Instruction-format pseudos about riscv-binutils-gdb HOT 12 CLOSED

I agree there's little reason to support the RVC formats. Might even just start with R and I and see what sticks.

from riscv-binutils-gdb.

https://sourceware.org/binutils/docs-2.29/as/s390-Directives.html .insn seems like a good model to follow

from riscv-binutils-gdb.

Stuffing it all in .insn will avoid namespace pollution, and can provide slightly better error messages.

from riscv-binutils-gdb.

Dear Experts

I recently just found my old RoCC prototype is not working anymore because of the toolchain have dropped the "Custom" instruction supporting. It was shocking to me and I feel this is really a upset. And by google and reading the issue riscv-collab/riscv-gnu-toolchain#190 I got to know it was decided to switch to the scheme filed by this issue. But if I understand it correctly until now this scheme is not yet really implemented in toolchain now?

That means, right now the GCC toolchain have no way to add the user-extension instructions (no custom supported, and no this new scheme implemented)?

Extension is the biggest weapon for RISC-V, when will we re-arm it? is there any plan or schedule for it?

Thanks
Bob

from riscv-binutils-gdb.

If someone has the time to prototype @sorear's proposal in GAS (something like .insn r, opcode, funct3, func7, rd, rs1, rs2), we would certainly review it and consider it for inclusion. @palmer-dabbelt and I just don't have time to do it ourselves right now. @kito-cheng maybe Andes also wants this feature, so you can get some time to do it?

from riscv-binutils-gdb.

Hummm, ya, it's useful to us, but it's not must to have item to us since we'll implement by riscv-opcode + binutils directly :P
Anyway, I'll put it into my personal TODO list.

from riscv-binutils-gdb.

@zhenbohu You can use the macro-based scheme, but not the assembler-based one. See here

riscv-collab/riscv-gnu-toolchain#190 (comment)

from riscv-binutils-gdb.

@palmer-dabbelt

Hi, Palmer

I have already read this before, but that scheme is not a good enough scheme, because it may need additional "mv" instructions to cause performance degradation. So the native support for extension instruction would be greatly better.

Thanks
Bob

from riscv-binutils-gdb.

Ya, I understand that -- it's just that this isn't a priority for us right now. If you're interested in hacking on the toolchain I can give you some pointers, it should be an easy thing to implement. If not, then hopefully we'll get some time to do this right but there's no timeline right now. Sorry!

from riscv-binutils-gdb.

@palmer-dabbelt

Thanks very much for your kind. Although I am very interested on it, but I am a purely hardware guy, not sure if I can really handle it. Maybe I will still wait for your official support on it in toolchain.

from riscv-binutils-gdb.

.insn directive is merged into upstream now :)

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=0e35537d754f1c687850d1caccb2d78d2e418391

from riscv-binutils-gdb.

Nice!

from riscv-binutils-gdb.