[CLML][CODEGEN] CLML native codegen utility

This util generates native CLML code given a DNN model.
It does import via tvmc, extracts clml_modules, get the json source and
finally generates clml_models.cc that holds source for various sub graphs.
cpp_clml tool has additional infrastructure to compile it as a standalong
binary that runs these models.

This PR adds symbol name to the generates json grpah.
Also, extends const_loader interface to get constant params.

apps/cpp_clml/CMakeLists.txt

@@ -0,0 +1,57 @@
+cmake_minimum_required(VERSION 3.13)
+
+project(clml_run VERSION 2.0)
+
+if(NOT DEFINED CMAKE_TOOLCHAIN_FILE)
+  message( FATAL_ERROR "CMAKE_TOOLCHAIN_FILE Not set, forcing exit. Suggested value: {ANDROID_NDK_PATH}/build/cmake/android.toolchain.cmake." )
+endif(NOT DEFINED CMAKE_TOOLCHAIN_FILE)
+
+if(NOT DEFINED ANDROID_ABI)
+  message( FATAL_ERROR "ANDROID_ABI Not set, forcing exit. Suggested value(s): arm64-v8a (64), armeabi-v7a (32)" )
+endif(NOT DEFINED ANDROID_ABI)
+
+if(NOT DEFINED CLML_SDK)
+  message( FATAL_ERROR "CLML_SDK Not set, forcing exit." )
+endif(NOT DEFINED CLML_SDK)
+
+# CMake/Android variables
+set( ANDROID_STL  c++_static CACHE STRING "Target Android STL") # default
+
+# Source variables
+set( OPENCL_INCLUDE_DIRS  ${CLML_SDK} CACHE PATH "filepath to OpenCL headers")
+set( ANDROID_SOURCE_TREE /path/to/android/au/ CACHE FILEPATH "optional filepath to the Android AU Tree, for building examples using ION Buffers") # tree required to build ION/DMA Buffer samples
+
+#c++ 11 is required
+set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_STANDARD_REQUIRED True)
+# set(CMAKE_CXX_FLAGS "-Wall -Werror")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
+
+#we do not want to pass -fno-exceptions
+if(${CMAKE_CXX_FLAGS} MATCHES "-fno-exceptions")
+  string(REGEX REPLACE "-fno-exceptions" "" CMAKE_CXX_FLAGS ${CMAKE_CXX_FLAGS})
+endif()
+
+#we do not want to pass -fno-rtti
+if(${CMAKE_CXX_FLAGS} MATCHES "-fno-rtti")
+  string(REGEX REPLACE "-fno-rtti" "" CMAKE_CXX_FLAGS ${CMAKE_CXX_FLAGS})
+endif()
+
+set(COMMON_SOURCE_FILES
+        clml_models.cc
+        clml_runner.cc
+        clml_runner.h
+        main.cc
+        ../../3rdparty/cnpy/cnpy.cpp
+        )
+
+include_directories(
+        src
+        ${OPENCL_INCLUDE_DIRS}
+        "../../3rdparty/dmlc-core/include"
+        "../../3rdparty/cnpy/"
+        )
+
+add_executable(clml_run ${COMMON_SOURCE_FILES})
+target_link_options(clml_run PRIVATE -Wl,--unresolved-symbols=ignore-in-shared-libs)
+target_link_libraries(clml_run ${CLML_SDK}/lib64/libOpenCL.so z)

