Hi Jan, With Mercury compiling to C, can you see a benefit to have an easy to use bridge to Mercury – say as an alternative to C, to create foreign predicates for SWI-Prolog programs, that take fully advantage of Mercury’s optimized compiling of code that is annotated with modes and types. thanks, …

We have a good bridge to C. You typically use that for things you cannot do in plain Prolog, such as accessing the OS. Another use case is stuff for which there is a nice C library and it is just a waste of time to redo all that in Prolog (e.g., the TLS (SSL) binding). Finally, there is stuff tha…

Hi Jan, With the Mercury approach in mind, what swi-prolog would need to become faster, is support for modes, types and those predicate signatures that indicate whether predicates are deterministic. From what I read, these capabilities enables the compiler to create more optimized routines. It wo…

[image] grossdan: From what I read, these capabilities enables the compiler to create more optimized routines. I don’t think there is consensus about that. You can probably find some old posts by (notably) Bart Demoen on these topics. Neither is there much consensus on the value of native …

Hi Jan, Thank you. I guess, to (re?) hash some of the arguments – compiling to C and then machine code is probably faster for equivalent code that otherwise would run via a VM. But, likely, much more importantly: when mode is provided, Mercury apparently generates separate routines per argumen…

The problem in a dynamic language as Prolog is that you may declare some property of a predicate, but the predicate still has to verify this is actually true rather than continuing with undefined behavior. You can win a little by doing the tests in a more optimal order, but typically that is just…

All good questions. No, i took it the benchmarks that were published, but didn’t actually check if they are dated … Dan

[image] anon95304481: One key in good performance is for example multi-argument indexing and last call optimization Multi-argument indexing helps not/little for these traditional benchmarks. The speedup you can get from last call optimization (SWI-Prolog gets the space advantage, but no sig…

I guess, my use case, requires close to system work – bit operations, memory locality, pointers, and little call overheads, that’s where C / C++ usually excels – perhaps this is a rather uncommon use case for prolog. Dan

[image] jan: The problem in a dynamic language as Prolog is that you may declare some property of a predicate, but the predicate still has to verify this is actually true rather than continuing with undefined behavior. Many years ago, I helped someone speed up their LISP program by 10x or so …

One very important factor that has yet to be mentioned in detail is support. For years Mercury was listed as a research language, it was only in the last year or so I noticed that it no longer states that. However I did experiment with it over a decade ago when it compiled to Microsoft CIL , the VM …

Bridge to Mercury

General

Topic		Replies	Views
Curious: How does Prolog "Byte code" compare to .NET IL (and some thoughts about enterprise ready systems) Help!	17	3416	March 10, 2021
SWI does some good garbage-collection/optimisation these days General	23	1150	October 28, 2021
Reasons for sticking with SWI, as opposed to Mercury General	4	728	February 1, 2023
Types with SWI-Prolog Split Topic discussion	32	4300	May 11, 2023
Optimizing deterministic code - directive to creating compiled code without choice points Help!	37	2076	January 22, 2020

Bridge to Mercury

Related topics