qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v2 0/5] disas: add libvixl to support A64 disass


From: Laurent Desnogues
Subject: Re: [Qemu-devel] [PATCH v2 0/5] disas: add libvixl to support A64 disassembly
Date: Tue, 4 Feb 2014 12:47:39 +0100

On Wed, Jan 29, 2014 at 9:51 PM, Peter Maydell <address@hidden> wrote:
> On 29 January 2014 20:01, Laurent Desnogues <address@hidden> wrote:
>> On Tue, Jan 28, 2014 at 12:45 PM, Peter Maydell
>> <address@hidden> wrote:
>>> Ping for review/testing/comments on this version, please?
>>
>> I still dislike the idea of importing so much code in particular for
>> something that is incomplete:  as far as I can see, AdvSIMD
>> instructions are not supported.  The very least that should be
>> done would be to add a file that gives a rough status of what is
>> and what is not implemented.
>
> I can add some text to a brief README file, sure.
>
>> And what if the vixl authors never implement AdvSIMD?  This
>> is the most difficult part of Aarch64 to disassemble (integer and
>> FP instructions are really easy to disassemble).  Will someone
>> add these and will vixl authors accept the changes or will we
>> then start diverging from vixl implementation?  Is vixl even still
>> supported or in development (no commit for 6 months)?
>
> As I understand the situation, it is supported but the model
> is more "we'll push out a release occasionally when we've
> done a big chunk of work" rather than a continuously updated
> public-facing git tree.
>
> This is no worse at all (in fact better) than the situation we have
> with the binutils disassemblers in the tree at the moment -- those
> are effectively totally unmaintained by their upstream as
> a result of the GPL2/GPL3 split.
>
> I agree that it would be nice if we supported the SIMD
> instructions in the disassembler. Adding them to vixl
> should be no worse than adding them to anything else,
> and I'd rather have a disassembler that supported at
> least the integer set than none at all.

That looks OK to me.

Thanks,

Laurent



reply via email to

[Prev in Thread] Current Thread [Next in Thread]