apps/cpp_clml/README.md

@@ -0,0 +1,145 @@
+<!--- Licensed to the Apache Software Foundation (ASF) under one -->
+<!--- or more contributor license agreements.  See the NOTICE file -->
+<!--- distributed with this work for additional information -->
+<!--- regarding copyright ownership.  The ASF licenses this file -->
+<!--- to you under the Apache License, Version 2.0 (the -->
+<!--- "License"); you may not use this file except in compliance -->
+<!--- with the License.  You may obtain a copy of the License at -->
+
+<!---   http://www.apache.org/licenses/LICENSE-2.0 -->
+
+<!--- Unless required by applicable law or agreed to in writing, -->
+<!--- software distributed under the License is distributed on an -->
+<!--- "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY -->
+<!--- KIND, either express or implied.  See the License for the -->
+<!--- specific language governing permissions and limitations -->
+<!--- under the License. -->
+
+# OpenCLML Debug Tool
+
+Tool to generate OpenCLML source file given a model from any framework and compile it as a native application that runs on Android target.
+This tool helps to debug or triage OpenCLML offloaded sub graphs as a standalone application.
+
+### Codegen
+
+Models can be downloaded from well known frameworks like Tensorflow, PyTorch, TFLite, Onnx ..etc.
+Assuming  ```resnet50.h5``` is a Keras ResNet50 model file, use the below command to generate a OpenCLML source for the model.
+
+```bash
+python3 scripts/clml_codegen.py resnet50.h5
+```
+
+Above command generates ```clml_models.cc``` and ```clml_params.npz```.
+```clml_models.cc``` contains cpp representation of all OpenCLML subgraphs offloaded by TVM compilation. This file will be used to build tool ```clml_run```.
+```clml_params.npz``` is a numpy dump of all params involved in all sub graphs of TVM module. This file to be copied to target.
+
+### Build Tool
+
+Copy the generated models source ```clml_models.cc``` under ```cpp_clml```.
+
+Below commands will compile the tool ```clml_run``` from generated source and other static dependents.
+
+```bash
+cmake -S . -B build_64 -D ANDROID_ABI=arm64-v8a -D CLML_SDK=<CLML SDK PATH> -D CMAKE_TOOLCHAIN_FILE=<ANDROID NDK PATH>/build/cmake/android.toolchain.cmake -D ANDROID_PLATFORM=latest
+cmake --build build_64
+```
+
+### Run the tool
+
+Copy ```clml_params.npz``` and ```clml_run``` to the target Android device
+
+```bash
+Android:/data/local/tmp $ ./clml_run --dump-meta
+Input         =
+Output        =
+Params        =
+DumpMeta      = 1
+.....
+Subgraph Name: tvmgen_default_clml_main_1
+    Input Count  : 1
+    Output Count : 1
+    Input MetaInfo
+        Input: tvmgen_default_clml_main_1_input_0
+            Dtype : float32
+            Shape : [1, 1, 1, 2048]
+    Output MetaInfo
+        Output: tvmgen_default_clml_main_1_layer_out_5
+            Dtype : float32
+            Shape : [1, 1000]
+
+Subgraph Name: tvmgen_default_clml_main_0
+    Input Count  : 1
+    Output Count : 1
+    Input MetaInfo
+        Input: tvmgen_default_clml_main_0_input_0
+            Dtype : float32
+            Shape : [1, 3, 230, 230]
+    Output MetaInfo
+        Output: tvmgen_default_clml_main_0_layer_out_406
+            Dtype : float32
+            Shape : [1, 2048, 1, 1]
+.....
+```
+
+The meta information above indicates that the ResNet50 model is partitioned such a way that there exists two OpenCLML subgraphs.
+
+Below command runs the models by setting the parameters from ```clml_params.npz```.
+
+```bash
+Android:/data/local/tmp $ ./clml_run --params=./clml_params.npz
+Input         =
+Output        =
+Params        = ./clml_params.npz
+DumpMeta      = 1
+......
+CLMLRunner Loading Params:./clml_params.npz
+CLMLRunner Loading Params:./clml_params.npz
+CLMLRunner::Run :tvmgen_default_clml_main_1
+CLMLRunner::Run :tvmgen_default_clml_main_0
+......
+```
+
+Below command can set the model inputs from ```input.npz```  and can output sub graph outputs to ```output.npz```.
+```input.npz``` should have numpy arrays for ```tvmgen_default_clml_main_1_input_0``` from sub graph ```tvmgen_default_clml_main_1``` and ```tvmgen_default_clml_main_0_input_0``` from sub graph ```tvmgen_default_clml_main_0```.
+
+```bash
+Android:/data/local/tmp $ ./clml_run --params=./clml_params.npz --input=./input.npz --output=./output.npz                                                                       <
+Input         = ./input.npz
+Output        = ./output.npz
+Params        = ./clml_params.npz
+DumpMeta      = 0
+Call Build Modules
+CLMLRunner Constructor: Input:./input.npz Output:./output.npz Params:./clml_params.npz
+CLML Target version:3
+CLMLRunner Loading Params:./clml_params.npz
+CLMLRunner Loading Inputs:./input.npz
+Set Input For:tvmgen_default_clml_main_1_input_0
+
+CLMLRunner Constructor: Input:./input.npz Output:./output.npz Params:./clml_params.npz
+CLML Target version:3
+CLMLRunner Loading Params:./clml_params.npz
+CLMLRunner Loading Inputs:./input.npz
+Set Input For:tvmgen_default_clml_main_0_input_0
+
+Loop Through the Modules
+CLMLRunner::Run :tvmgen_default_clml_main_1
+Saving Output:tvmgen_default_clml_main_1_layer_out_5
+CLMLRunner::Run :tvmgen_default_clml_main_0
+Saving Output:tvmgen_default_clml_main_0_layer_out_406
+......
+```
+
+The generated output file ```output.npz``` contains all the output from all sub modules.
+In this case it contains ```tvmgen_default_clml_main_1_layer_out_5``` for sub graph ```tvmgen_default_clml_main_1``` and ```tvmgen_default_clml_main_0_layer_out_406``` for sub graph ```tvmgen_default_clml_main_0``` as shown below.
+
+
+```bash
+Android:/data/local/tmp $ unzip -l output.npz
+Archive:  output.npz
+  Length      Date    Time    Name
+---------  ---------- -----   ----
+     4080  1980-00-00 00:00   tvmgen_default_clml_main_1_layer_out_5.npy
+     8272  1980-00-00 00:00   tvmgen_default_clml_main_0_layer_out_406.npy
+---------                     -------
+    12352                     2 files
+```