf      0x0B /* '110': message hop, simple padding, inner node */
+
+#define security        128
+#define capacity        (2*security)
+#define capacityInBytes (capacity/8)
+#define capacityInLanes (capacityInBytes/laneSize)
+#define rate            (1600-capacity)
+#define rateInBytes     (rate/8)
+#define rateInLanes     (rateInBytes/laneSize)
+
+#define ParallelSpongeFastLoop( Parallellism ) \
+    while ( inLen >= Parallellism * chunkSize ) { \
+        ALIGN(KeccakP1600times##Parallellism##_statesAlignment) unsigned char states[KeccakP1600times##Parallellism##_statesSizeInBytes]; \
+        unsigned char intermediate[Parallellism*capacityInBytes]; \
+        unsigned int localBlockLen = chunkSize; \
+        const unsigned char * localInput = input; \
+        unsigned int i; \
+        unsigned int fastLoopOffset; \
+        \
+        KeccakP1600times##Parallellism##_StaticInitialize(); \
+        KeccakP1600times##Parallellism##_InitializeAll(states); \
+        fastLoopOffset = KeccakP1600times##Parallellism##_12rounds_FastLoop_Absorb(states, rateInLanes, chunkSize / laneSize, rateInLanes, localInput, Parallellism * chunkSize); \
+        localBlockLen -= fastLoopOffset; \
+        localInput += fastLoopOffset; \
+        for ( i = 0; i < Parallellism; ++i, localInput += chunkSize ) { \
+            KeccakP1600times##Parallellism##_AddBytes(states, i, localInput, 0, localBlockLen); \
+            KeccakP1600times##Parallellism##_AddByte(states, i, suffixLeaf, localBlockLen); \
+            KeccakP1600times##Parallellism##_AddByte(states, i, 0x80, rateInBytes-1); \
+        } \
+        KeccakP1600times##Parallellism##_PermuteAll_12rounds(states); \
+        input += Parallellism * chunkSize; \
+        inLen -= Parallellism * chunkSize; \
+        ktInstance->blockNumber += Parallellism; \
+        KeccakP1600times##Parallellism##_ExtractLanesAll(states, intermediate, capacityInLanes, capacityInLanes ); \
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, intermediate, Parallellism * capacityInBytes) != 0) return 1; \
+    }
+
+#define ParallelSpongeLoop( Parallellism ) \
+    while ( inLen >= Parallellism * chunkSize ) { \
+        ALIGN(KeccakP1600times##Parallellism##_statesAlignment) unsigned char states[KeccakP1600times##Parallellism##_statesSizeInBytes]; \
+        unsigned char intermediate[Parallellism*capacityInBytes]; \
+        unsigned int localBlockLen = chunkSize; \
+        const unsigned char * localInput = input; \
+        unsigned int i; \
+        \
+        KeccakP1600times##Parallellism##_StaticInitialize(); \
+        KeccakP1600times##Parallellism##_InitializeAll(states); \
+        while(localBlockLen >= rateInBytes) { \
+            KeccakP1600times##Parallellism##_AddLanesAll(states, localInput, rateInLanes, chunkSize / laneSize); \
+            KeccakP1600times##Parallellism##_PermuteAll_12rounds(states); \
+            localBlockLen -= rateInBytes; \
+            localInput += rateInBytes; \
+           } \
+        for ( i = 0; i < Parallellism; ++i, localInput += chunkSize ) { \
+            KeccakP1600times##Parallellism##_AddBytes(states, i, localInput, 0, localBlockLen); \
+            KeccakP1600times##Parallellism##_AddByte(states, i, suffixLeaf, localBlockLen); \
+            KeccakP1600times##Parallellism##_AddByte(states, i, 0x80, rateInBytes-1); \
+        } \
+        KeccakP1600times##Parallellism##_PermuteAll_12rounds(states); \
+        input += Parallellism * chunkSize; \
+        inLen -= Parallellism * chunkSize; \
+        ktInstance->blockNumber += Parallellism; \
+        KeccakP1600times##Parallellism##_ExtractLanesAll(states, intermediate, capacityInLanes, capacityInLanes ); \
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, intermediate, Parallellism * capacityInBytes) != 0) return 1; \
+    }
+
+static unsigned int right_encode( unsigned char * encbuf, size_t value )
+{
+    unsigned int n, i;
+    size_t v;
+
+    for ( v = value, n = 0; v && (n < sizeof(size_t)); ++n, v >>= 8 )
+        ; /* empty */
+    for ( i = 1; i <= n; ++i )
+        encbuf[i-1] = (unsigned char)(value >> (8 * (n-i)));
+    encbuf[n] = (unsigned char)n;
+    return n + 1;
+}
+
+int KangarooTwelve_Initialize(KangarooTwelve_Instance *ktInstance, size_t outputLen)
+{
+    ktInstance->fixedOutputLength = outputLen;
+    ktInstance->queueAbsorbedLen = 0;
+    ktInstance->blockNumber = 0;
+    ktInstance->phase = ABSORBING;
+    return KeccakWidth1600_12rounds_SpongeInitialize(&ktInstance->finalNode, rate, capacity);
+}
+
+int KangarooTwelve_Update(KangarooTwelve_Instance *ktInstance, const unsigned char *input, size_t inLen)
+{
+    if (ktInstance->phase != ABSORBING)
+        return 1;
+
+    if ( ktInstance->blockNumber == 0 ) {
+        /* First block, absorb in final node */
+        unsigned int len = (inLen < (chunkSize - ktInstance->queueAbsorbedLen)) ? inLen : (chunkSize - ktInstance->queueAbsorbedLen);
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, input, len) != 0)
+            return 1;
+        input += len;
+        inLen -= len;
+        ktInstance->queueAbsorbedLen += len;
+        if ( (ktInstance->queueAbsorbedLen == chunkSize) && (inLen != 0) ) {
+            /* First block complete and more input data available, finalize it */
+            const unsigned char padding = 0x03; /* '110^6': message hop, simple padding */
+            ktInstance->queueAbsorbedLen = 0;
+            ktInstance->blockNumber = 1;
+            if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, &padding, 1) != 0)
+                return 1;
+            ktInstance->finalNode.byteIOIndex = (ktInstance->finalNode.byteIOIndex + 7) & ~7; /* Zero padding up to 64 bits */
+        }
+    }
+    else if ( ktInstance->queueAbsorbedLen != 0 ) {
+        /* There is data in the queue, absorb further in queue until block complete */
+        unsigned int len = (inLen < (chunkSize - ktInstance->queueAbsorbedLen)) ? inLen : (chunkSize - ktInstance->queueAbsorbedLen);
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->queueNode, input, len) != 0)
+            return 1;
+        input += len;
+        inLen -= len;
+        ktInstance->queueAbsorbedLen += len;
+        if ( ktInstance->queueAbsorbedLen == chunkSize ) {
+            unsigned char intermediate[capacityInBytes];
+            ktInstance->queueAbsorbedLen = 0;
+            ++ktInstance->blockNumber;
+            if (KeccakWidth1600_12rounds_SpongeAbsorbLastFewBits(&ktInstance->queueNode, suffixLeaf) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeSqueeze(&ktInstance->queueNode, intermediate, capacityInBytes) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, intermediate, capacityInBytes) != 0)
+                return 1;
+        }
+    }
+
+    #if defined(KeccakP1600times8_implementation) && !defined(KeccakP1600times8_isFallback)
+    #if defined(KeccakP1600times8_12rounds_FastLoop_supported)
+    ParallelSpongeFastLoop( 8 )
+    #else
+    ParallelSpongeLoop( 8 )
+    #endif
+    #endif
+
+    #if defined(KeccakP1600times4_implementation) && !defined(KeccakP1600times4_isFallback)
+    #if defined(KeccakP1600times4_12rounds_FastLoop_supported)
+    ParallelSpongeFastLoop( 4 )
+    #else
+    ParallelSpongeLoop( 4 )
+    #endif
+    #endif
+
+    #if defined(KeccakP1600times2_implementation) && !defined(KeccakP1600times2_isFallback)
+    #if defined(KeccakP1600times2_12rounds_FastLoop_supported)
+    ParallelSpongeFastLoop( 2 )
+    #else
+    ParallelSpongeLoop( 2 )
+    #endif
+    #endif
+
+    while ( inLen > 0 ) {
+        unsigned int len = (inLen < chunkSize) ? inLen : chunkSize;
+        if (KeccakWidth1600_12rounds_SpongeInitialize(&ktInstance->queueNode, rate, capacity) != 0)
+            return 1;
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->queueNode, input, len) != 0)
+            return 1;
+        input += len;
+        inLen -= len;
+        if ( len == chunkSize ) {
+            unsigned char intermediate[capacityInBytes];
+            ++ktInstance->blockNumber;
+            if (KeccakWidth1600_12rounds_SpongeAbsorbLastFewBits(&ktInstance->queueNode, suffixLeaf) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeSqueeze(&ktInstance->queueNode, intermediate, capacityInBytes) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, intermediate, capacityInBytes) != 0)
+                return 1;
+        }
+        else
+            ktInstance->queueAbsorbedLen = len;
+    }
+
+    return 0;
+}
+
+int KangarooTwelve_Final(KangarooTwelve_Instance *ktInstance, unsigned char * output, const unsigned char * customization, size_t customLen)
+{
+    unsigned char encbuf[sizeof(size_t)+1+2];
+    unsigned char padding;
+
+    if (ktInstance->phase != ABSORBING)
+        return 1;
+
+    /* Absorb customization | right_encode(customLen) */
+    if ((customLen != 0) && (KangarooTwelve_Update(ktInstance, customization, customLen) != 0))
+        return 1;
+    if (KangarooTwelve_Update(ktInstance, encbuf, right_encode(encbuf, customLen)) != 0)
+        return 1;
+
+    if ( ktInstance->blockNumber == 0 ) {
+        /* Non complete first block in final node, pad it */
+        padding = 0x07; /*  '11': message hop, final node */
+    }
+    else {
+        unsigned int n;
+
+        if ( ktInstance->queueAbsorbedLen != 0 ) {
+            /* There is data in the queue node */
+            unsigned char intermediate[capacityInBytes];
+            ++ktInstance->blockNumber;
+            if (KeccakWidth1600_12rounds_SpongeAbsorbLastFewBits(&ktInstance->queueNode, suffixLeaf) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeSqueeze(&ktInstance->queueNode, intermediate, capacityInBytes) != 0)
+                return 1;
+            if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, intermediate, capacityInBytes) != 0)
+                return 1;
+        }
+        --ktInstance->blockNumber; /* Absorb right_encode(number of Chaining Values) || 0xFF || 0xFF */
+        n = right_encode(encbuf, ktInstance->blockNumber);
+        encbuf[n++] = 0xFF;
+        encbuf[n++] = 0xFF;
+        if (KeccakWidth1600_12rounds_SpongeAbsorb(&ktInstance->finalNode, encbuf, n) != 0)
+            return 1;
+        padding = 0x06; /* '01': chaining hop, final node */
+    }
+    if (KeccakWidth1600_12rounds_SpongeAbsorbLastFewBits(&ktInstance->finalNode, padding) != 0)
+        return 1;
+    if ( ktInstance->fixedOutputLength != 0 ) {
+        ktInstance->phase = FINAL;
+        return KeccakWidth1600_12rounds_SpongeSqueeze(&ktInstance->finalNode, output, ktInstance->fixedOutputLength);
+    }
+    ktInstance->phase = SQUEEZING;
+    return 0;
+}
+
+int KangarooTwelve_Squeeze(KangarooTwelve_Instance *ktInstance, unsigned char * output, size_t outputLen)
+{
+    if (ktInstance->phase != SQUEEZING)
+        return 1;
+    return KeccakWidth1600_12rounds_SpongeSqueeze(&ktInstance->finalNode, output, outputLen);
+}
+
+int KangarooTwelve( const unsigned char * input, size_t inLen, unsigned char * output, size_t outLen, const unsigned char * customization, size_t customLen )
+{
+    KangarooTwelve_Instance ktInstance;
+
+    if (outLen == 0)
+        return 1;
+    if (KangarooTwelve_Initialize(&ktInstance, outLen) != 0)
+        return 1;
+    if (KangarooTwelve_Update(&ktInstance, input, inLen) != 0)
+        return 1;
+    return KangarooTwelve_Final(&ktInstance, output, customization, customLen);
+}

