Commit graph

19 commits

Author SHA1 Message Date
Vincent Ambo
4beee4cba7 refactor(tvix/glue): remove use of lazy_static
This is now supported in the standard library via std::sync::LazyLock, but
requires some manual shuffling around of code.

Change-Id: Ibb3be8458b8a8912ea04c9360d64c5cf914254d4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12609
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
2024-10-13 14:45:17 +00:00
Vincent Ambo
abff828ccc refactor(tvix/eval): remove use of imbl::OrdMap
Removes imbl::OrdMap in favour of an Rc over the standard library's BTreeMap,
which allows us to drop the imbl dependency completely.

In my local tests this is actually slightly faster for `hello` and `firefox`.

Change-Id: Ic9597ead4e98bf9530f290c6a94a3c5c3efd0acc
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12201
Reviewed-by: aspen <root@gws.fyi>
Tested-by: BuildkiteCI
2024-08-19 11:02:50 +00:00
Ilan Joselevich
f648f17ec3 feat(tvix): Jemalloc -> MiMalloc
Use the faster and newer MiMalloc memory allocator for all endpoints in
the workspace.

Change-Id: Ic60237284ed168e46ec6e8f28e2710bae4385c6f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12149
Tested-by: BuildkiteCI
Reviewed-by: aspen <root@gws.fyi>
2024-08-08 08:06:23 +00:00
Aspen Smith
8aa038a90e test(tvix/glue): Add a benchmark for firefox outPath
This is nice to test too - it's similar to hello, but runs for a lot
longer (like 7.5 seconds on my laptop) which means we get even better
stats for stuff.

Change-Id: I7935818f10a6d846d446e685b9263a72d7e2aabd
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12061
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: aspen <root@gws.fyi>
2024-07-30 08:58:49 +00:00
Yureka
67335c41b7 refactor(tvix): move service addrs into shared clap struct
Change-Id: I7cab29ecfa1823c2103b4c47b7d784bc31459d55
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12008
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: yuka <yuka@yuka.dev>
2024-07-22 13:36:08 +00:00
Yureka
8b77c7fcd7 refactor(tvix/store): use composition in tvix_store crate
Change-Id: Ie6290b296baba2b987f1a61c9bb4c78549ac11f1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11983
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: yuka <yuka@yuka.dev>
Tested-by: BuildkiteCI
2024-07-20 19:37:27 +00:00
Aspen Smith
dfe137786c refactor(tvix/eval): Builderize Evaluation
Make constructing of a new Evaluation use the builder pattern rather
than setting public mutable fields. This is currently a pure
refactor (no functionality has changed) but has a few advantages:

- We've encapsulated the internals of the fields in Evaluation, meaning
  we can change them without too much breakage of clients
- We have type safety that prevents us from ever changing the fields of
  an Evaluation after it's built (which matters more in a world where we
  reuse Evaluations).

More importantly, this paves the road for doing different things with
the construction of an Evaluation - notably, sharing certain things like
the GlobalsMap across subsequent evaluations in eg the REPL.

Fixes: b/262
Change-Id: I4a27116faac14cdd144fc7c992d14ae095a1aca4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11956
Tested-by: BuildkiteCI
Autosubmit: aspen <root@gws.fyi>
Reviewed-by: flokli <flokli@flokli.de>
2024-07-06 15:03:46 +00:00
Aspen Smith
d0ab3c8d15 feat(tvix): Switch to jemalloc
Switch tvix to using jemalloc as the default global allocator on
supported (eg, non-msvc) platforms. This gives a pretty reasonable
performance boost basically for free:

int                     time:   [24.361 µs 24.386 µs 24.418 µs]
                        change: [-19.355% -18.859% -18.527%] (p = 0.00 < 0.05)
                        Performance has improved.

merge small attrs       time:   [37.201 µs 37.328 µs 37.442 µs]
                        change: [-24.609% -24.266% -23.982%] (p = 0.00 < 0.05)
                        Performance has improved.

merge large attrs with small attrs
                        time:   [20.030 ms 20.135 ms 20.251 ms]
                        change: [-4.2811% -3.2549% -2.3807%] (p = 0.00 < 0.05)
                        Performance has improved.

hello outpath           time:   [967.91 ms 974.07 ms 983.33 ms]
                        change: [-5.5008% -4.4552% -3.4099%] (p = 0.00 < 0.05)
                        Performance has improved.

Change-Id: I6c6e6e3295ec2fca01ea28dc37bcb201cd811767
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10851
Tested-by: BuildkiteCI
Reviewed-by: sterni <sternenseemann@systemli.org>
Autosubmit: aspen <root@gws.fyi>
Reviewed-by: flokli <flokli@flokli.de>
2024-06-06 10:29:33 +00:00
Florian Klink
14766cfe1d refactor(tvix/store): drop calculate_nar from PathInfoService
This shouldn't be part of the PathInfoService trait.

Pretty much none of the PathInfoServices do implement it, and requiring
them to implement it means they also cannot make use of this calculation
already being done by other PathInfoServices.

Move it out into its own NarCalculationService trait, defined somewhere
at tvix_store::nar, and have everyone who wants to trigger nar
calculation use nar_calculation_service directly, which now is an
additional field in TvixStoreIO for example.

It being moved outside the PathInfoService trait doesn't prohibit
specific implementations to implement it (like the GRPC client for the
`PathInfoService` does.

This is currently wired together in a bit of a hacky fashion - as of
now, everything uses the naive implementation that traverses blob and
directoryservice, rather than composing it properly. I want to leave
that up to a later CL, dealing with other parts of store composition
too.

