about summary refs log tree commit diff
path: root/web
diff options
context:
space:
mode:
authorVincent Ambo <tazjin@tvl.su>2024-08-10T20·59+0300
committertazjin <tazjin@tvl.su>2024-08-19T11·02+0000
commitd6c57eb957abc9c9101779600e04b34209d5c436 (patch)
treee6ec5d95d200912c87e7343fba1e4aca086b4515 /web
parentddca074886196ba45c43646d04bd84618009159d (diff)
refactor(tvix/eval): ensure VM operations fit in a single byte r/8519
This replaces the OpCode enum with a new Op enum which is guaranteed to fit in a
single byte. Instead of carrying enum variants with data, every variant that has
runtime data encodes it into the `Vec<u8>` that a `Chunk` now carries.

This has several advantages:

* Less stack space is required at runtime, and fewer allocations are required
  while compiling.
* The OpCode doesn't need to carry "weird" special-cased data variants anymore.
* It is faster (albeit, not by much). On my laptop, results consistently look
  approximately like this:

  Benchmark 1: ./before -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
  Time (mean ± σ):      8.224 s ±  0.272 s    [User: 7.149 s, System: 0.688 s]
  Range (min … max):    7.759 s …  8.583 s    10 runs

  Benchmark 2: ./after -E '(import <nixpkgs> {}).firefox.outPath' --log-level ERROR --no-warnings
  Time (mean ± σ):      8.000 s ±  0.198 s    [User: 7.036 s, System: 0.633 s]
  Range (min … max):    7.718 s …  8.334 s    10 runs

  See notes below for why the performance impact might be less than expected.
* It is faster while at the same time dropping some optimisations we previously
  performed.

This has several disadvantages:

* The code is closer to how one would write it in C or Go.
* Bit shifting!
* There is (for now) slightly more code than before.

On performance I have the following thoughts at the moment:

In order to prepare for adding GC, there's a couple of places in Tvix where I'd
like to fence off certain kinds of complexity (such as mutating bytecode, which,
for various reaons, also has to be part of data that is subject to GC). With
this change, we can drop optimisations like retroactively modifying existing
bytecode and *still* achieve better performance than before.

I believe that this is currently worth it to pave the way for changes that are
more significant for performance.

In general this also opens other avenues of optimisation: For example, we can
profile which argument sizes actually exist and remove the copy overhead of
varint decoding (which does show up in profiles) by using more adequately sized
types for, e.g., constant indices.

Known regressions:

* Op::Constant is no longer printing its values in disassembly (this can be
  fixed, I just didn't get around to it, will do separately).

Change-Id: Id9b3a4254623a45de03069dbdb70b8349e976743
Reviewed-on: https://cl.tvl.fyi/c/depot/+/12191
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
Diffstat (limited to 'web')
-rw-r--r--web/tvixbolt/Cargo.lock7
-rw-r--r--web/tvixbolt/Cargo.nix82
2 files changed, 50 insertions, 39 deletions
diff --git a/web/tvixbolt/Cargo.lock b/web/tvixbolt/Cargo.lock
index c12647b97e57..995ea3e54d45 100644
--- a/web/tvixbolt/Cargo.lock
+++ b/web/tvixbolt/Cargo.lock
@@ -1583,6 +1583,7 @@ dependencies = [
  "tabwriter",
  "toml",
  "tvix-eval-builtin-macros",
+ "vu128",
 ]
 
 [[package]]
@@ -1637,6 +1638,12 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "49874b5167b65d7193b8aba1567f5c7d93d001cafc34600cee003eda787e483f"
 
 [[package]]
+name = "vu128"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c51a178c8f3f425d86542b14f3dce9e16e86bb86328e2293745e6744ebd62e11"
+
+[[package]]
 name = "wasi"
 version = "0.11.0+wasi-snapshot-preview1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
diff --git a/web/tvixbolt/Cargo.nix b/web/tvixbolt/Cargo.nix
index 68c34901a10b..5a79c9759ab7 100644
--- a/web/tvixbolt/Cargo.nix
+++ b/web/tvixbolt/Cargo.nix
@@ -4711,6 +4711,10 @@ rec {
             packageId = "tvix-eval-builtin-macros";
             rename = "builtin-macros";
           }
+          {
+            name = "vu128";
+            packageId = "vu128";
+          }
         ];
         devDependencies = [
           {
@@ -4852,6 +4856,17 @@ rec {
         ];
 
       };
+      "vu128" = rec {
+        crateName = "vu128";
+        version = "1.0.0";
+        edition = "2018";
+        sha256 = "049fsvml8rsyfj9j53ijhsxqcvp1x7fg651baj35shiziy61f6n5";
+        libPath = "vu128/vu128.rs";
+        authors = [
+          "John Millikin <john@john-millikin.com>"
+        ];
+
+      };
       "wasi" = rec {
         crateName = "wasi";
         version = "0.11.0+wasi-snapshot-preview1";
@@ -5973,52 +5988,41 @@ rec {
                 testPostRun
               ]);
           in
-          pkgs.runCommand "run-tests-${testCrate.name}"
-            {
-              inherit testCrateFlags;
-              buildInputs = testInputs;
-            } ''
-            set -e
+          pkgs.stdenvNoCC.mkDerivation {
+            name = "run-tests-${testCrate.name}";
 
-            export RUST_BACKTRACE=1
+            inherit (crate) src;
 
-            # recreate a file hierarchy as when running tests with cargo
+            inherit testCrateFlags;
 
-            # the source for test data
-            # It's necessary to locate the source in $NIX_BUILD_TOP/source/
-            # instead of $NIX_BUILD_TOP/
-            # because we compiled those test binaries in the former and not the latter.
-            # So all paths will expect source tree to be there and not in the build top directly.
-            # For example: $NIX_BUILD_TOP := /build in general, if you ask yourself.
-            # NOTE: There could be edge cases if `crate.sourceRoot` does exist but
-            # it's very hard to reason about them.
-            # Open a bug if you run into this!
-            mkdir -p source/
-            cd source/
+            buildInputs = testInputs;
 
-            ${pkgs.buildPackages.xorg.lndir}/bin/lndir ${crate.src}
+            buildPhase = ''
+              set -e
+              export RUST_BACKTRACE=1
 
-            # build outputs
-            testRoot=target/debug
-            mkdir -p $testRoot
+              # build outputs
+              testRoot=target/debug
+              mkdir -p $testRoot
 
-            # executables of the crate
-            # we copy to prevent std::env::current_exe() to resolve to a store location
-            for i in ${crate}/bin/*; do
-              cp "$i" "$testRoot"
-            done
-            chmod +w -R .
+              # executables of the crate
+              # we copy to prevent std::env::current_exe() to resolve to a store location
+              for i in ${crate}/bin/*; do
+                cp "$i" "$testRoot"
+              done
+              chmod +w -R .
 
-            # test harness executables are suffixed with a hash, like cargo does
-            # this allows to prevent name collision with the main
-            # executables of the crate
-            hash=$(basename $out)
-            for file in ${drv}/tests/*; do
-              f=$testRoot/$(basename $file)-$hash
-              cp $file $f
-              ${testCommand}
-            done
-          '';
+              # test harness executables are suffixed with a hash, like cargo does
+              # this allows to prevent name collision with the main
+              # executables of the crate
+              hash=$(basename $out)
+              for file in ${drv}/tests/*; do
+                f=$testRoot/$(basename $file)-$hash
+                cp $file $f
+                ${testCommand}
+              done
+            '';
+          };
       in
       pkgs.runCommand "${crate.name}-linked"
         {