摘要:
Provided is a parallel processor for supporting a floating-point operation. The parallel processor has a flexible structure for easy development of a parallel algorithm involving multimedia computing, requires low hardware cost, and consumes low power. To support floating-point operations, the parallel processor uses floating-point accumulators and a flag for floating-point multiplication. Using the parallel processor, it is possible to process a geometric transformation operation in a 3-dimensional (3D) graphics process at low cost. Also, the cost of a bus width for instructions can be minimized by a partitioned Single-Instruction Multiple-Data (SIMD) method and a method of conditionally executing instructions.
摘要:
According to the present invention, the amount of computation required for image recognition processing can be reduced by extracting only image recognition learning information for an object that may appear in a region having the geographical property of a current position and comparing the image recognition learning information with ambient-image information.
摘要:
Provided is a parallel processor for supporting a floating-point operation. The parallel processor has a flexible structure for easy development of a parallel algorithm involving multimedia computing, requires low hardware cost, and consumes low power. To support floating-point operations, the parallel processor uses floating-point accumulators and a flag for floating-point multiplication. Using the parallel processor, it is possible to process a geometric transformation operation in a 3-dimensional (3D) graphics process at low cost. Also, the cost of a bus width for instructions can be minimized by a partitioned Single-Instruction Multiple-Data (SIMD) method and a method of conditionally executing instructions.
摘要:
Provided are a multiple-gate MOS (metal oxide semiconductor) transistor and a method of manufacturing the same. The transistor includes a single crystalline active region having a channel region having an upper portion of a streamlined shape (∩) obtained by patterning an upper portion of a bulk silicon substrate with an embossed pattern, and having a thicker and wider area than the channel region; a nitride layer formed at both side surfaces of the single crystalline active region to expose an upper portion of the single crystalline active region at a predetermined height; and a gate electrode formed to be overlaid with the exposed upper portion of the single crystalline active region of the channel region.
摘要:
A multiple-single instruction multiple data (SIMD) processor and an arithmetic method using the same are disclosed. When various arithmetic operations should be individually carried out by SIMD arithmetic units, control right is sub-divided to perform the arithmetic operations, such that the time of the arithmetic operations can be shortened and the efficiency thereof can be raised. When sub-divided control is not required, the control right is withdrawn and the arithmetic operations are carried out using a minimum number of program memories and a minimum number of SIMD arithmetic units, such that memory and power consumption thereof can be reduced.
摘要:
Provided is a System on Chip (SoC) system for a multimedia system enabling high-speed transfer of a large amount of multimedia data and a processor to rapidly control a peripheral device. The SoC system includes a processor; a plurality of peripheral devices; a plurality of physically divided memories; a control bus for transferring a control signal from the processor to the peripheral devices and the memories; a data bus for transferring data between the processor, the peripheral devices and the memories; a bridge for coupling the control bus and the data bus to the processor; a plurality of memory controllers coupled to the control bus and controlling each of the memories; a Direct Memory Access (DMA) controller coupled to the data bus and the control bus and controlling data transfer between the peripheral devices and the memories; and a matrix switch coupled between the DMA controller and the memory controllers and enabling simultaneous multiple memory access.
摘要:
Provided are an apparatus and method for calculating a Sum of Absolute Differences (SAD) for motion estimation of a variable block capable of parallelly calculating SAD values with respect to a plurality of current frame macroblocks at a time. The apparatus includes a PE array unit including at least one Processing Element (PE) that is aligned in the form of a matrix, and parallelly calculating a SAD value of at least one pixel provided in a plurality of serial current frame macroblocks, a local memory including current frame macroblock data, reference frame macroblock data, and reference frame search area data, and transmitting the data to each PE that is provided in the PE array unit, and a controller for making a command for the data that are provided in the local memory to be transmitted corresponding to at least one pixel, on which each PE provided in the PE array unit performs calculation.