While investigating poor cold start performance on my GCP VM (32 cores,
130GB SSD), I found that writing to the various sqlite databases in
DENO_DIR was quite slow. The slowness seems to primarily be caused by
excessive latency from a number of `fsync()` calls.
The performance difference is best demonstrated by deleting the sqlite
databases from DENO_DIR while leaving the downloaded sources in place.
The benchmark (see notes below):
```
piscisaureus@bert-us:~/erofs/source$ export DENO_DIR=./.deno
piscisaureus@bert-us:~/erofs/source$ hyperfine --warmup 3 \
--prepare "rm -rf .deno/*_v1*" \
"deno run -A --cached-only demo.ts" \
"eatmydata deno run -A --cached-only demo.ts" \
"~/deno/target/release/deno run -A --cached-only demo.ts"
Benchmark 1: deno run -A --cached-only demo.ts
Time (mean ± σ): 1.174 s ± 0.037 s [User: 0.153 s, System: 0.184 s]
Range (min … max): 1.104 s … 1.212 s 10 runs
Benchmark 2: eatmydata deno run -A --cached-only demo.ts
Time (mean ± σ): 265.5 ms ± 3.6 ms [User: 138.5 ms, System: 135.1 ms]
Range (min … max): 260.6 ms … 271.2 ms 11 runs
Benchmark 3: ~/deno/target/release/deno run -A --cached-only demo.ts
Time (mean ± σ): 226.2 ms ± 9.2 ms [User: 136.7 ms, System: 93.3 ms]
Range (min … max): 218.8 ms … 247.1 ms 13 runs
Summary
~/deno/target/release/deno run -A --cached-only demo.ts ran
1.17 ± 0.05 times faster than eatmydata deno run -A --cached-only demo.ts
5.19 ± 0.27 times faster than deno run -A --cached-only demo.ts
```
Notes:
* Benchmark 1: unmodified Deno 1.43.6
* Benchmark 2: unmodified Deno 1.43.6 wrapped with `eatmydata` (which is
a tool to neuter `fsync()` calls)
* Benchmark 3: this PR applied on top of Deno 1.43.6
The script that got benchmarked:
```typescript
// demo.ts
import * as express from "npm:express@4.16.3";
import * as postgres from "https://deno.land/x/postgres/mod.ts";
let _dummy = [express, postgres]; // Force use of imports.
console.log("hello world");
```
This PR enables V8 code cache for ES modules and for `require` scripts
through `op_eval_context`. Code cache artifacts are transparently stored
and fetched using sqlite db and are passed to V8. `--no-code-cache` can
be used to disable.
---------
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This changes the lockfile to not store JSR specifiers in the "remote"
section. Instead a single JSR integrity is stored per package in the
lockfile, which is a hash of the version's `x.x.x_meta.json` file, which
contains hashes for every file in the package. The hashes in this file
are then compared against when loading.
Additionally, when using `{ "vendor": true }` in a deno.json, the files
can be modified without causing lockfile errors—the checksum is only
checked when copying into the vendor folder and not afterwards
(eventually we should add this behaviour for non-jsr specifiers as
well). As part of this change, the `vendor` folder creation is not
always automatic in the LSP and running an explicit cache command is
necessary. The code required to track checksums in the LSP would have
been too complex for this PR, so that all goes through deno_graph now.
The vendoring is still automatic when running from the CLI.
This was incorrectly being stored so the AOT cache for JSR specifiers
wasn't being used. I added a wrapper type to help prevent the API being
used incorrectly.
As title. This will help use the two independently from the other, which
will help in an upcoming deno doc PR where I need to parse the source
files with scope analysis.
This PR adds a new unstable "bring your own node_modules" (BYONM)
functionality currently behind a `--unstable-byonm` flag (`"unstable":
["byonm"]` in a deno.json).
This enables users to run a separate install command (ex. `npm install`,
`pnpm install`) then run `deno run main.ts` and Deno will respect the
layout of the node_modules directory as setup by the separate install
command. It also works with npm/yarn/pnpm workspaces.
For this PR, the behaviour is opted into by specifying
`--unstable-byonm`/`"unstable": ["byonm"]`, but in the future we may
make this the default behaviour as outlined in
https://github.com/denoland/deno/issues/18967#issuecomment-1761248941
This is an extremely rough initial implementation. Errors are
terrible in this and the LSP requires frequent restarts. Improvements
will be done in follow up PRs.
