Develop qnn zh

CMakeLists.txt

@@ -188,13 +188,16 @@ if(QNN)
         include(${ANDROID_NDK}/build/cmake/android.toolchain.cmake)
     endif()
     add_definitions(-DQNN_ENABLED)
-    # add_definitions(-DQNN_API)
+    # for debug
+    add_definitions(-DDEBUG)
 
     include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/Log)
     include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/PAL/include)
     include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/Utils)
     include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/WrapperUtils)
     include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/QNN)
+    include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN/Model)
+    include_directories(${PROJECT_SOURCE_DIR}/src/backends/QNN)
 
 
 

demo/qnn/qnn_test.cpp

@@ -4,31 +4,95 @@
 #include "Net.hpp"
 #include "Executor.hpp"
 #include "NetParameter.hpp"
+#include "QnnTypes.h"
 #include "express/Express.hpp"
 #include "tokenizers/BPE/Bpe.hpp"
 #include "backends/QNN/QNNBackend.hpp"
 #include "memory/SystemMemoryManager.hpp"
+#include "backends/QNN/op/QNNAdd.hpp"
 
 using namespace mllm;
 
 int main() {
-
     BackendConfig bnc;
 
     shared_ptr<MemoryManager> mm = nullptr;
     switch (bnc.memory) {
-        case BackendConfig::Memory_High:
-            mm = std::make_shared<SystemMemoryManager>();
-            break;
-        default:
-            mm = std::make_shared<SystemMemoryManager>();
-            break;
+    case BackendConfig::Memory_High:
+        mm = std::make_shared<SystemMemoryManager>();
+        break;
+    default:
+        mm = std::make_shared<SystemMemoryManager>();
+        break;
     }
 
     QNNBackend *qbn = new QNNBackend(mm);
-    qbn->init();
 
-    qbn->release();
-    delete  qbn;
+    // build graph
+    std::cout << "build graph" << std::endl;
+    // graph add node
+    uint32_t dimensions[] = {1, 2, 2, 2};
+    qbn->modelAddTensor("x", // Node Name
+                        (Qnn_Tensor_t){
+                            .version = QNN_TENSOR_VERSION_1,
+                            {.v1 = {
+                                 .id = 0,
+                                 .name = "x",
+                                 .type = QNN_TENSOR_TYPE_APP_WRITE,
+                                 .dataFormat = QNN_TENSOR_DATA_FORMAT_FLAT_BUFFER,
+                                 .dataType = QNN_DATATYPE_FLOAT_32,
+                                 .quantizeParams = {QNN_DEFINITION_UNDEFINED,
+                                                    QNN_QUANTIZATION_ENCODING_UNDEFINED,
+                                                    {.scaleOffsetEncoding = {.scale = 0.0000000000000000f, .offset = 0}}},
+                                 .rank = 4,
+                                 .dimensions = dimensions,
+                                 .memType = QNN_TENSORMEMTYPE_RAW,
+                                 {.clientBuf = {.data = nullptr,
+                                                .dataSize = 0}}}}});
+
+    float data[] = {1, 2, 3, 4, 5, 6, 7, 8};
+    qbn->modelAddTensor("y", // Node Name
+                        (Qnn_Tensor_t){
+                            .version = QNN_TENSOR_VERSION_1,
+                            {.v1 = {
+                                 .id = 0,
+                                 .name = "y",
+                                 .type = QNN_TENSOR_TYPE_STATIC,
+                                 .dataFormat = QNN_TENSOR_DATA_FORMAT_FLAT_BUFFER,
+                                 .dataType = QNN_DATATYPE_FLOAT_32,
+                                 .quantizeParams = {QNN_DEFINITION_UNDEFINED,
+                                                    QNN_QUANTIZATION_ENCODING_UNDEFINED,
+                                                    {.scaleOffsetEncoding = {.scale = 0.0000000000000000f, .offset = 0}}},
+                                 .rank = 4,
+                                 .dimensions = dimensions,
+                                 .memType = QNN_TENSORMEMTYPE_RAW,
+                                 {.clientBuf = {.data = data,
+                                                .dataSize = 32}}}}});
+
+    vector<Qnn_Tensor_t> outputs = {
+        (Qnn_Tensor_t){
+            .version = QNN_TENSOR_VERSION_1,
+            {.v1 = {
+                 .id = 0,
+                 .name = "add-output",
+                 .type = QNN_TENSOR_TYPE_APP_READ,
+                 .dataFormat = QNN_TENSOR_DATA_FORMAT_FLAT_BUFFER,
+                 .dataType = QNN_DATATYPE_FLOAT_32,
+                 .quantizeParams = {QNN_DEFINITION_UNDEFINED,
+                                    QNN_QUANTIZATION_ENCODING_UNDEFINED,
+                                    {.scaleOffsetEncoding = {.scale = 0.0000000000000000f, .offset = 0}}},
+                 .rank = 4,
+                 .dimensions = dimensions,
+                 .memType = QNN_TENSORMEMTYPE_RAW,
+                 {.clientBuf = {.data = nullptr,
+                                .dataSize = 0}}}}}};
+    qbn->graphAddNode("qnn-add", "ElementWiseAdd", {"x", "y"}, outputs, "qti.aisw");
+    // graph compile
+    std::cout << "graph compile" << std::endl;
+    qbn->graphFinilize();
+    // graph run
+    std::cout << "graph run" << std::endl;
+    qbn->graphExecute();
 
+    delete qbn;
 }

include/Types.hpp

@@ -7,7 +7,8 @@
 typedef enum {
     MLLM_CPU,
     MLLM_OPENCL,
-    MLLM_NNAPI
+    MLLM_NNAPI,
+    MLLM_QNN
 } BackendType;
 
 enum ErrorCode {

src/backends/QNN/CMakeLists.txt

@@ -1,13 +1,5 @@
-# include_directories("../../../include/QNN") 
-# include_directories(Log)
-# include_directories(PAL/src/linux)
-# include_directories(PAL/src/common)
-# include_directories(Utils)
-# include_directories(WrapperUtils)
-# include_directories(PAL/include)
-
 get_property(dirs DIRECTORY ${CMAKE_SOURCE_DIR} PROPERTY INCLUDE_DIRECTORIES)
-message(">>> include_dirs=${dirs}") #打印一下目录情况
+message(">>> include_dirs=${dirs}") #print include directories
 
 
 
@@ -29,6 +21,12 @@ file(GLOB MLLM_QNN_SRC
 
     ${CMAKE_CURRENT_LIST_DIR}/WrapperUtils/*.cpp
     ${CMAKE_CURRENT_LIST_DIR}/WrapperUtils/*.hpp
+
+    ${CMAKE_CURRENT_LIST_DIR}/Model/*.cpp
+    ${CMAKE_CURRENT_LIST_DIR}/Model/*.hpp
+
+    ${CMAKE_CURRENT_LIST_DIR}/op/*.cpp
+    ${CMAKE_CURRENT_LIST_DIR}/op/*.hpp
 
 )