Computer and methods for solving math functions

0Associated
Cases 
0Associated
Defendants 
0Accused
Products 
2Forward
Citations 
0
Petitions 
1
Assignment
First Claim
1. A computer comprising:
 an operations level that performs mathematical operations; and
a driver level comprising;
a first lookup table, wherein the first lookup table includes first data for calculating at least one mathematical function using a first level of accuracy; and
a second lookup table, wherein the second lookup table includes second data for calculating the at least one mathematical function using a second level of accuracy, wherein the first level of accuracy is greater than the second level of accuracy; and
a driver that executes the data from the first lookup table and the second lookup table, and wherein the driver sends instructions to the operations level;
wherein executing data in the second lookup table for solving the at least one function uses less energy than executing data in the first lookup table for solving the at least one function.
1 Assignment
0 Petitions
Accused Products
Abstract
Computers and methods for performing mathematical functions are disclosed. An embodiment of a computer includes an operations level and a driver level. The operations level performs mathematical operations. The driver level includes a first lookup table and a second lookup table, wherein the first lookup table includes first data for calculating at least one mathematical function using a first level of accuracy. The second lookup table includes second data for calculating the at least one mathematical function using a second level of accuracy, wherein the first level of accuracy is greater than the second level of accuracy. A driver executes either the first data or the second data depending on a selected level of accuracy.
5 Citations
View as Search Results
Processor for realizing at least two categories of functions  
Patent #
US 10,372,359 B2
Filed 05/10/2017

Current Assignee
ChengDu HaiCun IP Technology LLC

Sponsoring Entity
ChengDu HaiCun IP Technology LLC

Configurable processor with inpackage lookup table  
Patent #
US 10,445,067 B2
Filed 11/28/2018

Current Assignee
Hangzhou Haicun Information Technology Co. Ltd.

Sponsoring Entity
Hangzhou Haicun Information Technology Co. Ltd.

PROCESSING UNIT  
Patent #
US 20100332573A1
Filed 06/29/2010

Current Assignee
Fujitsu Limited

Sponsoring Entity
Fujitsu Limited

PROCESSING APPARATUS AND CONTROL METHOD THEREOF  
Patent #
US 20080228846A1
Filed 03/13/2008

Current Assignee
Fujitsu Limited

Sponsoring Entity
Fujitsu Limited

MATH CIRCUIT FOR ESTIMATING A TRANSCENDENTAL FUNCTION  
Patent #
US 20140222883A1
Filed 12/21/2011

Current Assignee
Intel Corporation

Sponsoring Entity
Intel Corporation

3 Claims
 1. A computer comprising:
an operations level that performs mathematical operations; and a driver level comprising; a first lookup table, wherein the first lookup table includes first data for calculating at least one mathematical function using a first level of accuracy; and a second lookup table, wherein the second lookup table includes second data for calculating the at least one mathematical function using a second level of accuracy, wherein the first level of accuracy is greater than the second level of accuracy; and a driver that executes the data from the first lookup table and the second lookup table, and wherein the driver sends instructions to the operations level; wherein executing data in the second lookup table for solving the at least one function uses less energy than executing data in the first lookup table for solving the at least one function.
 2. A computer for performing mathematical functions, the computer comprising:
an operations level; a driver level; and an instruction level; the operations level performing basic math operations; the driver level including; a first lookup table, wherein the first lookup table includes first data for calculating at least one mathematical function using a first level of accuracy; and a second lookup table, wherein the second lookup table includes second data for calculating the at least one mathematical function using a second level of accuracy, wherein the first level of accuracy is greater than the second level of accuracy; and a driver that executes the data from the first lookup table and the second lookup table, and wherein the driver sends instructions to the operation level and receives instructions from the instruction level; the instruction level including; a memory that stores instructions for solving complex functions; a controller that executes the instructions stored in the memory and transmits instructions to the driver; wherein the controller receives instructions indicating whether high accuracy or low energy is to be used in solving the function and wherein the controller instructs the driver to access the first lookup table when the function is to be solved using high accuracy and wherein the controller instructs the driver to access the second lookup table when the function is to be solved using low energy.
 3. A method for reducing energy consumption of a computer to solve math functions, the method comprising:
receiving a math function to be solved; receiving information as to whether the math function is to be solved using high energy or low energy; executing instructions for solving the math function using at least one derived math function; executing the derived math functions using data stored in a first lookup table if the math function is to be solved using high energy; and executing the derived math functions using data stored in a second lookup table if the math function is to be solved using low energy.
1 Specification
Some computers are used to perform complex math functions, such as cubed root and hyperbolic trigonometric functions. These complex math functions require many cycles to complete, which is time consuming and uses a lot of energy. The long time required to calculate complex functions creates a problem when the computer needs to calculate many complex functions in a short period. For example, if the computer is used to guide an aircraft, the computer may need to calculate pitch, yaw, and roll from a plurality of sensors using complex functions.
The high energy demand required to calculate complex functions becomes an issue with computers that are operating from limited power supplies, such as batteries. The calculation of the complex functions requires many cycles, which draws heavily on the batteries. In addition, the high energy demand creates heat, which may interfere with the operation of the computer.
Computers and methods for performing mathematical functions are disclosed. An embodiment of a computer includes an operations level and a driver level. The operations level performs mathematical operations. The driver level includes a first lookup table and a second lookup table, wherein the first lookup table includes first data for calculating at least one mathematical function using a first level of accuracy. The second lookup table includes second data for calculating the at least one mathematical function using a second level of accuracy, wherein the first level of accuracy is greater than the second level of accuracy. A driver executes the data from the first lookup table or the second lookup table and sends instructions to the operations level.
Computer architectures, sometimes referred to herein simply as computers, and methods of calculating mathematical functions are described herein. The computers and methods provide for calculating mathematical functions using different options. In the embodiments described herein, a first option calculates complex functions using high accuracy. In this option, many instructions have to be performed to calculate a single complex function, which uses a lot of energy. A second option calculates complex functions using lower accuracy. This second option does not require as many instructions as the high accuracy option, so the complex function can be solved faster, using uses less energy. High accuracy calculations require more instructions and therefore typically require more time to complete than low accuracy calculations. Low accuracy calculations use less energy than high accuracy calculations and may be used in situations where energy preservation is important, such as when the energy comes from a batterypowered source. These situations include situations where a lot of calculations are required on a batterypowered device. The low accuracy option uses less energy than the high accuracy option, so there is less drain on the battery.
In some embodiments, complex functions are broken down into instructions that use simpler functions. For example, the calculation of a hyperbolic sine may be performed by solving exponent functions and division and subtraction operations. In the embodiments described herein, the solutions to the exponent functions are stored in lookup tables. Thus, the computers do not need to calculate the hyperbolic sine function directly. Rather, the computers only need to solve the exponent functions and the subtraction and division operations, which are less time consuming and energy consuming than direct calculations of hyperbolic sine.
