Re: [Haskell-cafe] Compilers: Why do we need a core language?

newer
Portability of Safe Haskell...

older
Re: [Haskell-cafe] Compilers: Why...

citb＠lavabit.com

20 Nov 2012 20 Nov '12

10:49 a.m.

...

What would be the point in doing so?

Well, I don't know. Would it save some time? Why bother with a core language?

Show replies by date

AUGER Cédric

20 Nov 20 Nov

11:12 a.m.

New subject: Compilers: Why do we need a core language?

Le Tue, 20 Nov 2012 10:49:01 -0500 (EST), citb@lavabit.com a écrit :

...

...
What would be the point in doing so?

Well, I don't know. Would it save some time? Why bother with a core language?

The compilation process might be slightly faster, but I guess it wouldn't be much noticeable. Also I guess having a core language eases porting to new architectures, you "just" have to port a simple core language rather than porting a complex language. The semantics of the core language is also rather simple, so you can use it to explain and understand how it works, then for the high level part, you can give a semantics by compilation to the core language. Finally, adding a new feature seems easier and less error prone if you have to translate it into the core language rather than compiling it directly. I never studied the Haskell compiler so I do not know the details, but I think that having a core language is a good idea. It is also good to have some VM to porting purposes. If you have N source languages and M target machines, by providing N+M stuff, you get N×M compilers !

...

_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe

Richard O'Keefe

6:08 p.m.

New subject: Compilers: Why do we need a core language?

On 21/11/2012, at 4:49 AM, wrote:

...

...
What would be the point in doing so?

Well, I don't know. Would it save some time? Why bother with a core language?

For a high level language (and for this purpose, even Fortran 66 counts as "high level") you really don't _want_ a direct translation from source code to object code. You want to eliminate unused code and you want to do all sorts of analyses and improvements. It is *much* easier to do all that to a small core language than to the full source language.

Jacques Carette

22 Nov 22 Nov

7:56 a.m.

New subject: Compilers: Why do we need a core language?

On 20/11/2012 6:08 PM, Richard O'Keefe wrote:

...

On 21/11/2012, at 4:49 AM, wrote:

...
Well, I don't know. Would it save some time? Why bother with a core language? For a high level language (and for this purpose, even Fortran 66 counts as "high level") you really don't _want_ a direct translation from source code to object code. You want to eliminate unused code and you want to do all sorts of analyses and improvements. It is *much* easier to do all that to a small core language than to the full source language.

Actually, here I disagree. It might be much 'easier' for the programmers to do it for a small core language, but it may turn out to be much, much less effective. I 'discovered' this when (co-)writing a partial evaluator for Maple: we actually made our internal language *larger*, so that we could encode more invariants syntactically. This ended up making our jobs considerably easier, because we did not have to work so hard on doing fancy analyses to recover information that would otherwise have been completely obvious. Yes, there were a lot more cases, but each case was relatively easy; the alternative was a small number of extremely difficult cases. Jacques

Brandon Allbery

11:52 a.m.

New subject: Compilers: Why do we need a core language?

On Thu, Nov 22, 2012 at 7:56 AM, Jacques Carette wrote:

...

On 20/11/2012 6:08 PM, Richard O'Keefe wrote:

...
On 21/11/2012, at 4:49 AM, wrote:

Well, I don't know. Would it save some time? Why bother with a core

...
language?

For a high level language (and for this purpose, even Fortran 66 counts as "high level") you really don't _want_ a direct translation from source code to object code. You want to eliminate unused code and you want to do all sorts of analyses and improvements. It is *much* easier to do all that to a small core language than to the full source language.

Actually, here I disagree. It might be much 'easier' for the programmers to do it for a small core language, but it may turn out to be much, much less effective. I 'discovered' this when (co-)writing a partial evaluator for Maple:

You're still using a core language, though; just with a slightly different focus than Haskell's. I already mentioned gcc's internal language, which similarly is larger (semantically; syntactically it's sexprs). What combination is more appropriate depends on the language and the compiler implementation. -- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix/linux, openafs, kerberos, infrastructure http://sinenomine.net

Jacques Carette

23 Nov 23 Nov

8:39 a.m.

New subject: Compilers: Why do we need a core language?

On 22/11/2012 11:52 AM, Brandon Allbery wrote:

...

