Abstract: A method of assigning a buffer size in a video decoder includes the step of establishing a first buffer size for a scalable buffer. A video data stream is then processed with the scalable buffer configured to the first buffer size. A second buffer size is then selected for the scalable buffer. The video stream is then processed with the scalable buffer configured to the second buffer size. Memory utilization data characterizing memory performance during processing with the scalable buffer at the first buffer size and the second buffer size is then created. Afterwards, a buffer size is assigned for the scalable buffer in accordance with the memory utilization data.

Abstract: A method for pipelining variable length decode and inverse quantization operations in a hybrid motion-compensated and transform coded video decoder includes the step of mapping a new code word to a look-up table to retrieve a code word length, a zero-run length, and a quantized level. A new linear, zig-zagged position of a current coefficient is identified from the zero-run length and a previous zero-run length. The code word length is added to a current bitstream position to yield a new bitstream position. A quantization matrix coefficient from the new linear, zig-zagged position of the current coefficient is selected. The quantized level is multiplied by a predetermined value to produce a quantization product. In the case of inter block processing, a quantized level sign value is added to the quantization product. In the case of intra block processing, the quantization product does not include the quantization level sign.

Abstract: A computation unit employs a logarithmic number system that uses a logarithmic numerical representation that differs from an IEEE standard representation to improve the efficiency of computation, both by reducing the time expended in performing the computation and by reducing the size of the integrated circuit that performs the computation. The illustrative computation unit employs a numerical representation that is similar to the IEEE format except that the sign term is omitted. Thus only positive numbers are represented. The value of the mantissa is defined as a fractional number between zero and one. The logarithmic number system is a useful number system domain for multiplication, division, reciprocal, square root, and power computations using multiplication, division, and square root computation techniques described by following equations: A*B=Anti-log(log(A)+log(B)),  (3) A/B=Anti-log(log(A)−log(B)),  (4) B½=Anti-log(log(B)/2).

Abstract: An apparatus for decoding a Motion Compensated-Discrete Cosine Transform (MC-DCT) video stream includes an input port to receive an MC-DCT video stream with an associated hierarchy of data structures including a sequence data structure, a picture data structure, a slice data structure, and a macroblock data structure. A monitor processor splits the MC-DCT video stream into a set of video streams. A set of sub-processors processes the set of video streams. Each sub-processor has an assigned computational task performed on a specified hierarchical level of the associated hierarchy of data structures. Each sub-processor performs the assigned computational task with a designated data structure including all parameter data required at the specified hierarchical level.

Abstract: A computation unit computes a division operation Y/X by determining the value of a divisor reciprocal 1/X and multiplying the reciprocal by a numerator Y. The reciprocal 1/X value is determined using a quadratic approximation having a form: Ax2+Bx+C, where coefficients A, B, and C are constants that are stored in a storage or memory such as a read-only memory (ROM). The bit length of the coefficients determines the error in a final result. Storage size is reduced through use of “least mean square error”techniques in the determination of the coefficients that are stored in the coefficient storage. During the generation of partial products x2, Ax2, and Bx, the process of rounding is eliminated, thereby reducing the computational logic to implement the division functionality.

Abstract: A method of computing a square root or a reciprocal square root of a number in a computing device uses a piece-wise quadratic approximation of the number. The square root computation uses the piece-wise quadratic approximation in the form: squareroot(X)={overscore (A)}ix2+{overscore (B)}ix+{overscore (C)}i, in each interval i. The reciprocal square root computation uses the piece-wise quadratic approximation in the form: 1/squareroot(X)=Aix2+Bix+Ci, in each interval i. The coefficients {overscore (A)}i, {overscore (B)}i, and {overscore (C)}i, and Ai, Bi, and Ci are derived for the square root operation and for the reciprocal square root operation to reduce the least mean square error using a least squares approximation of a plurality of equally-spaced points within an interval. In one embodiment, 256 equally-spaced intervals are defined to represent the 23 bits of the mantissa.

Abstract: A parallel fixed-point square root and reciprocal square root computation uses the same coefficient tables as the floating point square root and reciprocal square root computation by converting the fixed-point numbers into a floating-point structure with a leading implicit 1. The value of a number X is stored as two fixed-point numbers. In one embodiment, the fixed-point numbers are converted to the special floating-point structure using a leading zero detector and a shifter. Following the square root computation or the reciprocal square root computation, the floating point result is shifted back into the two-entry fixed-point format. The shift count is determined by the number of leaded zeros detected during the conversion from fixed-point to floating-point format.