Verified Software: The Real Grand Challenge

Abstract

Abstract. This position paper addresses, and attempts to propose solutions for, critical issues in software engineering that need to be resolved before the Verified Software grand challenge as proposed by Professor Tony Hoare can be usefully exploited in industry to increase the assurance of software intensive systems. IntroductionThe following assumptions about programs and their correctness (which I refer to in the sequel as "assumptions") are implicit in the problem description of the Verifying Compiler Grand Challenge: 1) Associated with each program are types, assertions, and other annotations 1 that are readily available. 2) They are unassailable, inviolable, and invariant. 3) Their correctness is both necessary and sufficient for the correctness of the programs they annotate. In this paper, I argue that for programs that are intended to solve real-world problems, the subject of my research for more than fifteen years, none of these assumptions necessarily holds. I proceed to explain how this problem may be addressed, and conclude with what I think are more realistic expectations on the impact of the grand challenge problem and its solutions on real-world software development projects. Problem StatementFor programs whose behavior is easily specified as mathematical functions, it is conceivable that the assumptions are valid. An example of such a program is one that implements the 4-coloring algorithm for planar graphs. If we assume that program annotations can characterize the function being computed by the program, the proof of its correctness is probably derivable from the proof of correctness of the 4-coloring problem. However, even for such programs, the correctness of its annotations is often predicated upon extraneous factors in the program's execution environment, such as the word length of the processor, the size of the address space, or the amount of available memory. This is because 1 I shall loosely use the term "annotations" to refer to this redundant information.

Induced Alpha And Beta Electroencephalographic Rhythms Covary With Single-Trial Speech Intelligibility In Competition

“…Stimulus presentation was controlled using custom MATLAB (The MathWorks, Inc., Natick, MA) routines. EEG preprocessing was performed using the open-source software tools MNE-PYTHON (Gramfort et al, 2014) and SNAPsoftware (Bharadwaj, 2018). All further data analyses were performed using custom software in PYTHON (Python Software Foundation, Wilmington, DE).…”