[llvm-dev] RFC: AArch64 SVE Assembler/Disassembler patches

Alex Bradbury via llvm-dev llvm-dev at lists.llvm.org
Thu Oct 19 06:44:44 PDT 2017


On 19 October 2017 at 05:12, Sander De Smalen via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> Hi,
>
>
>
> Probably a lot of you are attending interesting talks at LLVM Dev meeting
> this week, so I hope this message isn't completely lost in all the
> excitement.
>
>
>
> In the past month we have carved off our changes to LLVM's
> assembler/disassembler that implement the AArch64 SVE instruction set [1].
> These changes are split these up into individual patches that purely focus
> on the assembler and disassembler and have no link to the IR (yet). We would
> like to start sharing these patches with upstream LLVM.
>
>
>
> I have made a best effort to split these changes into small, manageable
> patches (or patch sets) that can be reviewed and applied individually over
> time, where each patch/patch-set aims to add a new instruction (or variant
> or addressing mode for an instruction). Each patch-set has corresponding
> tests to cover the added instruction.
>
>
>
> A first set of patches that implement SVE (unpredicated) ADD/SUB
> instructions can be found in Phabricator:
>
> - https://reviews.llvm.org/D39087
>
> - https://reviews.llvm.org/D39088
>
> - https://reviews.llvm.org/D39089
>
> - https://reviews.llvm.org/D39090
>
> - https://reviews.llvm.org/D39091
>
>
>
> Please let me know if you have any comments or suggestions to make
> sharing/reviewing these patches easier. I'm open to any feedback to get
> patches in the right shape!

Hi Sander, I'm glad to see progress towards getting SVE support
upstream. Compared to codegen, I would hope it's really quite easy to
get these reviewed and merged (i.e. I don't see any reason why it
should be more difficult than the AArch64v8.3 additions). From a quick
skim of the patchset, it seems sensible (AArch64 regulars can review
the fine details - I added a few comments on the last patch). Was
there a specific aspect of the patches + approach taken that you were
looking for feedback on?

Best,

Alex


More information about the llvm-dev mailing list