[PATCH] RISC-V: Update Zihintpause extension version

Wed Feb 2 00:36:39 GMT 2022

On 2022/02/02 9:12, Palmer Dabbelt wrote:
> On Tue, 01 Feb 2022 05:46:56 PST (-0800), research_trasio@irq.a4lg.com wrote:
>> Sorry for no background on the first submission.  Probably I will have
>> to create cover letters even if the patch is very small.
>>
>> Yup, old 2.0 -> new 1.0 -> new 2.0.
>>
>> I list some RISC-V ISA Manual commits with author/committer dates:
>>
>> [Old 2.0 (initially proposed)]
>> https://github.com/riscv/riscv-isa-manual/commit/ea9410a6a5ea2b7cabfbef1ed0bf3507bb784c36 (2019-08-26 / 2020-10-17)
>> [New 1.0 (version number changed)]
>> https://github.com/riscv/riscv-isa-manual/commit/773a6c4cc9db7585d42ec732d5db24f930d1157a (2020-10-09 / 2020-10-17)
>> [New 1.0 (Ratified!?)]
>> https://github.com/riscv/riscv-isa-manual/commit/d8ab5c78c207d7138684fc92cc8f34c463975620 (2021-08-02 / 2021-08-02)
>> [New 2.0 (Ratified version corrected in a day)]
>> https://github.com/riscv/riscv-isa-manual/commit/7a58119dad5bd43e2171a26b56ef60f9591a1c9c (2021-08-02 / 2021-08-02)
>>
>> Latest draft still says that ratified version is "2.0":
>> https://github.com/riscv/riscv-isa-manual/releases/tag/draft-20220125-23a1255
>>
>> On 2022/02/01 11:20, Andrew Waterman wrote:
>>> On Mon, Jan 31, 2022 at 8:45 AM Palmer Dabbelt <palmer@dabbelt.com> wrote:
>>>>
>>>> On Sun, 30 Jan 2022 03:37:41 PST (-0800), binutils@sourceware.org wrote:
>>>>> Because ratified Zihintpause extension has a version number of 2.0
>>>>> (not 1.0), we should update the number.
>>>>
>>>> The ISA manual also says it's ratified so having it in the draft set is
>>>> kind of odd, but I'm not sure which spec version it should be included
>>>> in.  The commits saying it was ratified are in August 2021, but I can't
>>>> find a spec version that's listed as ratified and contains the new 2.0
>>>> (there was also an earlier 2.0, from before the 1.0).
>>>>
>>>> Not sure if this matters, though.  Looks like the only non-commentary
>>>> differents are: from the original 2.0 to 1.0, the text "No architectural
>>>> state is changed." was added; and from 1.0 to 2.0 the text "{\em fm}=0,
>>>> {\em rd}={\tt x0}, and {\em rs1}={\tt x0}" was added (to the encoding
>>>> section).  The second one isn't substantive, and while the first one
>>>> reads like a change to me IIRC we've been through that before and I just
>>>> have the definition of architectural state change wrong.
>>>
>>> The first one isn't substantive, either, because of the definition of
>>> HINT.  It's just a clarification.
>>
>> Although that version change is not substantive, this version
> 
> OK, sorry, I guess I'm just confused here.  No big deal, though -- if they're the same that's all that matters, I don't really care about the why part.
> 
>> number is copied to RISC-V attributes section
>> (e.g. with -march=rv32gc_zihintpause) so I thought it would be better to
>> be an actual ratified version.
>>
>> (Before the patch)
>> Contents of section .riscv.attributes:
>>  0000 415a0000 00726973 63760001 50000000  AZ...riscv..P...
>>  0010 05727633 32693270 315f6d32 70305f61  .rv32i2p1_m2p0_a
>>  0020 3270315f 66327032 5f643270 325f6332  2p1_f2p2_d2p2_c2
>>  0030 70305f7a 69637372 3270305f 7a696665  p0_zicsr2p0_zife
>>  0040 6e636569 3270305f 7a696869 6e747061  ncei2p0_zihintpa
>>  0050 75736531 70300008 010a0b             use1p0.....
>>
>> (After the patch [offset 0x53 is changed])
>> Contents of section .riscv.attributes:
>>  0000 415a0000 00726973 63760001 50000000  AZ...riscv..P...
>>  0010 05727633 32693270 315f6d32 70305f61  .rv32i2p1_m2p0_a
>>  0020 3270315f 66327032 5f643270 325f6332  2p1_f2p2_d2p2_c2
>>  0030 70305f7a 69637372 3270305f 7a696665  p0_zicsr2p0_zife
>>  0040 6e636569 3270305f 7a696869 6e747061  ncei2p0_zihintpa
>>  0050 75736532 70300008 010a0b             use2p0.....
> 
> Works for me, and given that there's no difference between the three I don't see any reason to worry about the spec versioning.
> 
> Looks like the copyright assignment might still be in limbo (I left my FSF keys at home, so I can't check right now), we can probably call this one trivial but I don't see any particular rush and given that a lot of your other patch sets are defiantly not trivial that has to get sorted out anyway.

As I mentioned in my latest (and big) Zfinx patch set (Part 1), my
copyright assignment to FSF is completed.
<https://sourceware.org/pipermail/binutils/2022-February/119570.html>

I didn't detailed it in the e-mail above but... it's completed in
2022-01-20 [gnu.org #1788718].

I agree that we don't have to rush about it.  As long as patches are
reviewed properly, that's fine.

Still, I hope that my Zfinx patchsets are sorted out in February so that
we can implement Zfh/Zfhmin on the tree with my Zfinx patchsets
either accepted, rejected or postponed.

Thanks,
Tsukasa

> 
> Thanks!
> 
>>
>> Thanks,
>> Tsukasa
>>
>>>
>>>>
>>>>>
>>>>> bfd/ChangeLog:
>>>>>
>>>>>       * elfxx-riscv.c (riscv_supported_std_z_ext): Update version
>>>>>       number of Zihintpause extension.
>>>>> ---
>>>>>  bfd/elfxx-riscv.c | 2 +-
>>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/bfd/elfxx-riscv.c b/bfd/elfxx-riscv.c
>>>>> index 9f52bb545ac..29755a6cb0a 100644
>>>>> --- a/bfd/elfxx-riscv.c
>>>>> +++ b/bfd/elfxx-riscv.c
>>>>> @@ -1183,7 +1183,7 @@ static struct riscv_supported_ext riscv_supported_std_z_ext[] =
>>>>>    {"zicsr",          ISA_SPEC_CLASS_20190608,        2, 0,  0 },
>>>>>    {"zifencei",               ISA_SPEC_CLASS_20191213,        2, 0,  0 },
>>>>>    {"zifencei",               ISA_SPEC_CLASS_20190608,        2, 0,  0 },
>>>>> -  {"zihintpause",    ISA_SPEC_CLASS_DRAFT,           1, 0,  0 },
>>>>> +  {"zihintpause",    ISA_SPEC_CLASS_DRAFT,           2, 0,  0 },
>>>>>    {"zfinx",          ISA_SPEC_CLASS_DRAFT,           1, 0,  0 },
>>>>>    {"zdinx",          ISA_SPEC_CLASS_DRAFT,           1, 0,  0 },
>>>>>    {"zqinx",          ISA_SPEC_CLASS_DRAFT,           1, 0,  0 },
>>>>>
>>>>> base-commit: 35b5767cf47169d11aa059fce0ed5b0fc213045d
>>>
> 