Browse Prior Art Database

Shift Count Adjustment Logic

IP.com Disclosure Number: IPCOM000108707D
Original Publication Date: 1992-Jun-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 2 page(s) / 60K

Publishing Venue

IBM

Related People

Cocanougher, D: AUTHOR [+2]

Abstract

The shift-count adjustment logic (SCAL) allows an unnormalized, unrounded exponent result to be used as an operand for data dependent instructions. (A data dependent instruction is a current instruction which uses for one of its operands data from the results of a previous instruction.) The SCAL eliminates the need for the unnormalized, unrounded exponent result to be first adjusted (normalized and rounded) in the write cycle, which in essence reduces a 'three-cycle' data dependent operation to a 'two-cycle' data dependent operation.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 82% of the total text.

Shift Count Adjustment Logic

       The shift-count adjustment logic (SCAL) allows an
unnormalized, unrounded exponent result to be used as an operand for
data dependent instructions.  (A data dependent instruction is a
current instruction which uses for one of its operands data from the
results of a previous instruction.)  The SCAL eliminates the need for
the unnormalized, unrounded exponent result to be first adjusted
(normalized and rounded) in the write cycle, which in essence reduces
a 'three-cycle' data dependent operation to a 'two-cycle' data
dependent operation.

      The figure illustrates a general block diagram of the exponent
subunit.  Instructions using the unnormalized, unrounded data from a
previous instruction select the data dependency path which is one of
the two inputs to the B and C data dependency 2:1 multiplexers
(muxes).

      Simultaneously, the SCAL interprets the LZA/LZD normalization/
rounding adjustment signals and generates the correct (adjusted)
output.  The 3:1 mux in the shift-count leg and the 2:1 mux in the
exponent result leg selects the correct adjustment depending on which
operands are data dependent.  The shift-count and exponent result are
now correctly adjusted.

      The 'two-cycle' data dependency completely eliminates the use
of the 'write cycle' by incorporating the data dependency path and
the SCAL as part of the multiply cycle. In essence, the 'write cycle'
and the 'multiply cycle' for all data dependen...