From e5a95816829966901c0f0f1480a509a833356bce Mon Sep 17 00:00:00 2001 From: Vincent Ambo Date: Tue, 13 Sep 2022 18:35:51 +0300 Subject: docs(web/tvl): rewrite blog post to be more "milestoney" Suggested by zimbatm. Change-Id: I5979cf820943dd44c8a759f226b340c37f9b0446 Reviewed-on: https://cl.tvl.fyi/c/depot/+/6572 Tested-by: BuildkiteCI Reviewed-by: eta --- web/tvl/blog/tvix-status-202209.md | 47 ++++++++++++++++++++++---------------- 1 file changed, 27 insertions(+), 20 deletions(-) (limited to 'web/tvl/blog/tvix-status-202209.md') diff --git a/web/tvl/blog/tvix-status-202209.md b/web/tvl/blog/tvix-status-202209.md index 8bf731f25c69..f617b5d1b464 100644 --- a/web/tvl/blog/tvix-status-202209.md +++ b/web/tvl/blog/tvix-status-202209.md @@ -28,24 +28,10 @@ and [sterni][]. ### So, what's working now? -The answer is *most things*! However, there are some unfinished and -important feature areas: +The answer is *most things*! You can enter many Nix language +expressions in [Tvixbolt][] and observe how they are evaluated. -1. The majority of Nix's `builtins` are not yet implemented (including - fundamental ones such as `import` and `derivation`). - -2. Recursive attribute sets (`rec`) are not yet implemented. This is - actually not because of the recursion in itself, but because of the - handling of nested keys (such as `a.b`), for which we are designing - a more efficient solution than what is currently in place. - -In both cases we have mostly figured out how to do the remaining work -and it is simply a question of time until we've done it. Progress is -steady and can of course be tracked [in the source][src] (viewer -without Javascript [here][src-noscript]). - -At the same time, we've already implemented some basics that are -hopefully going to have a huge impact further down, such as: +There's a lot of interesting stuff going on under-the-hood, notably: * The Tvix compiler is built to be able to emit warnings & errors without failing early, as well as retaining as much source @@ -66,7 +52,13 @@ hopefully going to have a huge impact further down, such as: differently based on whether it is empty, a `name/value` pair or a larger set. -### How does this all work? +We've also (constrained by the available features) run some initial +benchmarks against C++ Nix, and in most cases Tvix evaluation is an +order of magnitude faster. However, these benchmarks are in no way +indicative of real-life performance when evaluating something like +`nixpkgs` yet, so stay tuned for more information on performance. + +### How does it all work? Tvix's evaluator is implemented using a custom abstract machine with a very Nix-specific instruction set, as well as a compiler that @@ -89,8 +81,23 @@ towards). ### What's next for tvix-eval? -Apart from the missing language features outlined above, the next -steps are: +Despite everything, there are some unfinished and important feature +areas: + +1. The majority of Nix's `builtins` are not yet implemented (including + fundamental ones such as `import` and `derivation`). + +2. Recursive attribute sets (`rec`) are not yet implemented. This is + actually not because of the recursion in itself, but because of the + handling of nested keys (such as `a.b`), for which we are designing + a more efficient solution than what is currently in place. + +In both cases we have mostly figured out how to do the remaining work +and it is simply a question of time until we've done it. Progress is +steady and can of course be tracked [in the source][src] (viewer +without Javascript [here][src-noscript]). + +Apart from that, the next steps are: * Comprehensive benchmarking. We are standing up an infrastructure for continuous benchmarking to measure the impact of changes, and to be -- cgit 1.4.1