Age | Commit message (Collapse) | Author | Files | Lines |
|
When resolving a select expression (`attrs.name` or `attrs.name or
default`), if the set compiles to a constant attribute set (as is most
notably the case with `builtins`) we can backtrack and replace that
attribute set directly with the compiled value.
For something like `builtins.length`, this will directly emit an
`OpConstant` that leaves the `length` builtin on the stack.
Change-Id: I639654e065a06e8cfcbcacb528c6da7ec9e513ee
Reviewed-on: https://cl.tvl.fyi/c/depot/+/7957
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
|
|
This was reflowed in a funny way. Move the whole command into its own
line, to prevent it from happening.
Change-Id: Ifba4daf418487ca4c32586820071930d29020f42
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8026
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
|
|
Change-Id: If80194b5fdbf69512217bd4780e416e678045323
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8023
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
|
|
This fixes a very complicated bug (b/246). Evaluation
progresses *much* further after this, leading to several less
complicated bugs likely being uncovered by this
What was the problem?
=====================
Previously, when evaluating a thunk, we had a code path that looked
like this:
match *thunk {
ThunkRepr::Evaluated(Value::Thunk(ref inner_thunk)) => {
let inner_repr = inner_thunk.0.borrow().clone();
drop(thunk);
self.0.replace(inner_repr);
}
/* ... */
}
This code path created a copy of the inner `ThunkRepr` of a nested
thunk, and moved that copy into the `ThunkRepr` of the parent.
The effect of this was that the original `ThunkRepr` (unforced!) lived
on in the original thunk, without the memoization of the subsequent
forcing applying to it.
This had the result that Tvix would repeatedly evaluate these thunks
without ever memoizing them, if they occured repeatedly as shared
inner thunks. Most notably, this would *always* occur when
builtins.import was used.
What's the solution?
====================
I have completely rewritten `Thunk::force_trampoline_self` to make all
flows that can occur in it explicit. I have also removed the outer
loop inside of that function, and resorted to more use of trampolining
instead.
The function is now well-commented and it should be possible to read
it from top-to-bottom and get a general sense of what is going on,
though the trampolining itself (which is implemented in the VM) needs
to be at least partially understood for this.
What's the new problem(s)?
==========================
One new (known) problem is that we have to construct `Error` instances
in all error types here, but we do not have spans available in some
thunk-related situations. Due to b/238 we cannot ask the VM for an
arbitrary span from the callsite leading to the force. This means that
there are now code paths where, under certain conditions, causing an
evaluation error during thunk forcing will panic.
To fix this we will need to investigate and fix b/238, and/or add a
span tracking mechanism to thunks themselves.
What other impacts does this have?
==================================
With this commit, eval of nixpkgs mostly succeeds (things like stdenv
evaluate to the same hashes for us and C++ Nix, meaning we now
construct identical derivations without eval breaking).
Due to this we progress much further into nixpkgs, which lets us
uncover more additional bugs. For example, after this commit we can
quickly see that cl/7949 introduces some kind of behavioural issue and
should not be merged as-is (this was not apparent before).
Additionally, tvix-eval is now seemingly very fast. When doing
performance analysis of a nixpkgs eval, we now mostly see the code
path for shelling out to C++ Nix to add things to the store in there.
We still need those code paths, so we can not (yet) do a performance
analysis beyond that.
Change-Id: I738525bad8bc5ede5d8c737f023b14b8f4160612
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8012
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
|
|
We need to distinguish explicitly between the paths used for the
scanner, and the paths that populate the derivation inputs. The full
paths must be accessible from the result of the refscanner to populate
drv fields correctly.
This was previously hidden by debug changes that masked actual IO
operations with no-ops.
Change-Id: I037af6e6bbe2b573034d695f8779bee1b56bc125
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8022
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
|
|
Creates a cache of imported literal files (e.g.
`./default-builder.sh`) which avoids shelling out to Nix for each
instance of the same file.
Note that a better way to tackle this is to create memoizable thunks
for these expressions in the compiler, but we are lacking a little bit
of infrastructure for that at the moment.
Change-Id: Ibc062b20d81e97dd3986e734d225a744e1779fe7
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8015
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
|
|
Switch out the string-scanning algorithm used in the reference scanner.
