Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method for a scalable, high speed, 1R1W L0 hybrid cache

IP.com Disclosure Number: IPCOM000019327D
Publication Date: 2003-Sep-10
Document File: 3 page(s) / 40K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method for a scalable, high speed, 1R1W L0 hybrid cache. Benefits include improved performance and an improved design environment.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 54% of the total text.

Method for a scalable, high speed, 1R1W L0 hybrid cache

Disclosed is a method for a scalable, high speed, 1R1W L0 hybrid cache. Benefits include improved performance and an improved design environment.

Background

The L0 cache load to the address generation unit write-back bus and back around to the cache is a fundamental conventional architectural loop in microprocessors. It often sets the product’s frequency in 32-bit architecture designs. These designs require at least 20% extra design frequency in performance simulations to ensure adequate guard-banding for high volume manufacturing. Typically, L0 cache is designed as n-way associative, dual port register files (RF) for fast latency. However, in RF cache topology, the way selecting signals often becomes the timing limiter. In the contrary, a fully small signal arrays which can relax the way select signals required time, have limitations and concerns such as sense amp offset voltage scaling, low threshold voltage devices, and bit-line leakage

General description

The disclosed method is a scalable, high speed, dual port L0 hybrid cache. The method includes cache design topologies that have dual-ended full-swing local bit lines but incorporate small signal global bit lines for the read path. This arrangement enables the memory cell to be a standard register file cell, and the write path is single ended.

Advantages

The disclosed method provides advantages, including:


•         Improved performance due to reducing the silicon area by enabling, single-ended writes (one wire track) rather than dual-ended writes (two wire tracks)

•         Improved performance due to scalability in the write margin on store...