Benchmark improvements

* CI: Clean offline; display cleaned bytes; one artifact upload
* Add missing csv bench to iai benchmarks
* Generate flamegraphs for benchmarks
* Validate benchmark results separately

.github/actions/setup-rust/cleanup.sh

@@ -3,8 +3,11 @@ set -e
 
 echo Cleanup workspace build artifacts and extra target output
 
-# clean just the direct members of the current workspace, with cargo metadata should generalize to all rust projects
+# clean just the direct members of the current workspace, use cargo metadata to generalize to all rust projects
-cargo clean -p `cargo metadata --format-version 1 | jq -r '[.workspace_members[]|split(" ")|.[0]]|join(" ")'`
+cargo clean -p `cargo metadata --no-deps --offline --format-version 1 | jq -r '[.workspace_members[]|split(" ")|.[0]]|join(" ")'`
 
 # remove directories in /target/ that are not named `debug` or `release`
+before=`du -s target | awk '{print $1}'`
 find ./target -maxdepth 1 -type d ! -name debug ! -name release ! -name target -exec rm -r {} \;
+after=`du -s target | awk '{print $1}'`
+echo Deleted $(($before - $after)) bytes from target directory

.github/workflows/ci.yml

@@ -161,21 +161,19 @@ jobs:
           fail_on: nothing
           comment_mode: off
 
-      - run: cargo build --bench run_iai --release
+      - run: cargo build --all-targets --release
+      - run: cargo test --bench setup --release
       - run: cargo bench --bench run_iai
       - run: cargo bench --bench run_criterion
+      - run: cargo bench --bench run_criterion -- --profile-time 60
 
-      - name: Publish criterion benchmark results
+      - name: Publish benchmark results
         uses: actions/upload-artifact@v3
         with:
-          name: bench-criterion-results
+          name: benchmark-results
-          path: target/criterion/**/*
+          path: |
-
+            target/criterion/*
-      - name: Publish iai benchmark results
+            target/iai/*
-        uses: actions/upload-artifact@v3
-        with:
-          name: bench-iai-results
-          path: target/iai/scryer-prolog/run_iai/bench_group/*
 
   # Publish binaries when building for a tag
   release:

Cargo.lock

@@ -17,6 +17,19 @@ version = "1.0.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f26201604c87b1e01bd3d98f8d5d9a8fcbb815e8cedb41ffccbeb4bf593a35fe"
 
+[[package]]
+name = "ahash"
+version = "0.8.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "91429305e9f0a25f6205c5b8e0d2db09e0708a7a6df0f42212bb56c32c8ac97a"
+dependencies = [
+ "cfg-if",
+ "getrandom",
+ "once_cell",
+ "version_check",
+ "zerocopy",
+]
+
 [[package]]
 name = "aho-corasick"
 version = "1.0.2"
@@ -59,6 +72,12 @@ version = "0.5.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "23b62fc65de8e4e7f52534fb52b0f3ed04746ae267519eef2a83941e8085068b"
 
+[[package]]
+name = "arrayvec"
+version = "0.7.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "96d30a06541fbafbc7f82ed10c06164cfbd2c401138f6addd8404629c4b16711"
+
 [[package]]
 name = "assert_cmd"
 version = "1.0.8"
@@ -234,6 +253,12 @@ version = "0.3.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e3b5ca7a04898ad4bcd41c90c5285445ff5b791899bb1b0abdd2a2aa791211d7"
 
+[[package]]
+name = "bytemuck"
+version = "1.14.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "374d28ec25809ee0e23827c2ab573d729e293f281dfe393500e7ad618baa61c6"
+
 [[package]]
 name = "byteorder"
 version = "1.4.3"
@@ -380,6 +405,15 @@ version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e496a50fda8aacccc86d7529e2c1e0892dbd0f898a6b5645b5561b89c3210efa"
 
+[[package]]
+name = "cpp_demangle"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e8227005286ec39567949b33df9896bcadfa6051bccca2488129f108ca23119"
+dependencies = [
+ "cfg-if",
+]
+
 [[package]]
 name = "cpu-time"
 version = "1.0.0"
@@ -620,6 +654,15 @@ dependencies = [
  "num-order",
 ]
 
