Re: [Xen-devel] [PATCH v8 49/50] x86emul: add a SHA test case to the harness

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 21/06/2019 15:10, Jan Beulich wrote:
>>>> On 21.06.19 at 15:51, <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 15/03/2019 11:08, Jan Beulich wrote:
>>> +        /*
>>> +         * SHA256RNDS2
>>> +         *
>>> +         * SRC1 = { C0, D0, G0, H0 }
>>> +         * SRC2 = { A0, B0, E0, F0 }
>>> +         * XMM0 = W' = { ?, ?, WK1, WK0 }
>>> +         *
>>> +         * (NB that the notation again is not C-like, i.e. elements are 
>>> listed
>>> +         * high-to-low everywhere in this comment.)
>>> +         *
>>> +         * Ch(E,F,G) = (E & F) ^ (~E & G)
>>> +         * Maj(A,B,C) = (A & B) ^ (A & C) ^ (B & C)
>>> +         *
>>> +         * Σ0(A) = ROR2(A) ^ ROR13(A) ^ ROR22(A)
>>> +         * Σ1(E) = ROR6(E) ^ ROR11(E) ^ ROR25(E)
>> This looks like some encoding problems (and later on in the comment).  I
>> can't tell whether it is a consequence of the email or something present
>> in the underlying patch. 
> Oh - I forgot to properly enforce UTF8 while composing the mail.
> This is what it'll looks like when committed:
>
>         /*
>          * SHA256RNDS2
>          *
>          * SRC1 = { C0, D0, G0, H0 }
>          * SRC2 = { A0, B0, E0, F0 }
>          * XMM0 = W' = { ?, ?, WK1, WK0 }
>          *
>          * (NB that the notation again is not C-like, i.e. elements are listed
>          * high-to-low everywhere in this comment.)
>          *
>          * Ch(E,F,G) = (E & F) ^ (~E & G)
>          * Maj(A,B,C) = (A & B) ^ (A & C) ^ (B & C)
>          *
>          * Σ0(A) = ROR2(A) ^ ROR13(A) ^ ROR22(A)
>          * Σ1(E) = ROR6(E) ^ ROR11(E) ^ ROR25(E)

That reads much better.

>
>>> +static bool simd_check_avx_sha(void)
>>> +{
>>> +    return cpu_has_sha && cpu_has_avx;
>> I can't locate any [E]VEX encoding information for the SHA
>> instructions.  Is this a side effect of the rest of the test algorithm,
>> or am I missing something in the manual?
> There are only legacy forms as of now. Note the single sentence in
> the commit message:
>
> "Also use this for AVX512VL VPRO{L,R}{,V}D as well as some further shifts
>  testing."
>
> The AVX and AVX512F flavors of the test are good test cases for other
> insns; they are still only utilizing legacy encoded SHA insns.

Ok.  Acked-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel