Mnemo
ABOUT APPROACH ARCHITECTURE USAGE MUSES

About

The Mnemosyne project is led by GrammaTech in collaboration with UT Austin and MIT sponsored by DARPA and AFRL. Our work on Mnemosyne started in May 2020. If you want to chat you might find us on the Mnemosyne Gitter.

The Mnemosyne/docs git repository which holds all prior versions of this site—commit date is concurrent with publish date.

Goals

The goal for Mnemosyne is to provide an automated software development environment which is usable and enables developers to build better software faster.

Our efforts will be guided by the following measurable goals. These goals will be measured through our own use of Mnemosyne (i.e. dogfooding).

  1. Usability. How easily is Mnemosyne used. Is it difficult to identify, invoke, and then apply the results of the synthesis modules. We will attempt to continually evaluate usability in our own development using Mnemosyne. This is our most subjective metric.
  2. Quality. How good is the synthesized code returned by Mnemosyne. More generally what is the quality of entire software projects developed using Mnemosyne. Evaluation of this metric will leverage the many readily accessible tools for automated quality assessment of software projects; from linters and static analyzers to dynamic fuzzers.
  3. Scalability. There are two aspects to this question which we will attempt to measure independently. First, how do our individual synthesis modules scale against task size. From single expressions and statements, up to functions, and maybe eventually whole modules. At least initially we may use lines of code (LOC) of synthesized code as a proxy for complexity. Second is the question of how well Mnemosyne scales to the size of the overall project. As Mnemosyne relies on the software developer to decompose the top-level requirements into pieces which are tractable to the available synthesis modules the system should productively contribute to any scale of software project (or just to portions of a software project). However, it may still be the case that certain domains of software projects benefit more directly from the use of Mnemosyne for developer assistance.
  4. Automation. We will measure the overall impact of automation. We plan to augment our Argot-server to track the provenance of every character of code. We can then review this information to measure to which degree specific synthesis modules and the developer contributed to the code base of a project over the course of development.

Copyright (C) 2020 GrammaTech, Inc.

This material is based upon work supported by the US Air Force, AFRL/RIKE and DARPA under Contract No. FA8750-20-C-0208. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the US Air Force, AFRL/RIKE or DARPA.