Reference is made to
Examples of the complex mathematical functions include hyperbolic trigonometric functions, cubed root, and some power functions, such as 2^{x}. The instruction level 102 includes a main controller 110, which is sometimes referred to simply as a controller 110, and memory 112. The controller 110 has an input 114 that receives the abovedescribed functions and numbers. The controller 110 also has an output 116 that outputs the result of the calculation performed by the computer 100. The instruction level 102 may have different embodiments. For example, in some embodiments, the instruction level 102 is or includes an application specific instruction processor (ASIP).
The controller 110 has, or has access to, memory 112 by way of a port 113 on the controller 110 and a port 115 on the memory 112. The memory 112 stores code for solving the complex functions. In some embodiments, the controller 110 calls the memory 112 for instructions to solve a complex function, such as cosh(x). The memory 112 has the instructions stored for solving cosh(x), which, in the embodiments described herein is shown by equation (1) as follows:
The instruction level 102 communicates with the driver level 104 by way of ports 118 and 119 to solve some functions. In the example of equation (1), the driver level solves the exponent (exp) functions by using lookup tables as described in greater detail below. In some embodiments, the driver level 104 is a state machine. The driver level 104 includes a driver 120 and lookup tables 122. The driver 120 communicates with the controller 110 to receive the instructions from the memory 112 when the computer 100 is solving complex functions, such as equation (1). In some situations, simpler functions are solved that do not require the instructions stored in the memory 112, however, the driver 120 may still receive the instructions for solving these simpler functions from the controller 110.
The driver 120 communicates with the lookup tables 122. There are a plurality of lookup tables 122 that contain information or instructions for different baseline math functions as described below. In the embodiment of
As an example of the different lookup tables, the exponential (exp) function is used to solve the cosh(x) function using a fifteenth order Taylor series in the high accuracy solution and a third order polynomial curve fitting in the low accuracy solution. Polynomial curve fitting is sometimes referred to as polyfit. In the embodiment described herein, the relatively simple baseline functions accessed by the driver 120 and stored in the lookup tables 122 are shown in
The operations level 106 is sometimes referred to as being at the bottom level of the computer 100. In some embodiments, the operations level 106 is a hardware accelerator that performs basic floatingpoint operations such as addition, subtraction, and multiplication. The operations level 106 includes an operation module 130 that performs the operations. The operation module 130 communicates with the driver 120 by way of ports 132 and 134. The operation module 130 also communicates with the controller 110 by way of ports 136 and 138. In the embodiment of
As stated above, the operations level 106 performs low level operations using the operation module 130. In the embodiments described herein, the operation module 130 performs the following operations:
 floatingpoint addition and subtraction;
 fixedpoint addition and subtraction;
 floatingpoint multiplication;
 floatingpoint comparison;
 floatingpoint to integer casting, which is used to zeroout the fractional part in a floating point number;
 absolute (x) for a floating point number;
 shifting an operand to the right by 1 bit, which is required for inverse square root and reciprocal operations;
 addition of exponents of two floating point numbers, which is required for exp operations;
 masking exponents to 127, which is required for natural log operations;
 converting exponents to floating point numbers, which is required for natural log operations;
 negating computation outputs; and
 rounding results in internal format (29 bit mantissa) and standard format (23 bit mantissa).
Based on the functions that can be performed in the driver level 104 and the operations that can be performed in the operations level 106, the computer 100 is capable of solving a plurality of complex functions. An example embodiment of the complex functions that are performed by the computer 100 is shown by the table in
Having described the components of the computer 100, its operation will now be described using several different examples. Reference is made to
The operation module 130 executes the instruction by performing the operation. In this embodiment, two numbers are added together. As shown in
A second embodiment of the operation of the computer 100 is shown in
The controller 110 analyzes the instructions and determines that the function is performed at the driver level 104. Instructions are then sent to the driver 120 to perform the sine function. The instructions also indicate whether high accuracy or low accuracy is to be used. The driver 120 analyses the instructions and accesses the proper lookup table. If high accuracy is required, the LUT 124 is addressed and the sin(x) function is calculated using a sixteenth order Taylor series. If low accuracy or low energy is required, the LUT 126 is addressed and the sin(x) function is solved using the third order polyfit. In using either the Taylor series or the polyfit, the driver 120 needs to access operations, such as addition, in the operation module 130. In some embodiments, the driver 120 transmits values to the operation module 130 to be added with instructions that the sum is to be transmitted to the controller 110. The driver 120 also sends a signal to the controller 110 indicating that the result of the sin(x) function will be output by the operation module 130. When the result is received from the operation module 130, it is output by way of the output 116 of the controller 110.
In a third example, which is related to
The operation of the computer 100 solving a function is shown by the flowchart 200 of
The computer 100 provides for high accuracy and low accuracy solutions to complex math functions. In some embodiments, such as vehicle sensors, high accuracy calculations may be required. The LUT 124 stores the information necessary for the high accuracy calculations. Other embodiments do not need the high accuracy and may need to conserve energy. For example, some battery operated devices may operate using the low accuracy information in the LUT 126, which conserves energy.
While illustrative and presently preferred embodiments of harvesting devices have been described in detail herein, it is to be understood that the inventive concepts may be otherwise variously embodied and employed and that the appended claims are intended to be construed to include such variations except insofar as limited by the prior art.