[Quick Question] How to access the vector registers from .insn custom instruction directive?

[NazerkeT]

Hey everyone!

I was trying to add a new custom instruction to RISC-V Vector extension and using this directive for my earlier codes to test the tools:

asm __volatile__ (".insn r 0x6b, 0x0, 0x00, %0, %1, a1\n" : "=r"(a) : "r"(b) );

However, when I want to use v1 register instead of the a1, compiler says that I am using illegal operands. I guess the issue comes from using smth else instead of r after .insn. But atm, I cannot find info about that, can you please help with what I have to change in my inline assembly here?

I am using this command to run my application:

/opt/riscv-vector/bin/riscv64-unknown-elf-gcc -march=rv64gcv vec-mod.c

[angsch]

Better a late answer than no answer: Clobbering of vector registers such as v1 is currently not supported. kito-cheng answered in #365 that GCC 13 should support the usage of inline assembly with vector registers.