Abstract: A method comprising obtaining first, real, data to be processed. It is determined, based on a number of computation resources of a set of computation resources of a processing unit available for use during a processing cycle, to process at least a portion of the first data using a first subset of the set and to load second, artificial, data into a second subset of the set, disjoint from the first subset of the set, the second data comprising at least one artificial data element. In a processing cycle, at least the portion of the first data and the second data are loaded into first and second subsets of the set, respectively. The second subset is an artificially activated subset. The second data is inhibited from affecting output feature map data, which is generated based at least in part on the computational result.

Abstract: A method comprising obtaining first, real, data to be processed. It is determined, based on a number of computation resources of a set of computation resources of a processing unit available for use during a processing cycle, to process at least a portion of the first data using a first subset of the set and to load second, artificial, data into a second subset of the set, disjoint from the first subset of the set, the second data comprising at least one artificial data element. In a processing cycle, at least the portion of the first data and the second data are loaded into first and second subsets of the set, respectively. The second subset is an artificially activated subset. The second data is inhibited from affecting output feature map data, which is generated based at least in part on the computational result.

Abstract: A method for generating output feature map data during operation of neural network processing by a processing unit comprising a plurality of computation resources. The method comprises obtaining first, real, data to be processed and loading the first data into a set of the plurality of computation resources, causing the set of computation resources to generate a computational result, in a first processing cycle of the processing unit. A lack of real data for processing in a second processing cycle of the processing unit, which is subsequent to the first processing cycle, is detected. The method comprises obtaining second, artificial, data, loading the second data into an artificially activated set, of the set of computation resources, in the second processing cycle, inhibiting the second data from affecting the output feature map data, and generating the output feature map data based at least in part on the computational result.

Abstract: A method performed by a processing unit for generating an output feature map, the processing unit comprising an input feature map storage configured to store input feature map blocks. The input feature map storage is read by the processing unit to generate output feature map blocks. The method comprises sequentially loading input feature map blocks into the input feature map storage, using a first input feature map block stored in the input feature map storage to generate a partial computation for a first output feature map block, and reusing the first input feature map block stored in the input feature map storage to generate a partial computation for a second output feature map block without reloading the first input feature map block into the input feature map storage.

Abstract: A processing unit comprises a multiply-accumulate engine and a control unit. The engine comprises a plurality of dot product units, switching circuitry, a plurality of adders, and a plurality of accumulators. The switching circuitry, coupled between the dot product units and the adders, is configurable to selectively couple each of the adders to one of the plurality of dot product units. The adders are each associated with a respective accumulator of the plurality of accumulators. In a processing cycle, each of the dot product units is configured to output a product value, the control unit is operable to configure the switching circuitry such that each of the adders is coupled to a selected dot product unit of the plurality of dot product units, and each of the adders is configured to add the product value of the selected dot product unit to an accumulated value stored by the respective accumulator.

Abstract: A dot product array comprises dot product circuits each to process a respective pair of first and second input vectors to generate a respective dot product result. In a real number mode, each dot product result and vector element represents a respective real number. In a hypercomplex number mode, an input vector manipulation is applied to at least one of the first/second input vectors to be supplied to each dot product circuit, to cause the dot product array to generate hypercomplex dot product results each indicating a sum of hypercomplex products of corresponding pairs of hypercomplex numbers. In the hypercomplex number mode, respective subsets of elements of the first/second input vectors represent respective hypercomplex numbers, for which respective components are represented by different elements of the subset, and each hypercomplex dot product result comprises components represented by the dot product results generated by a corresponding group of at least two dot product circuits.

Abstract: A system for encoding and decoding a sequence of frames of video data. The system includes encoding processing circuitry configured to encode a sequence of source video frames using other source frames as reference frames. The encoding processing circuitry is also configured, when encoding a new source frame that has a different resolution to the resolution of a reference frame, to generate a scaled copy of the reference frame for the new source frame. The encoding processing circuitry is also configured to encode the new source frame using the scaled copy of the reference frame to provide output encoded video image data. The system also includes a video decoder configured to decode the encoded scaled copy of the reference frame and decode the encoded source frame using the decoded scaled copy of the reference frame to provide output video image data for the source frame.

Abstract: A processing unit is described that receives an instruction to perform a first operation on a first layer of a neural network, block dependency data, and an instruction to perform a second operation on a second layer of the neural network. The processing unit performs the first operation, which includes dividing the first layer into a plurality of input blocks, and operating on the input blocks to generate a plurality of output blocks. The processing unit then performs the second operation after the first operation has generated a set number of output blocks defined by the block dependency data.

Abstract: When encoding an array of data elements using an encoding process that generates an array of quantized frequency domain coefficients corresponding to an array of difference values between a source block of data elements of the array of data elements being encoded and a reference block of data elements, a bit count value for encoding an array of quantized frequency domain coefficients to be used to select whether to encode the source block in accordance with a particular set of encoding options is estimated by estimating a bit count value for encoding the locations of the non-zero quantized frequency domain coefficients in the array of quantized frequency domain coefficients based on the number of non-zero frequency domain coefficients in the array of quantized frequency domain coefficients.

Abstract: When encoding an array of data elements of a stream of arrays of data elements, a set of sample adaptive offset options to be used for encoding the array are selecting on the basis of a distortion value that is determined for a particular set of sample adaptive offset options. The distortion value is determined by generating one or more filtered reconstructed source blocks by applying a sample adaptive offset filter in accordance with the particular set of sample adaptive offset options to one or more reconstructed source blocks corresponding to a source block, and by then determining a distortion value for the particular set of sample adaptive offset options by comparing the one or more filtered reconstructed source blocks with the source block. When the particular set of sample adaptive offset options comprises an edge offset type, the distortion value is determined using a set of error values comprising fewer data elements than the source block of data elements.

