初始化

FaceDetect/.gitignore

@@ -0,0 +1,58 @@
+# Built application files
+*.apk
+*.ap_
+
+# Files for the ART/Dalvik VM
+*.dex
+
+# Java class files
+*.class
+
+# Generated files
+bin/
+gen/
+out/
+
+# Gradle files
+.gradle/
+build/
+
+# Local configuration file (sdk path, etc)
+local.properties
+
+# Proguard folder generated by Eclipse
+proguard/
+
+# Log Files
+*.log
+
+# Android Studio Navigation editor temp files
+.navigation/
+
+# Android Studio captures folder
+captures/
+
+# IntelliJ
+*.iml
+.idea/workspace.xml
+.idea/tasks.xml
+.idea/gradle.xml
+.idea/dictionaries
+.idea/libraries
+
+# Keystore files
+# Uncomment the following line if you do not want to check your keystore files in.
+#*.jks
+
+# External native build folder generated in Android Studio 2.2 and later
+.externalNativeBuild
+
+# Google Services (e.g. APIs or Firebase)
+google-services.json
+
+# Freeline
+freeline.py
+freeline/
+freeline_project_description.json
+app/official/
+.idea/

FaceDetect/CMakeLists.txt

@@ -0,0 +1,33 @@
+cmake_minimum_required(VERSION 3.4.1)
+
+set(lib_DIR ${CMAKE_SOURCE_DIR}/libs)
+include_directories(${CMAKE_SOURCE_DIR}/includes)
+
+set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fopenmp")
+set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fopenmp")
+set (CMAKE_LINKER_FLAGS "${CMAKE_LINKER_FLAGS} -fopenmp")
+
+set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -O2 -fvisibility=hidden -fomit-frame-pointer -fstrict-aliasing -ffunction-sections -fdata-sections -ffast-math")
+set (CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O2 -fvisibility=hidden -fvisibility-inlines-hidden -fomit-frame-pointer -fstrict-aliasing -ffunction-sections -fdata-sections -ffast-math")
+set (CMAKE_LINKER_FLAGS "${CMAKE_LINKER_FLAGS} -Wl,--gc-sections")
+
+add_library(libncnn STATIC IMPORTED)
+set_target_properties(libncnn PROPERTIES IMPORTED_LOCATION ${lib_DIR}/${ANDROID_ABI}/libncnn.a)
+
+add_library(libopencv_java3 SHARED IMPORTED)
+set_target_properties(libopencv_java3 PROPERTIES IMPORTED_LOCATION ${lib_DIR}/${ANDROID_ABI}/libopencv_java3.so)
+
+file(GLOB_RECURSE CPP_SRCS src/main/cpp/*.h
+        src/main/cpp/*.cpp
+        src/main/cpp/mtcnn.cpp
+        src/main/cpp/mtcnn_gpu.cpp
+        src/main/cpp/mtcnn_gpu_jni.cpp
+        src/main/cpp/mtcnn_jni.cpp)
+
+add_library(faceDetect SHARED ${CPP_SRCS} )
+
+find_library(log-lib log)
+find_library(jnigraphics-lib jnigraphics)
+find_library(vulkan-lib vulkan)
+
+target_link_libraries(faceDetect libncnn libopencv_java3 z ${log-lib} ${jnigraphics-lib} ${vulkan-lib})

