On Sun, Jul 26, 2020 at 11:40 PM wangtao (CH) <wangta...@huawei.com> wrote: > This is the patch to support ZFINX of RISC-V, which option is like > -march=rv32gc_zfinx. The ZFINX means f-registers in x-registers under RV-F > and RV-D extension. For more details, please refer to > https://github.com/riscv/riscv-zfinx/blob/master/Zfinx_spec.adoc. > This patch mainly adds the ZFINX option and abi constraints, and when it’s > under ZFINX, makes the f-registers as FIXED_REGs to avoid allocating > f-regsiters to pseudo registers. > And for binutils support, it has been done and I will send it to binutils-gdb > community to review later.
Normally I'd expect to see the binutils patch first, since the gcc patch can't be tested without the binutils patch. Looking at FSF Copyright assignments, I see that Huawei has corporate assignments for gcc and glibc, but I don't see one for binutils. If Huawei is writing the binutils patch, and we can't accept the binutils patch due to a missing copyright assignment, then that makes the gcc patch mostly useless. Current convention is that we only accept patches for ratified extensions, and zfinx is not ratified yet. It is still a proposed extension that may change in incompatible ways before it is ratified. It is good to have binutils/gcc patches so that we can test it, but they can't be on the master branch with current conventions. We can put them on a vendor branch in the FSF GCC tree. Or we can put them on a branch in the github.com/riscv trees. We do still need a copyright assignment from Huawei before we can use the github.com/riscv trees though, to avoid contaminating those trees with patches that can't be upstreamed. I haven't tried reviewing the patch yet. I took most of last week off, so this is now on my to do list and hopefully I can get to it soon. Jim
