2022-12-08 22:19:22 +01:00
|
|
|
{ depot, pkgs, lib, ... }:
|
|
|
|
|
2024-04-20 19:56:40 +02:00
|
|
|
(depot.tvix.crates.workspaceMembers.tvix-cli.build.override {
|
|
|
|
runTests = true;
|
|
|
|
}).overrideAttrs (finalAttrs: previousAttrs:
|
2023-05-22 15:46:40 +02:00
|
|
|
|
2024-04-20 19:56:40 +02:00
|
|
|
let
|
|
|
|
tvix-cli = finalAttrs.finalPackage;
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
|
|
|
|
benchmark-gnutime-format-string =
|
|
|
|
description:
|
|
|
|
"Benchmark: " +
|
|
|
|
(builtins.toJSON {
|
|
|
|
"${description}" = {
|
|
|
|
kbytes = "%M";
|
|
|
|
system = "%S";
|
|
|
|
user = "%U";
|
|
|
|
};
|
|
|
|
});
|
|
|
|
|
|
|
|
# You can run the benchmark with a simple `nix run`, like:
|
|
|
|
#
|
2024-04-20 19:56:40 +02:00
|
|
|
# nix-build -A tvix.cli.meta.ci.extraSteps.benchmark-nixpkgs-cross-hello-outpath
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
#
|
|
|
|
# TODO(amjoseph): store these results someplace more durable, like git trailers
|
|
|
|
#
|
|
|
|
mkExprBenchmark = { expr, description }:
|
|
|
|
let name = "tvix-cli-benchmark-${description}"; in
|
2024-04-20 19:56:40 +02:00
|
|
|
(pkgs.runCommand name { } ''
|
|
|
|
export SSL_CERT_FILE=${pkgs.cacert.out}/etc/ssl/certs/ca-bundle.crt
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
${lib.escapeShellArgs [
|
|
|
|
"${pkgs.time}/bin/time"
|
|
|
|
"--format" "${benchmark-gnutime-format-string description}"
|
|
|
|
"${tvix-cli}/bin/tvix"
|
|
|
|
"--no-warnings"
|
|
|
|
"-E" expr
|
|
|
|
]}
|
2024-04-20 19:56:40 +02:00
|
|
|
touch $out
|
|
|
|
'');
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
|
|
|
|
mkNixpkgsBenchmark = attrpath:
|
|
|
|
mkExprBenchmark {
|
|
|
|
description = builtins.replaceStrings [ ".drv" ] [ "-drv" ] attrpath;
|
|
|
|
expr = "(import ${pkgs.path} {}).${attrpath}";
|
|
|
|
};
|
|
|
|
|
2024-04-20 19:56:40 +02:00
|
|
|
# Constructs a Derivation invoking tvix-cli inside a build, ensures the
|
|
|
|
# calculated tvix output path matches what's passed in externally.
|
|
|
|
mkNixpkgsEvalTest = attrpath: expectedPath:
|
|
|
|
let
|
|
|
|
name = "tvix-eval-test-${builtins.replaceStrings [".drv"] ["-drv"] attrpath}";
|
|
|
|
in
|
|
|
|
(pkgs.runCommand name { } ''
|
|
|
|
export SSL_CERT_FILE=${pkgs.cacert.out}/etc/ssl/certs/ca-bundle.crt
|
|
|
|
TVIX_OUTPUT=$(${tvix-cli}/bin/tvix -E '(import ${pkgs.path} {}).${attrpath}')
|
|
|
|
EXPECTED='${/* the verbatim expected Tvix output: */ "=> \"${builtins.unsafeDiscardStringContext expectedPath}\" :: string"}'
|
|
|
|
|
|
|
|
echo "Tvix output: ''${TVIX_OUTPUT}"
|
|
|
|
if [ "$TVIX_OUTPUT" != "$EXPECTED" ]; then
|
|
|
|
echo "Correct would have been ''${EXPECTED}"
|
|
|
|
exit 1
|
|
|
|
fi
|
|
|
|
|
|
|
|
echo "Output was correct."