FaceDetect/build.gradle

@@ -0,0 +1,52 @@
+apply plugin: 'com.android.library'
+
+android {
+    compileSdkVersion 28
+
+    defaultConfig {
+        minSdkVersion 21
+        targetSdkVersion 28
+        versionCode 1
+        versionName "1.0"
+        testInstrumentationRunner "android.support.test.runner.AndroidJUnitRunner"
+        externalNativeBuild {
+            cmake {
+                arguments "-DANDROID_STL=c++_shared"
+                arguments "-DANDROID_TOOLCHAIN=clang", "-DANDROID_ARM_NEON=TRUE", "-DANDROID_PLATFORM=android-24"
+                abiFilters 'armeabi-v7a'
+            }
+        }
+        ndk {
+            abiFilters 'armeabi-v7a'
+            stl "gnustl_static"
+        }
+    }
+    compileOptions {
+        targetCompatibility JavaVersion.VERSION_1_8
+        sourceCompatibility JavaVersion.VERSION_1_8
+    }
+
+    buildTypes {
+        release {
+            minifyEnabled false
+            proguardFiles getDefaultProguardFile('proguard-android-optimize.txt'), 'proguard-rules.pro'
+        }
+    }
+
+    externalNativeBuild {
+        cmake {
+            path "CMakeLists.txt"
+        }
+    }
+
+    sourceSets {
+        main {
+            jniLibs.srcDirs = ['libs']
+        }
+    }
+
+}
+
+dependencies {
+    implementation fileTree(dir: 'libs', include: ['*.jar'])
+}

app/src/main/cpp/include/allocator.h → FaceDetect/includes/ncnn/allocator.h

@@ -56,20 +56,39 @@ static inline size_t alignSize(size_t sz, int n)
 
 static inline void* fastMalloc(size_t size)
 {
+#if _MSC_VER
+    return _aligned_malloc(size, MALLOC_ALIGN);
+#elif _POSIX_C_SOURCE >= 200112L || (__ANDROID__ && __ANDROID_API__ >= 17)
+    void* ptr = 0;
+    if (posix_memalign(&ptr, MALLOC_ALIGN, size))
+        ptr = 0;
+    return ptr;
+#elif __ANDROID__ && __ANDROID_API__ < 17
+    return memalign(MALLOC_ALIGN, size);
+#else
     unsigned char* udata = (unsigned char*)malloc(size + sizeof(void*) + MALLOC_ALIGN);
     if (!udata)
         return 0;
     unsigned char** adata = alignPtr((unsigned char**)udata + 1, MALLOC_ALIGN);
     adata[-1] = udata;
     return adata;
+#endif
 }
 
 static inline void fastFree(void* ptr)
 {
     if (ptr)
     {
+#if _MSC_VER
+        _aligned_free(ptr);
+#elif _POSIX_C_SOURCE >= 200112L || (__ANDROID__ && __ANDROID_API__ >= 17)
+        free(ptr);
+#elif __ANDROID__ && __ANDROID_API__ < 17
+        free(ptr);
+#else
         unsigned char* udata = ((unsigned char**)ptr)[-1];
         free(udata);
+#endif
     }
 }
 
@@ -100,40 +119,6 @@ static inline void fastFree(void* ptr)
 static inline int NCNN_XADD(int* addr, int delta) { int tmp = *addr; *addr += delta; return tmp; }
 #endif
 
-#ifdef _WIN32
-class Mutex
-{
-public:
-    Mutex() { InitializeSRWLock(&srwlock); }
-    ~Mutex() {}
-    void lock() { AcquireSRWLockExclusive(&srwlock); }
-    void unlock() { ReleaseSRWLockExclusive(&srwlock); }
-private:
-    // NOTE SRWLock is available from windows vista
-    SRWLOCK srwlock;
-};
-#else // _WIN32
-class Mutex
-{
-public:
-    Mutex() { pthread_mutex_init(&mutex, 0); }
-    ~Mutex() { pthread_mutex_destroy(&mutex); }
-    void lock() { pthread_mutex_lock(&mutex); }
-    void unlock() { pthread_mutex_unlock(&mutex); }
-private:
-    pthread_mutex_t mutex;
-};
-#endif // _WIN32
-
-class MutexLockGuard
-{
-public:
-    MutexLockGuard(Mutex& _mutex) : mutex(_mutex) { mutex.lock(); }
-    ~MutexLockGuard() { mutex.unlock(); }
-private:
-    Mutex& mutex;
-};
-
 class Allocator
 {
 public:

app/src/main/cpp/include/command.h → FaceDetect/includes/ncnn/command.h

@@ -29,8 +29,8 @@ namespace ncnn {
 class Command
 {
 public:
-    Command(const VulkanDevice* vkdev, uint32_t queue_index);
-    ~Command();
+    Command(const VulkanDevice* vkdev, uint32_t queue_family_index);
+    virtual ~Command();
 
 protected:
     int create_command_pool();
@@ -39,14 +39,11 @@ class Command
     // record issue
     int begin_command_buffer();
     int end_command_buffer();
-    int queue_submit();
-    int wait_fence();
+    int queue_submit_and_wait_fence();
 
 protected:
     const VulkanDevice* vkdev;
-    uint32_t queue_index;
-
-    VkQueue queue;
+    uint32_t queue_family_index;
 
