VFLOAT: The Northeastern Variable precision FLOATing point library

Contents

• What's New
• Introduction
• Floating Point Modules
• Download Area
• Archive Area
• Other Information

What's New

Major revision of VFLOAT available Sept 2010!

• All library components now include an asynchronous reset input.
• All library components now include a wait input to pause FP operations.
• The multiplier has variable latency depending on the bitwidth of the mantissa (user defined multiplier)
• fp_div and fp_sqrt are completely redesigned ! No preprocessing script required.
• Accumulator and Multiply-accumulate use latest version of VFLOAT components.
• All components should with Altera, although this has not been tested. Testing done with Xilinx tools.

Introduction

We have developed a library of fully parameterized hardware modules for variable precision floating-point arithmetic. The library includes modules for format control, arithmetic operators and conversion to and from any fixed-point format. All the modules have been implemented in VHDL and designed to run on different FPGA devices. They have been tested on reconfigurable computing engines from Annapolis Microsystems and Xilinx ML510.

An important feature of these modules is that denormalize, and normalize/round are decoupled from the arithmetic operations. This gives the designer full control over where and when to normalize floating point calculations, and results in savings of area in their hardware implementation. The currently implemented arithmetic operators are floating point add/sub, multiply, divide, square root, accumulate and multiply accumulate.

Please let us know if you are using the library by sending email to Prof. Leeser. We would also like to receive bug reports. We will do our best to keep the library up to date.

Back to Top

Floating Point Modules

• Fixed-point to floating-point conversion (unsigned)
• Fixed-point to floating-point conversion (signed)
• Floating-point to fixed-point conversion (unsigned)
• Floating-point to fixed-point conversion (signed)
• Floating-point adder/subtracter
• Floating-point multiplier
• Floating-point divider
• Floating-point square root
• Floating-point accumulator
• Floating-point multiply accumulator
• Floating-point round and normalizer

Click here for a graph of the dependences for each of the modules in the library.

Back to Top

Download Area

The revised library modules and examples can be downloaded here. Library Revision from Sept 28 2010.

Important! The new version of the library requires that you instantiate your own multipliers (mul, div, sqrt) and lookup tables (div and sqrt). Read the readme.txt file for details!

To use these modules, first unzip the zip file in directory fp_lib and then compile all the vhd files required for any operator or format controller. The dependencies of these modules can be found by clicking on the hyperlink on the left side of the dependency graph .

Below are some examples that implement basic single precision IEEE operations such as addition and multiplication.

Accumulator and Multiply-Accumulate are provided: The Multiply-Accumulate is based on an adder tree and the Accumulator is created using a single adder and feedback path. Clock latency for Multiply-Accumulate is fixed and it is variable for Accumulator.

Archive Area

If you have any questions or problems using the library, please contact Miriam Leeser

Back to Top

Other Information

Publications:

X. Wang and M. Leeser, VFloat: A Variable Precision Fixed- and Floating-Point Library for Reconfigurable Hardware , ACM Transactions on Reconfigurable Technology and Systems}. Vol. 3 No. 3, September 2010.

X. Wang and M. Leeser, A Truly Two Dimensional Systolic Array FPGA Implementation of QR Decomposition , ACM Transactions on Embedded Computer Systems. Vol. 9 No. 1, October 2009.

X. Wang, M. Leeser, and S. Braganza, Advanced Components in the Variable Precision Floating-Point Library, Field-programmable Custom Computing Machines (FCCM) 2006, April 2006, Napa CA.

X. Wang, M. Leeser, and H. Yu, A Parameterized Floating-Point Library Applied to Multispectral Image Clustering, 7th Annual MAPLD International Conference, September 2004, Washington DC.

X. Wang and M. Leeser,Variable Precision Floating Point Division and Square Root,8th Annual High Performance Embedded Computing Workshop,October 2004, Lexington, MA.

Pavle Belanovic and Miriam Leeser, A Library of Parameterized Floating Point Modules and Their Use.12th International Conference on Field Programmable Logic and Application. September 2002.

Pavle Belanovic Library of Parameterized Hardware Modules for Floating-Point Arithmetic with An Example Application M.S. Thesis, Dept of Electrical and Computer Engineering, Northeastern University, June 2002

People:

Faculty: Dr. Miriam Leeser

Graduate Students: Pavle Belanovic, Haiqian Yu, Sherman Braganza, Xiaojun Wang, Jainik Kathiara

Lab Visitors: Paolo Palana.

Reconfigurable Computing Lab

Terms:

ATTENTION: 

This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. 

 This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. 

You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

 

Back to Top

Maintained by Miriam Leeser

Reconfigurable Computing Lab, ECE Department,  Northeastern University Last revised: September 28, 2010.