src/crypto/randomx/panthera/KangarooTwelve.h

@@ -0,0 +1,89 @@
+/*
+Implementation by Ronny Van Keer, hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#ifndef _KangarooTwelve_h_
+#define _KangarooTwelve_h_
+
+#ifndef KeccakP1600_excluded
+
+#include <stddef.h>
+#include "align.h"
+#include "KeccakSpongeWidth1600.h"
+#include "Phases.h"
+
+typedef KCP_Phases KangarooTwelve_Phases;
+
+typedef struct {
+    KeccakWidth1600_12rounds_SpongeInstance queueNode;
+    KeccakWidth1600_12rounds_SpongeInstance finalNode;
+    size_t fixedOutputLength;
+    size_t blockNumber;
+    unsigned int queueAbsorbedLen;
+    KangarooTwelve_Phases phase;
+} KangarooTwelve_Instance;
+
+/** Extendable ouput function KangarooTwelve.
+  * @param  input           Pointer to the input message (M).
+  * @param  inputByteLen    The length of the input message in bytes.
+  * @param  output          Pointer to the output buffer.
+  * @param  outputByteLen   The desired number of output bytes.
+  * @param  customization   Pointer to the customization string (C).
+  * @param  customByteLen   The length of the customization string in bytes.
+  * @return 0 if successful, 1 otherwise.
+  */
+int KangarooTwelve(const unsigned char *input, size_t inputByteLen, unsigned char *output, size_t outputByteLen, const unsigned char *customization, size_t customByteLen );
+
+/**
+  * Function to initialize a KangarooTwelve instance.
+  * @param  ktInstance      Pointer to the instance to be initialized.
+  * @param  outputByteLen   The desired number of output bytes,
+  *                         or 0 for an arbitrarily-long output.
+  * @return 0 if successful, 1 otherwise.
+  */
+int KangarooTwelve_Initialize(KangarooTwelve_Instance *ktInstance, size_t outputByteLen);
+
+/**
+  * Function to give input data to be absorbed.
+  * @param  ktInstance      Pointer to the instance initialized by KangarooTwelve_Initialize().
+  * @param  input           Pointer to the input message data (M).
+  * @param  inputByteLen    The number of bytes provided in the input message data.
+  * @return 0 if successful, 1 otherwise.
+  */
+int KangarooTwelve_Update(KangarooTwelve_Instance *ktInstance, const unsigned char *input, size_t inputByteLen);
+
+/**
+  * Function to call after all the input message has been input, and to get
+  * output bytes if the length was specified when calling KangarooTwelve_Initialize().
+  * @param  ktInstance      Pointer to the hash instance initialized by KangarooTwelve_Initialize().
+  * If @a outputByteLen was not 0 in the call to KangarooTwelve_Initialize(), the number of
+  *     output bytes is equal to @a outputByteLen.
+  * If @a outputByteLen was 0 in the call to KangarooTwelve_Initialize(), the output bytes
+  *     must be extracted using the KangarooTwelve_Squeeze() function.
+  * @param  output          Pointer to the buffer where to store the output data.
+  * @param  customization   Pointer to the customization string (C).
+  * @param  customByteLen   The length of the customization string in bytes.
+  * @return 0 if successful, 1 otherwise.
+  */
+int KangarooTwelve_Final(KangarooTwelve_Instance *ktInstance, unsigned char *output, const unsigned char *customization, size_t customByteLen);
+
+/**
+  * Function to squeeze output data.
+  * @param  ktInstance     Pointer to the hash instance initialized by KangarooTwelve_Initialize().
+  * @param  data           Pointer to the buffer where to store the output data.
+  * @param  outputByteLen  The number of output bytes desired.
+  * @pre    KangarooTwelve_Final() must have been already called.
+  * @return 0 if successful, 1 otherwise.
+  */
+int KangarooTwelve_Squeeze(KangarooTwelve_Instance *ktInstance, unsigned char *output, size_t outputByteLen);
+
+#endif
+
+#endif

src/crypto/randomx/panthera/KeccakP-1600-SnP.h

@@ -0,0 +1,41 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+
+---
+
+Please refer to SnP-documentation.h for more details.
+*/
+
+#ifndef _KeccakP_1600_SnP_h_
+#define _KeccakP_1600_SnP_h_
+
+#define KeccakP1600_implementation      "64-bit reference implementation"
+#define KeccakP1600_stateSizeInBytes    200
+#define KeccakP1600_stateAlignment      8
+
+#ifdef KeccakReference
+void KeccakP1600_StaticInitialize( void );
+#else
+#define KeccakP1600_StaticInitialize()
+#endif
+void KeccakP1600_Initialize(void *state);
+void KeccakP1600_AddByte(void *state, unsigned char data, unsigned int offset);
+void KeccakP1600_AddBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
+void KeccakP1600_OverwriteBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length);
+void KeccakP1600_OverwriteWithZeroes(void *state, unsigned int byteCount);
+void KeccakP1600_Permute_Nrounds(void *state, unsigned int nrounds);
+void KeccakP1600_Permute_12rounds(void *state);
+void KeccakP1600_Permute_24rounds(void *state);
+void KeccakP1600_ExtractBytes(const void *state, unsigned char *data, unsigned int offset, unsigned int length);
+void KeccakP1600_ExtractAndAddBytes(const void *state, const unsigned char *input, unsigned char *output, unsigned int offset, unsigned int length);
+
+#endif

