of Microprocessors

MUSIC SYNTHESIS SOFTWARE 643
only for execution. In fact, each level should probably be broken down
further into many subroutines. While this book is not the place to discuss
modular and structured programming practices, it is dear that the major
levels need not all be coded in the same programming language. In fact, the
optimum balance between programming effort and music program speed
requires that different languages be used on different levels.
There is a little question that Levels 0 and 1 should be implemented in
assembly language for the machine being used. The reason, of course, is that
they execute at the sound sample rate of 8 ks/s to 50 ks/s. Furthermore, for
almost any personal computer, fixed-point (integer) arithmetic must be used
for the sample calculations. Most programmers are unfamiliar with the fine
points of fixed-point computation, having been spoiled by the floating-point
capabilities of large mainframe computers or high-level languages. Since
integer computation is vital to acceptable speed with microcomputers, it will
be discussed in detail later.
In isolated instances, the lower-level routines have actually been implemented
in the microcode ofa microprogrammed minicomputer. This essentially
means that new instructions are added to the computer's repertoire that
facilitate sample computation. What is actually happening is that the internal
machine registers and logical elements are being directly manipulated
by the microcode at maximum speed without the overhead of reading instructions
from main memory. Although two to three times faster,
microprogramming is even more obscure and difficult than assembly programming.
Also, only the larger minicomputers allow user microprogrammIng
anyway.
Dedicated hardware such as was discussed in the previous chapter can
be thought of as replacing Level 1 routines. If the dedicated hardware is
intelligent, Level 2 may also be replaced. In computer-controlled analog
systems, voltage-controlled oscillators, amplifiers, and filters perform the
functions of Level 1 routines, while envelope generators do the Level 2
functions.
High-Level Languages
Level 3 routines are best implemented in a higher-level language,
particularly in a delayed playback system. This level of the music system
typically handles a relatively small volume of data but does a lot of character
string scanning, data table maintenance, and decision making. Thus, only a
small portion of the computational effort is spent in these routines, while a
large portion of the programming effort is sunk into them. High-level
languages tend to minimize the programming effort required while simultaneously
making the programs easier to read and modify (if adequately
commented). High-level languages for microcomputer systems almost invariably
have features thar make it easy to link to assembly level routines.
The natural question, of course, is: Which high-level language should
be used in the upper levels of a music software system? Like microprocessor