+[[package]]
+name = "debugid"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "bef552e6f588e446098f6ba40d89ac146c8c7b64aade83c051ee00bb5d2bc18d"
+dependencies = [
+ "uuid",
+]
+
 [[package]]
 name = "derive_deref"
 version = "1.1.1"
@@ -711,6 +754,12 @@ version = "0.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c34f04666d835ff5d62e058c3995147c06f42fe86ff053337632bca83e42702d"
 
+[[package]]
+name = "equivalent"
+version = "1.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5443807d6dff69373d433ab9ef5378ad8df50ca6298caf15de6e52e24aaf54d5"
+
 [[package]]
 name = "errno"
 version = "0.3.3"
@@ -759,6 +808,18 @@ dependencies = [
  "windows-sys",
 ]
 
+[[package]]
+name = "findshlibs"
+version = "0.10.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "40b9e59cd0f7e0806cca4be089683ecb6434e602038df21fe6bf6711b2f07f64"
+dependencies = [
+ "cc",
+ "lazy_static",
+ "libc",
+ "winapi",
+]
+
 [[package]]
 name = "fnv"
 version = "1.0.7"
@@ -975,7 +1036,7 @@ dependencies = [
  "futures-sink",
  "futures-util",
  "http",
- "indexmap",
+ "indexmap 1.9.3",
  "slab",
  "tokio",
  "tokio-util",
@@ -1219,6 +1280,34 @@ dependencies = [
  "hashbrown 0.12.3",
 ]
 
+[[package]]
+name = "indexmap"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ad227c3af19d4914570ad36d30409928b75967c298feb9ea1969db3a610bb14e"
+dependencies = [
+ "equivalent",
+ "hashbrown 0.14.0",
+]
+
+[[package]]
+name = "inferno"
+version = "0.11.17"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c50453ec3a6555fad17b1cd1a80d16af5bc7cb35094f64e429fd46549018c6a3"
+dependencies = [
+ "ahash",
+ "indexmap 2.0.1",
+ "is-terminal",
+ "itoa",
+ "log",
+ "num-format",
+ "once_cell",
+ "quick-xml",
+ "rgb",
+ "str_stack",
+]
+
 [[package]]
 name = "instant"
 version = "0.1.12"
@@ -1300,7 +1389,7 @@ version = "0.7.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6607c62aa161d23d17a9072cc5da0be67cdfc89d3afb1e8d9c842bebc2525ffe"
 dependencies = [
- "arrayvec",
+ "arrayvec 0.5.2",
  "bitflags 1.3.2",
  "cfg-if",
  "ryu",
@@ -1414,6 +1503,15 @@ version = "2.6.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8f232d6ef707e1956a43342693d2a31e72989554d58299d7a88738cc95b0d35c"
 
+[[package]]
+name = "memmap2"
+version = "0.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "43a5a03cefb0d953ec0be133036f14e109412fa594edc2f77227249db66cc3ed"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "memoffset"
 version = "0.9.0"
@@ -1582,6 +1680,16 @@ dependencies = [
  "winapi",
 ]
 
+[[package]]
+name = "num-format"
+version = "0.4.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a652d9771a63711fd3c3deb670acfbe5c30a4072e664d7a3bf5a9e1056ac72c3"
+dependencies = [
+ "arrayvec 0.7.4",
+ "itoa",
+]
+
 [[package]]
 name = "num-modular"
 version = "0.6.1"
@@ -1898,6 +2006,28 @@ dependencies = [
  "plotters-backend",
 ]
 
+[[package]]
+name = "pprof"
+version = "0.13.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ef5c97c51bd34c7e742402e216abdeb44d415fbe6ae41d56b114723e953711cb"
+dependencies = [
+ "backtrace",
+ "cfg-if",
+ "criterion",
+ "findshlibs",
+ "inferno",
+ "libc",
+ "log",
+ "nix 0.26.4",
+ "once_cell",
+ "parking_lot 0.12.1",
+ "smallvec",
+ "symbolic-demangle",
+ "tempfile",
+ "thiserror",
+]
+
 [[package]]
 name = "ppv-lite86"
 version = "0.2.17"
@@ -1976,6 +2106,15 @@ dependencies = [
  "unicode-ident",
 ]
 
