We have introduced general matrix type with #50. Meanwhile, we still need following general matrix operations:
Addition/Subtraction
Multiplication
Transpose
Inversion
Unit matrix generation
Zero matrix generation
I wrote down as many operations as I can recall. There are possibly more :/ (I will keep updating the list)
It should be noted that we need to prioritize the operations for "general" size matrix.
Specialization or optimization for the specific dimensions should come later.
*) All above operations basically assume "one thread per one matrix (or track)" for device usage
We will need "one block per one matrix" operations as well some day
We have introduced general matrix type with #50. Meanwhile, we still need following general matrix operations:
I wrote down as many operations as I can recall. There are possibly more :/ (I will keep updating the list)
It should be noted that we need to prioritize the operations for "general" size matrix. Specialization or optimization for the specific dimensions should come later.
*) All above operations basically assume "one thread per one matrix (or track)" for device usage We will need "one block per one matrix" operations as well some day