Table of Contents
Advertisement
Nonstop Operation. When connected to special devices
(on a ready-resume basis), the computer can sometimes
become excessively delayed if the special device does not
respond quickly. A built-in watchdog timer assures that
the SIGMA 6 computer cannot be delayed for an exces-
sive length of time.
Real-Time Clocks. Many real-time functions must be timed
to occur at specific instants.
Other timing information is
also needed - elapsed time since a given event, for example,
or the current time of day. SIGMA 6 can contain two (or
four) real-time clocks with varying degrees of resolution
(1/60
second or
V8
mi lIisecond, for example) to meet these
needs. These clocks also allow easy handling of separate
time bases and relative time priorities.
Rapid Context Switching. When responding to a new set of
interrupt-initiated circumstances, a computer system must
preserve the current operating environment, for continuance
later, whi Ie setting up the new environment. This changing
of environments must be done quickly, with a minimum of
II
overhead II costs in time. In SIGMA 6, each one of up to
32 blocks of general-purpose arithmetic registers can, if
desired, be assigned to a specific environment. All rele-
vant information about the current environment (instructi on
address, current genera I regi s"er block, memory-protecti on
key, etc.) is kept in a 64-bit program status doubleword
(PSD). A single instruction stores the current PSD any-
where in memory and loads a new one from memory to es-
tablish a new environment, which includes information
identifying a new block of general-purpose registers. A
SIGMA 6 system can thus preserve and change its operating
environment completely through the execution of a single
i nstructi on.
Simultaneous I/O Channel Operation. The use of a multi-
plexor input/output processor (MIOP) or MIOP expansion
option permits up to 24 channels with standard-speed de-
vices to operate concurrently; the addition of more MIOPs
increases this throughput.
High-Speed Channel Operation. The use of the selector
input/output processor (SlOP) permits very high-speed data
transfer - up to one 32-bit word per memory cycle. To
meet special needs, data size can be 8 or 32 bits wide.
Memory Protection.
Both foreground (real-time) and back-
ground programs can be run concurrently ina SIGMA 6
system, because a foreground program is protected against
destructi on by on unchecked background program. Mem-
ory write-protection guarantees that protected areas of
memory can be written into only under predefined con-
ditions.
Under operating system control, the memory
access-protection feature also prevents accessing of mem-
ory for specified combinations of reading, writing, and
instruction acquisition.
Variable Precision Arithmetic. Much data encountered in
real-time systems are 16 bits or less. To permit this length
of data to be processed efficiently, SIGMA 6 provides half-
word arithmetic operations in addition to fullword oper-
ations.
Doubleword arithmetic operations (for extended
precision) are also included.
Direct Data Input/Output.
For handl ing asynchronous I/O,
a 32-bit word can be transferred directly to or from a
general-purpose register, so that an I/O channel need not
be occupied with relatively infrequent transmissions.
Interleave/Overlap.
To increase processing speeds, mem-
ory banks overlap cycles automatically wherever possible.
Core memory addresses can be interleaved modul0-2 or
modul0-4 on a bank basis to increase the probability of
overlapping.
GENERAL -PURPOSE FEATURES
General-purpose computing applications are characterized
primari Iy by an emphasis on computation and internal data
handling. Many operations are performed in floating-point
format and on strings of characters.
Other typical charac-
teristics include decimal arithmetic operations, the need to
convert binary numbers into decimal (for printing or display),
and considerable input/output at standard speeds. The
SIGMA 6 system includes the following general-purpose
computer features.
Floating-Point Hardware (optional). Floating-point in-
structions are avai lable in both short (32-bit) and long
(64-bit) formats.
Under program control, the user can
select optional zero checking, normalization, and signifi-
cance checking (which causes the computer to trap when a
post opera.tion shift of more than two hexadecimal places
occurs in the fraction of a floating-point number). The
significance checki ng feature permits the use of the short
floating-point format (for high processing speed and storage
economy) and the use of the ·Iong format when loss of
significance is detected.
Decimal Arithmetic Hardware.
Decimal arithmetic instruc-
tions operate on up to 31 digits plus sign.
This instruction
set also includes pack/unpack instructions (for converting to/
from the packed format of two digits per byte) and a general-
ized edit instruction (for zero suppression, check protection,
and formatting byte information with punctuation to displc:y
or print it).
Indirect Addressing. This feature provides for simple table
linkages and permits the user to keep data sections of
his program separate from procedure sections for ease of
maintenance.
Displacement Indexing. The technique of indexing by
means of a IIfloating
li
displacement permits the user to
access the desired unit of data without the need to con-
sider its size.
The index registers automatically align
themselves appropriately; thus, the same index register
can be used on arrays with different data sizes. For ex-
omple, in a matrix multiplication of any array of fullword,
single-precision, fixed-point numbers, the results can be
stored in a second array as double-precision numbers, using
the same index quantity for both arrays. If an index regis-
ter contains the value of k, then the user always accesses
the kth element, whether it is a byte, halfword, word, or
doublaword. Incrementing by various quantities according
to data size is not required; instead, incrementing is always
General-Purpose Features
5
Advertisement
Table of Contents