src/crypto/randomx/panthera/KeccakP-1600-reference.c

@@ -0,0 +1,402 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+
+---
+
+This file implements Keccak-p[1600] in a SnP-compatible way.
+Please refer to SnP-documentation.h for more details.
+
+This implementation comes with KeccakP-1600-SnP.h in the same folder.
+Please refer to LowLevel.build for the exact list of other files it must be combined with.
+*/
+
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include "brg_endian.h"
+#ifdef KeccakReference
+#include "displayIntermediateValues.h"
+#endif
+
+typedef unsigned char UINT8;
+typedef unsigned long long UINT64;
+typedef UINT64 tKeccakLane;
+
+#define maxNrRounds 24
+#define nrLanes 25
+#define index(x, y) (((x)%5)+5*((y)%5))
+
+#ifdef KeccakReference
+
+static tKeccakLane KeccakRoundConstants[maxNrRounds];
+static unsigned int KeccakRhoOffsets[nrLanes];
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_InitializeRoundConstants(void);
+void KeccakP1600_InitializeRhoOffsets(void);
+static int LFSR86540(UINT8 *LFSR);
+
+void KeccakP1600_StaticInitialize(void)
+{
+    if (sizeof(tKeccakLane) != 8) {
+        printf("tKeccakLane should be 64-bit wide\n");
+        abort();
+    }
+    KeccakP1600_InitializeRoundConstants();
+    KeccakP1600_InitializeRhoOffsets();
+}
+
+void KeccakP1600_InitializeRoundConstants(void)
+{
+    UINT8 LFSRstate = 0x01;
+    unsigned int i, j, bitPosition;
+
+    for(i=0; i<maxNrRounds; i++) {
+        KeccakRoundConstants[i] = 0;
+        for(j=0; j<7; j++) {
+            bitPosition = (1<<j)-1; /* 2^j-1 */
+            if (LFSR86540(&LFSRstate))
+                KeccakRoundConstants[i] ^= (tKeccakLane)1<<bitPosition;
+        }
+    }
+}
+
+void KeccakP1600_InitializeRhoOffsets(void)
+{
+    unsigned int x, y, t, newX, newY;
+
+    KeccakRhoOffsets[index(0, 0)] = 0;
+    x = 1;
+    y = 0;
+    for(t=0; t<24; t++) {
+        KeccakRhoOffsets[index(x, y)] = ((t+1)*(t+2)/2) % 64;
+        newX = (0*x+1*y) % 5;
+        newY = (2*x+3*y) % 5;
+        x = newX;
+        y = newY;
+    }
+}
+
+static int LFSR86540(UINT8 *LFSR)
+{
+    int result = ((*LFSR) & 0x01) != 0;
+    if (((*LFSR) & 0x80) != 0)
+        /* Primitive polynomial over GF(2): x^8+x^6+x^5+x^4+1 */
+        (*LFSR) = ((*LFSR) << 1) ^ 0x71;
+    else
+        (*LFSR) <<= 1;
+    return result;
+}
+
+#else
+
+static const tKeccakLane KeccakRoundConstants[maxNrRounds] =
+{
+    0x0000000000000001,
+    0x0000000000008082,
+    0x800000000000808a,
+    0x8000000080008000,
+    0x000000000000808b,
+    0x0000000080000001,
+    0x8000000080008081,
+    0x8000000000008009,
+    0x000000000000008a,
+    0x0000000000000088,
+    0x0000000080008009,
+    0x000000008000000a,
+    0x000000008000808b,
+    0x800000000000008b,
+    0x8000000000008089,
+    0x8000000000008003,
+    0x8000000000008002,
+    0x8000000000000080,
+    0x000000000000800a,
+    0x800000008000000a,
+    0x8000000080008081,
+    0x8000000000008080,
+    0x0000000080000001,
+    0x8000000080008008,
+};
+
+static const unsigned int KeccakRhoOffsets[nrLanes] =
+{
+     0,  1, 62, 28, 27, 36, 44,  6, 55, 20,  3, 10, 43, 25, 39, 41, 45, 15, 21,  8, 18,  2, 61, 56, 14
+};
+
+#endif
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_Initialize(void *state)
+{
+    memset(state, 0, 1600/8);
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_AddByte(void *state, unsigned char byte, unsigned int offset)
+{
+    assert(offset < 200);
+    ((unsigned char *)state)[offset] ^= byte;
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_AddBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length)
+{
+    unsigned int i;
+
+    assert(offset < 200);
+    assert(offset+length <= 200);
+    for(i=0; i<length; i++)
+        ((unsigned char *)state)[offset+i] ^= data[i];
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_OverwriteBytes(void *state, const unsigned char *data, unsigned int offset, unsigned int length)
+{
+    assert(offset < 200);
+    assert(offset+length <= 200);
+    memcpy((unsigned char*)state+offset, data, length);
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_OverwriteWithZeroes(void *state, unsigned int byteCount)
+{
+    assert(byteCount <= 200);
+    memset(state, 0, byteCount);
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600OnWords(tKeccakLane *state, unsigned int nrRounds);
+void KeccakP1600Round(tKeccakLane *state, unsigned int indexRound);
+static void theta(tKeccakLane *A);
+static void rho(tKeccakLane *A);
+static void pi(tKeccakLane *A);
+static void chi(tKeccakLane *A);
+static void iota(tKeccakLane *A, unsigned int indexRound);
+
+void KeccakP1600_Permute_Nrounds(void *state, unsigned int nrounds)
+{
+#if (PLATFORM_BYTE_ORDER != IS_LITTLE_ENDIAN)
+    tKeccakLane stateAsWords[1600/64];
+#endif
+
+#ifdef KeccakReference
+    displayStateAsBytes(1, "Input of permutation", (const unsigned char *)state, 1600);
+#endif
+#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
+    KeccakP1600OnWords((tKeccakLane*)state, nrounds);
+#else
+    fromBytesToWords(stateAsWords, (const unsigned char *)state);
+    KeccakP1600OnWords(stateAsWords, nrounds);
+    fromWordsToBytes((unsigned char *)state, stateAsWords);
+#endif
+#ifdef KeccakReference
+    displayStateAsBytes(1, "State after permutation", (const unsigned char *)state, 1600);
+#endif
+}
+
+void KeccakP1600_Permute_12rounds(void *state)
+{
+#if (PLATFORM_BYTE_ORDER != IS_LITTLE_ENDIAN)
+    tKeccakLane stateAsWords[1600/64];
+#endif
+
+#ifdef KeccakReference
+    displayStateAsBytes(1, "Input of permutation", (const unsigned char *)state, 1600);
+#endif
+#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
+    KeccakP1600OnWords((tKeccakLane*)state, 12);
+#else
+    fromBytesToWords(stateAsWords, (const unsigned char *)state);
+    KeccakP1600OnWords(stateAsWords, 12);
+    fromWordsToBytes((unsigned char *)state, stateAsWords);
+#endif
+#ifdef KeccakReference
+    displayStateAsBytes(1, "State after permutation", (const unsigned char *)state, 1600);
+#endif
+}
+
+void KeccakP1600_Permute_24rounds(void *state)
+{
+#if (PLATFORM_BYTE_ORDER != IS_LITTLE_ENDIAN)
+    tKeccakLane stateAsWords[1600/64];
+#endif
+
+#ifdef KeccakReference
+    displayStateAsBytes(1, "Input of permutation", (const unsigned char *)state, 1600);
+#endif
+#if (PLATFORM_BYTE_ORDER == IS_LITTLE_ENDIAN)
+    KeccakP1600OnWords((tKeccakLane*)state, 24);
+#else
+    fromBytesToWords(stateAsWords, (const unsigned char *)state);
+    KeccakP1600OnWords(stateAsWords, 24);
+    fromWordsToBytes((unsigned char *)state, stateAsWords);
+#endif
+#ifdef KeccakReference
+    displayStateAsBytes(1, "State after permutation", (const unsigned char *)state, 1600);
+#endif
+}
+
+void KeccakP1600OnWords(tKeccakLane *state, unsigned int nrRounds)
+{
+    unsigned int i;
+
+#ifdef KeccakReference
+    displayStateAsLanes(3, "Same, with lanes as 64-bit words", state, 1600);
+#endif
+
+    for(i=(maxNrRounds-nrRounds); i<maxNrRounds; i++)
+        KeccakP1600Round(state, i);
+}
+
+void KeccakP1600Round(tKeccakLane *state, unsigned int indexRound)
+{
+#ifdef KeccakReference
+    displayRoundNumber(3, indexRound);
+#endif
+
+    theta(state);
+#ifdef KeccakReference
+    displayStateAsLanes(3, "After theta", state, 1600);
+#endif
+
+    rho(state);
+#ifdef KeccakReference
+    displayStateAsLanes(3, "After rho", state, 1600);
+#endif
+
+    pi(state);
+#ifdef KeccakReference
+    displayStateAsLanes(3, "After pi", state, 1600);
+#endif
+
+    chi(state);
+#ifdef KeccakReference
+    displayStateAsLanes(3, "After chi", state, 1600);
+#endif
+
+    iota(state, indexRound);
+#ifdef KeccakReference
+    displayStateAsLanes(3, "After iota", state, 1600);
+#endif
+}
+
+#define ROL64(a, offset) ((offset != 0) ? ((((tKeccakLane)a) << offset) ^ (((tKeccakLane)a) >> (64-offset))) : a)
+
+static void theta(tKeccakLane *A)
+{
+    unsigned int x, y;
+    tKeccakLane C[5], D[5];
+
+    for(x=0; x<5; x++) {
+        C[x] = 0;
+        for(y=0; y<5; y++)
+            C[x] ^= A[index(x, y)];
+    }
+    for(x=0; x<5; x++)
+        D[x] = ROL64(C[(x+1)%5], 1) ^ C[(x+4)%5];
+    for(x=0; x<5; x++)
+        for(y=0; y<5; y++)
+            A[index(x, y)] ^= D[x];
+}
+
+static void rho(tKeccakLane *A)
+{
+    unsigned int x, y;
+
+    for(x=0; x<5; x++) for(y=0; y<5; y++)
+        A[index(x, y)] = ROL64(A[index(x, y)], KeccakRhoOffsets[index(x, y)]);
+}
+
+static void pi(tKeccakLane *A)
+{
+    unsigned int x, y;
+    tKeccakLane tempA[25];
+
+    for(x=0; x<5; x++) for(y=0; y<5; y++)
+        tempA[index(x, y)] = A[index(x, y)];
+    for(x=0; x<5; x++) for(y=0; y<5; y++)
+        A[index(0*x+1*y, 2*x+3*y)] = tempA[index(x, y)];
+}
+
+static void chi(tKeccakLane *A)
+{
+    unsigned int x, y;
+    tKeccakLane C[5];
+
+    for(y=0; y<5; y++) {
+        for(x=0; x<5; x++)
+            C[x] = A[index(x, y)] ^ ((~A[index(x+1, y)]) & A[index(x+2, y)]);
+        for(x=0; x<5; x++)
+            A[index(x, y)] = C[x];
+    }
+}
+
+static void iota(tKeccakLane *A, unsigned int indexRound)
+{
+    A[index(0, 0)] ^= KeccakRoundConstants[indexRound];
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_ExtractBytes(const void *state, unsigned char *data, unsigned int offset, unsigned int length)
+{
+    assert(offset < 200);
+    assert(offset+length <= 200);
+    memcpy(data, (unsigned char*)state+offset, length);
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_ExtractAndAddBytes(const void *state, const unsigned char *input, unsigned char *output, unsigned int offset, unsigned int length)
+{
+    unsigned int i;
+
+    assert(offset < 200);
+    assert(offset+length <= 200);
+    for(i=0; i<length; i++)
+        output[i] = input[i] ^ ((unsigned char *)state)[offset+i];
+}
+
+/* ---------------------------------------------------------------- */
+
+void KeccakP1600_DisplayRoundConstants(FILE *f)
+{
+    unsigned int i;
+
+    for(i=0; i<maxNrRounds; i++) {
+        fprintf(f, "RC[%02i][0][0] = ", i);
+        fprintf(f, "%08X", (unsigned int)(KeccakRoundConstants[i] >> 32));
+        fprintf(f, "%08X", (unsigned int)(KeccakRoundConstants[i] & 0xFFFFFFFFULL));
+        fprintf(f, "\n");
+    }
+    fprintf(f, "\n");
+}
+
+void KeccakP1600_DisplayRhoOffsets(FILE *f)
+{
+    unsigned int x, y;
+
+    for(y=0; y<5; y++) for(x=0; x<5; x++) {
+        fprintf(f, "RhoOffset[%i][%i] = ", x, y);
+        fprintf(f, "%2i", KeccakRhoOffsets[index(x, y)]);
+        fprintf(f, "\n");
+    }
+    fprintf(f, "\n");
+}

