Under What Conditions will My Application Give Reproducible Results?

Presented by: Martyn Corden, Intel Senior Software Engineer

Day: Wednesday
Date: December 12, 2018
Time: 11:00-12:00AM CST

Floating-point numbers are digital approximations of real numbers—they’re used to represent numbers much larger or smaller than the number of binary digits a computer’s memory register can be made to express.

But this finite-precision representation can lead to inherent uncertainty, as well as unavoidable errors, in the results of a floating-point computations.

For many applications, this is not a problem. But in cases where results must be exactly reproducible every time (like the critical functions used for finance, science & engineering, or quality control), the necessity of employing floating-point arithmetic must be tempered with the ability to deliver repeatable results.

Join senior software engineer Martyn Corden on December 12 to learn how this can be achieved for parallel applications using tools including the Intel® Compiler with OpenMP* and the free Intel® Performance Libraries.

Come with your software installed and your questions ready.

To register go to https://techdecoded.intel.io/webinar-registration/upcoming-webinars/?registration_source=appweb-dpd&cid=em-elq-42680&utm_source=elq&utm_medium=email&utm_campaign=42680&elq_cid=1112559