Dear all, My sincere apologies for the multiple emails. I forgot to say that the seminar will take place at 14:00. So: GSSI Library on Friday June 24, 2022 at 14:00 CET
Best,
– FT
From: Francesco Tudisco francesco.tudisco@gssi.it Date: Wednesday, 22 June 2022 at 12:07 To: Mathematics Professors math-profs@gssi.it, Mathematics Postdoc math-postdoc@gssi.it, Mathematics Researchers math-researchers@gssi.it, Mathematics Students math-students@gssi.it, Computer Science Professors cs-profs@gssi.it, Computer Science Postdoc cs-postdoc@gssi.it, Computer Science Researchers cs-researchers@gssi.it, Computer Science Students cs-students@gssi.it, nomads-list@gssi.it nomads-list@gssi.it Subject: [NOMADS] seminar - Massimiliano Fasi Dear all, You are all invited to the next NOMADS seminar on low-precision numerical computations, which will be given by Massimiliano Fasi from Durham University on Friday June 24, 2022.
The seminar will take place in the GSSI’s Library, but remote participation will also be possible via the zoom link: https://us02web.zoom.us/j/89668910844?pwd=TmVwQUFoNklUajAzMlEyMzB6ZVZzUT09
Below you can find the details about his talk. Looking forward to seeing you all on Friday!
– Francesco Tudisco
Massimiliano Fasi (Durham University, UK) https://www.durham.ac.uk/staff/massimiliano-fasi/
Title of the talk: CPFloat: A C Library for Emulating Low-Precision Arithmetic
Abstract: Low-precision floating-point arithmetic can be simulated via software by executing each arithmetic operation in hardware and rounding the result to the desired number of significant bits. For IEEE-compliant formats, rounding requires only standard mathematical library functions, but handling subnormals, underflow, and overflow demands special attention, and numerical errors can cause mathematically correct formulae to behave incorrectly in finite arithmetic. Moreover, the ensuing algorithms are not necessarily efficient, as the library functions these techniques build upon are typically designed to handle a broad range of cases and may not be optimized for the specific needs of floating-point rounding algorithms. CPFloat is a C library that offers efficient routines for rounding arrays of binary32 and binary64 numbers to lower precision. The software exploits the bit level representation of the underlying formats and performs only low-level bit manipulation and integer arithmetic, without relying on costly library calls. In numerical experiments the new techniques bring a considerable speedup (typically one order of magnitude or more) over existing alternatives in C, C++, and MATLAB. To the best of our knowledge, CPFloat is currently the most efficient and complete library for experimenting with custom low-precision floating-point arithmetic available in any language.
For more information, please see: https://num-gssi.github.io/seminar/