Architecture studies performed with IDaSS
A large number of IDaSS designs were not converted
(directly) into an ASIC layout. Some of the larger projects are listed
below:
Superscalar processors
These can execute more than a single instruction per clock. A
design has been made which can execute up to 4 instructions in parallel.
Distributed Programmable Logic Controller
This design allows the interconnection of up to 32 processing
units. Special hardware makes them behave as if they are executing
a single program. The target processing speed is 600 million operations
per second.
Dataflow processors
These simulate a network of interconnected processing
nodes. A superscalar version of the NEC uPD7281 device has been designed
incorporating 3 parallel calculating units and 4 parallel data token routing
and storage pipelines.
Input/output coprocessor
This design was intended to be capable of supporting a large number
of external devices by running programs for each of them (with hardware
multitasking). Using a single instruction, each of these programs can
start built-in Direct Memory Access hardware to transfer data at high speed,
after which the program simply continues.
Single chip multi-microprocessor
This design incorporates a set of identical RISC processors and a
number of shared co-processors for executing complex instructions.
Instruction and result routing matrices connect the two types of
processors. One of the co-processors performs the basic operations of
a multitasking kernel in hardware and dynamically assigns tasks to the
RISC cores (based upon task priorities and their status). Inter-task
communication is possible with hardware-implemented mailboxes and
semaphores. Click on the link labeled 'MuP' in
Bart Theelen's
homepage for more information.