Overview
Pipeline and vector processors often require simultaneous access to memory from multiple sources. An instruction pipeline may need to fetch an instruction and operand at the same time from different segments. Similarly, arithmetic pipelines usually require two or more operands to enter simultaneously.
Instead of using multiple memory buses for simultaneous access, memory can be partitioned into modules connected to common address and data buses, creating an elegant and efficient solution to this architectural challenge.
Multiple Module Memory Organization
Array
Array
Array
Array
A memory module is a memory array together with its own address and data registers. Each memory array has its own address register (AR) and data register (DR). The address registers receive information from a common address bus, and the data registers communicate with a bidirectional data bus.
How Memory Interleaving Works
Key Principle
The two least significant bits of the address distinguish between the modules. The modular system permits one module to initiate a memory access while other modules are in the process of reading or writing a word, and each module can honor a memory request independent of the state of other modules.
In an interleaved memory, different sets of addresses are assigned to different memory modules. For example, in a four-module memory system, even addresses may be in one module and odd addresses in another. When the number of modules is a power of 2, the least significant bits of the address select a memory module, and the remaining bits designate the specific location within the selected module.
Applications
Vector Processing
A vector processor using an n-way interleaved memory can fetch n operands from n different modules. By staggering memory access, the effective memory cycle time can be reduced significantly.
CPU Pipelining
A CPU with instruction pipeline can take advantage of multiple memory modules so that each segment in the pipeline can access memory independent of memory access from other segments.
Performance Benefits
- Enables simultaneous memory access from multiple sources without multiple memory buses
- Reduces effective memory cycle time by a factor close to the number of modules
- Allows independent module operation for enhanced parallelism
- Supports efficient pipeline and vector processor architectures
- Provides cost-effective solution for high-performance computing needs