Browse Prior Art Database

A FAST BRANCH CONTROL SCHEME

IP.com Disclosure Number: IPCOM000005598D
Original Publication Date: 1986-Oct-01
Included in the Prior Art Database: 2001-Oct-18
Document File: 1 page(s) / 51K

Publishing Venue

Motorola

Related People

Marty Hood: AUTHOR [+3]

Abstract

When designing the MC68020, we wanted to use a branch PLA in the microsequencing schema similar to that used on the MC68000110. However the circuit constraints placed on MC68020 were much more severe. The MC68020 was designed to have twice the operating frequency and the die is four times as large (much longer RC delays associated with some signals).

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 73% of the total text.

Page 1 of 1

0 M

MOTOROLA Technical Developments Volume 6 October 1986

A FAST BRANCH CONTROL SCHEME

by Marty Hood, Sob Thompson, and Doug MacGregor

   When designing the MC68020, we wanted to use a branch PLA in the microsequencing schema similar to that used on the MC68000110. However the circuit constraints placed on MC68020 were much more severe. The MC68020 was designed to have twice the operating frequency and the die is four times as large (much longer RC delays associated with some signals).

   The setup requirements associated with the inputs to the branch PLA which are determined in the ALlJ/ALUT are too early, necessitating a new control scheme. This new scheme was achieved through changes to the microcode and logic. This paper deals with the logic changes only since the microcode changes were clever recoding.

   The new logic prevents delays in sequencing of the micromachine during key flows, thus speeding condi- tional branching. This is achieved by using three-stateable CMOS buffers that drive directly onto the micro- address bus as opposed to the tree-decoders employed on older designs. The penalty for this additional speed is more silicon real estate. To avoid making this method prohibitively large, the inputs to the branch PLA were broken into three categories of speed. The faster inputs were fed into the traditional tree-decode PLA employed on the MC68000. The slower inputs were each given a three-stateable driver and decode logic to select them based on the branch co...