Bradley Lucier

I could help to some extent. I'd probably have to learn how to run code on simulated devices.

`(array-map f array)` returns a generalized array, with the new getter being the composition of `f` and `(array-getter array)`, which may later be manipulated by `array-copy`, `array-fold-left`, etc. So `array-copy`,...

> This probably performs well in benchmarks for small vectors, but you risk blowing the stack. I compared the speed of the current sample implementation of the SRFI and also...

Here's an experiment with `10,000 x 10,000` arrays: ``` heine:~> gsi Gambit v4.9.4-176-g2a0917c6 > (import (srfi 231)) > (define domain (make-interval '#(10000 10000))) > (define a (make-specialized-array domain f16-storage-class 1.))...

The example used 100,000,000 elements. Here's an example using `u16`s, so no boxing of intermediate values and no conversion routines: ``` > (define A (time (make-specialized-array domain u16-storage-class 1))) (time...

After doing more experiments, looking at space and time usage, and thinking about the different kind of applications an array SRFI might support, I'm more comfortable with `array-copy`, etc., being...

Here's a comparison between the two proposed implementations of `vector-map`, which give different answers. To be honest, I don't know which approach is "correct". ``` heine:~/lang/scheme/srfi-231/srfi-231-temp> cat vector-map-test.scm (let* ((cont1...

Just as a record, I include here test code for interactions of continuations and the various "continuation-safe" array constructors. These pass on Gambit's implementation and fail on Chibi's. I understand...

These failures are gone now.

I just tested this after your changes, and it still fails.