Open
Description
Add vector sequencer and execution support for the vector permutation instructions. At least 1 test should be added with an instruction from each category to verify proper sequencing. Some of these instruction types require reading up to 8 source registers to write to a single destination, which is not reasonable for sequencing. Need to set a reasonable limit on the number of source registers allowed per uop and come up with a creative way to sequence these more complex instructions. The solution may require that the uops be executed in order, which is ok.
- Vector Permutation Instructions: Integer Scalar Move Instructions
- Vector Permutation Instructions: Floating-Point Scalar Move Instructions
- Vector Permutation Instructions: Vector Slide Instructions
- Vector Permutation Instructions: Vector Register Gather Instructions
- Vector Permutation Instructions: Vector Compress Instruction
- Vector Permutation Instructions: Whole Vector Register Move
Metadata
Metadata
Assignees
Labels
No labels