TOPSTREAM™ Technology Overview
What is TOPSTREAM™?
Task Oriented Parallel Stream (TOPSTREAM™) is a next-generation microprocessor technology that models efficient distributed computing system. The TOPSTREAM™ distributed system consists of multiple autonomous processor cores that communicate through the on-chip hierarchical network. The processor cores interact with each other in order to achieve a common goal. The computer program that runs in the TOPSTREAM™ distributed system is organized as a Kahn Process Network (KPN), and the distributed programming is done through the process of software partitioning from legacy sequential programs.
Background technologies of TOPSTREAM™
TOPS Systems Corp. has started its R&D based on plenty of expertise on developing conventional microprocessors, such as x86 microprocessors, DSPs, distributed controllers, and Reconfigurable processors, and developed TOPSTREAM™ Architecture. We continue intensive R&D on core technologies for next-generation embedded systems.
Technologies of TOPS Systems Corporation.
We have developed a wide range of technologies. We have been filed more than 20 patents.
Ø Computer Architecture
■TOPSTREAM™ Architecture solves three major performance issues on computers
<Three Major Issues>
- ILP*1 Wall: Limitation on number of instructions that CPU can issue at a time
- Memory Wall: Speed gap between CPU and memory that increases at a rate of 50% a year
- Power Wall: CPU performance limited by Power Consumption
*1: Instruction Level Parallelism
<Overview of TOPSTREAM™ Architecture>
- Distributed Stream Processing
- Heterogeneous Multi-Core processor
- On-Chip scalable Extended Harvard Architecture bus with Distributed Arbitration
- Dual Instruction Set Architecture (RISC + Data Processing ISA) with 16-bit Instruction Length
- Inter-Core Shared Register-Banks
■TOPSTREAM™ Micro-Architecture drastically improves performance
<Overview of TOPSTREAM™ Micro-Architecture>
- Simple Instruction Execution Pipeline
- Independent execution of load-store for Stream I/O and computation for kernel processing
- Zero-Overhead Loop Mechanism
- Parallel Instruction Decoder
- Multi-Banked Multi-Register-Files (General Purpose Registers and Data Registers)
- Super Complex Operation Unit
- Extended Harvard Architecture bus (Instruction bus, Data bus, and System bus)
■Logic Design to improve efficiency for performance, energy, and cost
<Logic Design of TOPSTREAM™>
- Heterogeneous Multi-Core processor design platform
- Separation of Datapath and Control
- High Performance Complex Operation Unit
- Component-based RTL design
■TOPS Lib improves design efficiency
<Overview of TOPS_LIB>
- Parameterized Vectored Component Libraries
- Truth-Table based Function Libraries
Ø Software
■Object-Oriented Software technologies to solve issues on software for Multi-Core processors
<Overview of Object-Oriented Software technologies>
- Static Object Mapping
- Dynamic Object Mapping
- Micro ORB Engine
Ø Software Development Environment
■Compiler Technology to highly utilize resources on Multi-Core processors
<Overview of Compiler Technologies>
- Scheduling Algorithms for Stream Processing
- Optimization Algorithms for Microprocessor Architecture dependent optimization
■TS-ISIM Instruction Set Simulator to accelerate Software test
<Overview of TS-ISIM Instruction Set Simulator>
- Ultra-Fast ISS for Multi-Core processor
- Object-Oriented design to support Multiple Instruction Set Architecture
Ø Simulation Technologies
■Simulation technologies to enable ultra-fast simulation speed
<Overview of Simulation Technologies>
- Cycle based Simulation by static scheduling
- Ultra High Speed HW/SW Simulation based on TLM
- Automatic Conversion from RTL to Cycle based C model
Ø 3D-Chip Stack Technologies
■COOL Interconnect Technologies to enable scalable 3D-chip stacking
Ø Application Technologies
■Application Technologies to enable Next-Generation Information Systems
<Overview of Application Technologies>
- 3D recognition technology based on stream input from a single camera inputs
- Dynamic Cryptography based Communication System
|