Have you any information on Prolog’s performance in the 2025 “Billion Nested Loop Iteration” test? Three hundred different programming languages are said to have taken part in. [My apologies if a thread already exists on this issue]

Using :- set_prolog_flag(optimize, true). on the original function reduces execution time by 50%. Replacing the original code with your code, keeping optimize on, reduces another 30% of the execution time. Wait: the original code has a 37% distortion in profiler time, and the new code has 2%. Thi…

“What would be interesting would be a backtracking heavy benchmark” … and why not! It is not just a question of relative advantage, but also one of higher level of computational capability.

[image] anon19431257: Maybe nb_setarg wouldn’t change much in this case, it’s the loops themselves Well, profiling shows that nb_setarg is 22% of the whole runtime, so pretty important I would say. [image] anon19431257: I have written some plain Prolog (below) and I still get performance…

[image] abaljeu: Twice as slow as python is not excellent. I wonder how Prolog compares with Haskell? :slight_smile:

[image] AQ2017: and why not! It is not just a question of relative advantage, but also one of higher level of computational capability No sure how to read this. Prolog is designed to search for solutions (variable bindings) that satisfy some logic based specification. Pattern matching ba…

[image] jan: Few algorithms based arrays and notably updating arrays can be represented comfortably and efficient in Prolog. Richard O’Keefe implemented FFTs in Prolog: Learning to implement data structures & algorithms in Prolog - #4 by peter.ludemann

[image] anon19431257: BTW, how does 100 millions become 1 billion? LOL. It’s explained in the README of the project: Changes to the benchmarks compared to legacy runner … loops: The inner loop is now 10k, again to allow slower languages to complete more runs. It was 100k before .

Perhaps a move away from plain number crunching to logic intensive number crunching? Even better, provide Prolog with the missing “low-level” extra capabilities? … Sounds like F1 racing.

Hi, with your code as is, this is the result I get with the same query as you. I compared this to a naive “count nested loops” query: 101 ?- time( loops(40, R) ). % 200,051,123 inferences, 5.729 CPU in 5.742 seconds (100% CPU, 34922005 Lips) R = 198000. 101 ?- time( aggregate_all(count, ( between(…

So, the benchmark/challenge is measuring 100 million recursive call where each call does a load_memory + add + store_memory operation. This took you 5.7 sec. You are comparing that with: 100 million backtracking + a simple non-backtracking counter. that took 5 sec 100 million backtracking. that t…

Prolog's performance on "Billion Nested Loop Iteration"

General

peter.ludemann January 12, 2026, 9:10pm 13

Richard O’Keefe implemented FFTs in Prolog: Learning to implement data structures & algorithms in Prolog - #4 by peter.ludemann

Topic		Replies	Views
Six ways to iterate in Prolog General	44	14849	June 15, 2019
Porting the SWI-Prolog benchmark suite: comparing 8 Prolog systems Nice to know benchmark , performance , comparison	23	2776	November 28, 2023
Database micro-benchmark (Discussion) Discussion	22	946	April 25, 2020
Prolog program is much slower than Haskell -- why? General discussion	20	2348	February 24, 2023
Thinking In Prolog - for procedural programmers Help!	17	3501	July 21, 2019

Prolog's performance on "Billion Nested Loop Iteration"

Related topics