Abstract: A set of encoding options to use when encoding an array of data elements is selected based on a bit count value and a distortion value for that set of encoding options. The distortion value is determined from a set of error values that represents the difference between a set of frequency domain coefficients and a set of de-quantised coefficients. The set of frequency domain coefficients are generated by applying only a subset of row transformations or only a subset of column transformations. A set of quantised coefficients are generated by quantising only a subset of the set of frequency domain coefficients. This arrangement provides a more efficient way of selecting encoding options, but without a detrimental reduction in the efficacy of the selection process.

Abstract: In a data processing system that comprises a memory 8 comprising N memory banks 11, a memory controller is configured to store one or more N data unit×N data unit arrays of data in the memory 8 such that each data unit in each row of each N×N data unit array is stored in a different memory bank of the N memory banks 11, and such that each data unit in each column of each N×N data unit array is stored in a different memory bank of the N memory banks 11.

Abstract: A method of video encoding is provided. The method includes downscaling a block of a frame of a video to generate a downscaled block. The method includes downscaling a reference block of a reference frame of the video to generate a downscaled reference block. The method includes processing the downscaled block to calculate first compression efficacy data indicative of a first efficacy of encoding the block using intra-frame prediction. The method includes processing the downscaled block and the downscaled reference block to calculate second compression efficacy data indicative of a second efficacy of encoding the block using inter-frame prediction. The method includes based on at least one of the first compression efficacy data or the second compression efficacy data, determining that the block is to be encoded using inter-frame prediction. Other methods of video encoding and a video encoder system are also provided.

Abstract: When encoding an array of data elements of a stream of arrays of data elements, a set of sample adaptive offset options to be used for encoding the array are selecting on the basis of a distortion value that is determined for a particular set of sample adaptive offset options. The distortion value is determined by generating one or more filtered reconstructed source blocks by applying a sample adaptive offset filter in accordance with the particular set of sample adaptive offset options to one or more reconstructed source blocks corresponding to a source block, and by then determining a distortion value for the particular set of sample adaptive offset options by comparing the one or more filtered reconstructed source blocks with the source block. When the particular set of sample adaptive offset options comprises an edge offset type, the distortion value is determined using a set of error values comprising fewer data elements than the source block of data elements.

Abstract: When generating intra-frame encoding cost measures for evaluating intra-frame encoding of a frame of video image data, a measure of the cost of intra-frame encoding a block of the frame is determined by determining a separate cost measure for each sub-block of a set of plural sub-blocks making up the block to be encoded, and then determining the measure of the cost of encoding the block to be encoded as the sum of the determined cost measures for each of the sub-blocks.

Abstract: A method of video encoding is provided. The method includes downscaling a block of a frame of a video to generate a downscaled block. The method includes downscaling a reference block of a reference frame of the video to generate a downscaled reference block. The method includes processing the downscaled block to calculate first compression efficacy data indicative of a first efficacy of encoding the block using intra-frame prediction. The method includes processing the downscaled block and the downscaled reference block to calculate second compression efficacy data indicative of a second efficacy of encoding the block using inter-frame prediction. The method includes based on at least one of the first compression efficacy data or the second compression efficacy data, determining that the block is to be encoded using inter-frame prediction. Other methods of video encoding and a video encoder system are also provided.

Abstract: When encoding an array of data elements using an encoding process that generates an array of quantized frequency domain coefficients corresponding to an array of difference values between a source block of data elements of the array of data elements being encoded and a reference block of data elements, a bit count value for encoding an array of quantized frequency domain coefficients to be used to select whether to encode the source block in accordance with a particular set of encoding options is estimated by estimating a bit count value for encoding the locations of the non-zero quantized frequency domain coefficients in the array of quantized frequency domain coefficients based on the number of non-zero frequency domain coefficients in the array of quantized frequency domain coefficients.

Abstract: When generating intra-frame encoding cost measures for evaluating intra-frame encoding of a frame of video image data, a measure of the cost of intra-frame encoding a block of the frame is determined by determining a separate cost measure for each sub-block of a set of plural sub-blocks making up the block to be encoded, and then determining the measure of the cost of encoding the block to be encoded as the sum of the determined cost measures for each of the sub-blocks.

Abstract: To perform motion estimation for a video frame block to be encoded, a difference measure is determined for each of a plurality of reference frame block positions at a first, coarser resolution. The determined difference measures are then used estimate difference measures for reference frame blocks at positions at a second resolution that is finer than the first resolution. The estimated second, finer position resolution difference measures are then used to select a set of reference frame block positions for which to determine “full” difference measures. The determined “full” difference measures for each of the selected reference frame block positions are then used to select the reference frame block position to use when encoding the frame block and a motion vector corresponding to that reference frame block position is associated with and encoded for the frame block being encoded.

Abstract: A set of encoding options to use when encoding an array of data elements is selected based on a bit count value and a distortion value for that set of encoding options. The distortion value is determined from a set of error values that represents the difference between a set of frequency domain coefficients and a set of de-quantised coefficients. The set of frequency domain coefficients are generated by applying only a subset of row transformations or only a subset of column transformations. A set of quantised coefficients are generated by quantising only a subset of the set of frequency domain coefficients. This arrangement provides a more efficient way of selecting encoding options, but without a detrimental reduction in the efficacy of the selection process.