The OpenLDK is very interesting - it looks like it “compiles” to the vintage procedural dialect within CL (eg TAGBODY etc.) I wonder if someone’s ever bypassed the “procedural Lisp” level and just used a CL implementation’s internal assembler interactively, though. (IIRC both SBCL and CCL expose theirs.)
I did that to write simd routines for sbcl: https://github.com/sbcl/sbcl/blob/master/src/code/arm64-simd...
Probably the best way of writing assembly, can evaluate the function immediately, use macros and any other code to emit instructions, even can print register values (instruction-level stepping would be even better, but too much work).
TAGBODY/GO are broadly used in advanced Lisp macros. If you expand a non-trivial extended LOOP invocation you'd likely see some.
If you compile to an implemenation's assembler (even where that possible) you don't really compile into Lisp anymore. And really the Lisp compiler is going to do a better job at generating machine code.
If you are interested in this, you might also be interested to learn that I also got clojure running on SBCL via OpenLDK. See https://github.com/atgreen/cl-clojure.
Regarding LLM-usage, the bulk of OpenLDK was written without the use of LLMs. But recently I let Claude loose on the code to fix a few remaining problems blocking kawa. Claude also upleveled the Java support from Java 8 to Java 21.
What a blast from the past! I too remember with pleasure the days working with Anthony Green (and others) at Cygnus. I like to boast that (apart from Java) Kawa Scheme is the oldest compiler-based language implementation still available for the JVM.
RMS itself being a diehard Scheme and Elisp user said that he found Java elegant over C. This was OFC long before Go and when C++ was king in the 90's.
On Java itself, when CLOS, a dog-ancient system for Common Lisp it's enough to support the Java class/method/object system by itself tells a lot on how great CL can be, even with SBCL which is the top tier free (as in freedom) interpreter/compiler out there.
On performance, well, who knows; remember that PyPy itself back in the day was written in Python itself and it ran things much faster than the vanilla Python interpreter.
Kawa predates Clojure by a decade. (Kawa work started 1996; Clojure's initial release was 2007.) Also, Clojure isn't really focused on high performance, while that has always been a priority for Kawa, which generates bytecode similar to Java, especially if you include suitable type annotations. (It is likely Clojure have have improved in this respect - they have a lot more people working on it.)
Agreed, "Java" was an oversimplification. It's actually JVM bytecode. It's still strange in my book to use an object soup runtime for something (Lisp/Scheme) that feels closer to functional languages to me.
Clojure is a different case, because when you already are on the JVM anyway, then Clojure is still infinitely better than no Lisp at all. It's not the same as putting the JVM somewhere where it wasn't before and where it's not actually needed.
The Computer Abstractions book/course for Scheme had some kind of VM written in Java where you had to write an assembler in Scheme as the final 'biggie' project.
Well, GNU Kawa is named after the Polish word for coffee (going with a play on Java rather than a play on Scheme like Guile and Larceny EDIT: and Gambit went with).
I am a fan.
reply