| Re: [eigen] Matrix support in Clang |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Matrix support in Clang
- From: Mark Borgerding <mark@xxxxxxxxxxxxxx>
- Date: Tue, 18 Feb 2020 09:41:59 -0500
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/simple; d=borgerding.net; s=20191115-ca3xhr5z; t=1582036920; bh=Ocw/1rFyT7rSHaPOo/YeaEf3piOrAk/8YPulc5LDVuA=; h=Subject:To:From:Date:From; b=bOD5kntBgtNui/kem5RfYdqIKl4S5+hX2YTxvQ8mjiaknZddJnPUHrWQfDD5/j1zC jWvLIPJjmHr2Q4tPmpztCjmriUhNihUyu0QC/TCwo8BM69INqQOhbKxcLbuGPgM8eb byluJifXFCJaDYweJAdr+NSEvMD8HlUmFStlrEm0=
Caveat: I'm an interested but mostly ignorant observer.
It seems to me like the clang/llvm developers may want to work on a
`eigen_clang` fork in order to utilize the testing and benchmark code of
Eigen.
Showcasing against baseline Eigen would provide an important data point
for the "so what?" question that all projects must face.
It would also provide a testing baseline. I'm sure there would need to
be additional tests specific to the domain of compilers.
I think that baseline Eigen vs. the clang/llvm matrix support could be
the sort of friendly rivalry that benefits both sides (and subsequently
all users).
-- Mark Borgerding
On 2/17/20 5:45 PM, Florian Hahn wrote:
> Hi,
>
>> On 12 Feb 2020, at 14:28, Christoph Hertzberg <chtz@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:
>>
>> Hi, sorry for the late reply.
>>
>> This certainly looks interesting (I just glanced over it so far), but I guess for Eigen this could only serve as "inspiration" for better code, since we need to stay compiler-independent.
>>
>> At the moment, I don't think making different code-paths just to support this, is a good idea. Although, eventually this could be handled like a different "architecture”.
> Great, thanks for taking a look! It makes perfect sense to wait until support is more widely available and has matured.
>
> Cheers,
> Florian
>
>
>
>