Conform to CFU Task Group instruction encodings

[google]

Conform to CFU Task Group instruction encodings

[tcal-x]

Copied from the old private repo:

Currently we hand the CFU the high 20b of the instruction raw, allowing for any interpretation. We only use the "custom0" opcode space. The register contents specified by the fields rs1 and rs2 are always read and passed to the CFU.

funct7 = instr[31:25]
funct3 = instr[14:12]

The task group defines R-type instructions in "custom0", with {funct7,funct3} as the 10b function_id.
It is likely some of the R-type instruction function_id opcode space will be reserved for "standard" CFU instructions such as set_state and read_state.

The task group also defines an immediate format in "custom1" space, with funct3==3'b000, function_id=instr[23:20], and an 8b immediate in instr[31:24], which is sign-extended and passed as the second operand to the CFU.

Link to picture of encoding:
https://user-images.githubusercontent.com/63030309/110150278-ec1e3800-7d93-11eb-8abe-55df8cc43fc8.png