UPLC evaluation slows down (for high usage of nested lambdas?)

[nielstron]

What Git revision are you using?

aiken v0.0.28

What operating system are you using, and which version?

• Linux / Ubuntu
• Linux / Other
• macOS
• Windows

Describe what the problem is?

I was trying to compare the computation cost of different fibonacci computation algorithms. With https://github.com/ImperatorLang/eopsin I wrote both a recursive and an imperative version.

I compiled them with eopsin, then run aiken uplc eval <version>.uplc (con data #02).
I get the following output:

(venv) ➜  eopsin-lang git:(dev) ✗ time aiken uplc eval fib_iter.uplc "(con data #03)"

Result
------

(con data #02)


Costs
-----
cpu: 44619695
memory: 129880

Budget
------
cpu: 9955380305
memory: 13870120

aiken uplc eval fib_iter.uplc "(con data #03)"  2.22s user 0.31s system 99% cpu 2.537 total
(venv) ➜  eopsin-lang git:(dev) ✗ time aiken uplc eval fib_rec.uplc "(con data #03)" 

Result
------

(con data #02)


Costs
-----
cpu: 39405624
memory: 113058

Budget
------
cpu: 9960594376
memory: 13886942

aiken uplc eval fib_rec.uplc "(con data #03)"  0.00s user 0.00s system 96% cpu 0.009 total

As you can see, the imperative version takes 2.5 s while the recursive version takes 0.009 seconds, even though both take roughly the same amount of CPU budget and Memory Budget. Evaluating the same with my python implementation of the uplc interpreter terminates quickly for both versions. I suggest there is some inefficiency lurking in the interpreter.

What should be the expected behavior?

The evaluation of two scripts with the same CPU and Memory budget should take roughly the same amount of time.

[MicroProofs]

#193
Probably caused by this issue