src/crypto/randomx/panthera/KeccakSponge-common.h

@@ -0,0 +1,35 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#ifndef _KeccakSpongeCommon_h_
+#define _KeccakSpongeCommon_h_
+
+#include <string.h>
+#include "align.h"
+
+#define KCP_DeclareSpongeStructure(prefix, size, alignment) \
+    ALIGN(alignment) typedef struct prefix##_SpongeInstanceStruct { \
+        unsigned char state[size]; \
+        unsigned int rate; \
+        unsigned int byteIOIndex; \
+        int squeezing; \
+    } prefix##_SpongeInstance;
+
+#define KCP_DeclareSpongeFunctions(prefix) \
+    int prefix##_Sponge(unsigned int rate, unsigned int capacity, const unsigned char *input, size_t inputByteLen, unsigned char suffix, unsigned char *output, size_t outputByteLen); \
+    int prefix##_SpongeInitialize(prefix##_SpongeInstance *spongeInstance, unsigned int rate, unsigned int capacity); \
+    int prefix##_SpongeAbsorb(prefix##_SpongeInstance *spongeInstance, const unsigned char *data, size_t dataByteLen); \
+    int prefix##_SpongeAbsorbLastFewBits(prefix##_SpongeInstance *spongeInstance, unsigned char delimitedData); \
+    int prefix##_SpongeSqueeze(prefix##_SpongeInstance *spongeInstance, unsigned char *data, size_t dataByteLen);
+
+#endif

src/crypto/randomx/panthera/KeccakSponge.inc

@@ -0,0 +1,311 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#define JOIN0(a, b)                     a ## b
+#define JOIN(a, b)                      JOIN0(a, b)
+
+#define Sponge                          JOIN(prefix, _Sponge)
+#define SpongeInstance                  JOIN(prefix, _SpongeInstance)
+#define SpongeInitialize                JOIN(prefix, _SpongeInitialize)
+#define SpongeAbsorb                    JOIN(prefix, _SpongeAbsorb)
+#define SpongeAbsorbLastFewBits         JOIN(prefix, _SpongeAbsorbLastFewBits)
+#define SpongeSqueeze                   JOIN(prefix, _SpongeSqueeze)
+
+#define SnP_stateSizeInBytes            JOIN(SnP, _stateSizeInBytes)
+#define SnP_stateAlignment              JOIN(SnP, _stateAlignment)
+#define SnP_StaticInitialize            JOIN(SnP, _StaticInitialize)
+#define SnP_Initialize                  JOIN(SnP, _Initialize)
+#define SnP_AddByte                     JOIN(SnP, _AddByte)
+#define SnP_AddBytes                    JOIN(SnP, _AddBytes)
+#define SnP_ExtractBytes                JOIN(SnP, _ExtractBytes)
+
+int Sponge(unsigned int rate, unsigned int capacity, const unsigned char *input, size_t inputByteLen, unsigned char suffix, unsigned char *output, size_t outputByteLen)
+{
+    ALIGN(SnP_stateAlignment) unsigned char state[SnP_stateSizeInBytes];
+    unsigned int partialBlock;
+    const unsigned char *curInput = input;
+    unsigned char *curOutput = output;
+    unsigned int rateInBytes = rate/8;
+
+    if (rate+capacity != SnP_width)
+        return 1;
+    if ((rate <= 0) || (rate > SnP_width) || ((rate % 8) != 0))
+        return 1;
+    if (suffix == 0)
+        return 1;
+
+    /* Initialize the state */
+    SnP_StaticInitialize();
+    SnP_Initialize(state);
+
+    /* First, absorb whole blocks */
+#ifdef SnP_FastLoop_Absorb
+    if (((rateInBytes % (SnP_width/200)) == 0) && (inputByteLen >= rateInBytes)) {
+        /* fast lane: whole lane rate */
+        size_t j;
+        j = SnP_FastLoop_Absorb(state, rateInBytes/(SnP_width/200), curInput, inputByteLen);
+        curInput += j;
+        inputByteLen -= j;
+    }
+#endif
+    while(inputByteLen >= (size_t)rateInBytes) {
+        #ifdef KeccakReference
+        displayBytes(1, "Block to be absorbed", curInput, rateInBytes);
+        #endif
+        SnP_AddBytes(state, curInput, 0, rateInBytes);
+        SnP_Permute(state);
+        curInput += rateInBytes;
+        inputByteLen -= rateInBytes;
+    }
+
+    /* Then, absorb what remains */
+    partialBlock = (unsigned int)inputByteLen;
+    #ifdef KeccakReference
+    displayBytes(1, "Block to be absorbed (part)", curInput, partialBlock);
+    #endif
+    SnP_AddBytes(state, curInput, 0, partialBlock);
+
+    /* Finally, absorb the suffix */
+    #ifdef KeccakReference
+    {
+        unsigned char delimitedData1[1];
+        delimitedData1[0] = suffix;
+        displayBytes(1, "Block to be absorbed (last few bits + first bit of padding)", delimitedData1, 1);
+    }
+    #endif
+    /* Last few bits, whose delimiter coincides with first bit of padding */
+    SnP_AddByte(state, suffix, partialBlock);
+    /* If the first bit of padding is at position rate-1, we need a whole new block for the second bit of padding */
+    if ((suffix >= 0x80) && (partialBlock == (rateInBytes-1)))
+        SnP_Permute(state);
+    /* Second bit of padding */
+    SnP_AddByte(state, 0x80, rateInBytes-1);
+    #ifdef KeccakReference
+    {
+        unsigned char block[SnP_width/8];
+        memset(block, 0, SnP_width/8);
+        block[rateInBytes-1] = 0x80;
+        displayBytes(1, "Second bit of padding", block, rateInBytes);
+    }
+    #endif
+    SnP_Permute(state);
+    #ifdef KeccakReference
+    displayText(1, "--- Switching to squeezing phase ---");
+    #endif
+
+    /* First, output whole blocks */
+    while(outputByteLen > (size_t)rateInBytes) {
+        SnP_ExtractBytes(state, curOutput, 0, rateInBytes);
+        SnP_Permute(state);
+        #ifdef KeccakReference
+        displayBytes(1, "Squeezed block", curOutput, rateInBytes);
+        #endif
+        curOutput += rateInBytes;
+        outputByteLen -= rateInBytes;
+    }
+
+    /* Finally, output what remains */
+    partialBlock = (unsigned int)outputByteLen;
+    SnP_ExtractBytes(state, curOutput, 0, partialBlock);
+    #ifdef KeccakReference
+    displayBytes(1, "Squeezed block (part)", curOutput, partialBlock);
+    #endif
+
+    return 0;
+}
+
+/* ---------------------------------------------------------------- */
+/* ---------------------------------------------------------------- */
+/* ---------------------------------------------------------------- */
+
+int SpongeInitialize(SpongeInstance *instance, unsigned int rate, unsigned int capacity)
+{
+    if (rate+capacity != SnP_width)
+        return 1;
+    if ((rate <= 0) || (rate > SnP_width) || ((rate % 8) != 0))
+        return 1;
+    SnP_StaticInitialize();
+    SnP_Initialize(instance->state);
+    instance->rate = rate;
+    instance->byteIOIndex = 0;
+    instance->squeezing = 0;
+
+    return 0;
+}
+
+/* ---------------------------------------------------------------- */
+
+int SpongeAbsorb(SpongeInstance *instance, const unsigned char *data, size_t dataByteLen)
+{
+    size_t i, j;
+    unsigned int partialBlock;
+    const unsigned char *curData;
+    unsigned int rateInBytes = instance->rate/8;
+
+    if (instance->squeezing)
+        return 1; /* Too late for additional input */
+
+    i = 0;
+    curData = data;
+    while(i < dataByteLen) {
+        if ((instance->byteIOIndex == 0) && (dataByteLen >= (i + rateInBytes))) {
+#ifdef SnP_FastLoop_Absorb
+            /* processing full blocks first */
+            if ((rateInBytes % (SnP_width/200)) == 0) {
+                /* fast lane: whole lane rate */
+                j = SnP_FastLoop_Absorb(instance->state, rateInBytes/(SnP_width/200), curData, dataByteLen - i);
+                i += j;
+                curData += j;
+            }
+            else {
+#endif
+                for(j=dataByteLen-i; j>=rateInBytes; j-=rateInBytes) {
+                    #ifdef KeccakReference
+                    displayBytes(1, "Block to be absorbed", curData, rateInBytes);
+                    #endif
+                    SnP_AddBytes(instance->state, curData, 0, rateInBytes);
+                    SnP_Permute(instance->state);
+                    curData+=rateInBytes;
+                }
+                i = dataByteLen - j;
+#ifdef SnP_FastLoop_Absorb
+            }
+#endif
+        }
+        else {
+            /* normal lane: using the message queue */
+            partialBlock = (unsigned int)(dataByteLen - i);
+            if (partialBlock+instance->byteIOIndex > rateInBytes)
+                partialBlock = rateInBytes-instance->byteIOIndex;
+            #ifdef KeccakReference
+            displayBytes(1, "Block to be absorbed (part)", curData, partialBlock);
+            #endif
+            i += partialBlock;
+
+            SnP_AddBytes(instance->state, curData, instance->byteIOIndex, partialBlock);
+            curData += partialBlock;
+            instance->byteIOIndex += partialBlock;
+            if (instance->byteIOIndex == rateInBytes) {
+                SnP_Permute(instance->state);
+                instance->byteIOIndex = 0;
+            }
+        }
+    }
+    return 0;
+}
+
+/* ---------------------------------------------------------------- */
+
+int SpongeAbsorbLastFewBits(SpongeInstance *instance, unsigned char delimitedData)
+{
+    unsigned int rateInBytes = instance->rate/8;
+
+    if (delimitedData == 0)
+        return 1;
+    if (instance->squeezing)
+        return 1; /* Too late for additional input */
+
+    #ifdef KeccakReference
+    {
+        unsigned char delimitedData1[1];
+        delimitedData1[0] = delimitedData;
+        displayBytes(1, "Block to be absorbed (last few bits + first bit of padding)", delimitedData1, 1);
+    }
+    #endif
+    /* Last few bits, whose delimiter coincides with first bit of padding */
+    SnP_AddByte(instance->state, delimitedData, instance->byteIOIndex);
+    /* If the first bit of padding is at position rate-1, we need a whole new block for the second bit of padding */
+    if ((delimitedData >= 0x80) && (instance->byteIOIndex == (rateInBytes-1)))
+        SnP_Permute(instance->state);
+    /* Second bit of padding */
+    SnP_AddByte(instance->state, 0x80, rateInBytes-1);
+    #ifdef KeccakReference
+    {
+        unsigned char block[SnP_width/8];
+        memset(block, 0, SnP_width/8);
+        block[rateInBytes-1] = 0x80;
+        displayBytes(1, "Second bit of padding", block, rateInBytes);
+    }
+    #endif
+    SnP_Permute(instance->state);
+    instance->byteIOIndex = 0;
+    instance->squeezing = 1;
+    #ifdef KeccakReference
+    displayText(1, "--- Switching to squeezing phase ---");
+    #endif
+    return 0;
+}
+
+/* ---------------------------------------------------------------- */
+
+int SpongeSqueeze(SpongeInstance *instance, unsigned char *data, size_t dataByteLen)
+{
+    size_t i, j;
+    unsigned int partialBlock;
+    unsigned int rateInBytes = instance->rate/8;
+    unsigned char *curData;
+
+    if (!instance->squeezing)
+        SpongeAbsorbLastFewBits(instance, 0x01);
+
+    i = 0;
+    curData = data;
+    while(i < dataByteLen) {
+        if ((instance->byteIOIndex == rateInBytes) && (dataByteLen >= (i + rateInBytes))) {
+            for(j=dataByteLen-i; j>=rateInBytes; j-=rateInBytes) {
+                SnP_Permute(instance->state);
+                SnP_ExtractBytes(instance->state, curData, 0, rateInBytes);
+                #ifdef KeccakReference
+                displayBytes(1, "Squeezed block", curData, rateInBytes);
+                #endif
+                curData+=rateInBytes;
+            }
+            i = dataByteLen - j;
+        }
+        else {
+            /* normal lane: using the message queue */
+            if (instance->byteIOIndex == rateInBytes) {
+                SnP_Permute(instance->state);
+                instance->byteIOIndex = 0;
+            }
+            partialBlock = (unsigned int)(dataByteLen - i);
+            if (partialBlock+instance->byteIOIndex > rateInBytes)
+                partialBlock = rateInBytes-instance->byteIOIndex;
+            i += partialBlock;
+
+            SnP_ExtractBytes(instance->state, curData, instance->byteIOIndex, partialBlock);
+            #ifdef KeccakReference
+            displayBytes(1, "Squeezed block (part)", curData, partialBlock);
+            #endif
+            curData += partialBlock;
+            instance->byteIOIndex += partialBlock;
+        }
+    }
+    return 0;
+}
+
+/* ---------------------------------------------------------------- */
+
+#undef Sponge
+#undef SpongeInstance
+#undef SpongeInitialize
+#undef SpongeAbsorb
+#undef SpongeAbsorbLastFewBits
+#undef SpongeSqueeze
+#undef SnP_stateSizeInBytes
+#undef SnP_stateAlignment
+#undef SnP_StaticInitialize
+#undef SnP_Initialize
+#undef SnP_AddByte
+#undef SnP_AddBytes
+#undef SnP_ExtractBytes

