This repository consists of a hardware design for accelerating vector arithmetic (Level 1 BLAS) operations for the posit number format.
Both the hardware description of the accelerator as well as a C++ software library is provided. The library can be used in order to easily pass vectors to the FPGA and perform the selected operation on it.
Currently, the following vector operations are supported:
- Vector aggregation (vector dot product, vector sum)
- Vector (scalar) addition
- Vector (scalar) subtraction
- Vector (scalar) multiplication
git clone https://github.com/lvandam/posit_blas_hdl.git posit_blas_hdl
cd posit_blas_hdl
git submodule update --init
Available at https://cmake.org/download/, or perhaps through the OS package manager
For some accuracy measurements I am using the Boost C++ library (version 1.67, but perhaps later versions also work).
A quick tutorial on how to install it can be found at https://www.boost.org/doc/libs/1_67_0/more/getting_started/unix-variants.html. Perhaps it is also available through the package manager.
The following instructions are extracted from https://github.com/apache/arrow/tree/master/cpp
sudo apt-get install cmake libboost-dev libboost-filesystem-dev libboost-system-dev
git clone https://github.com/apache/arrow.git arrow
git checkout f8cd36a # To ensure compatibility
cd arrow/cpp
mkdir release
cd release
cmake .. -DCMAKE_BUILD_TYPE=Release
make unittest
sudo make install
git clone https://github.com/open-power/snap.git snap
git checkout 0313b46 # To ensure compatibility
Set the SNAP_ROOT environment variable, i.e. by putting this in a .bashrc file:
export SNAP_ROOT=<PATH_TO_SNAP_FOLDER>
Then
cd snap
make snap_config
make software
git clone https://github.com/lvandam/fletcher.git fletcher # Forked from repo by Johan Peltenburg (compatibility)
cd fletcher/runtime
mkdir build && cd build
cmake .. -DPLATFORM_SNAP=ON
make
sudo make install
cd sw
mkdir -p build && cd build
cmake ..
make
sudo make install
sudo capi-flash-script.sh bitfiles/XXXX.bin
Build and run one of the example programs located in examples/.
Make sure to set the parameters in src/defines.hpp according to the posit configuration used!
cd examples/gram
mkdir -p build && cd build
cmake ..
make
./gram
This content is developed as part of a research project at the Computer Engineering lab at Delft University of Technology. If any of this is of use to you, please include the following reference in your related work:
Laurens van Dam, Johan Peltenburg, Zaid Al-Ars, and H. Peter Hofstee. 2019. An Accelerator for Posit Arithmetic Targeting Posit Level 1 BLAS Routines and Pair-HMM. In Proceedings of the Conference for Next Generation Arithmetic 2019 (CoNGA’19). Association for Computing Machinery, New York, NY, USA, Article 5, 1–10. DOI:https://doi.org/10.1145/3316279.3316284