Hi Pippijn.

> I am trying to build the alice compiler but it fails at  
> bootstrapping. Is
> there a way to skip bootstrapping alltogether?

Only by installing the pre-built packages. (You could also build  
everything but the bootstrap yourself and then take the .alc (Alice  
component) files from any of the available packages, since they are  
platform-independent. But you would need to recreate the proper  
directory structure manually.)

Any particular reason you are building it yourself?

> Building, installing and
> running alice 1.4 works fine.
>
>   http://www.xinutec.org/~pippijn/files/up/make.log
>
> the error is:
>
>   ../compiler/backend-common/NJ/SimplifyRec.sml:127.11-128.44 Error:
>   operator and operand don't agree [tycon mismatch]
>     operator domain: pat fld_info * lab * pat
>     operand:         pat fld_info * lab * exp
>     in expression:
>       Fld (info,label,exp)

That is really weird. Looking at the sources I cannot see any possible  
reason for this error. The compiler does not seem to treat the Fld  
constructor polymorphic enough in this (and the other) case.  
Unfortunately, I don't have access to SML/NJ 110.69 at the moment, but  
I suspect it is a problem with one of the more recent updates of NJ.  
The only thing I can suggest right now is downgrading (e.g. to  
110.0.7, which is known to work).


> Another question: are there any plans on improving alice's speed? In  
> my
> benchmark, it is about 10 times slower than SML/NJ. SML/NJ compiles  
> its
> bytecode to native code on the fly, just like seam (I believe). Maybe
> using a different jit library would help (libjit can do  
> optimisations on
> its own intermediate code). Alice is a nice research language with
> interesting features but it also being more than 20 times slower than
> (dynamically typed) common lisp (sbcl) is not good.

Not quite, SML/NJ is an offline (optimizing) native code compiler, not  
a jit compiler (even though it works in interactive sessions).

In any case, we are aware that the performance of Alice is not great,  
but given our limited resources, performance and optimization has  
never been a high priority for us. Also, the Alice project is in  
hibernation for the time being, so I'm sorry to say that there are no  
plans for improving performance at the moment.

Cheers,
- Andreas


_______________________________________________
alice-users mailing list
[email protected]
http://www.ps.uni-sb.de/mailman/listinfo/alice-users

Reply via email to