Large-scale matrix restructuring and matrix-scalar operations
Embodiments of apparatuses and methods for copying and operating on matrix elements are described. In embodiments, an apparatus includes a hardware instruction decoder to decode a single instruction and execution circuitry, coupled to hardware instruction decoder, to perform one or more operations corresponding to the single instruction. The single instruction has a first operand to reference a base address of a first representation of a source matrix and a second operand to reference a base address of second representation of a destination matrix. The one or more operations include copying elements of the source matrix to corresponding locations in the destination matrix and filling empty elements of the destination matrix with a single value.
1. An apparatus comprising:
a hardware instruction decoder to decode a single instruction having a first operand to reference a base address of a first representation of a source matrix, a second operand to reference a base address of second representation of a destination matrix, and a third operand to specify one of a plurality of fill values, wherein the plurality of fill values includes a zero and a one; and
execution circuitry to perform one or more operations corresponding to the single instruction, including:
copying elements of the source matrix to same element locations in the destination matrix; and
filling empty element locations of the destination matrix with the one of the plurality of fill values.
2. The apparatus of claim 1 , wherein the single instruction also has a fourth operand having a first field to specify a first number of rows of the source matrix.
3. The apparatus of claim 2 , wherein the fourth operand also has a second field to specify a second number of rows of the destination matrix.
4. The apparatus of claim 1 , wherein the single instruction also has a fourth operand having a first field to specify a first number of columns of the source matrix.
5. The apparatus of claim 4 , wherein the fourth operand also has a second field to specify a second number of columns of the destination matrix.
6. The apparatus of claim 1 , wherein the single instruction is also to specify a size of each element.
7. The apparatus of claim 1 , wherein the single instruction is also to specify an operation to perform on each element of the source matrix before copying to the destination matrix.
8. The apparatus of claim 7 , wherein the operation is to be performed on each element of the source matrix using a scalar value to be specified by the instruction.
9. A method comprising:
decoding, by a hardware instruction decoder, a single instruction having a first operand to reference a base address of a first representation of a source matrix, a second operand to reference a base address of second representation of a destination matrix, and a third operand to specify one of a plurality of fill values, wherein the plurality of fill values includes a zero and a one; and
performing, by execution circuitry, one or more operations corresponding to the single instruction, including:
copying elements of the source matrix to same element locations in the destination matrix; and
filling empty element locations of the destination matrix with the one of the plurality of fill values.
10. The method of claim 9 , wherein the single instruction also has a fourth operand having a first field to specify a first number of rows of the source matrix.
11. The method of claim 10 , wherein the fourth operand also has a second field to specify a second number of rows of the destination matrix.
12. The method of claim 9 , wherein the single instruction also has a fourth operand having a first field to specify a first number of columns of the source matrix.
13. The method of claim 12 , wherein the fourth operand also has a second field to specify a second number of columns of the destination matrix.
14. The method of claim 9 , wherein the single instruction is also to specify a size of each element.
15. The method of claim 9 , wherein the single instruction is also to specify an operation to perform on each element of the source matrix before copying to the destination matrix.