The construction of aho-corasick automata made up the vast majority of
runtime when evaluating nixpkgs previously. While the actual scanning
with a constructed automaton is relatively fast, we almost never scan
for the same set of strings twice and the cost is not worth it.
An algorithm that better matches our needs is the Wu-Manber multiple
string match algorithm, which works efficiently on *long* and *random*
strings of the *same length*, which describes store paths (up to their
hash component).
This switches the refscanner crate to a Rust implementation[0][1] of
this algorithm.
This has several implications:
1. This crate does not provide a way to scan streams. I'm not sure if
this is an inherent problem with the algorithm (probably not, but
it would need buffering). Either way, related functions and
tests (which were actually unused) have been removed.
2. All strings need to be of the same length. For this reason, we
truncate the known paths after their hash part (they are still
unique, of course).
3. Passing an empty set of matches, or a match that is shorter than
the length of a store path, causes the crate to panic. We safeguard
against this by completely skipping the refscanning if there are no
known paths (i.e. when evaluating the first derivation of an eval),
and by bailing out of scanning a string that is shorter than a
store path.
On the upside, this reduces overall runtime to less 1/5 of what it was
before when evaluating `pkgs.stdenv.drvPath`.
[0]: Frankly, it's a random, research-grade MIT-licensed
crate that I found on Github:
https://github.com/jneem/wu-manber
[1]: We probably want to rewrite or at least fork the above crate, and
add things like a three-byte wide scanner. Evaluating large
portions of nixpkgs can easily lead to more than 65k derivations
being scanned for.
Change-Id: I08926778e1e5d5a87fc9ac26e0437aed8bbd9eb0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8017
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
|
|
The README was very sparse before and we've actually had people email
us (as it says to contact us) just to ask what Tvix *is*. This should
answer some questions!
Change-Id: I0f248cb060eccfe086468afed1d648652b35dfd1
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8018
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
|
|
... not just a TODO.
Most use-cases of unsafeDiscardStringContext are for cases where a
string is processed in some ways and no longer contains a "physical"
reference, but still has its context attached in C++ Nix.
We don't need to do this. This does diverge in behaviour in use-cases
related to build scheduling, but that whole behaviour will be
different in Tvix.
Change-Id: I4056d4c09f62d44d6bd52b791db03fe5556672b5
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8016
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
|
|
... instead of a BTreeMap, as we do not need ordering guarantees here.
HashMaps are noticeably faster here (especially as we've been sorting
essentially random data!).
Change-Id: Ie92d74286df9f763c04c9b226ef1066ee8484c13
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8014
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
|
|
This aids in debugging quite substantially.
Change-Id: Ic43232aa6165ae1c3db7ac2701938e1dfeeb418c
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8013
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Autosubmit: tazjin <tazjin@tvl.su>
|
|
Change-Id: Ie6bb2df16f79f7d977a7e95187a6b81e3ac0108f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8011
Reviewed-by: tazjin <tazjin@tvl.su>
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
|
|
Change-Id: I08b2dc2393819c4f0d3871fe05bdbd29e89a6d75
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8010
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
Reviewed-by: tazjin <tazjin@tvl.su>
|
|
Change-Id: I594a6652e2efe7aa6e35c7cdd84fc3097660614f
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8009
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Autosubmit: tazjin <tazjin@tvl.su>
|
|
It looks like this needs to be set for the tvix pipeline to succeed.
It was set to `canon` for `tvl-kit` (not sure if manually, or some
autodetection previously did it for us that's not present anymore).
Anyways, this sets it to how it's set in the web interface, to hopefully
fix it.
Change-Id: Ic3eb60e3f421fa949a84dcdaa928823ff45f679a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8008
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Autosubmit: flokli <flokli@flokli.de>
|
|
Pushes a mirror of tvix to github/tvlfyi/tvix
Change-Id: I86e7c1b1dc1fcc1012fa72234588b26ac2554056
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8007
Autosubmit: tazjin <tazjin@tvl.su>
Reviewed-by: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
|
|
Change-Id: I4e81694b9686f977a6590c5e1703a4ef413b0cf4
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8003
Autosubmit: flokli <flokli@f |