Graphics processing unit and graphics processing system

Abstract

A device for controlling the table capacity smaller comprises a table of logarithms and a table of exponents preserving values of a logarithmic function and exponential function with a base of the second power, a multiplier, a shift unit for shifting the input value by a proper integer when the domain of logarithm is not included in the input value range of the table of logarithms, a logarithm addition unit for adding the shift amount to the value referred to in the table of logarithms, an exponent subtraction unit for subtracting a proper integer L from the input value when the domain of the exponential function is not included in the input value range of the table of exponents, and an exponent shift unit for shifting the referred value in the table of exponents the subtraction amount. A high-speed processing unit comprises a converter for converting floating-point data to fixed-point data having a bit of decimal point. By doing this, a structured graphics system can perform high-speed processing with a small table capacity.

Description

[0001] This application is a continuation of application Ser. No. 10 / 391,235, filed Mar. 19, 2003, which is a continuation of application Ser. No. 09 / 521,597 filed Mar. 9, 2000, which is a divisional of application Ser. No. 09 / 009,588 filed Jan. 20, 1998, the content of which is incorporated herein by reference in its entirety.BACKGROUND OF THE INVENTION [0002] Conventionally, with respect to the multiplication for obtaining XN for two digital numerical data X and N, as described in Haruhiko Okumura, “Latest Algorithm Dictionary in C language”, p. 105 and 106, p. 162 and 163, and p. 304 published from Gijutsu Hyoronsha, Feb. 25, 1991, a method that a technique of obtaining a logarithm and an exponent by loop calculation by performing expansion into power series or expansion into continued fraction for the logarithmic function and exponential function is used, and the logarithm of X is calculated on a software basis, and the result is multiplied by N, and finally the exponent of the ...

AI Technical Summary

Benefits of technology

[0014] An object of the present invention is to provide a device for using no loop calculation, referring to a small-capacity table, and performing power calculation at high speed.

[0015] To accomplish this object, the present invention provides a logarithm calculation unit for calculating and outputting a logarithmic value of X for an input value of X by referring to a table of logarithms, a multiplier for multiplying an output of the logarithm calculation unit by another value of N, and an exponent calculation unit for calculating and outputting an exponential value for an output of the multiplier by referring to a table of exponents and so as to reduce the table capacity more, the present invention provides a logarithm shift unit for multiplying an input of the logarithm calculation unit by 2L for an appropriate integer L when the aforementioned input value is not included in the input value range of the table of logarithms and a logarithm adding unit for setting the multiplication result as input of the table of logarithms and referring to the table of logarithms and then adding L to the referred value and setting it as output of the logarithm calculation unit in the logarithm calculation unit and provides an exponent subtraction unit for subtracting an appropriates integer M from, input of the exponent calculation unit when the aforementioned input value is not included in the input value range of the table of exponents and an exponent shift unit for setting the subtraction result as input of the table of exponents and referring to the table of exponents and then multiplying the referred value by 2−M and setting it as output of the exponent calculation unit in the exponent calculation unit.

[0017] Namely, according to the present invention, when power calculation of XN is to be performed for input values of X and N, a logarithm logaX with a base of “a” is obtained by referring to the table of logarithms, and logaX×N is calculated by the multiplier, and alogaX×N=XN which is the logaX×N power of a is calculated by referring to the table of exponents. Since the present invention does not perform loop calculation, high speed calculation is possible. By dividing the table into two parts such as the table of logarithms and table of exponents, inputs of the tables can be unified to one and the capacity of each table can be reduced.

[0019] By doing this, even if input values of the logarithm calculation unit and exponent calculation unit are not included in the input value ranges of the tables of logarithms and exponents, power calculation can be performed. Therefore, the input value range of the tables of logarithms and exponents can be limited and the table capacity can be reduced.

[0020] Another object of the present invention is to provide a graphics processing unit for decreasing the reduction in the processing speed necessary to transform a floating-point number to a fixed-point number.

[0022] According to this graphics processing unit of the present invention, the data converter for converting floating-point data to fixed-point data is installed between the processor and the rendering unit, so that the processing of the processor and the transformation from floating-point data to fixed-point data can be performed on a pipeline basis and high-speed graphics processing can be performed.

Problems solved by technology

As mentioned above, in the conventional examples, a problem arises that in the former case, loop calculation is generated and speeding-up of the process is difficult and in the latter case, since the two data X and N are inputted to the table, the gradient of the input values to the table is the product of the gradation number of X and the gradation number of N, accordingly the table capacity is increased.

The reason is that it is difficult to perform operations to be performed by the rendering unit by a floating-point number from a viewpoint of circuit scale and operation speed.

Since the geometry unit transforms from a floating-point number to a fixed-point number as mentioned above, the burden of the geometry unit is increased and since the computing element of the geometry unit is used for transformation from a floating-point number to a fixed-point number during this period, the geometry unit cannot perform the intrinsic process such as coordinate transformation and the throughput of the whole graphics processing unit reduces.

Images