Method and apparatus for implementing adaptive forward differencing using integer arithmetic

An adaptive forward differencing apparatus for generating curves for display on a computer graphic display device, said apparatus comprising a central processing unit (CPU), memory, and input/out means comprising the computer graphics display device, said computer graphics display device comprising a matrix of pixels, said curves displayed on the computer graphics display device by actuating certain of the pixels, said certain pixels to be actuated being identified by pixel data comprising sets of coordinate values which identify locations of pixels in the matrix, said apparatus further comprising:

• means for receiving a plurality of data values representative of said pixels to be activated to generate the curve;
• interpreting means connected to the receiving means for interpreting said data values and representing said curve as a set of parametric functions, each parametric function representative of a coordinate value of the set of coordinate values;
• forward difference means connected to the interpreting means to calculate the output of the parametric functions for a range of parameter values, said output of the forward difference means comprising a plurality of results, said results comprising sets of coordinate values which identify locations of pixels in the matrix, said results calculated on an incremental basis wherein a parameter value is equal to the sum of a prior parameter value, said prior parameter value being the parameter value used to calculate a prior adjacent set of coordinate values, and the parametric step size comprising;
• translating means for translating each parametric function according to the parameter value, said parameter value equal to the sum of the parametric step size and the prior parameter value, into a forward difference basis comprising forward difference coefficients, each forward difference coefficient corresponding to a different order derivative of each parametric function such that a first forward difference coefficient corresponds to the value of the parametric function, a parametric function such that a first forward difference coefficient corresponds to the value of the parametric function, a second forward difference coefficient corresponds to a first order derivative of the parametric function and succeeding forward difference coefficients respectively correspond to succeeding order derivatives of the parametric function;
• a plurality of registers connected to the translating means for storing each of the forward difference coefficients of each parametric function;
• register control means for controlling the format of the forward difference coefficients stored in the plurality of registers, the first forward difference coefficient of each parametric function being stored in a first format having a fixed predetermined number of fractional bits, and succeeding forward difference coefficients being stored in formats having a number of fractional bits equal to the fixed predetermined number of fractional bits plus a number of guard bits, said number of guard bits being equal to a tessellation number multiplied by a multiplication factor, said multiplication factor having a value zero for the first forward difference coefficient and incremented by one for each succeeding forward difference coefficient, the highest order coefficient retaining the same multiplication factor as the preceding lower order coefficient;
• calculating means connected to the plurality of registers for calculating a result of each parametric function for a predetermined parameter value using the forward difference basis comprising the forward difference coefficients, each result representing a coordinate value of the set of coordinate values;
• adjusting means for modifying the parametric step size;
• a frame buffer for receiving and storing pixel data;
• and
• a pixel filter connected to the forward difference means and the frame buffer comprising an input means connected to the forward difference means to receive a set of coordinate values output by the forward difference means, a first output means connected to the forward difference means to input control signals to control the parametric step size and the set of coordinate values computed, and a second output means for outputing pixel data to the frame buffer, said pixel filter comprising;
• means for comparing a first set of coordinate values to an adjacent second set of coordinate values, said first set of coordinate values being the prior adjacent set of coordinate values output to the frame buffer;
• if the value of the difference between the first set of coordinate values and the second set of coordinate values is greater than a range of difference values acceptable as a uniform increment,
• a first control signal, output by the first output means to the forward difference means, to increase the parametric step size and re-compute coordinate values of the second set of coordinates using a parameter value equal to the sum of the parameter value used to compute the first set of coordinate values and the increased parametric step size, said register control means increasing the number of guard bits in the succeeding registers by decreasing the tessellation number by a value of one;
• and
• if the value of the difference between the first set of coordinate values and the adjacent second set of coordinate values is less than the range of difference values accepted as a uniform increment,
• a second control signal, output by said first output means to the forward difference means, to decreas the parametric step size and re-compute coordinate values of the second set of coordinates using a parameter value equal to the sum of the parameter value used to compute the first set of coordinate values and the decreased parametric step size;
• and
• if the value of the difference between the first set of coordinate values and the adjacent second set of coordinate values is within the range of difference values acceptable as a uniform increment, means for performing a forward step operation along the curve comprising,
• said second output means outputing pixel data comprising the second set of coordinate values to the frame buffer;
• and
• a third control signal output by said first output means to the forward difference means to set the prior parameter value to be the parameter value used to compute the second set of coordinate values and to compute coordinate values of a next set of coordinates adjacent to the second set of coordinates using a parameter value equal to the sum of the prior parameter value and the parametric step size, and setting the second set of coordinates to be the first set of coordinates and the next set of coordinates to be the second set of coordinates;
• a graphic display control device connected to the frame buffer and the graphics display device to read the pixel data from the frame buffer and actuate the pixels on the display device at the locations identified by the pixel data;
• wherein the pixels actuated on the computer graphics display device to generate a curve are spaced apart by uniform increments.

