Re: [Qemu-devel] [RISU v2 00/11] Support for i386/x86_64 with vector extensions

[Alex Bennée]

Jan Bobek <address@hidden> writes:

> This patch series adds support for i386 and x86_64 architectures to
> RISU.  Notably, vector registers (SSE, AVX, AVX-512) are supported for
> verification of the apprentice. This is V2 of the series posted in
> [1].
>
> I decided not to drop the register definitions from the second patch
> as suggested by Alex Bennée [4], but replaced them in the fourth patch
> instead. This keeps the second and third patches code-motion only.
>
> I wasn't 100% sure how to acknowledge Richard's contributions in some
> of the patches, and eventually decided to include a Suggested-by:
> line. Let me know if that's (not) acceptable.
>
>  -Jan
>
> Changes in V2:
>   - included Richard Henderson's fix-ups [2] and vector register
>     support [3]
>   - replaced the magic numbers for XSAVE feature sets with symbolic
>     constants
>   - symbolic names ("sse", "avx", "avx512") as well as numbers are
>     accepted for the parameter --xfeatures

I'm not sure where my test went wrong but I guess it's around xfeatures.
The code says required argument but risu doesn't seem to stop me not
specifying it. I suspect we should default to the most minimal x86_64 we
can and explicitly enable extra features.

Storing xfeat in the stream is a good idea so people don't mix up their
dumps but we probably need more validation when running in master mode
that the feature you have enabled is actually available on the
processor. Otherwise you'll potentially end up generating test streams
on HW with no support and just get a bunch of undef noise ;-)

However the series is looking pretty good so far. Looking forward to the
next iteration.

--
Alex Bennée