src/crypto/randomx/panthera/KeccakSpongeWidth1600.c

@@ -0,0 +1,54 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#include "KeccakSpongeWidth1600.h"
+
+#ifdef KeccakReference
+    #include "displayIntermediateValues.h"
+#endif
+
+#ifndef KeccakP1600_excluded
+    #include "KeccakP-1600-SnP.h"
+
+    #define prefix KeccakWidth1600
+    #define SnP KeccakP1600
+    #define SnP_width 1600
+    #define SnP_Permute KeccakP1600_Permute_24rounds
+    #if defined(KeccakF1600_FastLoop_supported)
+        #define SnP_FastLoop_Absorb KeccakF1600_FastLoop_Absorb
+    #endif
+        #include "KeccakSponge.inc"
+    #undef prefix
+    #undef SnP
+    #undef SnP_width
+    #undef SnP_Permute
+    #undef SnP_FastLoop_Absorb
+#endif
+
+#ifndef KeccakP1600_excluded
+    #include "KeccakP-1600-SnP.h"
+
+    #define prefix KeccakWidth1600_12rounds
+    #define SnP KeccakP1600
+    #define SnP_width 1600
+    #define SnP_Permute KeccakP1600_Permute_12rounds
+    #if defined(KeccakP1600_12rounds_FastLoop_supported)
+        #define SnP_FastLoop_Absorb KeccakP1600_12rounds_FastLoop_Absorb
+    #endif
+        #include "KeccakSponge.inc"
+    #undef prefix
+    #undef SnP
+    #undef SnP_width
+    #undef SnP_Permute
+    #undef SnP_FastLoop_Absorb
+#endif

src/crypto/randomx/panthera/KeccakSpongeWidth1600.h

@@ -0,0 +1,31 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#ifndef _KeccakSpongeWidth1600_h_
+#define _KeccakSpongeWidth1600_h_
+
+#include "KeccakSponge-common.h"
+
+#ifndef KeccakP1600_excluded
+    #include "KeccakP-1600-SnP.h"
+    KCP_DeclareSpongeStructure(KeccakWidth1600, KeccakP1600_stateSizeInBytes, KeccakP1600_stateAlignment)
+    KCP_DeclareSpongeFunctions(KeccakWidth1600)
+#endif
+
+#ifndef KeccakP1600_excluded
+    #include "KeccakP-1600-SnP.h"
+    KCP_DeclareSpongeStructure(KeccakWidth1600_12rounds, KeccakP1600_stateSizeInBytes, KeccakP1600_stateAlignment)
+    KCP_DeclareSpongeFunctions(KeccakWidth1600_12rounds)
+#endif
+
+#endif

src/crypto/randomx/panthera/Phases.h

@@ -0,0 +1,22 @@
+/*
+Implementation by Ronny Van Keer, hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#ifndef _Phases_h_
+#define _Phases_h_
+
+typedef enum {
+    NOT_INITIALIZED,
+    ABSORBING,
+    FINAL,
+    SQUEEZING
+} KCP_Phases;
+
+#endif

src/crypto/randomx/panthera/align.h

@@ -0,0 +1,32 @@
+/*
+Implementation by the Keccak Team, namely, Guido Bertoni, Joan Daemen,
+Michaël Peeters, Gilles Van Assche and Ronny Van Keer,
+hereby denoted as "the implementer".
+
+For more information, feedback or questions, please refer to our website:
+https://keccak.team/
+
+To the extent possible under law, the implementer has waived all copyright
+and related or neighboring rights to the source code in this file.
+http://creativecommons.org/publicdomain/zero/1.0/
+*/
+
+#ifndef _align_h_
+#define _align_h_
+
+/* on Mac OS-X and possibly others, ALIGN(x) is defined in param.h, and -Werror chokes on the redef. */
+#ifdef ALIGN
+#undef ALIGN
+#endif
+
+#if defined(__GNUC__)
+#define ALIGN(x) __attribute__ ((aligned(x)))
+#elif defined(_MSC_VER)
+#define ALIGN(x) __declspec(align(x))
+#elif defined(__ARMCC_VERSION)
+#define ALIGN(x) __align(x)
+#else
+#define ALIGN(x)
+#endif
+
+#endif

src/crypto/randomx/panthera/brg_endian.h

@@ -0,0 +1,143 @@
+/*
+ ---------------------------------------------------------------------------
+ Copyright (c) 1998-2008, Brian Gladman, Worcester, UK. All rights reserved.
+
+ LICENSE TERMS
+
+ The redistribution and use of this software (with or without changes)
+ is allowed without the payment of fees or royalties provided that:
+
+  1. source code distributions include the above copyright notice, this
+     list of conditions and the following disclaimer;
+
+  2. binary distributions include the above copyright notice, this list
+     of conditions and the following disclaimer in their documentation;
+
+  3. the name of the copyright holder is not used to endorse products
+     built using this software without specific written permission.
+
+ DISCLAIMER
+
+ This software is provided 'as is' with no explicit or implied warranties
+ in respect of its properties, including, but not limited to, correctness
+ and/or fitness for purpose.
+ ---------------------------------------------------------------------------
+ Issue Date: 20/12/2007
+ Changes for ARM 9/9/2010
+*/
+
+#ifndef _BRG_ENDIAN_H
+#define _BRG_ENDIAN_H
+
+#define IS_BIG_ENDIAN      4321 /* byte 0 is most significant (mc68k) */
+#define IS_LITTLE_ENDIAN   1234 /* byte 0 is least significant (i386) */
+
+#if 0
+/* Include files where endian defines and byteswap functions may reside */
+#if defined( __sun )
+#  include <sys/isa_defs.h>
+#elif defined( __FreeBSD__ ) || defined( __OpenBSD__ ) || defined( __NetBSD__ )
+#  include <sys/endian.h>
+#elif defined( BSD ) && ( BSD >= 199103 ) || defined( __APPLE__ ) || \
+      defined( __CYGWIN32__ ) || defined( __DJGPP__ ) || defined( __osf__ )
+#  include <machine/endian.h>
+#elif defined( __linux__ ) || defined( __GNUC__ ) || defined( __GNU_LIBRARY__ )
+#  if !defined( __MINGW32__ ) && !defined( _AIX )
+#    include <endian.h>
+#    if !defined( __BEOS__ )
+#      include <byteswap.h>
+#    endif
+#  endif
+#endif
+#endif
+
+/* Now attempt to set the define for platform byte order using any  */
+/* of the four forms SYMBOL, _SYMBOL, __SYMBOL & __SYMBOL__, which  */
+/* seem to encompass most endian symbol definitions                 */
+
+#if defined( BIG_ENDIAN ) && defined( LITTLE_ENDIAN )
+#  if defined( BYTE_ORDER ) && BYTE_ORDER == BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( BYTE_ORDER ) && BYTE_ORDER == LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( _BIG_ENDIAN ) && defined( _LITTLE_ENDIAN )
+#  if defined( _BYTE_ORDER ) && _BYTE_ORDER == _BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( _BYTE_ORDER ) && _BYTE_ORDER == _LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( _BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( _LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN ) && defined( __LITTLE_ENDIAN )
+#  if defined( __BYTE_ORDER ) && __BYTE_ORDER == __BIG_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( __BYTE_ORDER ) && __BYTE_ORDER == __LITTLE_ENDIAN
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( __BIG_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+#if defined( __BIG_ENDIAN__ ) && defined( __LITTLE_ENDIAN__ )
+#  if defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __BIG_ENDIAN__
+#    define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#  elif defined( __BYTE_ORDER__ ) && __BYTE_ORDER__ == __LITTLE_ENDIAN__
+#    define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#  endif
+#elif defined( __BIG_ENDIAN__ )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#elif defined( __LITTLE_ENDIAN__ )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#endif
+
+/*  if the platform byte order could not be determined, then try to */
+/*  set this define using common machine defines                    */
+#if !defined(PLATFORM_BYTE_ORDER)
+
+#if   defined( __alpha__ ) || defined( __alpha ) || defined( i386 )       || \
+      defined( __i386__ )  || defined( _M_I86 )  || defined( _M_IX86 )    || \
+      defined( __OS2__ )   || defined( sun386 )  || defined( __TURBOC__ ) || \
+      defined( vax )       || defined( vms )     || defined( VMS )        || \
+      defined( __VMS )     || defined( _M_X64 )
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+
+#elif defined( AMIGA )    || defined( applec )    || defined( __AS400__ )  || \
+      defined( _CRAY )    || defined( __hppa )    || defined( __hp9000 )   || \
+      defined( ibm370 )   || defined( mc68000 )   || defined( m68k )       || \
+      defined( __MRC__ )  || defined( __MVS__ )   || defined( __MWERKS__ ) || \
+      defined( sparc )    || defined( __sparc)    || defined( SYMANTEC_C ) || \
+      defined( __VOS__ )  || defined( __TIGCC__ ) || defined( __TANDEM )   || \
+      defined( THINK_C )  || defined( __VMCMS__ ) || defined( _AIX )       || \
+      defined( __s390__ ) || defined( __s390x__ ) || defined( __zarch__ )
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+
+#elif defined(__arm__)
+# ifdef __BIG_ENDIAN
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+# else
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+# endif
+#elif 1     /* **** EDIT HERE IF NECESSARY **** */
+#  define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
+#elif 0     /* **** EDIT HERE IF NECESSARY **** */
+#  define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
+#else
+#  error Please edit lines 132 or 134 in brg_endian.h to set the platform byte order
+#endif
+
+#endif
+
+#endif