+[[package]]
+name = "quick-xml"
+version = "0.26.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f50b1c63b38611e7d4d7f68b82d3ad0cc71a2ad2e7f61fc10f1328d917c93cd"
+dependencies = [
+ "memchr",
+]
+
 [[package]]
 name = "quote"
 version = "1.0.33"
@@ -2158,6 +2297,15 @@ dependencies = [
  "winreg",
 ]
 
+[[package]]
+name = "rgb"
+version = "0.8.37"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "05aaa8004b64fd573fc9d002f4e632d51ad4f026c2b5ba95fcb6c2f32c2c47d8"
+dependencies = [
+ "bytemuck",
+]
+
 [[package]]
 name = "ring"
 version = "0.16.20"
@@ -2342,7 +2490,7 @@ dependencies = [
  "git-version",
  "hostname",
  "iai-callgrind",
- "indexmap",
+ "indexmap 1.9.3",
  "lazy_static",
  "lexical",
  "libc",
@@ -2354,6 +2502,7 @@ dependencies = [
  "num-order",
  "ordered-float",
  "phf 0.9.0",
+ "pprof",
  "predicates-core",
  "proc-macro2",
  "quote",
@@ -2637,6 +2786,12 @@ version = "0.9.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "6980e8d7511241f8acf4aebddbb1ff938df5eebe98691418c4468d0b72a96a67"
 
+[[package]]
+name = "stable_deref_trait"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a8f112729512f8e442d81f95a8a7ddf2b7c6b8a1a6f509a95864142b30cab2d3"
+
 [[package]]
 name = "static_assertions"
 version = "1.1.0"
@@ -2649,6 +2804,12 @@ version = "1.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9e08d8363704e6c71fc928674353e6b7c23dcea9d82d7012c8faf2a3a025f8d0"
 
+[[package]]
+name = "str_stack"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9091b6114800a5f2141aee1d1b9d6ca3592ac062dc5decb3764ec5895a47b4eb"
+
 [[package]]
 name = "string_cache"
 version = "0.8.7"
@@ -2706,6 +2867,29 @@ version = "2.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "81cdd64d312baedb58e21336b31bc043b77e01cc99033ce76ef539f78e965ebc"
 
+[[package]]
+name = "symbolic-common"
+version = "12.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "405af7bd5edd866cef462e22ef73f11cf9bf506c9d62824fef8364eb69d4d4ad"
+dependencies = [
+ "debugid",
+ "memmap2",
+ "stable_deref_trait",
+ "uuid",
+]
+
+[[package]]
+name = "symbolic-demangle"
+version = "12.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2bcd041ccfb77d9c70639efcd5b804b508ac7a273e9224d227379e225625daf9"
+dependencies = [
+ "cpp_demangle",
+ "rustc-demangle",
+ "symbolic-common",
+]
+
 [[package]]
 name = "syn"
 version = "1.0.109"
@@ -3047,6 +3231,12 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "711b9620af191e0cdc7468a8d14e709c3dcdb115b36f838e601583af800a370a"
 
+[[package]]
+name = "uuid"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "88ad59a7560b41a70d191093a945f0b87bc1deeda46fb237479708a1d6b6cdfc"
+
 [[package]]
 name = "vcpkg"
 version = "0.2.15"
@@ -3352,3 +3542,23 @@ name = "xmlparser"
 version = "0.13.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4d25c75bf9ea12c4040a97f829154768bbbce366287e2dc044af160cd79a13fd"
+
+[[package]]
+name = "zerocopy"
+version = "0.7.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8cd369a67c0edfef15010f980c3cbe45d7f651deac2cd67ce097cd801de16557"
+dependencies = [
+ "zerocopy-derive",
+]
+
+[[package]]
+name = "zerocopy-derive"
+version = "0.7.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c2f140bda219a26ccc0cdb03dba58af72590c53b22642577d88a927bc5c87d6b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.37",
+]

Cargo.toml

@@ -115,9 +115,20 @@ serial_test = "2.0.0"
 iai-callgrind = "0.7.3"
 criterion = "0.5.1"
 
+[target.'cfg(not(target_os = "windows"))'.dev-dependencies]
+pprof = { version = "0.13.0", features = ["criterion", "flamegraph"] }
+
 [patch.crates-io]
 modular-bitfield = { git = "https://github.com/mthom/modular-bitfield" }
 
+[profile.bench]
+lto = true
+opt-level = 3
+
+[profile.release]
+lto = true
+opt-level = 3
+
 [[bench]]
 name = "run_criterion"
 harness = false

benches/README.md

@@ -4,16 +4,16 @@ The `benches` directory contains benchmarks that test scryer-prolog performance.
 
