Tier 1 architectures

2022-09-22 Thread Simon Peyton Jones
Ben, Matthew, Moritz, and friends

Is this wiki page about architectures still accurate?
https://gitlab.haskell.org/ghc/ghc/-/wikis/platforms

For example, ARM is not Tier 1, or "apple silicon".

Yet I know some of our developers have invested lots of effort in other
architectures, so maybe those efforts are not reflected here.

Relevant is Moritz's post about 32-bit architectures

.

We should in due course add Javascript and Web Assembly as Tier 1 back ends?

Are we saying "if your customer bases uses Tier 2 architectures, you can't
rely on GHC from one release to the next"?  I wonder if there are companies
for which Tier-2 architectures are mission-critical.  Mis-aligned
expectations cause upset.

I mention all this because it is relevant to our stability guarantees.
Every time we release we should point to this list.

Simon
___
ghc-devs mailing list
ghc-devs@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs


Re: Tier 1 architectures

2022-09-22 Thread Ben Gamari
Simon Peyton Jones  writes:

> Ben, Matthew, Moritz, and friends
>
> Is this wiki page about architectures still accurate?
> https://gitlab.haskell.org/ghc/ghc/-/wikis/platforms
>
Hi Simon,

Indeed there were a few inaccuracies on that page; I have fixed
these and generally cleaned up the page.

> For example, ARM is not Tier 1, or "apple silicon".
>
> Yet I know some of our developers have invested lots of effort in other
> architectures, so maybe those efforts are not reflected here.
>
Fairly recently there has been work on RISC-V (rv64) and PowerPC
(ppc64le), as well as some work on s390x via LLVM. However, I wouldn't
consider any of these Tier 1.

> Relevant is Moritz's post about 32-bit architectures
> 
> .
>
> We should in due course add Javascript and Web Assembly as Tier 1 back ends?
>
Indeed, that is the plan although 9.6 will rather ship these as Tier 2
targets.

> Are we saying "if your customer bases uses Tier 2 architectures, you can't
> rely on GHC from one release to the next"?  I wonder if there are companies
> for which Tier-2 architectures are mission-critical.  Mis-aligned
> expectations cause upset.
>
I have heard that some people are using amd64/FreeBSD, although that can very
nearly be promoted to a Tier 1 now. Bodigrim once mentioned that he was
considering deploying Haskell on s390x, although I'm not sure what
became of that. Otherwise I would be quite surprised if any commercial
customers are relying on any of the other Tier 2 or Tier 3 platforms.

> I mention all this because it is relevant to our stability guarantees.
> Every time we release we should point to this list.
>
My sense is that this list should ideally rather live in the users guide
since it changes from release to release.

Cheers,

- Ben


signature.asc
Description: PGP signature
___
ghc-devs mailing list
ghc-devs@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs


Re: Tier 1 architectures

2022-09-22 Thread Moritz Angermann
Thanks Ben!

Just FYI: We do have folks actively deploying to iOS and Android ๐Ÿ™ˆ at
simplex chat.

I do agree that we want this to be in the user guide though; as itโ€™s quite
version dependent.

Cheers,
  Moritz

On Fri, 23 Sep 2022 at 3:22 AM, Ben Gamari  wrote:

> Simon Peyton Jones  writes:
>
> > Ben, Matthew, Moritz, and friends
> >
> > Is this wiki page about architectures still accurate?
> > https://gitlab.haskell.org/ghc/ghc/-/wikis/platforms
> >
> Hi Simon,
>
> Indeed there were a few inaccuracies on that page; I have fixed
> these and generally cleaned up the page.
>
> > For example, ARM is not Tier 1, or "apple silicon".
> >
> > Yet I know some of our developers have invested lots of effort in other
> > architectures, so maybe those efforts are not reflected here.
> >
> Fairly recently there has been work on RISC-V (rv64) and PowerPC
> (ppc64le), as well as some work on s390x via LLVM. However, I wouldn't
> consider any of these Tier 1.
>
> > Relevant is Moritz's post about 32-bit architectures
> > <
> https://discourse.haskell.org/t/running-project-built-on-raspberry-pi-with-cabal-gives-weird-errors/2429/6
> >
> > .
> >
> > We should in due course add Javascript and Web Assembly as Tier 1 back
> ends?
> >
> Indeed, that is the plan although 9.6 will rather ship these as Tier 2
> targets.
>
> > Are we saying "if your customer bases uses Tier 2 architectures, you
> can't
> > rely on GHC from one release to the next"?  I wonder if there are
> companies
> > for which Tier-2 architectures are mission-critical.  Mis-aligned
> > expectations cause upset.
> >
> I have heard that some people are using amd64/FreeBSD, although that can
> very
> nearly be promoted to a Tier 1 now. Bodigrim once mentioned that he was
> considering deploying Haskell on s390x, although I'm not sure what
> became of that. Otherwise I would be quite surprised if any commercial
> customers are relying on any of the other Tier 2 or Tier 3 platforms.
>
> > I mention all this because it is relevant to our stability guarantees.
> > Every time we release we should point to this list.
> >
> My sense is that this list should ideally rather live in the users guide
> since it changes from release to release.
>
> Cheers,
>
> - Ben
> ___
> ghc-devs mailing list
> ghc-devs@haskell.org
> http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
>
___
ghc-devs mailing list
ghc-devs@haskell.org
http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs