build_and_setup.md

Build and Setup

Dependencies

A number of packages are required to build a minimal AvA development and test environment. To use full components and benchmarks, one should follow the instructions per specification or benchmark to setup the machine.

The following script will check whether Ubuntu 18.04 is running and install the necessary dependencies.

sudo ./tools/install_dependencies.sh

The following instructions are tested on Ubuntu 18.04 (Linux 4.15) with GCC 7.5.0, Python 3.6.9, Boost 1.71.x, cmake 3.19.1 and Protobuf 3.0-3.9.

Configuration

Clone AvA's base repository and initiailize it:

git clone https://github.com/utcs-scea/ava.git
cd ava
git submodule update --init --recursive
./generate.py -s demo
cd ..
mkdir build
cd build
cmake ../ava

The above commands initialize a cmake build directory with all flags turned OFF. AvA supports a handful of accelerators and APIs, and the merge of the old codes into the new build system (v2.0) is still ongoing.

generate.py downloads the customized AvA-LLVM module and generates codes from the provided specifications. To see the supported specifications, please run:

./generate.py -h

Please visit Generate Code for more information.

API framework	Version	Status
CUDA driver	10.1	TESTED
CUDA runtime	10.1	TESTED
Demo	0.0.1	TESTED
GTI	4.5.0.3	TESTED
HIP		NO
NCSDK v2	2.10.01	TESTED
ONNXruntime CUDA	10.1	TESTED*
OpenCL	1.2	TESTED
QuickAssist	1.7	TESTED
TensorFlow CUDA	10.1	TESTED
TensorFlow C	2.3.1	TESTED
Test	0.1	TESTED

* Upstream changes (ava-serverless) have not been merged.

AvA manager	Status
Demo	TESTED
Galvanic	NO
Katana	NO
Legacy (r1.0)	TESTED
Nvidia GPU	NO

This tutorial shows how to configure and build the minimal AvA demo API remoting system. To build and use other supported virtualized APIs, please reference respective documentations. To virtualize a new API with AvA, please reference the virtualize_new_api tutorial.

In the build directory, run:

ccmake .

Then turn on AVA_GEN_DEMO_SPEC and AVA_MANAGER_DEMO and press c to reconfigure the build.

One can run this without using ccmake to configure the build again:

cmake ../ava -DAVA_GEN_DEMO_SPEC=ON -DAVA_MANAGER_DEMO=ON

Build and Run

Build the demo system simply by:

make -j`nproc`
make install

This generates the guestlib and API server codes in ava/cava/demo_nw and installs their binary files to build/install/demo. The demo manager is compiled and installed into build/install/bin.

ls install/demo/bin
ls install/demo/lib
ls install/bin

AvA's base repository includes a tiny standalone demo program. If compilation fails, make sure the Makefile has the correct path to the build folder.

pushd .
cd ../ava/cava/samples/demo/test_program/
make
popd

Start the demo manager by

./install/bin/demo_manager --worker_path install/demo/bin/worker

Add AvA configuration file:

sudo mkdir -p /etc/ava
sudo tee /etc/ava/guest.conf <<EOF
channel = "TCP";
manager_address = "0.0.0.0:3333";
gpu_memory = [1024L];
EOF

The guestlib assumes by default the configuration file is at /etc/ava/guest.conf, but any path can be used as long as the environment variable AVA_CONFIG_FILE_PATH is exported and contains the absolute path to the configuration file.

Run the test program by

LD_LIBRARY_PATH=install/demo/lib ../ava/cava/samples/demo/test_program/test

The demo spec implements and annotates an int ava_test_api(int x) API which returns x+1 to the caller. In most cases, the API is defined in a separate shared library and the library is linked to the generated API server. But for the sake of simplicity, ava_test_api is just defined in the spec.

The demo program simply calls an int ava_test_api(int) API which is forwarded to the API server. The API server prints a message RECEIVED AVA_TEST_API(9999) and sends a response to the demo program. The demo program then prints a message RECEIVED AVA_TEST_API = 10000.