Preparing your result...
Loading...
Press Esc to dismiss this message

Operand History Table (01-Dec-1984)

Thumbnail
IP.com Prior Art Database Disclosure (Source: IPCOM)
Disclosure Number IPCOM000044351D dated 01-Dec-1984
Added to Prior Art Database on 05-Feb-2005
Disclosed by: IBM (TDB 12-84 p.3815-3816)
Country: United States
Related People
Emma, PG - AUTHOR [+4] [-4]
Yorktown
Pomerene, JH - AUTHOR
Yorktown
Puzak, TR - AUTHOR
Yorktown
Rechtschaffen, RN - AUTHOR
Yorktown
Sparacio, FJ - AUTHOR
Yorktown
Disclosure File: 2 pages / 65.1 KB / English (United States) / 1 attachment(s)

Operand pre-fetching for faster instruction executions is described. Trace analysis has shown that many of the instructions that require an operand compute the same operand address each time that the instruction is executed. Consider the instruction format below: (Image Omitted) The operand address is computed by adding the contents of the base register to the contents of the index register to the value of the displacement field supplied by the instruction. That is, operand address = C(B) + C(X) + displacement. This calculation is normally performed at address generation (AGEN) time. The operands are then fetched from the cache or memory. The time needed to fetch the operands depends on the speed of the cache or memory. Fig. 1 represents an idealized, cycle by cycle description of an instruction passing through a pipeline.

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 64% of the total text.

Page 1 of 2

Operand History Table

Operand pre-fetching for faster instruction executions is described. Trace analysis has shown that many of the instructions that require an operand compute the same operand address each time that the instruction is executed. Consider the instruction format below:

(Image Omitted)

The operand address is computed by adding the contents of the base register to the contents of the index register to the value of the displacement field supplied by the instruction. That is, operand address = C(B) + C(X) + displacement. This calculation is normally performed at address generation (AGEN) time. The operands are then fetched from the cache or memory. The time needed to fetch the operands depends on the speed of the cache or memory. Fig. 1 represents an idealized, cycle by cycle description of an instruction passing through a pipeline. The decode, AGEN, operand fetch (C1,C2), and execution phases are represented. Fig. 2 shows a mechanism that can fetch operands for instructions with constant operand addresses before the instructions are even decoded. The operand history table (O.H.T.) holds the instruction address and operand address of each instruction that computes the same operand address each time it is executed. Each entry in the O.H.T. contains two addresses. The first is the instruction address. The second entry is the address of the operand for that instruction. The mechanism shown in Fig. 2 operates as follows: Each instruction address fetched b...

(Source: IPCOM)
First page image
(Source: IPCOM)