How to configure the number of vector registers compile can use?

Dear developers,

Is it possible to configure the number of vector registers during compilation? Because I want to make some experiments where I can limit the number of vector registers the system can use. For instance, the clang++ has options like -ffixed-x9 to reserve the x9 register (AArch64/RISC-V only).

For the RISC-V K1 core, does it have the similar flags or options? Where can I find it?
Any materials about it would be appreciated.

Thanks

  1. If you just want to configure the number of registers, you can may use the option:riscv-v-register-bit-width-lmul;
  2. If you want to constrain certain vector registers, it seems that is not currently supported.
2 Likes

Thank you for replying.