|
|
|
|
touch $out
|
|
|
|
'');
|
|
|
|
|
|
|
|
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
benchmarks = {
|
|
|
|
benchmark-hello = (mkNixpkgsBenchmark "hello.outPath");
|
|
|
|
benchmark-cross-hello = (mkNixpkgsBenchmark "pkgsCross.aarch64-multiplatform.hello.outPath");
|
|
|
|
benchmark-firefox = (mkNixpkgsBenchmark "firefox.outPath");
|
|
|
|
benchmark-cross-firefox = (mkNixpkgsBenchmark "pkgsCross.aarch64-multiplatform.firefox.outPath");
|
2023-12-09 07:30:37 +01:00
|
|
|
# Example used for benchmarking LightSpan::Delayed in commit bf286a54bc2ac5eeb78c3d5c5ae66e9af24d74d4
|
|
|
|
benchmark-nixpkgs-attrnames = (mkExprBenchmark { expr = "builtins.length (builtins.attrNames (import ${pkgs.path} {}))"; description = "nixpkgs-attrnames"; });
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
};
|
2024-04-20 19:56:40 +02:00
|
|
|
|
|
|
|
evalTests = {
|
|
|
|
eval-nixpkgs-stdenv-drvpath = (mkNixpkgsEvalTest "stdenv.drvPath" pkgs.stdenv.drvPath);
|
|
|
|
eval-nixpkgs-stdenv-outpath = (mkNixpkgsEvalTest "stdenv.outPath" pkgs.stdenv.outPath);
|
|
|
|
eval-nixpkgs-hello-outpath = (mkNixpkgsEvalTest "hello.outPath" pkgs.hello.outPath);
|
|
|
|
eval-nixpkgs-firefox-outpath = (mkNixpkgsEvalTest "firefox.outPath" pkgs.firefox.outPath);
|
|
|
|
eval-nixpkgs-firefox-drvpath = (mkNixpkgsEvalTest "firefox.drvPath" pkgs.firefox.drvPath);
|
|
|
|
eval-nixpkgs-cross-stdenv-outpath = (mkNixpkgsEvalTest "pkgsCross.aarch64-multiplatform.stdenv.outPath" pkgs.pkgsCross.aarch64-multiplatform.stdenv.outPath);
|
|
|
|
eval-nixpkgs-cross-hello-outpath = (mkNixpkgsEvalTest "pkgsCross.aarch64-multiplatform.hello.outPath" pkgs.pkgsCross.aarch64-multiplatform.hello.outPath);
|
|
|
|
};
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
in
|
|
|
|
{
|
2023-02-02 22:49:56 +01:00
|
|
|
meta = {
|
2024-04-20 19:56:40 +02:00
|
|
|
ci.targets = (builtins.attrNames benchmarks) ++ (builtins.attrNames evalTests);
|
2023-02-02 22:49:56 +01:00
|
|
|
};
|
feat(tvix/cli): add macrobenchmark
This commit adds a simple MVP benchmark, built on our nix
infrastructure instead of cargo. It simply runs `tvix-eval` inside
of GNU time, and prints the three essential statistics in a short
JSON blob.
You can run the benchmark with a simple `nix run`, like:
nix run -f . tvix.cli.benchmark-hello
nix run -f . tvix.cli.benchmark-firefox
nix run -f . tvix.cli.benchmark-cross-firefox
Currently these blobs are stored only in the CI logs, which I'm sure
get garbage-collected at some point. We should be putting them in
the git trailers, but that can wait for a future CL.
I tried using `cargo bench` for this but found it incredibly
frustrating. Maybe I'm doing it wrong. It seems to be designed for
microbenchmarks only, and very hard to control. It kept building
all sorts of unnecessary stuff (like the tests), and unlike
crate2nix it was doing all the builds on only a single machine
instead of using more than one machine. Worse, for that single
machine it kept picking my laptop instead of my fast servers! It
seems excessively cargo-flavored for such a straightforward task.
Benchmark: {"hello.outPath":{"kbytes":"244736","system":"0.36","user":"2.76"}}
Benchmark: {"firefox.outPath":{"kbytes":"1506736","system":"2.38","user":"32.01"}}
Benchmark: {"pkgsCross.aarch64-multiplatform.firefox.outPath":{"kbytes":"11334548","system":"10.70","user":"107.07"}}
Change-Id: I85bc046ec551360284d7ecfc81a03914f0085909
Reviewed-on: https://cl.tvl.fyi/c/depot/+/10216
Autosubmit: Adam Joseph <adam@westernsemico.com>
Reviewed-by: grfn <grfn@gws.fyi>
Reviewed-by: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
2023-12-09 05:15:31 +01:00
|
|
|
|
2024-04-20 19:56:40 +02:00
|
|
|
# Expose benchmarks and evalTests as standard CI targets.
|
|
|
|
passthru = benchmarks // evalTests;
|
2023-02-02 22:49:56 +01:00
|
|
|
})
|