Abstract: Aspects include a compute array of a processor with mixed-precision numerical linear algebra support. A first precision and a first shape of a first input matrix and a second precision and a second shape of a second input matrix to the compute array are determined. A plurality of linear algebra operations is repeated in parallel within the compute array to update a result matrix in an accumulator register based on the first input matrix, the second input matrix, and a number of rank updates of the result matrix to store in the accumulator register.

Abstract: Aspects include a compute array of a processor with mixed-precision numerical linear algebra support. A first precision and a first shape of a first input matrix and a second precision and a second shape of a second input matrix to the compute array are determined. A plurality of linear algebra operations is repeated in parallel within the compute array to update a result matrix in an accumulator register based on the first input matrix, the second input matrix, and a number of rank updates of the result matrix to store in the accumulator register.

Abstract: Aspects include a compute array of a processor with mixed-precision numerical linear algebra support. A first precision and a first shape of a first input matrix and a second precision and a second shape of a second input matrix to the compute array are determined. A number of rank updates of a result matrix to store in an accumulator register having a predetermined size are determined, where the number of rank updates is based on the first precision and the first shape of the first input matrix, the second precision and the second shape of the second input matrix, and the predetermined size of the accumulator register. A plurality of linear algebra operations is repeated in parallel within the compute array to update the result matrix in the accumulator register based on the first input matrix, the second input matrix, and the number of rank updates.

Abstract: Embodiments of the present invention are directed to a new instruction set extension and a method for providing 3D lane predication for matrix operations. In a non-limiting embodiment of the invention, a first input matrix having m rows and k columns and a second input matrix having k rows and n columns are received by a compute array of a processor. A three-dimensional predicate mask having an M-bit row mask, an N-bit column mask, and a K-bit rank mask is generated. A result matrix of up to m rows, up to n columns, and up to k rank updates is determined based on the first input matrix, the second input matrix, and the predicate mask.

Abstract: Embodiments of the present invention are directed to a new instruction set extension and a method for providing 3D lane predication for matrix operations. In a non-limiting embodiment of the invention, a first input matrix having m rows and k columns and a second input matrix having k rows and n columns are received by a compute array of a processor. A three-dimensional predicate mask having an M-bit row mask, an N-bit column mask, and a K-bit rank mask is generated. A result matrix of up to m rows, up to n columns, and up to k rank updates is determined based on the first input matrix, the second input matrix, and the predicate mask.

Abstract: Aspects include a compute array of a processor with mixed-precision numerical linear algebra support. A first precision and a first shape of a first input matrix and a second precision and a second shape of a second input matrix to the compute array are determined. A number of rank updates of a result matrix to store in an accumulator register having a predetermined size are determined, where the number of rank updates is based on the first precision and the first shape of the first input matrix, the second precision and the second shape of the second input matrix, and the predetermined size of the accumulator register. A plurality of linear algebra operations is repeated in parallel within the compute array to update the result matrix in the accumulator register based on the first input matrix, the second input matrix, and the number of rank updates.

Abstract: Status indicators are provided. An operation is performed in a current operating mode of the processor, in which the current operating mode is dynamically selectable from a plurality of supported operating modes. A first set of status indicators is set, based on performing the operation in the current operating mode. A second set of status indicators is set, based on performing the operation in the current operating mode, but the setting is to reflect at least one status condition for another operating mode supported by the processor. At least one of the first set of status indicators or the second set of status indicators is used in further processing.

Abstract: A set machine instruction is provided that has associated therewith a result location to be used with a set operation. The set machine instruction is executed, which includes checking contents of a selected field, and determining, based on the checking, whether the contents of the selected field indicate a first condition, a second condition or a third condition represented in one data type. The result location is set to a value based on the determining, wherein the value, based on the setting, is of a data type different from the one data type and represents a result of a previously executed instruction. The result of the previously executed instruction being one of the first condition, the second condition or the third condition.

Abstract: A machine instruction is provided that has associated therewith a result location to be used for a set operation, a first source, a second source, and an operation select field configured to specify a plurality of selectable operations. The machine instruction is executed, which includes obtaining the first source, the second source, and a selected operation, and performing the selected operation on the first source and the second source to obtain a result in one data type. That result is quantized to a value in a different data type, and the value is placed in the result location.

Abstract: A machine instruction is provided that has associated therewith a result location to be used for a set operation, a first source, a second source, and an operation select field configured to specify a plurality of selectable operations. The machine instruction is executed, which includes obtaining the first source, the second source, and a selected operation, and performing the selected operation on the first source and the second source to obtain a result in one data type. That result is quantized to a value in a different data type, and the value is placed in the result location.