 Benchmarks are run via two harnesses:
 
+* `criterion` - criterion performs statistical analysis of benchmark runs and is
+  great for benchmarking locally.
 * `iai-callgrind` - this runs the benchmark with callgrind, which is able to
   precisely track the number of instructions executed during the run. This is
   especially helpful in a public CI runner context where neighboring VMs can
-  cause a very high wall time variance. This means that it doesn't track wall
+  cause a very high wall time variance. While instructions executed is only
-  time which is what we really care about, but it is a good tradeoff for CI
+  correlated with the desired metric (wall time), this is a good tradeoff for CI
-  where tracking runtime is unreliable.
+  where that metric is unreliable.
-* `criterion` - criterion performs statistical analysis of benchmark runs and is
-  great for benchmarking locally.
 
-Run them using the following commands:
+Run benchmarks with the following commands:
 
 ```
 cargo bench --bench run_criterion
@@ -21,6 +21,9 @@ cargo bench --bench run_criterion
 # run a particular criterion benchmark
 cargo bench --bench run_criterion -- <benchmark_name>
 
+# run in profiling mode which outputs flamegraphs. Set profile time in seconds:
+cargo bench --bench run_criterion -- --profile-time <time>
+
 # to run iai, you need valgrind installed and to install iai-callgrind-runner
 # at the same version as is in Cargo.toml:
 cargo install iai-callgrind-runner --version 0.7.3
@@ -29,7 +32,7 @@ cargo bench --bench run_iai
 ```
 
 For consistency, both runners -- `run_iai.rs` and `run_criterion.rs` -- import
-the same setup code from `benches.rs`.
+the same setup code from `setup.rs`.
 
 ## Setup
 
@@ -50,45 +53,43 @@ This design is meant to suppoort defining lots of benchmarks.
 
 To add a new benchmark:
 
-* Add a new file `benches/[module].pl` that contains setup code. Import
+* Add a new file `benches/[module].pl` that contains setup prolog code. Import
   libraries, define predicates, etc.
-* Add a new section in `setup.rs::benches()` that refers to it and add some
+* Add a new section in `setup.rs::prolog_benchmarks()` that refers to to the
-  benchmarks.
+  file and write a query to be benchmarked.
+* If the query mutates the machine, then use `Strategy::Fresh` so the criterion
+  benchmark will recreate a new machine for each benchmark run, otherwise use
+  `Strategy::Reuse` which has lower overhead. (This is not used by the iai
+  benchmark because it only runs once anyway.)
 
 Some tips:
 
 * The goal of benchmarking is to know if a library or engine change improved
   performance or not.
-* Once a benchmark is defined and named, avoid changing it's definition. In
+* Once a benchmark is defined and named, avoid changing it's definition. If a
-  general, if a benchmark needs to change to be more useful, give the new
+  benchmark needs to change to be more useful, give the new definition a new
-  definition a new name. This will prevent charts from showing wild changes in
+  name instead. This will prevent charts from showing wild changes in
   performance just because the definition changed (see previous).
 * Aim for queries to execute in less than 0.5s realtime. Longer runtimes make it
   easier for humans to see big differences, but benchmarks either run 10x slower
   (iai) or execute repeatedly to attain statistical significance (criterion) and
-  in both cases queries that take longer become cumbersome to run.
+  in both cases benchmarking queries that take longer than about 0.5s are
+  cumbersome to run.
 * Consider that the library runtime actually parses the text output of the top
   level. So don't use custom outputs or it will fail to parse. Also keep the
   output small so it doesn't just benchmark the ouput parsing code.
 * DO test the output of the benchmark run, we don't want to count broken
   benchmarks.
-* Because a query may run against the same machine multiple times, don't
-  permanently mutate the state of the engine with the query since that will
-  taint subsequent runs. (Benchmarking assertz et al is desirable, but will
-  require some adjustments to how the machine is set up for runs.)
 
