Division is notoriously difficult to implement efficiently in hardware. Ercegovac and Lang are globally recognized experts in this specific subfield. They explain:
is a Distinguished Professor Emeritus at UCLA. He earned his PhD from the University of Illinois and has specialized in digital arithmetic, digital design, and computer architecture for decades. A Fellow of the IEEE since 2003 for his contributions, he has been deeply involved in organizing the IEEE Symposia on Computer Arithmetic since 1978. His ongoing research includes online arithmetic algorithms, approximate arithmetic, and arithmetic for machine learning architectures.
Unlike introductory textbooks that only scratch the surface of binary addition, this text delves deep into the mathematical optimization of hardware. It bridges the gap between high-level algorithmic theory and low-level gate implementation, focusing heavily on trade-offs involving: Hardware area (gate count/cost) Power consumption Interconnect complexity 2. Core Concepts Covered in the Book
Modern deep learning hardware relies heavily on specialized, quantized arithmetic (like FP8, INT4, or Brain Floating Point). Understanding Ercegovac and Lang's methodology allows designers to build custom, error-tolerant arithmetic blocks. digital arithmetic by ercegovac and lang pdf
The book is structured to guide readers from basic number systems to advanced arithmetic units:
Because both authors have had illustrious careers in academia (with Miloš D. Ercegovac at UCLA and Tomás Lang at UC Irvine), various lecture notes, chapter supplements, and errata sheets are hosted legally on university websites. These resources often provide a comprehensive summary of the formulas and architectures found in the book.
[Inputs] ──> [Radix-4 Booth Encoding] ──> [Partial Product Generation] │ [Final Sum] <── [High-Speed Vector Adder] <── [Wallace/Dadda Tree (CSA)] The Radix-4 SRT Division Design Space He earned his PhD from the University of
Finally, a : Some websites offer "free PDF" downloads but should be approached with extreme caution. These sources often host unauthorized copies and may pose significant security risks, including viruses, malware, and copyright infringement. Supporting the authors and publisher by acquiring the book legally ensures you receive a complete, correct, and virus-free file.
: Multiplicative division methods like Newton-Raphson and Goldschmidt iteration, which double the precision bits at every step. Function Evaluation
in a radix-2 system), hardware implementations can eliminate carry-propagation chains, unlocking massive parallelism. 2. Key Topics Covered in the Framework Unlike introductory textbooks that only scratch the surface
Introduction to redundant number systems where digits can take negative values (e.g.,
In the realm of computer architecture, digital arithmetic forms the bedrock upon which all computational processing is built. Whether it is a simple microcontroller executing basic additions or a state-of-the-art graphics processing unit (GPU) performing billions of floating-point operations per second, the efficiency of the underlying arithmetic circuits determines the performance of the entire system.
: Contains over 250 exercises to reinforce technical concepts.