CORDIC
en.wikipedia.org
en.wikipedia.org
(The motivation was that CORDIC with Walther's 1971 improvements performs great for trigonometric/hyperbolic/arc-functions, Euclidean distance, sqrt and also allows (bit-serial) division and multiplication - but still slow compared to array multipliers. This stuff was used in 1970s multifunction calculators (when area.)
Since I decided to not do a PhD or do followup publications, my work proved the concept worked, then pretty much vanished into the conf literature.
The concept is straightforward; if anyone out there wants to synthesize it on a Xilinx or similar, or do a soft-macro (Verilog/VHDL), and get some performance numbers, drop me a line.
Also, Timmerman, Hahn, Hosticka did some good innovation in the 1990s on scaling, pipelining and latency. [1] But for most use-cases, you don't pipeline unless you really need to trade higher throughput off against minimal gatecount/silicon area and latency.
And in the 1990s, there was interest in CORDIC for matrix triangularization and SVD.
[0]: "Hybrid multiplier/CORDIC unit", IEEE ICASSP-99 https://ieeexplore.ieee.org/abstract/document/758297
[1]: https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=cord...
I'm no longer doing hardware, I said if other people want to synthesize the concept, it's pretty easy to do (I implemented it in VHDL, MatLab and C. Or Verilog), and I'd love to see some performance numbers; they can drop me a line if they have any questions.
AFAIK, CORDIC hardware (undeservedly) became a backwater by the 1990s, except in specialized areas like angle sensors, navigation or maybe SVD decomposition.
https://code.activestate.com/recipes/576792-polar-to-rectang...
https://alchitry.com/projects/gpu http://ijsetr.com/uploads/435216IJSETR3253-390.pdf
>"CORDIC uses simple shift-add operations for several computing tasks such as the calculation of trigonometric, hyperbolic and logarithmic functions, real and complex multiplications, division, square-root calculation, solution of linear systems, eigenvalue estimation, singular value decomposition, QR factorization and many others."