Browse Prior Art Database

Off-Peak Optimization for Very-High-Level-Languages

IP.com Disclosure Number: IPCOM000102300D
Original Publication Date: 1990-Nov-01
Included in the Prior Art Database: 2005-Mar-17
Document File: 2 page(s) / 91K

Publishing Venue

IBM

Related People

Brackenbury, IF: AUTHOR

Abstract

To improve performance of frequently used Very-High-Level-Language (VHLL) programs, without sacrificing interactive speed of their development and revision, does not require additional work of the programmer or additional costs.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Off-Peak Optimization for Very-High-Level-Languages

       To improve performance of frequently used
Very-High-Level-Language (VHLL) programs, without sacrificing
interactive speed of their development and revision, does not require
additional work of the programmer or additional costs.

      The scheme relates to the optimization of VHLL programs using
mechanisms already implemented for: incremental archiving of data;
optimizing off-peak garbage collection services; and the gradual
improvement of graphics picture quality.  In the latter case the
picture is first drawn coarsely and fast, but when left alone, the
graphics are recalculated and put up a better picture.  Off-peak
workstation processing power is used to optimize VHLL programs, e.g.,
in the evening, at night and at lunchtime. It allows fast
interpreting and debugging followed by batch compilation for best
object code.  No operating system modifications are needed.

      For best interactive performance when developing or modifying
VHLL programs, the performance of the interpreter or compiler is the
focus.  There is seldom time for significant levels of optimization.
In traditional software development the programmer optimized his
functionally complete program by manually invoking and tending an
optimizing compiler - or optimizing mode of his usual compiler.  Few
have the patience or affordable time to explicitly optimize VHLL
programs.  Good VHLL tools emphasize ease-of-use, compile with speed
and also produce fast run-time code.  These contradictory criteria
are the measures against which VHLL is judged.

      This disclosure capitalizes on the widespread move to
workstations away from timesharing systems, in particular for the
professional who uses programming as a "means to an end".
Workstations, unlike timeshared resources, are used in bursts and
remain idle for most of their power-on time. The scheme consists of a
local software service process running continuously at low priority.
It maintains lists of all VHLL programs on the workstation, and for
each program remembers the number of times it has been executed
without being modified.  The raw data for calculating this
information is available through the time-stamp a...