Disclosed embodiments relate to executing a vector multiplication instruction. In one example, a processor includes fetch circuitry to fetch the vector multiplication instruction having fields for an opcode, first and second source identifiers, and a destination identifier, decode circuitry to decode the fetched instruction, execution circuitry to, on each of a plurality of corresponding pairs of fixed-sized elements of the identified first and second sources, execute the decoded instruction to generate a double-sized product of each pair of fixed-sized elements, the double-sized product being represented by at least twice a number of bits of the fixed size, and generate an unsigned fixed-sized result by rounding the most significant fixed-sized portion of the double-sized product to fit into the identified destination.

An integrated circuit for allowing a band of an external memory to be effectively used in processing a layer algorithm is disclosed. One aspect of the present disclosure relates to an integrated circuit including a first arithmetic part including a first arithmetic unit and a first memory, wherein the first arithmetic unit performs an operation and the first memory stores data for use in the first arithmetic unit and a first data transfer control unit that controls transfer of data between the first memory and a second memory of a second arithmetic part including a second arithmetic unit, wherein the second arithmetic part communicates with an external memory via the first arithmetic part.

The present disclosure provides a computation device. The computation device is configured to perform a machine learning computation, and includes an operation unit, a controller unit, and a conversion unit. The storage unit is configured to obtain input data and a computation instruction. The controller unit is configured to extract and parse the computation instruction from the storage unit to obtain one or more operation instructions, and to send the one or more operation instructions and the input data to the operation unit. The operation unit is configured to perform operations on the input data according to one or more operation instructions to obtain a computation result of the computation instruction. In the examples of the present disclosure, the input data involved in machine learning computations is represented by fixed-point data, thereby improving the processing speed and efficiency of training operations.

An instruction is executed in round-for-reround mode wherein the permissible resultant value that is closest to and no greater in magnitude than the infinitely precise result is selected. If the selected value is not exact and the units digit of the selected value is either 0 or 5, then the digit is incremented by one and the selected value is delivered. In all other cases, the selected value is delivered.

The present disclosure provides a computation device and method. The device may include an input module configured to acquire input data; a model generation module configured to construct an offline model according to an input network structure and weight data; a neural network operation module configured to generate a computation instruction based on the offline model and cache the computation instruction, and compute the data to be processed based on the computation instruction to obtain a computation result; and an output module configured to output a computation result. The device and method may avoid the overhead caused by running an entire software architecture, which is a problem in a traditional method.

Embodiments detailed herein relate to arithmetic operations of float-point values. An exemplary processor includes decoding circuitry to decode an instruction, where the instruction specifies locations of a plurality of operands, values of which being in a floating-point format. The exemplary processor further includes execution circuitry to execute the decoded instruction, where the execution includes to: convert the values for each operand, each value being converted into a plurality of lower precision values, where an exponent is to be stored for each operand; perform arithmetic operations among lower precision values converted from values for the plurality of the operands; and generate a floating-point value by converting a resulting value from the arithmetic operations into the floating-point format and store the floating-point value.

In one disclosed embodiment, a processor includes a first execution unit and a second execution unit, a register file, and a data path including a plurality of lanes. The data path and the register file are arranged so that writing to the register file by the first execution unit and by the second execution unit is allowed over the data path, reading from the register file by the first execution unit is allowed over the data path, and reading from the register file by the second execution unit is not allowed over the data path. The processor also includes a power control circuit configured to, when a transfer of data between the register file and either of the first and second execution units uses less than all of the lanes, power down the lanes of the data path not used for the transfer of the data.

A method of converting a data stored in a memory from a first format to a second format is disclosed. The method includes extending a number of bits in the data stored in a double data rate (DDR) memory by one bit to form an extended data. The method further includes determining whether the data stored in the DDR is signed or unsigned data. Moreover, responsive to determining that the data is signed, a sign value is added to the most significant bit of the extended data and the data is copied to lower order bits of the extended data. Responsive to determining that the data is unsigned, the data is copied to lower order bits of the extended data and the most significant bit is set to an unsigned value, e.g., zero. The extended data is stored in an on-chip memory (OCM) of a processing tile of a machine learning computer array.

Embodiments of systems, methods, and apparatuses for heterogeneous computing are described. In some embodiments, a hardware heterogeneous scheduler dispatches instructions for execution on one or more plurality of heterogeneous processing elements, the instructions corresponding to a code fragment to be processed by the one or more of the plurality of heterogeneous processing elements, wherein the instructions are native instructions to at least one of the one or more of the plurality of heterogeneous processing elements.

Techniques related to executing a plurality of instructions by a processor comprising receiving a first instruction for execution on an instruction execution pipeline, beginning execution of the first instruction, receiving one or more second instructions for execution on the instruction execution pipeline, the one or more second instructions associated with a higher priority task than the first instruction, storing a register state associated with the execution of the first instruction in one or more registers of a capture queue associated with the instruction execution pipeline, copying the register state from the capture queue to a memory, determining that the one or more second instructions have been executed, copying the register state from the memory to the one or more registers of the capture queue, and restoring the register state to the instruction execution pipeline from the capture queue.