Apparatuses and related methods for overflow detection and clamping with parallel operand processing

Patent metadata

Field	Value
Publication number	US-9256577-B2
Application number	US-201313758577-A
Country	US
Kind code	B2
Filing date	Feb 4, 2013
Priority date	Aug 30, 2000
Publication date	Feb 9, 2016
Grant date	Feb 9, 2016

Abstract

A method and apparatus for overflow detection and clamping with parallel operand processing for fixed-point multipliers is disclosed. The disclosure predicts when a multiplication of a number of operands will exceed a pre-determined number of bits based upon the fixed-point format of the operands. The prediction is performed in parallel with the multiplication of the operands. The multiplication need not be completed in full, but only to the extent to determine whether overflow exists. If an overflow detection occurs, clamping is instituted. The parallel operation of the overflow detection and the multiplication provides a faster clamping circuit than would otherwise be available from a serial multiplication followed by a clamping analysis.

First claim

What is claimed is:

1. An apparatus, comprising:
multiplication circuitry configured to generate a multiplication result by at least partially multiplying a first operand and a second operand;
overflow detection circuitry configured to generate a clamp bit indicative of a prediction that the multiplication result will exceed a maximum representable positive or negative value of an output, wherein the multiplication circuitry and the overflow detection circuitry are configured to operate independently and substantially in parallel with each other; and
selection circuitry operably coupled to the multiplication circuitry and the overflow detection circuitry, the selection circuitry configured to select between the multiplication result and a clamp value as the output generated responsive to whether an overflow situation is determined by logically OR-ing a most significant bit of the multiplication result and the clamp bit.

2. The apparatus of claim 1 , wherein the selection circuitry includes a multiplexor configured to select between the multiplication result and the clamp value based on a clamping signal received by the multiplexor.

3. The apparatus of claim 2 , wherein the selection circuitry further includes an OR-gate operably coupled to the multiplexor, the overflow detection circuitry, and the multiplication circuitry, wherein the OR-gate is configured to generate the clamping signal responsive to perform the logically OR-ing of the clamp bit and the most significant bit of the multiplication result.

4. The apparatus of claim 2 , wherein the clamp value is hardwired into the multiplexor.

5. The apparatus of claim 1 , wherein the multiplication result is a partial multiplication result that is less than full precision.

6. The apparatus of claim 5 , wherein the multiplication result has a precision that is two times a desired range for the output.

7. The apparatus of claim 1 , wherein each of the first operand and the second operand are represented as fixed-point numbers having a sign portion, an integer portion, and a fractional portion.

8. The apparatus of claim 7 , wherein the selection circuitry is configured to output the clamp value as either a positive value or a negative value by logically XOR-ing the sign portion of the first operand and the sign portion of the second operand.

9. The apparatus of claim 7 , wherein the prediction is based, at least in part, on the overflow detection circuitry performing a sum of leading zeros in the integer portion of the first operand and leading zeros in the integer portion of the second operand.

10. The apparatus of claim 7 , wherein the prediction is based, at least in part, on the overflow detection circuitry performing a sum of leading ones in the integer portion of the first operand and leading ones in the integer portion of the second operand.

11. The apparatus of claim 7 , wherein:
if both the first operand and the second operand are positive, the prediction is based on the overflow detection circuitry summing leading zeros in the integer portion of the first operand and leading zeros in the integer portion of the second operand;
if both the first operand and the second operand are negative, the prediction is based on the overflow detection circuitry summing leading ones in the integer portion of the first operand and leading ones in the integer portion of the second operand; and
if the first operand and the second operand have different signs, the prediction is based on the overflow detection circuitry summing leading ones in the integer portion of a negative operand and leading zeros in the integer portion of a positive operand.

12. The apparatus of claim 1 , wherein the overflow detection circuitry is configured to assert the clamp bit responsive to determining that both the first operand and the second operand have only zeros after leading ones.

13. The apparatus of claim 1 , further comprising:
a first register operably coupled with the overflow detection circuit and the multiplication circuitry, the first register configured to receive the first operand; and
a second register operably coupled with the overflow detection circuit and the multiplication circuitry, the second register configured to receive the second operand.

14. The apparatus of claim 1 , further comprising:
a clamp bit register coupled between the overflow detection circuitry and the selection circuitry, wherein the clamp bit register is configured to store the clamp bit; and
a multiplier result register coupled between the multiplication circuitry and the selection circuitry, wherein the multiplier result register is configured to store the multiplication result.

15. The apparatus of claim 1 , further comprising a clamping system that includes the multiplication circuitry, the overflow detection circuitry, and the multiplexor.

16. The apparatus of claim 1 , wherein the clamp value is equal to the maximum representable positive or negative value of the output.

17. A method of processing operands in an arithmetic operation, the method comprising:
generating a product by performing at least a partial multiplication of a plurality of operands;
predicting overflow of the product to generate a clamp bit independently and substantially in parallel with performing the at least a partial multiplication; and
generating a final result by selecting between the product and an overflow value responsive to logically OR-ing a most significant bit of the product and the clamp bit.

18. The method of claim 17 , further comprising receiving the plurality of operands in multiplication circuitry and overflow prediction circuitry, the plurality of operands having a fixed-point format.

19. The method of claim 18 , wherein generating a product includes generating a product to have a number of integer bits that is at least one bit greater than a number of integer bits of the final result.

20. The method of claim 17 , wherein predicting overflow of the product includes overflow detection circuitry asserting the clamp bit if overflow of the product is predicted.

21. The method of claim 20 , wherein selecting between the product and the overflow value responsive to logically OR-ing includes:
selecting the product if neither the clamp bit nor the most significant bit of the product is asserted; and
selecting the overflow value if at least one of the clamp bit and the most significant bit of the product is asserted.

22. The method of claim 20 , wherein selecting between the product and the overflow value responsive to logically OR-ing includes:
generating a clamping signal by logically OR-ing the clamp bit and the most significant bit of the product; and
inputting the clamping signal into a multiplexor that selects either the product and the overflow value based on a logic value of the clamping signal.

23. A method of processing operands in an arithmetic operation, the method comprising:
generating a product by performing at least a partial multiplication of a plurality of operands;
predicting an overflow prediction of the product independently and substantially in parallel with performing the at least a partial multiplication; and
clamping a final result to an overflow value if an overflow state is determined by logically OR-ing a most significant bit of the product and a clamp bit generated from the overflow prediction.

24. The method of claim 23 , wherein clamping a final result to the overflow value includes selection

Assignees

• Micron Technology Inc

Inventors

• Bentz Ole

Classifications

• G06F7/49921 Primary

  Saturation, i.e. clipping the result to a minimum or maximum value · CPC title

• G06F17/10 Primary

  Complex mathematical operations {(function generation by table look-up G06F1/03; evaluation of elementary functions by calculation G06F7/544)} · CPC title