Need the accuracy, resolution and tolerance values of Newton-Raphson inverse algorithm and Maclaurin series.

Hi Prasad,

Thanks for your post.

To get higher precision results, Newton-Raphson interpolation would be recommended.

For RSQRxP instructions to calculate 1/sqrt(v), Newton-Raphson interpolation would improve the mantissa precision to 2-16, and one additional interpolation could improve the mantissa precision to 2-24 for single precision and to 2-32 for double-precision.

May be, you could check the app. report for C66x implementation to check the impact on the error performance, kindly refer section 3.1.5 and table 4 for 1/sqrt(x) calculation results from the below report:

www.ti.com/.../sprabg7.pdf

I think you should read a bit further on in the application report and there's the Newton-Raphson interpolation used which would give much better results on the c66x.

Here's some background on the topic:

betterexplained.com/.../understanding-quakes-fast-inverse-square-root

Please check the inverse square root demo on the above article to find approximation roots of any function and can keep iterating the method to get closer and closer to the root.

I think, you can also look for an alternative to use IQMath Library from TI as below:

focus.ti.com/.../sprc542.html

Thanks & regards,
Sivaraj K

-------------------------------------------------------------------------------------------------------

Please click the Verify Answer button on this post if it answers your question.

-------------------------------------------------------------------------------------------------------