@Thomas Klemm -> CORDIC Article

[pito]

Quote:What is the most efficient approach in terms of speed and accuracy between CORDIC and Taylor?
Or may be CORDIC was the right choice for old calculators with little ROM?

The cordic was the choice with early calculators which had 1-4kBytes of rom (and fully different architecture as it is today, ie. my HP-25 with 2kx10bit of rom - see the disassembled rom at http://www.jacques-laporte.org/Woodstock...s/hp25.txt). I did some benchmarks in past on cortex M3 (32bit cordic vs. single precision) where I set the cordic's N-iterations such I get similar results as with single precision and the timing difference was not dramatic. The Taylor does not need too many terms to get required precision.

You may try the 32bit cordic (ready code to run):
http://www.dcs.gla.ac.uk/~jhw/cordic/

Today the cordic algorithms are mostly used in FPGA or ASIC designs - there are sources in VHDL/Verilog available - it is the implementation in hardware which is _very_ fast.

Also mind the WP34s is using decNumber library with arbitrary lenght decimal floating point (set to maybe 34 digits or more for trigo), which is not the same as the IEEE double precision (or single precision).

On the ARM a single precision (32bit, 6-7digits precision) calculation is typically twice faster than the double precision (64bit, 14-15digits), and a 34digits decimal floating point is maybe 10-15x slower than the double precision..