System and method for floating-point computation

• US 6,327,604 B1
• Filed: 05/22/2000
• Issued: 12/04/2001
• Est. Priority Date: 12/09/1997
• Status: Expired due to Term

• Current Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Steele, Guy L. Jr.
• Primary Examiner(s): Mai, Tan V.
• Application Number:
  US09/576,310
• Time in Patent Office: 561 Days
• Field of Search: 708/204-205, 708/495-497, 708/510
• US Class Current: 708/495
• CPC Class Codes: G06F 7/483 Computations with numbers represented by a non-linear combination of denominational numbers, e.g. rational numbers, logarithmic number system or floating-point numbers G06F7/4806, G06F7/4824, G06F7/49, G06F7/491, G06F7/544 take precedence
• CPC Class Codes: G06F 7/49936 Normalisation mentioned as feature only

Abstract

A system is disclosed for performing floating point computation in connection with numbers in a base floating point representation (such as the representation defined in IEEE Std. 754) that defines a plurality of formats, including a normalized format and a de-normalized format, using a common floating point representation that defines a unitary normalized format. The system includes a base to common representation converter, a processor and a common to base representation converter. The base to common representation converter converts numbers from the base floating point representation to the common floating point representation, so that all numbers involved in a computation will be expressed in the unitary normalized format. The processor is configured to perform a mathematical operation of at least one predetermined type in connection with the converted numbers generated by the base to common representation converter to generate a floating point result in the common representation. The common to base representation converter converts numbers from the common floating point representation selectively to either the normalized or de-normalized format of the base representation.

Claims

1. A common to base representation converter for converting floating point numbers to a base floating point representation that defines a plurality of formats, including a normalized format and a de-normalized format, from a common representation defining a unitary normalized format, the unitary normalized format being used both for numbers that, in the base floating point representation, were in the normalized format and numbers that, in the base floating point representation, were in the de-normalized format, each floating point number in said base floating point representation and said common floating point representation including an exponent field and a fraction field, the common to base representation converter comprising:

A. an exponent processor configured to generate from the common floating point representation exponent field the base floating point representation exponent field and fraction de-normalization information; and

B. a fraction processor configured to selectively generate from the common floating point representation fraction field and the fraction de-normalization information a base floating point representation fraction field in one of the normalized format or the de-normalized format.

2. A common to base representation converter as defined in claim 1 in which, for each number in the common floating point representation, the common floating point representation fraction field comprises a predetermined number of digits extending from a most significant digit position to a least significant digit position, the fraction processor being configured to generate the base floating point representation fraction field by shifting the digits of the fraction field of the base common floating point representation fraction field toward less significant digit locations, the number of digit locations being selected in response to the fraction de-normalization information.

3. A method of performing floating point computation in connection with numbers in a base floating point representation that defines a plurality of formats, including a normalized format and a de-normalized format, the method comprising the steps of:

A. converting said numbers to a common floating point representation defining a unitary normalized format, thereby generating converted numbers; and

B. performing a mathematical operation of at least one predetermined type in connection with the converted numbers.

4. A method as defined in claim 3 in which the base floating point representation is defined by an Institute of Electrical And Electronic Engineers Standard 754 (1985).

5. A method as defined in claim 3 in which each floating point number in said base floating point representation and said common floating point representation includes an exponent field and a fraction field, said base to common representation conversion step comprising the steps of:

A. generating from the base floating point representation fraction field a normalized common floating point representation fraction field and fraction normalization information in connection therewith; and

B. generating the common floating point representation exponent field from the base floating point representation exponent field and the fraction normalization information.

6. A method as defined in claim 3 further including a common to base representation conversion step for converting numbers in the common floating point representation selectively to the normalized format or the de-normalized format of said base floating point representation.

7. A method as defined in claim 6 in which each floating point number in said base floating point representation and said common floating point representation includes an exponent field and a fraction field, the base to common representation conversion step comprising the steps of:

A. generating from the common floating point representation exponent field from the base floating point representation exponent field and fraction normalization information; and

B. selectively generating from the common floating point representation fraction field and the fraction normalization information a base floating point representation fraction field in one of the normalized format or the de-normalized format.

8. A method as defined in claim 3 in which said processing step comprises the steps of:

A. performing at least one type of arithmetic operation in connection with the converted numbers provided thereto, thereby to generate processed converted numbers; and

B. generating from the processed converted numbers respective rounded result values.

9. A method as defined in claim 8 in which each said processed converted number includes an exponent field having an exponent value and a fraction field having a fraction value, the fraction value including a series of digits in a series of digit positions, the rounding step including the step of selecting one of said digit positions at which rounding is to occur for a respective one of said processed converted numbers from the exponent value of the respective one of said processed converted numbers.

10. A method of converting floating point numbers to a base floating point representation that defines a plurality of formats, including a normalized format and a denormalized format, from a common representation defining a unitary normalize format, the unitary normalized format being used both for numbers that, in the base floating point representation, were in the normalized format and numbers that, in the base floating point representation, were in the de-normalized format, each floating point number in said base floating point representation and said common floating point representation including an exponent field and a fraction field, the method comprising the steps of:

A. generating from the common floating point representation exponent field the base floating point representation exponent field and fraction de-normalization information; and

B. selectively generating from the common floating point representation fraction field and the fraction de-normalization information a base floating point representation fraction field in one of the normalized format or the de-normalized format.

11. A method as defined in claim 10 in which, for each number in the common floating point representation, the common floating point representation fraction field comprises a predetermined number of digits extending from a most significant digit position to a least significant digit position, the fraction processing step including the step of generating the base floating point representation fraction field by shifting the digits of the fraction field of the base common floating point representation fraction field toward less significant digit locations, the number of digit locations being selected in response to the fraction de-normalization information.