https://bugs.kde.org/show_bug.cgi?id=414268
--- Comment #6 from ahashmi <assad.has...@linaro.org> ---
Thanks for reviewing/testing Mark Wielaard and Peter Maydell.
> There are two machines this seems to work correctly on:
> Arch and hwcaps: ARM64, LittleEndian, v8.1-rdm-atomics
> Arch and hwcaps: ARM64, LittleEndian, v8.2-dpcvap-rdm-atomics-fp16-vfp16
>
> But one where I get Arch and hwcaps: ARM64, LittleEndian, v8.1 and then the
> impossible happens:
>
> VEX: SQRDML{AH|SH} instructions are mandatory from v8.1 onwards.
> Found: v8.1
> Cannot continue. Good-bye
Fixed.
Just goes to show I need to test on a wider range of hardware before
submitting!
> The right way to test fields in these ID registers is to look at the entire
> 4-bit number and test that it is greater than or equal to the value
> representing the feature you're looking for. (See D13.1.3 "Principles of the
> ID
> scheme for fields in ID registers" in the v8A Arm ARM.)
Fixed. Good spot, it was staring me in the face!
> For A-profile my recommendation is as far as you possibly can to avoid having
> any "is this v8.1?" or "is this v8.2?" tests and instead to stick strictly to
> "is feature X present?". The architecture permits a v8.x implementation to
I thought about this disconnect between version and supported feature set when
implementing the initial patch and considered omitting versions altogether. I
decided against it because I don't yet have enough experience with Valgrind,
particularly the decode-to-IR logic and the tools pipeline. On the basis that
the version is set when its mandatory option(s) are established and it may
simplify things down the line, I prefer to leave this for now. If it transpires
versions are redundant, I'll remove them at the earliest sign.
--
You are receiving this mail because:
You are watching all bug changes.