Abstract: A set machine instruction is provided that has associated therewith a result location to be used with a set operation. The set machine instruction is executed, which includes checking contents of a selected field, and determining, based on the checking, whether the contents of the selected field indicate a first condition, a second condition or a third condition represented in one data type. The result location is set to a value based on the determining, wherein the value, based on the setting, is of a data type different from the one data type and represents a result of a previously executed instruction. The result of the previously executed instruction being one of the first condition, the second condition or the third condition.

Abstract: A method is disclosed for loading a vector with a processor. The method includes obtaining, by the processor, a variable-length vector load instruction. The method also includes determining that the vector load instruction specifies a vector register for a target, a memory address, and a length, wherein the memory address and the length are each specified in at least a general purpose register. The method also includes determining whether data should be loaded into the vector register using big endian byte-ordering or little endian byte-ordering. The method further includes loading data from memory into the vector register, wherein if the length is less than a length of the vector register, setting one or more residue bytes in the vector register to a pad value, wherein the residue bytes are determined based on the determined byte-ordering.

Abstract: A method is disclosed for storing vector data into memory with a processor. The method includes obtaining, by the processor, a variable-length vector store instruction. The method also includes determining that the vector store instruction specifies a vector register for a source, a memory address, and a length, where the memory address and the length are each specified in at least a general purpose register. The method also includes determining whether data should be stored into memory at the memory address using big endian byte-ordering or little endian byte-ordering. The method further includes storing data from the vector register into memory, where if the length is less than a length of the vector register, storing only the data from the vector register specified by the length.

Abstract: Embodiments relate to vector memory access instructions for big-endian (BE) element ordered computer code and little-endian (LE) element ordered computer code. An aspect includes determining a mode of a computer system comprising one of a BE mode and an LE mode. Another aspect includes determining a code type comprising one of BE code and LE code. Another aspect includes determining a data type of data in a main memory that is associated with the object code comprising one of BE data and LE data. Another aspect includes based on the mode, code type, and data type, inserting a memory access instruction into the object code to perform a memory access associated with the vector in the object code, such that the memory access instruction performs element ordering of elements of the vector, and data ordering within the elements of the vector, in accordance with the determined mode, code type, and data type.

Abstract: A Set Boolean machine instruction is provided that has associated therewith a result location to be used for a set Boolean operation and a mask. The mask is configured to test a plurality of types of conditions, including simple conditions and composite conditions. The machine instruction is executed, and the executing includes performing a first logical operation between the mask and contents of a selected field to obtain an output. The mask indicates a condition to be tested, and the condition is one type of condition of the plurality of types of conditions. The executing further includes performing a second logical operation on the output to obtain a first value represented as one data type, and placing a result in the result location based on the first value. The result including a second a value of another data type, the other data type being different from the one data type.

Abstract: A Set Boolean machine instruction is provided that has associated therewith a result location to be used for a set Boolean operation and a mask. The mask is configured to test a plurality of types of conditions, including simple conditions and composite conditions. The machine instruction is executed, and the executing includes performing a first logical operation between the mask and contents of a selected field to obtain an output. The mask indicates a condition to be tested, and the condition is one type of condition of the plurality of types of conditions. The executing further includes performing a second logical operation on the output to obtain a first value represented as one data type, and placing a result in the result location based on the first value. The result including a second a value of another data type, the other data type being different from the one data type.

Abstract: A parallel processing method, system, and/or computer program product for performing data parallel wide accesses on an unstructured text is provided. The parallel processing includes creating a pointer that points to a beginning of the unstructured text and loading into a vector register a string segment of the unstructured text based on the pointer. Then, access permissions of a first byte of the string segment are automatically tested. In turn, a determination is made as to whether the string segment includes an end indication, and a remaining portion of the unstructured text is validated by accessing and loading a last character identified by the end indication into the vector register when the string segment is determined to include the end indication.

Abstract: A machine instruction to find a condition location within registers, such as vector registers. The machine instruction has associated therewith a register to be examined and a result location. The register includes a plurality of elements. In execution, the machine instruction counts a number of contiguous elements of the plurality of elements of the register having a particular value in a selected location within the contiguous elements. Other locations within the contiguous elements are ignored for the counting. The counting provides a count placed in the result location.

Abstract: A machine instruction to find a condition location within registers, such as vector registers. The machine instruction has associated therewith a register to be examined and a result location. The register includes a plurality of elements. In execution, the machine instruction counts a number of contiguous elements of the plurality of elements of the register having a particular value in a selected location within the contiguous elements. Other locations within the contiguous elements are ignored for the counting. The counting provides a count placed in the result location.

Abstract: A machine instruction to find a condition location within registers, such as vector registers. The machine instruction has associated therewith a register to be examined and a result location. The register includes a plurality of elements. In execution, the machine instruction counts a number of contiguous elements of the plurality of elements of the register having a particular value in a selected location within the contiguous elements. Other locations within the contiguous elements are ignored for the counting. The counting provides a count placed in the result location.