Browse Prior Art Database

Speed-Independent Capture Cycle Setting Algorithm for Chip to Chip Communication

IP.com Disclosure Number: IPCOM000020237D
Original Publication Date: 2003-Nov-04
Included in the Prior Art Database: 2003-Nov-04
Document File: 2 page(s) / 50K

Publishing Venue

IBM

Abstract

Disclosed is a method to set registers based on system timing information and system speed. This method would typically be used to facilitate chip to chip communication, but could be used for internal chip settings as well.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

  Speed-Independent Capture Cycle Setting Algorithm for Chip to Chip Communication

   When communicating between chips using a 'dumb' (untagged data), synchronous, pumped interface, a capture cycle register is programmed to receive the correct data on the correct internal clock cycle. This type of interface, and thus the capture cycle register, is used pervasively throughout IBM's product line and elsewhere. Two examples of this type of register are: (1) a register in a memory controller which selects capture cycle of read data from memory, (2) a register we call a target cycle register used for all communication between chips in the Central Electronic Complex (CEC) of recently developed server lines.

     Generally, as the speed of the interface changes, a different capture cycle must be programmed into the register. In addition, the capture cycle is dependent on other settings in the chips. Thus, a large number of different capture cycle settings may be required. In previous solutions, a software table would be used to look up a capture cycle based on the speed of the interface. This has three drawbacks: (1) the tables can be very large for systems with a large number of interfaces and/or a large range of speed, (2) the tables must be updated whenever another configuration register that affects the capture cycle is changed, (3) one must know the range of speeds one would like to run the interfaces before encoding the table.

     Rather than encode specific capture cycles in a table, a speed and configuration-dependent formula is incorporated into code and the relevant parameters required to evaluate the formula are passed in as cycle-time and configuration-independent values.

     This new approach substantially reduces the size of the look-up table - the table size for comparable information is reduced by more than a factor of 100 for our applications.

     The approach allows a machine to be operated through a range of frequencies without foreknowledge of the required range - in a previous machine generation, we repeatedly updated the tables to reflect new frequency ranges; this is now eliminated

     The approach is flexible to adjust to other register settings without requiring a new table - previously, we also updated tables several times when related register settings were updated.

     Here are the basic steps to implement this invention. This description uses a specific example to illustrate an application. A wide variety of other implementations can be used to set other speed dependent registers.

[A] A general formula/algorithm for a...