Adds an experimental unstable built-in package manager to Deno, but it is
currently not usable because the registry infrastructure hasn't been
setup and it points to a non-existent url by default. The default
registry url can be configured via the `DENO_REGISTRY_URL` environment
variable.
Renames the unstable `deno_modules` directory and corresponding settings
to `vendor` after feedback. Also causes the vendoring of the
`node_modules` directory which can be disabled via
`--node-modules-dir=false` or `"nodeModulesDir": false`.
This changes the design of the manifest.json file to have a separate
"folders" map for mapping hashed directories. This allows, for example,
to add files in a folder like `http_localhost_8000/#testing_5de71/` and
have them be resolved automatically as long as their remaining
components are identity-mappable to the file system (not hashed). It
also saves space in the manifest.json file by only including the hashed
directory instead of each descendant file.
```
// manifest.json
{
"folders": {
"https://localhost/NOT_MAPPABLE/": "localhost/#not_mappable_5cefgh"
},
"modules": {
"https://localhost/folder/file": {
"headers": {
"content-type": "application/javascript"
}
},
}
}
// folder structure
localhost
- folder
- #file_2defn (note: I've made up the hashes in these examples)
- #not_mappable_5cefgh
- mod.ts
- etc.ts
- more_files.ts
```
Code run within Deno-mode and Node-mode should have access to a
slightly different set of globals. Previously this was done through a
compile time code-transform for Node-mode, but this is not ideal and has
many edge cases, for example Node's globalThis having a different
identity than Deno's globalThis.
This commit makes the `globalThis` of the entire runtime a semi-proxy.
This proxy returns a different set of globals depending on the caller's
mode. This is not a full proxy, because it is shadowed by "real"
properties on globalThis. This is done to avoid the overhead of a full
proxy for all globalThis operations.
The globals between Deno-mode and Node-mode are now properly segregated.
This means that code running in Deno-mode will not have access to Node's
globals, and vice versa. Deleting a managed global in Deno-mode will
NOT delete the corresponding global in Node-mode, and vice versa.
---------
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
Co-authored-by: Aapo Alasuutari <aapo.alasuutari@gmail.com>
Partially supersedes #19016.
This migrates `spawn` and `spawn_blocking` to `deno_core`, and removes
the requirement for `spawn` tasks to be `Send` given our single-threaded
executor.
While we don't need to technically do anything w/`spawn_blocking`, this
allows us to have a single `JoinHandle` type that works for both cases,
and allows us to more easily experiment with alternative
`spawn_blocking` implementations that do not require tokio (ie: rayon).
Async ops (+~35%):
Before:
```
time 1310 ms rate 763358
time 1267 ms rate 789265
time 1259 ms rate 794281
time 1266 ms rate 789889
```
After:
```
time 956 ms rate 1046025
time 954 ms rate 1048218
time 924 ms rate 1082251
time 920 ms rate 1086956
```
HTTP serve (+~4.4%):
Before:
```
Running 10s test @ http://localhost:4500
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 68.78us 19.77us 1.43ms 86.84%
Req/Sec 68.78k 5.00k 73.84k 91.58%
1381833 requests in 10.10s, 167.36MB read
Requests/sec: 136823.29
Transfer/sec: 16.57MB
```
After:
```
Running 10s test @ http://localhost:4500
2 threads and 10 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 63.12us 17.43us 1.11ms 85.13%
Req/Sec 71.82k 3.71k 77.02k 79.21%
1443195 requests in 10.10s, 174.79MB read
Requests/sec: 142921.99
Transfer/sec: 17.31MB
```
Suggested-By: alice@ryhl.io
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This removes `ProcState` and replaces it with a new `CliFactory` which
initializes our "service structs" on demand. This isn't a performance
improvement at the moment for `deno run`, but might unlock performance
improvements in the future.
1. Breaks up functionality within `ProcState` into several other structs
to break out the responsibilities (`ProcState` is only a data struct
now).
2. Moves towards being able to inject dependencies more easily and have
functionality only require what it needs.
3. Exposes `Arc<T>` around the "service structs" instead of it being
embedded within them. The idea behind embedding them was to reduce the
verbosity of needing to pass around `Arc<...>`, but I don't think it was
exactly working and as we move more of these structs to be more
injectable I don't think the extra verbosity will be a big deal.
Fast-follow on #18401 -- the reason that some tests were panicking in
the `CacheDB` `impl Drop` was that the cache itself was being dropped
during panic and the runtime may or may not still exist at that point.
We can reduce the actual tokio runtime testing to where it's needed.
In addition, we return the journal mode to `TRUNCATE` to avoid the risk
of data corruption.
