Re: [PATCH RFC 0/2] RISC-V: T-Head vector handling

[Heiko Stuebner]

Hi Palmer,

Am Mittwoch, 15. März 2023, 06:29:41 CET schrieb Palmer Dabbelt:
> On Tue, 28 Feb 2023 13:54:33 PST (-0800), heiko@xxxxxxxxx wrote:
> > From: Heiko Stuebner <heiko.stuebner@xxxxxxxx>
> >
> > As is widely known the T-Head C9xx cores used for example in the
> > Allwinner D1 implement an older non-ratified variant of the vector spec.
> >
> > While userspace will probably have a lot more problems implementing
> > support for both, on the kernel side the needed changes are actually
> > somewhat small'ish and can be handled via alternatives somewhat nicely.
> >
> > With this patchset I could run the same userspace program (picked from
> > some riscv-vector-test repository) that does some vector additions on
> > both qemu and a d1-nezha board. On both platforms it ran sucessfully and
> > even produced the same results.
> >
> >
> > As can be seen in the todo list, there are 2 places where the changed
> > SR_VS location still needs to be handled in the next revision
> > (assembly + ALTERNATIVES + constants + probably stringify resulted in
> >  some grey hair so far already)
> >
> >
> > ToDo:
> > - follow along with the base vector patchset
> > - handle SR_VS access in _save_context and _secondary_start_sbi
> >
> >
> > Heiko Stuebner (2):
> >   RISC-V: define the elements of the VCSR vector CSR
> >   RISC-V: add T-Head vector errata handling
> >
> >  arch/riscv/Kconfig.erratas           |  13 +++
> >  arch/riscv/errata/thead/errata.c     |  32 ++++++
> >  arch/riscv/include/asm/csr.h         |  31 +++++-
> >  arch/riscv/include/asm/errata_list.h |  62 +++++++++++-
> >  arch/riscv/include/asm/vector.h      | 139 +++++++++++++++++++++++++--
> >  5 files changed, 261 insertions(+), 16 deletions(-)
> 
> I have no opposition to calling the T-Head vector stuff an errata 
> against V, the RISC-V folks have already made it quite apparent that 
> anything goes here.  I would like to get the standard V uABI sorted out 
> first, though, as there's still a lot of moving pieces there.

yeah, that's the reason the series is an RFC and is based on the main
vector series and I fully expect the main support to land first :-) .


> It's kind 
> of hard here as T-Head got thrown under the bus, but I'm not sure what 
> else to do about it.

Thankfully on the kernel-side the differences to implemeent both "at the
same time" are not that huge - userspace of course will need to figure
out their own solution.


Heiko