On Thu, Nov 22, 2012 at 7:56 AM, Jacques Carette mailto:carette@mcmaster.ca> wrote:

On 20/11/2012 6:08 PM, Richard O'Keefe wrote:

On 21/11/2012, at 4:49 AM, mailto:citb@lavabit.com> wrote:

Well, I don't know. Would it save some time? Why bother with a core language?

For a high level language (and for this purpose, even Fortran 66 counts as "high level") you really don't _want_ a direct translation from source code to object code. You want to eliminate unused code and you want to do all sorts of analyses and improvements. It is *much* easier to do all that to a small core language than to the full source language.

Actually, here I disagree. It might be much 'easier' for the programmers to do it for a small core language, but it may turn out to be much, much less effective. I 'discovered' this when (co-)writing a partial evaluator for Maple:

You're still using a core language, though; just with a slightly different focus than Haskell's. I already mentioned gcc's internal language, which similarly is larger (semantically; syntactically it's sexprs). What combination is more appropriate depends on the language and the compiler implementation.

Right, we agree: it is not 'core language' I disagreed with, it is 'smaller core language'. And we also agree that smaller/larger depends on the eventual application. But 'smaller core language' is so ingrained as "conventional wisdom" that I felt compelled to offer evidence against this bit of folklore. Jacques

Mike Meyer

9:59 a.m.

New subject: Compilers: Why do we need a core language?

Jacques Carette wrote:

...

On 22/11/2012 11:52 AM, Brandon Allbery wrote:

...
On Thu, Nov 22, 2012 at 7:56 AM, Jacques Carette
...
mailto:carette@mcmaster.ca> wrote:

On 20/11/2012 6:08 PM, Richard O'Keefe wrote:

On 21/11/2012, at 4:49 AM, mailto:citb@lavabit.com> wrote:

Well, I don't know. Would it save some time? Why bother with a core language?

For a high level language (and for this purpose, even Fortran 66 counts as "high level") you really don't _want_ a direct translation from source code to object code. You want to eliminate unused code and you want to do all sorts of analyses and improvements. It is *much* easier to do all that to a small core language than to the full source language.

Actually, here I disagree. It might be much 'easier' for the programmers to do it for a small core language, but it may turn out to be much, much less effective. I 'discovered' this when (co-)writing a partial evaluator for Maple:

You're still using a core language, though; just with a slightly different focus than Haskell's. I already mentioned gcc's internal language, which similarly is larger (semantically; syntactically it's

...
sexprs). What combination is more appropriate depends on the language and the compiler implementation.

Right, we agree: it is not 'core language' I disagreed with, it is 'smaller core language'. And we also agree that smaller/larger depends on the eventual application. But 'smaller core language' is so ingrained as "conventional wisdom" that I felt compelled to offer evidence against this bit of folklore.

I don't think you're evidence contacts that bit of folklore. But as stated it's vague. In particular, is "smaller" relative to the full source language, or is it absolute (in which case you should compile to a RISC architecture and optimize that :-)? Since the latter seems silly, I have to ask if your core language for Maple was larger than Maple? -- Sent from my Android tablet with K-9 Mail. Please excuse my swyping.

Brandon Allbery

10:17 a.m.

New subject: Compilers: Why do we need a core language?

On Fri, Nov 23, 2012 at 9:59 AM, Mike Meyer wrote:

...

I have to ask if your core language for Maple was larger than Maple?

I can think of several reasons why it would make sense, especially if the core language is much harder for a human to write for but much easier to optimize and/or generate machine code for. -- brandon s allbery kf8nh sine nomine associates allbery.b@gmail.com ballbery@sinenomine.net unix/linux, openafs, kerberos, infrastructure http://sinenomine.net

Jacques Carette

10:45 a.m.

New subject: Compilers: Why do we need a core language?

On 23/11/2012 9:59 AM, Mike Meyer wrote:

...

[...] I have to ask if your core language for Maple was larger than Maple?

Yes. Maple 10 had 62 cases in its AST, we had 75 (p.13 of [1]) Jacques [1] http://www.cas.mcmaster.ca/~carette/publications/scp_MaplePE.pdf

4561

Age (days ago)

4564

Last active (days ago)

List overview

Download

8 comments

6 participants

participants (6)

AUGER Cédric
Brandon Allbery
citb＠lavabit.com
Jacques Carette
Mike Meyer
Richard O'Keefe