llvm dev - Mar 2015 - [LLVMdev] Introduction of AArch64/ARM v8.1 extension

Hi guys,

I'm going to upstream newly-developed support for AArch64 and ARM backends
for v8.1 architecture extension.

Briefly it is described on
http://community.arm.com/groups/processors/blog/2014/12/02/the-armv8-a-archi
tecture-and-its-ongoing-development .

The complete technical manual is available to shortlist of ARM partners but
, unfortunately, official public update of ARM Architecture Reference Manual
for this extension is not issued yet.

This way, my patches going to be an early-access code for the architecture
design, that is strongly believed to be final.

The whole bunch of patches will comprise:

-          Some refactoring,  

-          5 separate small extensions to AArch64 backend (RDMA, Atomic,
PAN, VHE, LOR)

-          2 separate small extensions for ARM backend (RDMA, PAN),

-          Introducing of architecture name (armv8.1-a) and generic cpu name
for llvm (generic-armv8.1-a)

-          Introducing of architecture name and generic cpu name for clang.

I've already submitted 

-          One patch for one extension for AArch64 backend
(http://reviews.llvm.org/D7998) and

-          One patch for one extension for ARM backend
(http://reviews.llvm.org/D7999)

 

The following patches will be based on previous, in a chain.

 

Would you please drop me a hint, whether should I provide the following
patches 

-          one-by one, every time waiting for approval, optional re-submit
and commit,

-          all patches at a time in a chain, based one upon other, or

-          merge all the patches in one huge one

 

Thanks, 

Vladimir
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150302/58ab2b91/attachment.html>

Hi Vladimir,

It'll be good to see this. Some interesting extensions there! The
basic outline looks pretty reasonable.

On 2 March 2015 at 03:09, Vladimir Sukharev <vladimir.sukharev at arm.com>
wrote:
> Would you please drop me a hint, whether should I provide the following
> patches
>
> -          one-by one, every time waiting for approval, optional re-submit
> and commit,
>
> -          all patches at a time in a chain, based one upon other, or
Either of these could work, though the first probably makes things
easier for reviewers unless the reasons for the decisions don't become
apparent until later.
> -          merge all the patches in one huge one
This isn't a good idea.

Cheers.

Tim.