and the problem of large numbers eating small numbers_loongknown’s blog

class=”markdown_views prism-atom-one-light”> In-depth understanding of floating-point number implementation principles, number range and precision, and the problem of large numbers eating small numbers Floating point number implementation principle Normalized representation Denormalized representation (including 0) Special values ​​(INF, NaN) Indicates range Accuracy Big numbers eat small numbers Principle Solution Option 1: Use higher precision data types Scheme 2: Segmented sum Option Three: Kahan Summation Reference After reading this article, you will have a deep understanding of the principles of floating-point number implementation: normalized representation, non-normalized representation, +/- 0, INF, NAN, floating-point number representation range and precision. And will also figure out the principle of big numbers eating small numbers and the corresponding solutions. A few days ago, a colleague and I discussed the precision of floating-point numbers and the problem of large numbers eating decimals. I just took this opportunity to write this article on the implementation principle of floating-point numbers and the problem of large numbers eating decimals, which is also convenient for later people. See similar questions. Floating point number implementation principle Here we take IEEE-754 single-precision floating-point numbers as an example, and double-precision floating-point numbers are similar. Single-precision floating-point numbers are stored in computers as 32 bits. These 32…