src/crypto/randomx/panthera/insecure_memzero.h

@@ -0,0 +1 @@
+#define insecure_memzero(buf, len) /* empty */

src/crypto/randomx/panthera/sha256.c

@@ -0,0 +1,653 @@
+/*-
+ * Copyright 2005-2016 Colin Percival
+ * Copyright 2016-2018 Alexander Peslyak
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <assert.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "insecure_memzero.h"
+#include "sysendian.h"
+
+#include "sha256.h"
+
+#ifdef __ICC
+/* Miscompile with icc 14.0.0 (at least), so don't use restrict there */
+#define restrict
+#define static_restrict static
+#elif defined(_MSC_VER)
+#define restrict
+#define static_restrict
+#elif __STDC_VERSION__ >= 199901L
+/* Have restrict */
+#define static_restrict static restrict
+#elif defined(__GNUC__)
+#define restrict __restrict
+#define static_restrict static __restrict
+#else
+#define restrict
+#define static_restrict
+#endif
+
+/*
+ * Encode a length len*2 vector of (uint32_t) into a length len*8 vector of
+ * (uint8_t) in big-endian form.
+ */
+static void
+be32enc_vect(uint8_t * dst, const uint32_t * src, size_t len)
+{
+
+	/* Encode vector, two words at a time. */
+	do {
+		be32enc(&dst[0], src[0]);
+		be32enc(&dst[4], src[1]);
+		src += 2;
+		dst += 8;
+	} while (--len);
+}
+
+/*
+ * Decode a big-endian length len*8 vector of (uint8_t) into a length
+ * len*2 vector of (uint32_t).
+ */
+static void
+be32dec_vect(uint32_t * dst, const uint8_t * src, size_t len)
+{
+
+	/* Decode vector, two words at a time. */
+	do {
+		dst[0] = be32dec(&src[0]);
+		dst[1] = be32dec(&src[4]);
+		src += 8;
+		dst += 2;
+	} while (--len);
+}
+
+/* SHA256 round constants. */
+static const uint32_t Krnd[64] = {
+	0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5,
+	0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
+	0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3,
+	0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
+	0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc,
+	0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
+	0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7,
+	0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
+	0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13,
+	0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
+	0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3,
+	0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
+	0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5,
+	0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
+	0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208,
+	0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
+};
+
+/* Elementary functions used by SHA256 */
+#define Ch(x, y, z)	((x & (y ^ z)) ^ z)
+#define Maj(x, y, z)	((x & (y | z)) | (y & z))
+#define SHR(x, n)	(x >> n)
+#define ROTR(x, n)	((x >> n) | (x << (32 - n)))
+#define S0(x)		(ROTR(x, 2) ^ ROTR(x, 13) ^ ROTR(x, 22))
+#define S1(x)		(ROTR(x, 6) ^ ROTR(x, 11) ^ ROTR(x, 25))
+#define s0(x)		(ROTR(x, 7) ^ ROTR(x, 18) ^ SHR(x, 3))
+#define s1(x)		(ROTR(x, 17) ^ ROTR(x, 19) ^ SHR(x, 10))
+
+/* SHA256 round function */
+#define RND(a, b, c, d, e, f, g, h, k)			\
+	h += S1(e) + Ch(e, f, g) + k;			\
+	d += h;						\
+	h += S0(a) + Maj(a, b, c);
+
+/* Adjusted round function for rotating state */
+#define RNDr(S, W, i, ii)			\
+	RND(S[(64 - i) % 8], S[(65 - i) % 8],	\
+	    S[(66 - i) % 8], S[(67 - i) % 8],	\
+	    S[(68 - i) % 8], S[(69 - i) % 8],	\
+	    S[(70 - i) % 8], S[(71 - i) % 8],	\
+	    W[i + ii] + Krnd[i + ii])
+
+/* Message schedule computation */
+#define MSCH(W, ii, i)				\
+	W[i + ii + 16] = s1(W[i + ii + 14]) + W[i + ii + 9] + s0(W[i + ii + 1]) + W[i + ii]
+
+/*
+ * SHA256 block compression function.  The 256-bit state is transformed via
+ * the 512-bit input block to produce a new state.
+ */
+static void
+SHA256_Transform(uint32_t state[static_restrict 8],
+    const uint8_t block[static_restrict 64],
+    uint32_t W[static_restrict 64], uint32_t S[static_restrict 8])
+{
+	int i;
+
+	/* 1. Prepare the first part of the message schedule W. */
+	be32dec_vect(W, block, 8);
+
+	/* 2. Initialize working variables. */
+	memcpy(S, state, 32);
+
+	/* 3. Mix. */
+	for (i = 0; i < 64; i += 16) {
+		RNDr(S, W, 0, i);
+		RNDr(S, W, 1, i);
+		RNDr(S, W, 2, i);
+		RNDr(S, W, 3, i);
+		RNDr(S, W, 4, i);
+		RNDr(S, W, 5, i);
+		RNDr(S, W, 6, i);
+		RNDr(S, W, 7, i);
+		RNDr(S, W, 8, i);
+		RNDr(S, W, 9, i);
+		RNDr(S, W, 10, i);
+		RNDr(S, W, 11, i);
+		RNDr(S, W, 12, i);
+		RNDr(S, W, 13, i);
+		RNDr(S, W, 14, i);
+		RNDr(S, W, 15, i);
+
+		if (i == 48)
+			break;
+		MSCH(W, 0, i);
+		MSCH(W, 1, i);
+		MSCH(W, 2, i);
+		MSCH(W, 3, i);
+		MSCH(W, 4, i);
+		MSCH(W, 5, i);
+		MSCH(W, 6, i);
+		MSCH(W, 7, i);
+		MSCH(W, 8, i);
+		MSCH(W, 9, i);
+		MSCH(W, 10, i);
+		MSCH(W, 11, i);
+		MSCH(W, 12, i);
+		MSCH(W, 13, i);
+		MSCH(W, 14, i);
+		MSCH(W, 15, i);
+	}
+
+	/* 4. Mix local working variables into global state. */
+	state[0] += S[0];
+	state[1] += S[1];
+	state[2] += S[2];
+	state[3] += S[3];
+	state[4] += S[4];
+	state[5] += S[5];
+	state[6] += S[6];
+	state[7] += S[7];
+}
+
+static const uint8_t PAD[64] = {
+	0x80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+};
+
+/* Add padding and terminating bit-count. */
+static void
+SHA256_Pad(SHA256_CTX * ctx, uint32_t tmp32[static_restrict 72])
+{
+	size_t r;
+
+	/* Figure out how many bytes we have buffered. */
+	r = (ctx->count >> 3) & 0x3f;
+
+	/* Pad to 56 mod 64, transforming if we finish a block en route. */
+	if (r < 56) {
+		/* Pad to 56 mod 64. */
+		memcpy(&ctx->buf[r], PAD, 56 - r);
+	} else {
+		/* Finish the current block and mix. */
+		memcpy(&ctx->buf[r], PAD, 64 - r);
+		SHA256_Transform(ctx->state, ctx->buf, &tmp32[0], &tmp32[64]);
+
+		/* The start of the final block is all zeroes. */
+		memset(&ctx->buf[0], 0, 56);
+	}
+
+	/* Add the terminating bit-count. */
+	be64enc(&ctx->buf[56], ctx->count);
+
+	/* Mix in the final block. */
+	SHA256_Transform(ctx->state, ctx->buf, &tmp32[0], &tmp32[64]);
+}
+
+/* Magic initialization constants. */
+static const uint32_t initial_state[8] = {
+	0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A,
+	0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19
+};
+
+/**
+ * SHA256_Init(ctx):
+ * Initialize the SHA256 context ${ctx}.
+ */
+void
+SHA256_Init(SHA256_CTX * ctx)
+{
+
+	/* Zero bits processed so far. */
+	ctx->count = 0;
+
+	/* Initialize state. */
+	memcpy(ctx->state, initial_state, sizeof(initial_state));
+}
+
+/**
+ * SHA256_Update(ctx, in, len):
+ * Input ${len} bytes from ${in} into the SHA256 context ${ctx}.
+ */
+static void
+_SHA256_Update(SHA256_CTX * ctx, const void * in, size_t len,
+    uint32_t tmp32[static_restrict 72])
+{
+	uint32_t r;
+	const uint8_t * src = in;
+
+	/* Return immediately if we have nothing to do. */
+	if (len == 0)
+		return;
+
+	/* Number of bytes left in the buffer from previous updates. */
+	r = (ctx->count >> 3) & 0x3f;
+
+	/* Update number of bits. */
+	ctx->count += (uint64_t)(len) << 3;
+
+	/* Handle the case where we don't need to perform any transforms. */
+	if (len < 64 - r) {
+		memcpy(&ctx->buf[r], src, len);
+		return;
+	}
+
+	/* Finish the current block. */
+	memcpy(&ctx->buf[r], src, 64 - r);
+	SHA256_Transform(ctx->state, ctx->buf, &tmp32[0], &tmp32[64]);
+	src += 64 - r;
+	len -= 64 - r;
+
+	/* Perform complete blocks. */
+	while (len >= 64) {
+		SHA256_Transform(ctx->state, src, &tmp32[0], &tmp32[64]);
+		src += 64;
+		len -= 64;
+	}
+
+	/* Copy left over data into buffer. */
+	memcpy(ctx->buf, src, len);
+}
+
+/* Wrapper function for intermediate-values sanitization. */
+void
+SHA256_Update(SHA256_CTX * ctx, const void * in, size_t len)
+{
+	uint32_t tmp32[72];
+
+	/* Call the real function. */
+	_SHA256_Update(ctx, in, len, tmp32);
+
+	/* Clean the stack. */
+	insecure_memzero(tmp32, 288);
+}
+
+/**
+ * SHA256_Final(digest, ctx):
+ * Output the SHA256 hash of the data input to the context ${ctx} into the
+ * buffer ${digest}.
+ */
+static void
+_SHA256_Final(uint8_t digest[32], SHA256_CTX * ctx,
+    uint32_t tmp32[static_restrict 72])
+{
+
+	/* Add padding. */
+	SHA256_Pad(ctx, tmp32);
+
+	/* Write the hash. */
+	be32enc_vect(digest, ctx->state, 4);
+}
+
+/* Wrapper function for intermediate-values sanitization. */
+void
+SHA256_Final(uint8_t digest[32], SHA256_CTX * ctx)
+{
+	uint32_t tmp32[72];
+
+	/* Call the real function. */
+	_SHA256_Final(digest, ctx, tmp32);
+
+	/* Clear the context state. */
