Floating-Point Supportive Pipeline for Emulated Shared Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing emulated shared memory architectures face challenges in optimizing pipeline architecture for efficient execution of floating-point operations, leading to increased latency and complexity in circuit layout.
Innovation Solution
A processor architecture with interleaved inter-thread pipelines, featuring separate parallel branches for integer and floating-point operations, where floating-point units are located operatively in parallel with the memory access segment, allowing simultaneous execution of operations and reducing latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If floating-point units are integrated into the pipeline sequence, then floating-point operations can be performed, but pipeline latency increases and throughput decreases
Solution Approach 1:
The patent introduces a parallel pipeline branch dedicated to floating-point operations, running concurrently with the integer pipeline. This dimensional separation allows floating-point and integer operations to execute simultaneously without interfering with each other's throughput, effectively adding a new execution dimension to the processor architecture.
Solution Approach 2:
The processor pipeline is segmented into two independent parallel branches: an integer pipeline and a floating-point pipeline. Each branch handles its respective operation type independently, allowing the system to process integer and floating-point operations simultaneously without the latency penalties of sequential processing.
2Speed
If separate pipeline branches for integer and floating-point operations are implemented, then operation execution speed increases, but device complexity increases
Solution Approach 1:
The parallel pipeline branches share common infrastructure components including the instruction fetch unit, instruction decode unit, register files, and output write-back mechanisms. This universal sharing of components reduces the overall complexity increase while maintaining the speed benefits of parallel execution paths.
Solution Approach 2:
The patent merges the integer and floating-point pipelines at key stages such as instruction fetching, decoding, and result write-back. This consolidation allows the processor to benefit from parallel execution while avoiding the complexity of completely separate instruction pipelines, as both operation types share the same control and storage infrastructure.
Data Source
AI summary
A processor architecture arrangement for emulated shared memory (ESM) architectures, including a number of multithreaded processors each provided with interleaved inter-thread pipeline and a plurality of functional units for carrying out arithmetic and logical operations on data, wherein the pipeline includes at least two operatively parallel pipeline branches, first pipeline branch includes a first sub-group of said plurality of functional units, such as ALUs (arithmetic logic unit), arranged for carrying out integer operations, and second pipeline branch includes a second, non-overlapping sub-group of said plurality of functional units, such as FPUs (floating point unit), arranged for carrying out floating point operations, and further wherein one or more of the functional units of at least said second sub-group arranged for floating point operations are located operatively in parallel with the memory access segment of the pipeline.