Systems, methods, and apparatuses for tile transpose

A non-transitory machine-readable medium having program code stored thereon which, when executed by a machine, causes the machine to perform operations, comprising:

• receiving, by a matrix operations accelerator coupled to or including at least one memory, a matrix load and transpose instruction having fields for an opcode to indicate a matrix load and transpose operation, a source matrix tile operand to indicate a location of a source matrix tile in the at least one memory, and a destination matrix tile operand to indicate one or more registers in which to store a corresponding transposed result matrix tile, the source matrix tile having a first arrangement of rows and columns of a plurality of source matrix data elements and the corresponding transposed result matrix tile having a second arrangement of the rows and columns of the plurality of source matrix data elements;
• and
• executing, by the matrix operations accelerator, the matrix load and transpose instruction to read the plurality of source matrix data elements from the at least one memory and transpose the plurality of source matrix data elements in the one or more registers to generate the corresponding transposed result matrix tile, wherein transposing the plurality of source matrix data elements comprises transforming each row or column of the plurality of source matrix data elements into a corresponding column or row, respectively, of the corresponding transposed result matrix tile.