Change-Id: I18d07ea4301d4a07651b8218bc5fe95e4e307208
Reviewed-on: https://cl.tvl.fyi/c/depot/+/11619
Reviewed-by: Connor Brewster <cbrewster@hey.com>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
2024-05-11 13:33:59 +00:00
aspen
d74c68025b revert(tvix): don't use Rc::clone explicitly
This reverts commit d9565a4d0a.

Reason for revert: this was intentional - putting Rc::clone instead of
.clone is a common Rust idiom, and makes it explicit that we're cloning
a shared reference, not an underlying resource

Change-Id: I41a5f323ee35d7025dc7bb02f7d5d05d0051798d
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10995
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
2024-02-21 18:36:11 +00:00
Ryan Lahfa
7388078630 feat(tvix/eval): implement builtins.filterSource
We add a new set of builtins called `import_builtins`, which
will contain import-related builtins, such as `builtins.path` and
`builtins.filterSource`. Both can import paths into the store, with
various knobs to alter the result, e.g. filtering, renaming, expected
hashes.

We introduce `filtered_ingest` which will drive the filtered ingestion
via the Nix function via the generator machinery, and then we register
the root node to the path info service inside the store.

`builtins.filterSource` is very simple, `builtins.path` is a more
complicated model requiring the same logic albeit more sophisticated
with name customization, file ingestion method and expected SHA-256.

Change-Id: I1083f37808b35f7b37818c8ffb9543d9682b2de2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10654
Autosubmit: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
2024-02-20 14:16:36 +00:00
Florian Klink
d9565a4d0a refactor(tvix): don't use Rc::clone explicitly
All these functions have an Rc of Rc<TvixStoreIO> or Rc<dyn StoreIO>, so
we can call io.clone() directly.

Change-Id: I96a5bcb9af4aca93892cb72bcfaf14540da52381
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10989
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
2024-02-20 12:24:13 +00:00
Aspen Smith
0db46dacea feat(tvix/glue): Init fetcher builtins
Initialize a new empty builtins module `fetcher_builtins`, which will
contain the builtins which fetch URLs from the internet:

* fetchurl
* fetchGit
* fetchTarball
* fetchTree (maybe? this is experimental)

These builtins are all implemented in CPP nix at:
https://github.com/NixOS/nix/blob/2.20.2/src/libexpr/primops/fetchTree.cc

These builtins are added to the evaluation context using a similar
mechanism to the derivation builtins, and have been added everywhere
derivation builtins were previously being added.

Change-Id: I133b91cc9560f23028621414537f712e7bd8a825
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10974
Autosubmit: aspen <root@gws.fyi>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
2024-02-19 16:51:10 +00:00
Florian Klink
a720278843 fix(tvix/glue/benches/eval): add impure_builtins
Impure builtins need to be explicitly added. This was missed from
cl/10640.

Change-Id: I1f8a84aa02f1cf8a26247960d6eb15895e09c610
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10667
Tested-by: BuildkiteCI
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: flokli <flokli@flokli.de>
2024-01-20 00:33:04 +00:00
Florian Klink
501827db59 refactor(tvix/glue): add BuildService to TvixStoreIO
TvixStoreIO triggers builds whenever IO into a not-yet-built store path
is requested, if it knows how to build that path.

Change-Id: If30e9db6be2f2a30cbc9d0576f357f3ecfa0d35a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10645
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
2024-01-18 14:30:23 +00:00
Ryan Lahfa
12ae96cff2 feat(tvix/glue): use TvixStoreIO as derivation builtin state
We propagate a `TvixStoreIO` as the `state` of our derivation-specific
builtins in the glue crate.

The evaluators `io_handle` itself is using a Rc<dyn EvalIO>.

An earlier version of TvixStoreIO was also introducing generics over the
different internal services themselves, but we opted for instead
hardcoding this to Arc<dyn …> for the sake of less macro voodoo.

Change-Id: I535c476f06b840858fa3070c4a237ece47f7a15b
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10636
Reviewed-by: raitobezarius <tvl@lahfa.xyz>
Autosubmit: raitobezarius <tvl@lahfa.xyz>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
2024-01-18 14:28:00 +00:00
Ryan Lahfa
099ca6b7c0 feat(tvix/glue): contextful derivation
We calculate the input context by performing the union of context
over all input of the derivation.

Then, we just pass the rest of it to the remaining machinery.

Finally, we re-emit an `outPath` and a `drvPath` containing the expected
contexts.

Change-Id: I74905fb258b5bee8b08d1208c9eb87f51b92a890
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10436
Autosubmit: raitobezarius <tvl@lahfa.xyz>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2024-01-03 23:23:33 +00:00
Florian Klink
4fba57c2c9 refactor(tvix/eval): remove code and location from struct
Instead, it's passed in the evaluate/compile_only functions, which feels
more naturally. It lets us set up the Evaluation struct long before
we actually feed it with data to evaluate.

Now that Evaluation::new() would be accepting an empty list of
arguments, we can simply implement Default, making things a bit more
idiomatic.

Change-Id: I4369658634909a0c504fdffa18242a130daa0239
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10475
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
2023-12-31 13:15:24 +00:00
Florian Klink
8018313b68 feat(tvix/glue): add nixpkgs eval benchmark
This adds a criterion.rs-based testbench into tvix-glue.

It can be invoked by running `cargo bench` from inside the `tvix-glue`
crate.
`target/criterion/report/index.html` contains nice graphs.
It's able to diff against the previous run, so you can invoke `cargo
bench` before and after a certain change to reason about the impact in
evaluation performance.

Currently, we need to create a bunch of Evaluator resources inside the
benchmark loop itself, which is a bit annoying, as it leaks into the
things we benchmark.

This should become better with b/262.

Fixes b/322.

Change-Id: I91656a308887baa1d459ed54d58baae919a4aaf2
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10245
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
2023-12-12 10:27:51 +00:00