From 522593291751c89db2cab0db7e6522973894cafd Mon Sep 17 00:00:00 2001 From: Marshall Lochbaum Date: Sat, 20 Aug 2022 18:15:05 -0400 Subject: Update comments on CBQN performance: most things implemented natively --- docs/running.html | 2 +- running.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/running.html b/docs/running.html index 7f6d411e..947d3acf 100644 --- a/docs/running.html +++ b/docs/running.html @@ -87,7 +87,7 @@

C sources are kept in the CBQN repository, but it also depends on bytecode from the BQN sources here. Running make gets a working copy right away with saved bytecode. Then to use the latest bytecode, call $ ./BQN genRuntime …/BQN, where …/BQN points to this repository, and run make again.

CBQN is developed on Linux, and as-is will only run on Unix-like systems (including macOS). To run on Windows, WSL has the best support but there are also native builds based on each of Cygwin and Mingw here.

genRuntime can also be run with another BQN implementation (the Node.js one works but takes up to a minute), and plain ./genRuntime uses your system's bqn executable. I symlink …/CBQN/BQN to ~/bin/bqn so I can easily use CBQN for scripting.

-

CBQN uses the self-hosted runtime to achieve full primitive coverage, and implements specific primitives or parts of primitives natively to speed them up. This means primitives with native support—including everything used by the compiler—are fairly fast while others are much slower.

+

CBQN has native support for most primitive functionality and falls back to the self-hosted runtime to fill the gaps. The most important operations are fast, and it's almost always possible to write code that sticks to them. However, some cases, particularly those that deal with multiple axes, are much slower (although still fine for most use cases).

dzaima/BQN

dzaima/BQN is an implementation in Java created by modifying the existing dzaima/APL, and should be easy to run on desktop Linux and Android. It was historically the main implementation, but is now updated only to stay up to date with language changes. Major missing functionality is dyadic Depth (⚇) and set functions ⊐⊒∊⍷ with rank >1, and there are various small differences from the BQN spec, mostly to do with rank, handling of atoms, fills, and headers. It uses UTF-16 instead of UTF-32, so that characters like 𝕩 don't behave correctly.

To get an executable that works like CBQN, make a script with the following contents. Scripts may use #! /usr/bin/env dbqn to run with dzaima/BQN specifically, but this is rare now (in this repository, only test/dzaima does it).

diff --git a/running.md b/running.md index 42cb4b1d..91523ddd 100644 --- a/running.md +++ b/running.md @@ -47,7 +47,7 @@ CBQN is developed on Linux, and as-is will only run on Unix-like systems (includ `genRuntime` can also be run with another BQN implementation (the Node.js one works but takes up to a minute), and plain `./genRuntime` uses your system's `bqn` executable. I symlink `…/CBQN/BQN` to `~/bin/bqn` so I can easily use CBQN for scripting. -CBQN uses the self-hosted runtime to achieve full primitive coverage, and implements specific primitives or parts of primitives natively to speed them up. This means primitives with native support—including everything used by the compiler—are fairly fast while others are much slower. +CBQN has native support for most primitive functionality and falls back to the self-hosted runtime to fill the gaps. The most important operations are fast, and it's almost always possible to write code that sticks to them. However, some cases, particularly those that deal with multiple axes, are much slower (although still fine for most use cases). ### dzaima/BQN -- cgit v1.2.3