Lisp as an Alternative to Java
Published on Mar 29, 2020 by Sachin.
I recently read an article by Erann Gat titled Lisp as an Alternative to Java where Lisp was compared to C, C++ & Java with respect to the performance, execution time, & memory usage. This post summarizes the study.
Highlights
- 16 programs written by 14 programmers(all self-selected)
- 12(Common Lisp) + 4(Scheme)
- Lisp performance is comparable to or better than C++ in execution speed.
- due to lower variability.
- and reduces project risk.
- Execution of the fastest C & C++ programs were faster than the fastest Lisp program.
Comparison
| Lisp | C, C++ | Java | |
|---|---|---|---|
| Development time | 2 to 8.5 hrs. | 3 to 25 hrs. | 4 to 63 hrs. |
| Avg. Programmer’s experience | 6.2 yrs. | 9.6 yrs. | 7.7 yrs. |
| LoC | 51 to 182 | 107 to 164 | 107 to 164 |
| LoC(mean) | 119 | 277 | 277 |
| LoC(median) | 134 | 244 | 244 |
| LoC(standard deviation) | 10 | - | - |
| Execution time | Fast | Faster | - |
| Runtime | - Better than C & C++ | ||
| - Much better than Java | |||
| Runtime(mean) | 41 sec. | 165 sec. | - |
| Runtime(media) | 30 sec. | 54 sec. | - |
| Runtime(standard deviation) | 11 sec. | 77 sec. | - |
| Memory consumption1 | - Higher than C & C++ | ||
| - Comparable to Java |
Speculations
- Why does Lisp seem to do so well?
- At time Lisp was considered slow because the compilers were unavailable & immature, but modern Lisp compilers are mature, stable and high quality.
- Much faster debug cycle: As opposed to most compilers which recompiles the entire file when any part is changed. This re-links the entire program, may also require stopping which results in the loss of state. Lisp compiler on the other hand is inherently designed to be incremental & interactive. Functions are individually compiled and linked to the running programs, so the state is preserved and used in the next run.
- Smaller size of Lisp code: Lisp does not require type declaration which save LoC and it provide first class functions to write complex algorithms in a very few lines.
- If the above is true, why isn’t Lisp used more than this?
- Designed as a language for AI, it failed to deliver the promise.
- A myth that Lisp is big & slow.
Footnotes:
1
Lisp & Java performs memory management using garbage collector, hence the allocated memory from the Operating System might not be used by the application. The memory consumption of Lisp is the addition of the memory used by Lisp development environment, compiler, & development libraries.