lkml.org 
[lkml]   [2022]   [Nov]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v3 2/3] RISC-V: uapi: add HWCAP for Bitmanip/Scalar Crypto
On Thu, Nov 24, 2022 at 05:54:00PM +0000, Conor Dooley wrote:
> On 24/11/2022 17:34, Samuel Ortiz wrote:
> > On Thu, Nov 24, 2022 at 05:20:37PM +0000, Conor Dooley wrote:
> >> On 24/11/2022 17:12, Samuel Ortiz wrote:
> >>> [You don't often get email from sameo@rivosinc.com. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
> >>>
> >>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> >>>
> >>> On Thu, Nov 24, 2022 at 11:55:01AM +0000, Conor Dooley wrote:
> >>>> On Thu, Nov 24, 2022 at 11:47:30AM +0100, Samuel Ortiz wrote:
> >>>>
> >>>>> Patch #1 is definitely needed regardless of which interface we pick for
> >>>>> exposing the ISA strings to userspace.
> >>>>
> >>>> I took another look at #1, and I feel more confused about what
> >>>> constitutes canonical order than I did before! If you know better than
> >>>> I, and you probably do since you're interested in these 6 month old
> >>>> patches, some insight would be appreciated!
> >>>
> >>> Assuming we don't go with hwcap, I dont think the order of the
> >>> riscv_isa_ext_id enum matters that much?
> >>
> >> The chief put it in canonical order so that's good enough for me!
> >>
> >>>
> >>> iiuc we're building the cpuinfo string from the riscv_isa_ext_data
> >>> array, and I think the current code is incorrect:
> >>>
> >>> static struct riscv_isa_ext_data isa_ext_arr[] = {
> >>> __RISCV_ISA_EXT_DATA(sscofpmf, RISCV_ISA_EXT_SSCOFPMF),
> >>> __RISCV_ISA_EXT_DATA(sstc, RISCV_ISA_EXT_SSTC),
> >>> __RISCV_ISA_EXT_DATA(svinval, RISCV_ISA_EXT_SVINVAL),
> >>> __RISCV_ISA_EXT_DATA(svpbmt, RISCV_ISA_EXT_SVPBMT),
> >>> __RISCV_ISA_EXT_DATA(zicbom, RISCV_ISA_EXT_ZICBOM),
> >>> __RISCV_ISA_EXT_DATA(zihintpause, RISCV_ISA_EXT_ZIHINTPAUSE),
> >>> __RISCV_ISA_EXT_DATA("", RISCV_ISA_EXT_MAX),
> >>> };
> >>>
> >>> zicbom and zihintpause should come before supervisor level extensions.
> >>> I'm going to send a patch for that.
> >>
> >> idk, Palmer explicitly re-ordered this:
> >> https://lore.kernel.org/linux-riscv/20220920204518.10988-1-palmer@rivosinc.com/
> >>
> >> By my reading of the isa manual, what Palmer did is correct as
> >> those are not "Additional Standard Extensions". /shrug
> >
> > Hmm, by their name (Z[a-b]+) they are Additional Standard Extensions.
> > What am I missing?
>
> Right, and this is where I get confused. Zam and Ztso *are* Additional
> Standard Extensions, I think we can agree on that one? For those
> extensions:
> \chapter{``Ztso'' Standard Extension for Total Store Ordering, v0.1}
> \chapter{``Zam'' Standard Extension for Misaligned Atomics, v0.1}
>
> They're also called out specifically in the table:
> https://github.com/riscv/riscv-isa-manual/blob/master/src/naming.tex#L147
>
> For Zihintpause however:
> \chapter{``Zihintpause'' Pause Hint, Version 2.0}
>
> See what I mean? I looked at the specs for the bitmanip stuff and for
> crypto, which both never mention being standard.

I *think* this is because Zihintpause, bitmap and crypto are ratified
but not yet part of an official spec (non-draft) release?


> That table has the caption:
> > The table also defines the canonical order in which extension names
> > must appear in the name string, with top-to-bottom in table
> > indicating first-to-last in the name string.
>
> It only calls out Zicsr, Zifencei, Zam and Ztso are being permitted
> before Sdef, but as I said I am not a specs person, so perhaps some
> of the extensions in question are intended to go there but have not
> yet been merged into the isa manual doc. Zihintpause *is* in the
> isa manual though but not specifically called out.
>
> Anyways, hopefully that at least helps with my line of thinking!
It does, thanks. It's a little confusing, I agree.

Cheers,
Samuel.

\
 
 \ /
  Last update: 2022-11-24 19:10    [W:0.048 / U:2.588 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site