Customer Reviews

Synthesis of Arithmetic Circuits: FPGA, ASIC and Embedded Systems

5 star:		(4)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

There's a lot to like here. It goes over all the low-level stuff you could hope for, including creative number system, carry-lookahead, Booth encodings, and SRT division. It addresses some of the needs of crypto people, with discussion of finite-field arithmetic. It even gives enough intro to residue number systems for the desperate developer to gain a toehold - 10,000 digit addition or subtraction can be done in a few-digit time, as long as the expense of getting into and out of RNS are amortized.

That's all good for someone who can't trust their synthesis tools for good carry chains, or for someone headed way into the weirdness. The ranges where I live get distressingly little attention. If you need a dot product of two vectors, this will do a great job on the multiply and add steps as long as you can work out all the pipelining implications for yourself, but those were never the problem - it's the parallelism (how many multiplies can you run? how deep is your adder tree? or do you have something better?). It's the memory bottleneck (what do you mean you read "a word" from memory? I want 100). It's the numbers that number-crunchers use, i.e. IEEE 754, which get a moment of mention at the beginning and at the end. Those start turning strange with NaNs, signed zeroes, and denorms, then go totally off the rails when things like Intel (not always IEEE) compliance arise from the deep.

This could be a good text for a mid-level practitioner or student, fluent with logic design but blissfully ignorant of numerical analysis. If that's your trajectory, you'll spend some amount of time where this book lives. Then you'll advance, and it will no longer serve you. That's not a criticism, since every level has its own needs, but the prospective buyer should weigh needs to be met against needs that this meets. Not all readers will find a match.

-- wiredweird

In the part dedicated to general algorithms, very interesting new presentations or generalizations, made this work attractive at the theoretical point of view. Extensions of booth algorithms and generalizations to base B operation make the work innovative at the mathematical point of view. At the implementation level there is very good and innovative ideas towards special applications in FPGA (mainly Xilinx oriented). It would have been desirable to cope with some other technology, but the book may be considered self containing anyway.

The presentation of arithmetic theory and applications is innovative. Some of the topics are inedited; they present new approaches for both algorithmic and implementation aspects. It is a very interesting reference book for what refer to computer arithmetic in general and special purpose arithmetic circuit in particular.

That's an exciting and useful book in all synthesis manner: almost no gate-level circuits inside, as in modern EDA tools it don't need to.

A lot of algorithms (eg. log, sin, sqr...) which is beyond fast adders or one-cycle multipliers that can be easily found in many DSP hardware books. In fact, we make and sells a DSP state-machine chips in almost a million pcs that certain arithmetic circuit blocks is inspired by the book.

This book is quite original in its presentation. The selection of implementations is of interest.
The theoretical foundations are sound and presented in a well organized way.
The applications cope with the actual technology: especially in what concerns programmable devices.
It is a good book for advanced students and a must have tool for the professional designer.