In the world of computer engineering and VLSI design, certain texts transcend mere "reference books" to become foundational pillars. Among these, Digital Arithmetic by and Tomás Lang stands as a towering achievement. For over two decades, it has been the definitive resource for understanding how numbers are represented, manipulated, and calculated in digital systems.
You can try searching for the PDF online, but I couldn't find a freely available version. However, you can try checking the following sources:
The book "Digital Arithmetic" by Ercegovac and Lang provides a comprehensive coverage of digital arithmetic, which is a fundamental aspect of computer design and digital systems. The book focuses on the principles and techniques of digital arithmetic, including the representation of numbers, arithmetic operations, and algorithms for performing these operations.
Why "Digital Arithmetic" is Relevant to Modern AI and Hardware
Digital Arithmetic by Miloš Ercegovac and Tomás Lang remains a cornerstone text for anyone seeking to understand or build high-performance computing hardware. By providing a rigorous mathematical framework alongside practical implementation constraints, it equips engineers with the tools necessary to innovate in the eras of edge computing, specialized silicon, and artificial intelligence. digital arithmetic by ercegovac and lang pdf
: Detailed study of SRT division (named after Sweeney, Robertson, and Tocher).
Here is a brief summary of each chapter in the book:
: Combining small ROM tables with interpolation circuits for fast function approximation. 3. Finding and Accessing the PDF Legally
It is reasonable to ask: With ChatGPT and Copilot generating Verilog code for adders, why study digital arithmetic from a 2004 textbook? In the world of computer engineering and VLSI
: Professor Lang’s academic archives contain detailed papers on SRT division and CORDIC architectures.
If you are asking for the PDF because you cannot afford the $100 price tag, then use the legal alternatives: library loans, used copies, or institutional access. The knowledge inside is worth ten times that price.
The concepts and techniques presented in "Digital Arithmetic" have numerous applications in various fields, including:
Section C — Design and analysis (30 marks) 11. (8) Carry-lookahead adder design - For a 16-bit adder using 4-bit carry-lookahead blocks, draw the carry generate/propagate equations and compute worst-case gate-level carry delay assuming: - AND/OR gate delay = 1 unit - XOR delay = 2 units - Give numeric delay to produce final sum bits. 12. (8) Divider hardware cost vs. latency trade-offs - Compare non-restoring, restoring, and SRT division algorithms in terms of hardware complexity (qualitative), per-iteration operations, and latency for an n-bit divider. Provide a small table summarizing complexities for n-bit result. 13. (8) Error analysis for truncated multiplier - For an n×n binary multiplier where only the top k most significant partial-product rows are kept (truncation), derive an upper bound for absolute truncation error as a function of n and k. Provide a numeric example for n=16, k=12. 14. (6) Practical implementation note - Recommend three practical microarchitectural techniques (brief bullet points) from Ercegovac & Lang to improve throughput of a multiply unit in an ASIC implementation, with one sentence justification each. You can try searching for the PDF online,
The you need (e.g., maximum clock speed vs. lowest power consumption).
This discipline is crucial because each new generation of processors or digital systems creates new arithmetic design challenges. Designers, researchers, and graduate students turn to this field to find solid solutions to these problems. It plays a central role in the design of general-purpose CPUs, GPUs, and the embedded systems that power everything from smartphones to self-driving cars.
Machine learning accelerators rely heavily on low-precision, high-throughput arithmetic (like FP8, INT8, and bfloat16). The mathematical principles of quantization and fused multiply-add (FMA) units detailed in this book are crucial for designing tensor processing units.