NewProjectChecklist

• What precisely will be your contribution?
  • I will design a C++ class to replace doubles that people can use in their numerical applicationsto verify that they aren't loosing accuracy due to errors due to their use of floating point
• What question are you answering?
  • Do numerical analysts need to be worried about rounding errors in practice? How often and where?
• Why should anyone care?
  • Rounding error can be the major contribution of error in a simulation.
  • it's effects are ignored because it's hard to analyze theoretically.
  • If you don't profile your code with my class, you have no way of debugging where your rounding error is.
• What larger question does this address?
  • Can we improve how real numbers and their representations are represented by computers>?
• Who would be interested in this research?
  • People who practice NA, systems people, people that develop floating point circuits.
• What work is similar to what you've suggested?
  • NEED MORE

• What will be your new result?
  • A vindication that rounding error isn't that important, or a clear warning that rounding error is very important
• What new knowledge have you contributed that the reader can use elsewhere?
  • Profiling an application is an important step of debugging.
  • A library they can use to test their own applications
• What previous work do you build on? What do you provide a superior alternative to?
  • theory of significant digits
  • design and rational of floating point systems
  • NEED MORE
• How is your result different from and better than this prior work?
  • We can point you to the precise line number where your code is malfunctioning. We can tell you what formulas you need to rewrite.
  • NEED MORE
• What, precisely and in detail, is your new result?
  • A class library to test the accuracy of floating point iterations

Evaluation chart

• dependent theory
  • Profiling in debugging
  • the design of floating point systems
• how well explored is the topic.
  • NEED MORE
• estimated effort
  • for community
    • easy
  • for you
    • easy, if not a bit harder
• research versus development
  • development in the implementation
  • research in the application to real numerical codes.
• impact - hihg medium low
  • medium. Not an earth shattering result, but it should see use.
• prerequisites - dependencies amongst your categories.
  • none
• human resources
  • low
• is it state of the art?
  • NEED MORE

• Summarize your project idea in 30 seconds? Read it aloud and time yourself.
  • Rounding error is often assumed to be negligable compared to truncation error when analyzing the error of a numerical code. However, for most codes rounding error can dominate the error of the calculations. I propose a C++ class library aids in the detection and quantification of rounding error. I will use this library to analyze known results and applications.