+	insecure_memzero(ctx, sizeof(SHA256_CTX));
+
+	/* Clean the stack. */
+	insecure_memzero(tmp32, 288);
+}
+
+/**
+ * SHA256_Buf(in, len, digest):
+ * Compute the SHA256 hash of ${len} bytes from ${in} and write it to ${digest}.
+ */
+void
+SHA256_Buf(const void * in, size_t len, uint8_t digest[32])
+{
+	SHA256_CTX ctx;
+	uint32_t tmp32[72];
+
+	SHA256_Init(&ctx);
+	_SHA256_Update(&ctx, in, len, tmp32);
+	_SHA256_Final(digest, &ctx, tmp32);
+
+	/* Clean the stack. */
+	insecure_memzero(&ctx, sizeof(SHA256_CTX));
+	insecure_memzero(tmp32, 288);
+}
+
+/**
+ * HMAC_SHA256_Init(ctx, K, Klen):
+ * Initialize the HMAC-SHA256 context ${ctx} with ${Klen} bytes of key from
+ * ${K}.
+ */
+static void
+_HMAC_SHA256_Init(HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen,
+    uint32_t tmp32[static_restrict 72], uint8_t pad[static_restrict 64],
+    uint8_t khash[static_restrict 32])
+{
+	const uint8_t * K = _K;
+	size_t i;
+
+	/* If Klen > 64, the key is really SHA256(K). */
+	if (Klen > 64) {
+		SHA256_Init(&ctx->ictx);
+		_SHA256_Update(&ctx->ictx, K, Klen, tmp32);
+		_SHA256_Final(khash, &ctx->ictx, tmp32);
+		K = khash;
+		Klen = 32;
+	}
+
+	/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
+	SHA256_Init(&ctx->ictx);
+	memset(pad, 0x36, 64);
+	for (i = 0; i < Klen; i++)
+		pad[i] ^= K[i];
+	_SHA256_Update(&ctx->ictx, pad, 64, tmp32);
+
+	/* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
+	SHA256_Init(&ctx->octx);
+	memset(pad, 0x5c, 64);
+	for (i = 0; i < Klen; i++)
+		pad[i] ^= K[i];
+	_SHA256_Update(&ctx->octx, pad, 64, tmp32);
+}
+
+/* Wrapper function for intermediate-values sanitization. */
+void
+HMAC_SHA256_Init(HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen)
+{
+	uint32_t tmp32[72];
+	uint8_t pad[64];
+	uint8_t khash[32];
+
+	/* Call the real function. */
+	_HMAC_SHA256_Init(ctx, _K, Klen, tmp32, pad, khash);
+
+	/* Clean the stack. */
+	insecure_memzero(tmp32, 288);
+	insecure_memzero(khash, 32);
+	insecure_memzero(pad, 64);
+}
+
+/**
+ * HMAC_SHA256_Update(ctx, in, len):
+ * Input ${len} bytes from ${in} into the HMAC-SHA256 context ${ctx}.
+ */
+static void
+_HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void * in, size_t len,
+    uint32_t tmp32[static_restrict 72])
+{
+
+	/* Feed data to the inner SHA256 operation. */
+	_SHA256_Update(&ctx->ictx, in, len, tmp32);
+}
+
+/* Wrapper function for intermediate-values sanitization. */
+void
+HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void * in, size_t len)
+{
+	uint32_t tmp32[72];
+
+	/* Call the real function. */
+	_HMAC_SHA256_Update(ctx, in, len, tmp32);
+
+	/* Clean the stack. */
+	insecure_memzero(tmp32, 288);
+}
+
+/**
+ * HMAC_SHA256_Final(digest, ctx):
+ * Output the HMAC-SHA256 of the data input to the context ${ctx} into the
+ * buffer ${digest}.
+ */
+static void
+_HMAC_SHA256_Final(uint8_t digest[32], HMAC_SHA256_CTX * ctx,
+    uint32_t tmp32[static_restrict 72], uint8_t ihash[static_restrict 32])
+{
+
+	/* Finish the inner SHA256 operation. */
+	_SHA256_Final(ihash, &ctx->ictx, tmp32);
+
+	/* Feed the inner hash to the outer SHA256 operation. */
+	_SHA256_Update(&ctx->octx, ihash, 32, tmp32);
+
+	/* Finish the outer SHA256 operation. */
+	_SHA256_Final(digest, &ctx->octx, tmp32);
+}
+
+/* Wrapper function for intermediate-values sanitization. */
+void
+HMAC_SHA256_Final(uint8_t digest[32], HMAC_SHA256_CTX * ctx)
+{
+	uint32_t tmp32[72];
+	uint8_t ihash[32];
+
+	/* Call the real function. */
+	_HMAC_SHA256_Final(digest, ctx, tmp32, ihash);
+
+	/* Clean the stack. */
+	insecure_memzero(tmp32, 288);
+	insecure_memzero(ihash, 32);
+}
+
+/**
+ * HMAC_SHA256_Buf(K, Klen, in, len, digest):
+ * Compute the HMAC-SHA256 of ${len} bytes from ${in} using the key ${K} of
+ * length ${Klen}, and write the result to ${digest}.
+ */
+void
+HMAC_SHA256_Buf(const void * K, size_t Klen, const void * in, size_t len,
+    uint8_t digest[32])
+{
+	HMAC_SHA256_CTX ctx;
+	uint32_t tmp32[72];
+	uint8_t tmp8[96];
+
+	_HMAC_SHA256_Init(&ctx, K, Klen, tmp32, &tmp8[0], &tmp8[64]);
+	_HMAC_SHA256_Update(&ctx, in, len, tmp32);
+	_HMAC_SHA256_Final(digest, &ctx, tmp32, &tmp8[0]);
+
+	/* Clean the stack. */
+	insecure_memzero(&ctx, sizeof(HMAC_SHA256_CTX));
+	insecure_memzero(tmp32, 288);
+	insecure_memzero(tmp8, 96);
+}
+
+/* Add padding and terminating bit-count, but don't invoke Transform yet. */
+static int
+SHA256_Pad_Almost(SHA256_CTX * ctx, uint8_t len[static_restrict 8],
+    uint32_t tmp32[static_restrict 72])
+{
+	uint32_t r;
+
+	r = (ctx->count >> 3) & 0x3f;
+	if (r >= 56)
+		return -1;
+
+	/*
+	 * Convert length to a vector of bytes -- we do this now rather
+	 * than later because the length will change after we pad.
+	 */
+	be64enc(len, ctx->count);
+
+	/* Add 1--56 bytes so that the resulting length is 56 mod 64. */
+	_SHA256_Update(ctx, PAD, 56 - r, tmp32);
+
+	/* Add the terminating bit-count. */
+	ctx->buf[63] = len[7];
+	_SHA256_Update(ctx, len, 7, tmp32);
+
+	return 0;
+}
+
+/**
+ * PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, c, buf, dkLen):
+ * Compute PBKDF2(passwd, salt, c, dkLen) using HMAC-SHA256 as the PRF, and
+ * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
+ */
+void
+PBKDF2_SHA256(const uint8_t * passwd, size_t passwdlen, const uint8_t * salt,
+    size_t saltlen, uint64_t c, uint8_t * buf, size_t dkLen)
+{
+	HMAC_SHA256_CTX Phctx, PShctx, hctx;
+	uint32_t tmp32[72];
+	union {
+		uint8_t tmp8[96];
+		uint32_t state[8];
+	} u;
+	size_t i;
+	uint8_t ivec[4];
+	uint8_t U[32];
+	uint8_t T[32];
+	uint64_t j;
+	int k;
+	size_t clen;
+
+	/* Sanity-check. */
+	assert(dkLen <= 32 * (size_t)(UINT32_MAX));
+
+	if (c == 1 && (dkLen & 31) == 0 && (saltlen & 63) <= 51) {
+		uint32_t oldcount;
+		uint8_t * ivecp;
+
+		/* Compute HMAC state after processing P and S. */
+		_HMAC_SHA256_Init(&hctx, passwd, passwdlen,
+		    tmp32, &u.tmp8[0], &u.tmp8[64]);
+		_HMAC_SHA256_Update(&hctx, salt, saltlen, tmp32);
+
+		/* Prepare ictx padding. */
+		oldcount = hctx.ictx.count & (0x3f << 3);
+		_HMAC_SHA256_Update(&hctx, "\0\0\0", 4, tmp32);
+		if ((hctx.ictx.count & (0x3f << 3)) < oldcount ||
+		    SHA256_Pad_Almost(&hctx.ictx, u.tmp8, tmp32))
+			goto generic; /* Can't happen due to saltlen check */
+		ivecp = hctx.ictx.buf + (oldcount >> 3);
+
+		/* Prepare octx padding. */
+		hctx.octx.count += 32 << 3;
+		SHA256_Pad_Almost(&hctx.octx, u.tmp8, tmp32);
+
+		/* Iterate through the blocks. */
+		for (i = 0; i * 32 < dkLen; i++) {
+			/* Generate INT(i + 1). */
+			be32enc(ivecp, (uint32_t)(i + 1));
+
+			/* Compute U_1 = PRF(P, S || INT(i)). */
+			memcpy(u.state, hctx.ictx.state, sizeof(u.state));
+			SHA256_Transform(u.state, hctx.ictx.buf,
+			    &tmp32[0], &tmp32[64]);
+			be32enc_vect(hctx.octx.buf, u.state, 4);
+			memcpy(u.state, hctx.octx.state, sizeof(u.state));
+			SHA256_Transform(u.state, hctx.octx.buf,
+			    &tmp32[0], &tmp32[64]);
+			be32enc_vect(&buf[i * 32], u.state, 4);
+		}
+
+		goto cleanup;
+	}
+
+generic:
+	/* Compute HMAC state after processing P. */
+	_HMAC_SHA256_Init(&Phctx, passwd, passwdlen,
+	    tmp32, &u.tmp8[0], &u.tmp8[64]);
+
+	/* Compute HMAC state after processing P and S. */
+	memcpy(&PShctx, &Phctx, sizeof(HMAC_SHA256_CTX));
+	_HMAC_SHA256_Update(&PShctx, salt, saltlen, tmp32);
+
+	/* Iterate through the blocks. */
+	for (i = 0; i * 32 < dkLen; i++) {
+		/* Generate INT(i + 1). */
+		be32enc(ivec, (uint32_t)(i + 1));
+
+		/* Compute U_1 = PRF(P, S || INT(i)). */
+		memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
+		_HMAC_SHA256_Update(&hctx, ivec, 4, tmp32);
+		_HMAC_SHA256_Final(T, &hctx, tmp32, u.tmp8);
+
+		if (c > 1) {
+			/* T_i = U_1 ... */
+			memcpy(U, T, 32);
+
+			for (j = 2; j <= c; j++) {
+				/* Compute U_j. */
+				memcpy(&hctx, &Phctx, sizeof(HMAC_SHA256_CTX));
+				_HMAC_SHA256_Update(&hctx, U, 32, tmp32);
+				_HMAC_SHA256_Final(U, &hctx, tmp32, u.tmp8);
+
+				/* ... xor U_j ... */
+				for (k = 0; k < 32; k++)
+					T[k] ^= U[k];
+			}
+		}
+
+		/* Copy as many bytes as necessary into buf. */
+		clen = dkLen - i * 32;
+		if (clen > 32)
+			clen = 32;
+		memcpy(&buf[i * 32], T, clen);
+	}
+
+	/* Clean the stack. */
+	insecure_memzero(&Phctx, sizeof(HMAC_SHA256_CTX));
+	insecure_memzero(&PShctx, sizeof(HMAC_SHA256_CTX));
+	insecure_memzero(U, 32);
+	insecure_memzero(T, 32);
+
+cleanup:
+	insecure_memzero(&hctx, sizeof(HMAC_SHA256_CTX));
+	insecure_memzero(tmp32, 288);
+	insecure_memzero(&u, sizeof(u));
+}

