Floating-point multiplicator and method of compatible double-prepcision and double-single precision computing

A floating-point multiplier and single-precision technology, which is applied to calculations using number system representations and calculations using non-contact manufacturing equipment, can solve problems such as reducing the execution time of double-precision operations and hardware waste

Active Publication Date: 2006-09-13
LOONGSON TECH CORP
View PDF1 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Either in order to improve the performance of double-precision multiplication, use two 53-bit × 27-bit multiplication trees to compress double-precision and reduce the execution time of double-precision operations, but this structure only uses one 53-bit for single-precision multiplication × 27-bit compression tree, another 53-bit × 27-bit compression tree is in an idle state, and the hardware is wasted at this time (see Chinese patent 01141498.7)

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Floating-point multiplicator and method of compatible double-prepcision and double-single precision computing
  • Floating-point multiplicator and method of compatible double-prepcision and double-single precision computing
  • Floating-point multiplicator and method of compatible double-prepcision and double-single precision computing

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0081] Example 1: Floating-point double-precision multiplication (expressed in hexadecimal) 0x3ff0000000000000×0x3fffffffffffffff.

[0082] 1. After the 53-bit mantissa passes through the first selector 2, it outputs 0x10000000000000.

[0083] 2. First use the base-2 Potts encoder 3 to encode the 53-bit mantissa of the multiplier 0x3ff0000000000000 (implicit 1 is added in front of the 52 bits) 0x10000000000000 using the above encoding table to be encoded as {1, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0 , -0, -0, -0, -0, -0, -0, -0}, and then multiply the encoded result with the mantissa 0x1fffffffffffff (53 bits) of the multiplicand 0x3fffffffffffffff to obtain 27 partial products, 0 ~26 partial products are all 0, and the 27th partial product is 0x1fffffffffffff.

[0084] 3. The 0th to 12th partial products are sent to the 13:2 compression tree 4 for compression, the 13th to 26th partial products are sent to the 14:2 compression tree 5 f...

example 2

[0090] Example 2: Floating-point double-single-precision multiplication (expressed in hexadecimal) 0x3f8000003f800000×0x3fffffff3fffffff.

[0091] 1. The mantissa of the double-single-precision multiplier 0x3f8000003f800000 is filled into 53 bits 0x02000000800000 through the double-single-precision continuous encoding filling unit 1 .

[0092] 2. Through the first selector 2, select and output 53 bits 0x02000000800000.

[0093] 3. Using base 2 Potts encoder 3, the encoding becomes {0, 1, -2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, -2, 0 ,0,0,0,0,0,0,0,0,0,0}. Multiply the coded result with the two mantissas of the multiplicand 0x3fffffff 3fffffff to get 27 partial products, the 0th to 12th partial products are the multiplication results of the first single precision, and the 13th to 27th are the second single precision the result of the multiplication.

[0094] 4. For the first single precision, send 0 to 12 partial products to the 13:2 compression tree 4 for compression to obt...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

A floating-point multiplier consists of single accuracy continuous coding - packing unit, the first selector, base 2 Pots coder, 13 : 2 compression tree, 14 : 2 compression tree, 48 bit adder, 4 : 2 compression composite tree, the second selector, 106 bit adder, 48 bit single accuracy round - off and normalization unit, 106 bit double accuracy round - off and normalization unit, double accuracy and double accuracy index processing unit . The said floating-point multiplier can multiplex double accuracy data channel for carrying out calculation to obtain two single accuracy multiplication results of floating-point at each beat.

Description

technical field [0001] The invention relates to a floating-point multiplier in a microprocessor, in particular to a floating-point multiplier and a calculation method compatible with double precision and double single precision. The present invention can make the 64-bit floating-point multiplier able to execute double-precision floating-point multiplication instructions and double single-precision (MUL.PS in the MIPS V instruction set, the suffix PS is the abbreviation of pairedsingle, see MIPS V instruction set, MIPS is Microprocessor withoutInterlocked Piped Stages for short) floating-point multiplication instructions. Background technique [0002] With the popularization of scientific computing and multimedia, data calculation requires higher precision. In order to improve the performance of the multiplier, the current microprocessor structure uses hardware to support the double-precision and single-precision formats in the IEEE754 standard. Point multiplier (Intelx86 im...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F7/52
Inventor 齐子初张戈胡伟武
Owner LOONGSON TECH CORP
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap