Doubts about branch Boom prediction decoding

Archive RISC-V
1

According to the riscv manual, ret and call are two pseudo instructions respectively.
ret is jalr, x0, x1, 0.
call is {auipc x1, offset[31:12]; jalr x1, x1, offset[11:0]}.

In this link boom decode， use the following codes to judge ret and call instruction.
io.out.is_call := (cs_is_jal || cs_is_jalr) && GetRd(io.inst) === RA
io.out.is_ret := cs_is_jalr && GetRs1(io.inst) === BitPat(“b00?01”) && GetRd(io.inst) === X0

Why not judge the instruction like this
io.out.is_call := (cs_is_jalr) && GetRd(io.inst) === RA
io.out.is_ret := cs_is_jalr && GetRs1(io.inst) === BitPat(“b00001”) && GetRd(io.inst) === X0

2

Because if the call is to a function located within +/- 1 MB of the PC then jal is used not the two instruction auipc;jalr sequence.

3

How to interpret the ret instruction?
The rs1 written in the manual is x1 instead of x5.

4

x1 is the ABI return address register. x5 is the alternate link register, intended to be used for millicode routines. Hardware that has a call/return stack needs to support both x1 and x5. This is mentioned in the ISA documentation in the JAL and JALR descriptions.

5

Thank you for your answer