Methods, systems and computer program products are provided for automated runtime configuration for dataflows to automatically select or adapt a runtime environment or resources to a dataflow plan prior to execution. Metadata generated for dataflows indicates dataflow information, such as numbers and types of sources, sinks and operations, and the amount of data being consumed, processed and written. Weighted dataflow plans are created from unweighted dataflow plans based on metadata. Weights that indicate operation complexity or resource consumption are generated for data operations. A runtime environment or resources to execute a dataflow plan is/are selected based on the weighted dataflow and/or a maximum flow. Preferences may be provided to influence weighting and runtime selections.

Embodiments described herein are generally directed to an improved vector normalization instruction. An embodiment of a method includes responsive to receipt by a GPU of a single instruction specifying a vector normalization operation to be performed on V vectors: (i) generating V squared length values, N at a time, by a first processing unit, by, for each N sets of inputs, each representing multiple component vectors for N of the vectors, performing N parallel dot product operations on the N sets of inputs. Generating V sets of outputs representing multiple normalized component vectors of the V vectors, N at a time, by a second processing unit, by, for each N squared length values of the V squared length values, performing N parallel operations on the N squared length values, wherein each of the N parallel operations implement a combination of a reciprocal square root function and a vector scaling function.

An apparatus to facilitate machine learning matrix processing is disclosed. The apparatus comprises a memory to store matrix data one or more processors to execute an instruction to examine a message descriptor included in the instruction to determine a type of matrix layout manipulation operation that is to be executed, examine a message header included in the instruction having a plurality of parameters that define a two-dimensional (2D) memory surface that is to be retrieved, retrieve one or more blocks of the matrix data from the memory based on the plurality of parameters and a register file including a plurality of registers, wherein the one or more blocks of the matrix data is stored within a first set of the plurality of registers.

An apparatus and method for propagating conditionally evaluated values are disclosed. For example, a method according to one embodiment comprises: reading each value contained in an input mask register, each value being a true value or a false value and having a bit position associated therewith; for each true value read from the input mask register, generating a first result containing the bit position of the true value; for each false value read from the input mask register following the first true value, adding the vector length of the input mask register to a bit position of the last true value read from the input mask register to generate a second result; and storing each of the first results and second results in bit positions of an output register corresponding to the bit positions read from the input mask register.

A processor of an aspect includes packed data registers, and a decode unit to decode an instruction. The instruction may indicate a first source packed data to include at least four data elements, to indicate a second source packed data to include at least four data elements, and to indicate a destination storage location. An execution unit is coupled with the packed data registers and the decode unit. The execution unit, in response to the instruction, is to store a result packed data in the destination storage location. The result packed data may include at least four indexes that may identify corresponding data element positions in the first and second source packed data. The indexes may be stored in positions in the result packed data that are to represent a sorted order of corresponding data elements in the first and second source packed data.

A processor includes: an address generating unit that, when an instruction decoded by a decoding unit is an instruction to execute arithmetic processing on a plurality of operand sets each including a plurality of operands that are objects of the arithmetic processing, in parallel a plurality of times, generates an address set corresponding to each of the operand sets of the arithmetic processing for each time, based on a certain address displacement with respect to the plurality of operands included in each of the operand sets; a plurality of instruction queues that hold the generated address sets corresponding to the respective operand sets, in correspondence to respective processing units; and a plurality of processing units that perform the arithmetic processing in parallel on the operand sets obtained based on the respective address sets outputted by the plurality of instruction queues.

A method is described that includes executing a convolutional neural network layer on an image processor having an array of execution lanes and a two-dimensional shift register. The executing of the convolutional neural network includes loading a plane of image data of a three-dimensional block of image data into the two-dimensional shift register. The executing of the convolutional neural network also includes performing a two-dimensional convolution of the plane of image data with an array of coefficient values by sequentially: concurrently multiplying within the execution lanes respective pixel and coefficient values to produce an array of partial products; concurrently summing within the execution lanes the partial products with respective accumulations of partial products being kept within the two dimensional register for different stencils within the image data; and, effecting alignment of values for the two-dimensional convolution within the execution lanes by shifting content within the two-dimensional shift register array.

In one embodiment, a method for tensor data distribution using a direct-memory access agent includes generating, by a first controller, source addresses indicating locations in a source memory where portions of a source tensor are stored. A second controller may generate destination addresses indicating locations in a destination memory where portions of a destination tensor are to be stored. The direct-memory access agent receives a source address generated by the first controller and a destination address generated by the second controller and determines a burst size. The direct-memory access agent may issue a read request comprising the source address and the burst size to read tensor data from the source memory and may store the tensor data into an alignment buffer. The direct-memory access agent then issues a write request comprising the destination address and the burst size to write data from the alignment buffer into the destination memory.

A streaming engine employed in a digital data processor may specify a fixed read-only data stream defined by plural nested loops. An address generator produces address of data elements for the nested loops. A steam head register stores data elements next to be supplied to functional units for use as operands. A stream template register independently specifies a linear address or a circular address mode for each of the nested loops.

A method of implementing a processor architecture and corresponding system includes operands of a first size and a datapath of a second size. The second size is different from the first size. Given a first array of registers and a second array of registers, each register of the first and second arrays being of the second size, selecting a first register and corresponding second register from the first array and the second array, respectively, to perform operations of the first size. This allows a user, who is interfacing with the hardware processor through software, to provide data of the datapath bit-width instead of the register bit-width. Advantageously, the user is agnostic to the size of the registers.