     VkCommandPool command_pool;
     VkCommandBuffer command_buffer;
@@ -72,12 +69,18 @@ class VkCompute : public Command
 
     void record_pipeline(const Pipeline* pipeline, const std::vector<VkMat>& bindings, const std::vector<vk_constant_type>& constants, const VkMat& m);
 
-    int submit();
+    void record_write_timestamp(uint32_t query);
 
-    int wait();
+    int submit_and_wait();
 
     int reset();
 
+#if NCNN_BENCHMARK
+    int create_query_pool(uint32_t query_count);
+
+    int get_query_pool_results(uint32_t first_query, uint32_t query_count, std::vector<uint64_t>& results);
+#endif // NCNN_BENCHMARK
+
 protected:
     // record pipeline things
     void record_bind_pipeline(VkPipeline pipeline);
@@ -95,6 +98,10 @@ class VkCompute : public Command
     void record_prepare_transfer_barrier(const VkMat& m);
     void record_prepare_compute_barrier(const VkMat& m);
 
+#if NCNN_BENCHMARK
+    void reset_query_pool();
+#endif // NCNN_BENCHMARK
+
 protected:
     // recording issue
     void copy_buffer(VkBuffer src, size_t src_offset, VkBuffer dst, size_t dst_offset, size_t size);
@@ -108,6 +115,9 @@ class VkCompute : public Command
     void compute_transfer_barrier(VkBuffer buffer, size_t offset, size_t size);
     void compute_compute_barrier(VkBuffer buffer, size_t offset, size_t size);
     void transfer_transfer_barrier(VkBuffer buffer, size_t offset, size_t size);
+#if NCNN_BENCHMARK
+    void write_timestamp(uint32_t query);
+#endif // NCNN_BENCHMARK
 
 protected:
     // delayed record
@@ -126,6 +136,7 @@ class VkCompute : public Command
         // 7=compute-transfer barrier
         // 8=compute-compute barrier
         // 9=transfer-transfer barrier
+        // 10=write timestamp
         int type;
 
         union
@@ -140,12 +151,20 @@ class VkCompute : public Command
         struct { VkBuffer buffer; size_t offset; size_t size; } compute_transfer_barrier;
         struct { VkBuffer buffer; size_t offset; size_t size; } compute_compute_barrier;
         struct { VkBuffer buffer; size_t offset; size_t size; } transfer_transfer_barrier;
+#if NCNN_BENCHMARK
+        struct { uint32_t query; } write_timestamp;
+#endif // NCNN_BENCHMARK
         };
 
         std::vector<VkBufferCopy> regions;
         std::vector<vk_constant_type> constants;
     };
     std::vector<record_type> delayed_records;
+
+#if NCNN_BENCHMARK
+    uint32_t query_count;
+    VkQueryPool query_pool;
+#endif // NCNN_BENCHMARK
 };
 
 class VkTransfer : public Command
@@ -154,13 +173,9 @@ class VkTransfer : public Command
     VkTransfer(const VulkanDevice* vkdev);
     ~VkTransfer();
 
-    void record_upload(const Mat& src, VkMat& dst);
-
-    void record_download(const VkMat& src, Mat& dst);
+    void record_upload(const Mat& src, VkMat& dst, const Option& opt);
 
-    int submit();
-
-    int wait();
+    int submit_and_wait();
 
 public:
     VkAllocator* weight_vkallocator;
@@ -178,9 +193,6 @@ class VkTransfer : public Command
     // delayed record
     struct record_type
     {
-        // 0=upload
-        // 1=download
-        int type;
         size_t size;
         Mat mat;
         VkMat vkmat;