src/crypto/randomx/panthera/sha256.h

@@ -0,0 +1,129 @@
+/*-
+ * Copyright 2005-2016 Colin Percival
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#ifndef _SHA256_H_
+#define _SHA256_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * Use #defines in order to avoid namespace collisions with anyone else's
+ * SHA256 code (e.g., the code in OpenSSL).
+ */
+#define SHA256_Init libcperciva_SHA256_Init
+#define SHA256_Update libcperciva_SHA256_Update
+#define SHA256_Final libcperciva_SHA256_Final
+#define SHA256_Buf libcperciva_SHA256_Buf
+#define SHA256_CTX libcperciva_SHA256_CTX
+#define HMAC_SHA256_Init libcperciva_HMAC_SHA256_Init
+#define HMAC_SHA256_Update libcperciva_HMAC_SHA256_Update
+#define HMAC_SHA256_Final libcperciva_HMAC_SHA256_Final
+#define HMAC_SHA256_Buf libcperciva_HMAC_SHA256_Buf
+#define HMAC_SHA256_CTX libcperciva_HMAC_SHA256_CTX
+
+/* Context structure for SHA256 operations. */
+typedef struct {
+	uint32_t state[8];
+	uint64_t count;
+	uint8_t buf[64];
+} SHA256_CTX;
+
+/**
+ * SHA256_Init(ctx):
+ * Initialize the SHA256 context ${ctx}.
+ */
+void SHA256_Init(SHA256_CTX *);
+
+/**
+ * SHA256_Update(ctx, in, len):
+ * Input ${len} bytes from ${in} into the SHA256 context ${ctx}.
+ */
+void SHA256_Update(SHA256_CTX *, const void *, size_t);
+
+/**
+ * SHA256_Final(digest, ctx):
+ * Output the SHA256 hash of the data input to the context ${ctx} into the
+ * buffer ${digest}.
+ */
+void SHA256_Final(uint8_t[32], SHA256_CTX *);
+
+/**
+ * SHA256_Buf(in, len, digest):
+ * Compute the SHA256 hash of ${len} bytes from ${in} and write it to ${digest}.
+ */
+void SHA256_Buf(const void *, size_t, uint8_t[32]);
+
+/* Context structure for HMAC-SHA256 operations. */
+typedef struct {
+	SHA256_CTX ictx;
+	SHA256_CTX octx;
+} HMAC_SHA256_CTX;
+
+/**
+ * HMAC_SHA256_Init(ctx, K, Klen):
+ * Initialize the HMAC-SHA256 context ${ctx} with ${Klen} bytes of key from
+ * ${K}.
+ */
+void HMAC_SHA256_Init(HMAC_SHA256_CTX *, const void *, size_t);
+
+/**
+ * HMAC_SHA256_Update(ctx, in, len):
+ * Input ${len} bytes from ${in} into the HMAC-SHA256 context ${ctx}.
+ */
+void HMAC_SHA256_Update(HMAC_SHA256_CTX *, const void *, size_t);
+
+/**
+ * HMAC_SHA256_Final(digest, ctx):
+ * Output the HMAC-SHA256 of the data input to the context ${ctx} into the
+ * buffer ${digest}.
+ */
+void HMAC_SHA256_Final(uint8_t[32], HMAC_SHA256_CTX *);
+
+/**
+ * HMAC_SHA256_Buf(K, Klen, in, len, digest):
+ * Compute the HMAC-SHA256 of ${len} bytes from ${in} using the key ${K} of
+ * length ${Klen}, and write the result to ${digest}.
+ */
+void HMAC_SHA256_Buf(const void *, size_t, const void *, size_t, uint8_t[32]);
+
+/**
+ * PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, c, buf, dkLen):
+ * Compute PBKDF2(passwd, salt, c, dkLen) using HMAC-SHA256 as the PRF, and
+ * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
+ */
+void PBKDF2_SHA256(const uint8_t *, size_t, const uint8_t *, size_t,
+    uint64_t, uint8_t *, size_t);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* !_SHA256_H_ */

src/crypto/randomx/panthera/sysendian.h

@@ -0,0 +1,94 @@
+/*-
+ * Copyright 2007-2014 Colin Percival
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#ifndef _SYSENDIAN_H_
+#define _SYSENDIAN_H_
+
+#include <stdint.h>
+
+/* Avoid namespace collisions with BSD <sys/endian.h>. */
+#define be32dec libcperciva_be32dec
+#define be32enc libcperciva_be32enc
+#define be64enc libcperciva_be64enc
+#define le32dec libcperciva_le32dec
+#define le32enc libcperciva_le32enc
+
+static inline uint32_t
+be32dec(const void * pp)
+{
+	const uint8_t * p = (uint8_t const *)pp;
+
+	return ((uint32_t)(p[3]) + ((uint32_t)(p[2]) << 8) +
+	    ((uint32_t)(p[1]) << 16) + ((uint32_t)(p[0]) << 24));
+}
+
+static inline void
+be32enc(void * pp, uint32_t x)
+{
+	uint8_t * p = (uint8_t *)pp;
+
+	p[3] = x & 0xff;
+	p[2] = (x >> 8) & 0xff;
+	p[1] = (x >> 16) & 0xff;
+	p[0] = (x >> 24) & 0xff;
+}
+
+static inline void
+be64enc(void * pp, uint64_t x)
+{
+	uint8_t * p = (uint8_t *)pp;
+
+	p[7] = x & 0xff;
+	p[6] = (x >> 8) & 0xff;
+	p[5] = (x >> 16) & 0xff;
+	p[4] = (x >> 24) & 0xff;
+	p[3] = (x >> 32) & 0xff;
+	p[2] = (x >> 40) & 0xff;
+	p[1] = (x >> 48) & 0xff;
+	p[0] = (x >> 56) & 0xff;
+}
+
+static inline uint32_t
+le32dec(const void * pp)
+{
+	const uint8_t * p = (uint8_t const *)pp;
+
+	return ((uint32_t)(p[0]) + ((uint32_t)(p[1]) << 8) +
+	    ((uint32_t)(p[2]) << 16) + ((uint32_t)(p[3]) << 24));
+}
+
+static inline void
+le32enc(void * pp, uint32_t x)
+{
+	uint8_t * p = (uint8_t *)pp;
+
+	p[0] = x & 0xff;
+	p[1] = (x >> 8) & 0xff;
+	p[2] = (x >> 16) & 0xff;
+	p[3] = (x >> 24) & 0xff;
+}
+
+#endif /* !_SYSENDIAN_H_ */