In a system for rendering a parametric function, a method for rendering the parametric function using adaptive forward differencing and integer arithmetic, wherein the output values of the parametric function comprise a plurality of results which are within a range of valid output values identified by a predetermined heuristic function, said output values calculated in predetermined increments according to a parameter value, said parameter value incremented by an amount equal to a parametric step size wherein the parameter value is equal to the sum of a prior parameter value used to calculate a prior output value and the parametric step size, said results being calculated in predetermined increments by adjusting the parametric step size, said method comprising the steps of:

• translating the parametric function into a forward difference basis comprising forward difference coefficients, each forward difference coefficient corresponding to a different order derivative of the parametric function such that a first forward difference coefficient corresponds to the value of the parametric function, a second forward difference coefficient corresponds to a first order derivative of the parametric function and succeeding forward difference coefficients correspond to succeeding order derivatives of the parametric function;
• storing the forward difference coefficients in plurality of registers;
• controlling the format of the forward difference coefficients stored in the plurality of registers, the first forward difference coefficient being stored in a first format having a fixed predetermined number of fractional bits and succeeding forward difference coefficients being stored in formats having a number of fractional bits equal to the fixed predetermined number of fractional bits plus a number of guard bits equal to the tessellation number multiplied by a multiplication factor, said multiplication factor being zero for the first forward difference coefficient and incremented by one for each succeeding forward difference coefficient, the highest order coefficient retaining the same multiplication factor value as the preceding lower order coefficient;
• calculating a result of the fixed parametric function for a parameter value using the forward difference basis comprising the forward difference coefficients stored in the plurality of registers;
• comparing the result with the heuristic function to determine if the result is within the range of valid output values;
• if the result is less than the range of valid output values, increasing the parametric step size by decreasing the number of guard bits in the registers by increasing the tessellation number by a predetermined adjustment value and recalculating a result of the parametric function according to an adjusted parameter value equal to the sum of the prior parameter value and the increased parametric step size;
• if the result is greater than the range of value output values, decreasing the parametric step size by increasing the number of guard bits in the registers by decreasing the tessellation number by the predetermined adjustment value and recalculating a result of the parametric function according to an adjusted parameter value equal to the sum of the prior parameter value and the decreased parametric step size;
• if the result is within the range of valid output values, outputting the result as an output value of the parametric function and advancing to the calculation of the next output value by setting the prior parameter value to be equal to the parameter value used to calculate the result output as an output value, updating the parameter value to be equal to the sum of the prior parameter value and the parametric step size and updating the forward difference coefficients stored in the registers according to the updated parameter value;
• whereby the parametric function is rendered for a plurality of parameter values within a range of parameter values and the first coefficient consistently maintains the fixed predetermined number of fractional bits resulting in increased precision of the results calculated.

An apparatus for rendering a third order polynomial parametric function using adaptive forward differencing and integer arithmetic, said third order polynomial parametric function being of the form:

• f(t)=At.sup.3 +Bt.sup.2 +Ct+D, wherein the output of the parametric function rendered comprises a plurality of output values which are rendered in uniform increments, said output values calculated in uniform increments according to a parameter value incremented by an amount equal to a parametric step size, wherein the parameter value is equal to the sum of a prior parameter value used to calculate a prior output value and the parametric step size, said output values being calculated to be rendered in uniform increments by adjusting the parametric step size, said apparatus comprising:
• receiving means for receiving an initial parameter value and parametric step size and the parametric function to be rendered, wherein said initial parameter value is used to initialize the prior parameter value and said initial parametric step size is used to initialize the parametric step size;
• translating means connected to the receiving means for translating the parametric function according to the parameter value, said parameter value equal to the sum of the parametric step size and the prior parameter value, into a forward difference basis comprising forward difference coefficients according to the following equations:
• d=D;
• c=C+((B+(A>>n))>>n);
• b=2B+(6A>>n);
• a=6A>>n;
• wherein each forward difference coefficient corresponds to a different order derivative of the parametric function, d represents a first forward difference coefficient corresponding to the value of the parametric function, c represents the second forward difference coefficient corresponding to the first order derivative of the parametric function, b represents a third forward difference coefficient corresponding to the second order derivative of the parametric function, a represents a fourth forward difference coefficient corresponding to the third order derivative of the parametric function, n represents a tessellation number and >>
• represent a right shift of n bits,
• a plurality of registers connected to the translating means for storing each of the forward difference coefficients;
• register control means for controlling the format of the forward difference coefficients stored in the plurality of registers, the first forward difference coefficient being stored in a first format having a fixed predetermined number of fractional bits, and succeeding forward difference coefficients being, stored in formats having a number of fractional bits equal to the fixed predetermined number of fractional bits plus a number of guard bits equal to the tessellation number multiplied by a multiplication factor, said multiplication having a value of zero for the first forward difference coefficient and incremented by one for each succeeding forward difference coefficient, the highest order coefficient retaining the same multiplication factor as the preceding lower order coefficient, whereby the first forward difference coefficient is stored having 16 fractional bits, the second forward difference coefficient is stored having 16+n fractional bits, where n is the tessellation number, and the third and fourth forward difference coefficients are stored having 16+2n fractional bits;
• calculating means connected to the plurality of registers for calculating a result of the parametric function for the parameter value using the forward difference basis comprising the forward difference coefficients;
• a comparator connected to the calculating means for comparing a prior output value to the result output by the calculating means;
• if the value of the difference between the prior output value and the result is greater than a range of difference values acceptable as a uniform increment,
• a first control signal, output by the first output means to the translating means and register control means, to increase the parametric step size and re-compute the result using a parameter value equal to the sum of the prior parameter value and the increased parametric step size, by respectively updating the forward difference coefficients and format of the forward difference coefficients stored in the plurality of registers according to the following equations:
• a'=a>>1;
• b'=b-a';
• c'=c-(b>>(n+2));
• d'=d;
• and
• n'=n+1
• wherein d' represents the updated first forward difference coefficient, c', b' and a respectively represent the updated second, third and fourth forward difference coefficients, and n' represents the updated tessellation number, and the number of guard bits in the second forward difference register is increased by one, in the third forward difference register by two and in the fourth forward difference register by two;
• if the value of the difference between the prior output value and the result is less than the range of difference values accepted as a uniform increment,
• a second control signal, output by said first output means to the translating means and register control means, to decrease the parametric step size and re-compute coordinate values of the second set of coordinates using a parameter value equal to the sum of the prior parameter value and the decreased parametric step size, by respectively updating the forward difference coefficients and format of the forward difference coefficients stored in the plurality of registers according to the following equations:
• d'=d;
• c'=c+(b>>(n+1));
• b'=b+a;
• a'=a<<1;
• and
• n'=n-1
• whereby the number of guard bits in the second forward difference register is decreased by one, in the third forward difference register by two and in the fourth forward difference register by two;
• if the value of the difference between the prior output value and the result is within the range of difference values acceptable as a uniform increment, means for advancing to the calculation of the next output value by performing a forward step operation comprising,
• said second output means outputing the result as an output value;
• and
• a third control signal output by said first output means to the translating means, register control means and comparator to set the prior output value to be the result output as an output value and the prior parameter value to be the parameter value used to calculate the result output as an output value, and to compute a next output valve a uniform increment difference in value to the result output as an output value, said next output value corrupted using a parameter value equal to the sum of the prior parameter value and the parametric step size, whereby the result determined using the parameter value equal to the sum of the prior parameter value and the parametric step size is compared to the prior output value to determine the next output value, said translating means and register control means respectively updating the forward difference coefficients and format of the forward difference coefficients stored in the plurality of registers according to the following equations;
• d'=d+(c>>n);
• c'=c+(b>>n);
• b'=b+a;
• whereby the first forward difference coefficient consistently has the fixed predetermined number of fractional bits with increased precision due to the use of guard bits with respect to the succeeding forward difference coefficients.