[Embench] New Embench FP suite repository

Jeremy Bennett jeremy.bennett at embecosm.com
Wed Jul 22 20:47:02 CEST 2020

Hash: SHA1

On 22/07/2020 19:24, Ray Simar wrote:
> Hi all,
> This all sounds great.  I wanted to ask about the long term
> architecture of the repository.  I can see that this repository
> might include a number of things.  For example, it would certainly
> have floating-point benchmarks and the necessary data to test
> them.
> But I wonder about other content.  For example, we might have adoc
> files to document the philosophy behind the benchmarks, rules for
> running them etc.  And, at some point there might be make files to
> build them etc., though I could see these residing elsewhere.

(moving to mailing list discussion)

Hi Ray,

For embench-iot we went down the python route, because we perceived
that Makefile technology was not flexible enough for all the ways you
might want to both build and run the benchmarks. The original BEEBS
was makefile based, and really painful to use and maintain.

There is a hope that we will move to a Western Digital platform for
building and running everything.

We have a documentation directory for embench-iot, and have generally
stuck to MarkDown, since it seems so ubiquitous, and everyone knows
how to use it. I have used Asciidoc - the RISC-V Foundation likes it,
and it is supported by GitHub. The problem is it has something of a
learning curve for new contributors. Plus the tool support for
previewing before you commit is poor. But would be interesting to hear
what others think.

One thing we did establish was a strict discipline on git commit
messages, following GNU ChangeLog principles. If this is to be a long
term project, maintaining good records is essential - the embench-iot
repo has examples.

> Now, I know I am new to this effort, so I want to be sure I do
> things with the repository that are consistent with the long term
> vision.  We could begin to flesh out the repository with the stubs
> for this now.  My experience is that it is best to rought out some
> of this now, rather than later.  But, I’ll follow your lead on
> this.

Embench is stil quite new, so we are establishing how we do things.
Input from the wider community is welcome, and you have the
opportunity to try new things with the floating point Embench project.

Best wishes,


> Thanks for the guidance!
> All the best, Ray
>> On Jul 22, 2020, at 12:30 PM, Jeremy Bennett
>> <jeremy.bennett at embecosm.com
>> <mailto:jeremy.bennett at embecosm.com>>
> On 22/07/2020 17:07, Ray Simar wrote:
>>>> Sounds great!  I accepted the invitation to embench.  I’m
>>>> also going
> to use this as an opportunity to try out some new (to me) GitHub
> tools.  This time I’ll be using sublime merge.  Any other
> recommendations are welcome.
> Hi Ray,
> I've set things up, so you can't commit direct to master, you have
> to submit a pull request.
> Recommended workflow is to fork the repository into your personal
> account, then submit pull requests to the embench repo from your
> personal fork. Even for your own work - you can approve your own
> pull request. That way the official repo does not get cluttered
> with a lot of development branches.
> There is an informal convention that you name your own work
> branches prefixed by your user name, and if it is fixing an issue,
> the issue number. For example raysimar-issue65-fix. That way the
> maintainer (which is you) can easily see for each pull request
> whose work it is and what it relates to.
> I'm a mid-level git user, but I have staff who are gurus, so if
> you run into problems, one way or another I can probably get an
> answer :-)
> Thanks very much for your support - great to have a new person
> driving part of the project. Let me know if I can help.
> Best wishes,
> Jeremy
>>>> All the best, Ray
>>>>> On Jul 21, 2020, at 11:03 AM, Jeremy Bennett
> <jeremy.bennett at embecosm.com <mailto:jeremy.bennett at embecosm.com>>
> wrote:
>>>> On 20/07/2020 22:05, Ray Simar wrote:
>>>>>>> Hi Jeremy,
>>>>>>> Try raysimar for my GitHub handle.
>>>>>>> FYI, I am usual a casual user of GitHub, so don’t be
>>>>>>> afraid to give
>>>> me pointer!
>>>> Hi Ray,
>>>> You should have access as a maintainer of the repository.
>>>> I've also made Paolo Savini a maintainer for convenience.
>>>> General way of working is for you to create your own fork of
>>>> the repository, make changes on a branch within your fork
>>>> (typically including your id in the name of the branch), then
>>>> submit a pull-request when you are ready to incorporate your
>>>> commits into the main repository. You, Paolo or me can
>>>> approve merge requests in the repo.
>>>> Best wishes,
>>>> Jeremy
>>>>>>> All the best, Ray
>>>>>>>> On Jul 20, 2020, at 11:47 AM, Jeremy Bennett
>>>> <jeremy.bennett at embecosm.com
>>>> <mailto:jeremy.bennett at embecosm.com>> wrote:
>>>>>>> Hi Ray,
>>>>>>> I've created a new repository for the Embench floating
>>>>>>> point benchmark suite.
>>>>>>> https://github.com/embench/embench-fp
>>>>>>> Could you let me know your GitHub handle and I'll add
>>>>>>> you as administrator.
>>>>>>> Best wishes,
>>>>>>> Jeremy

- -- 
Cell: +44 (7970) 676050
SkypeID: jeremybennett
Twitter: @jeremypbennett
Email: jeremy.bennett at embecosm.com
Web: www.embecosm.com
PGP key: 1024D/BEF58172FB4754E1 2009-03-20


More information about the Embench mailing list