 ## CI
 
 Both benchmark harnesses are run in `.github/workflows/ci.yaml` in the `report`
 job, and the results are published as build artifacts.
 
-A future action may consume the build artifacts and publish a report using the
-results.
-
 ## Todo
 
 - [ ] Currently, the execution time to load a module is not benchmarked. It
   would be nice to have at least one benchmark for loading a module (probably a
   big one).
-- [ ] Write a new action that consumes the test and benchmark results and plots
+- [ ] Write a new action that downloads the test and benchmark results
-  them over time and publishes a report (github pages?).
+  artifacts, plots them over time, and publishes a report to github pages.

benches/run_criterion.rs

@@ -1,5 +1,8 @@
 use criterion::{criterion_group, criterion_main, BatchSize, Criterion};
 
+#[cfg(not(target_os = "windows"))]
+use pprof::criterion::{Output, PProfProfiler};
+
 mod setup;
 
 fn bench_criterion(c: &mut Criterion) {
@@ -13,9 +16,21 @@ fn bench_criterion(c: &mut Criterion) {
     }
 }
 
+#[cfg(not(target_os = "windows"))]
+fn config() -> Criterion {
+    Criterion::default()
+        .sample_size(20)
+        .with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)))
+}
+
+#[cfg(target_os = "windows")]
+fn config() -> Criterion {
+    Criterion::default().sample_size(20)
+}
+
 criterion_group!(
-    name = bench_group;
+    name = benches;
-    config = Criterion::default().sample_size(10);
+    config = config();
     targets = bench_criterion
 );
-criterion_main!(bench_group);
+criterion_main!(benches);

benches/run_iai.rs

@@ -1,21 +1,18 @@
 use iai_callgrind::{library_benchmark, library_benchmark_group, main};
+use scryer_prolog::machine::parsed_results::QueryResolution;
 
 mod setup;
 
 #[library_benchmark]
-#[bench::normal(setup::prolog_benches()["count_edges_short"].setup())]
+#[bench::count_edges(setup::prolog_benches()["count_edges"].setup())]
-fn bench_edges(mut run: impl FnMut()) {
+#[bench::numlist(setup::prolog_benches()["numlist"].setup())]
-    run();
+#[bench::csv_codename(setup::prolog_benches()["csv_codename"].setup())]
-}
+fn bench(mut run: impl FnMut() -> QueryResolution) -> QueryResolution {
-
+    run()
-#[library_benchmark]
-#[bench::normal(setup::prolog_benches()["numlist_short"].setup())]
-fn bench_numlist(mut run: impl FnMut()) {
-    run();
 }
 
 library_benchmark_group!(
-    name = bench_group;
+    name = benches;
-    benchmarks = bench_edges, bench_numlist
+    benchmarks = bench
 );
-main!(library_benchmark_groups = bench_group);
+main!(library_benchmark_groups = benches);

benches/setup.rs

@@ -2,7 +2,7 @@ use std::{collections::BTreeMap, fs, path::Path};
 
 use maplit::btreemap;
 use scryer_prolog::machine::{
-    parsed_results::{QueryMatch, QueryResolution, Value},
+    parsed_results::{QueryResolution, Value},
     Machine,
 };
 
@@ -10,20 +10,13 @@ pub fn prolog_benches() -> BTreeMap<&'static str, PrologBenchmark> {
     [
         (
             "count_edges",                       // name of the benchmark
-            "benches/edges.pl",                  // name of the prolog module file to load
+            "benches/edges.pl", // name of the prolog module file to load. use the same file in multiple benchmarks
-            "independent_set_count(aa, Count).", // query to benchmark in the context of the loaded module
+            "independent_set_count(ky, Count).", // query to benchmark in the context of the loaded module. consider making the query adjustable to tune the run time to ~0.1s
-            Strategy::Reuse,
-            btreemap! { "Count" => Value::try_from("211954906".to_string()).unwrap(), }, // list of expected bindings
-        ),
-        (
-            "count_edges_short",
-            "benches/edges.pl", // use the same file in multiple benchmarks
-            "independent_set_count(ky, Count).", // consider making the query adjustable to tune the run time to ~0.1s
             Strategy::Reuse,
             btreemap! { "Count" => Value::try_from("2869176".to_string()).unwrap() },
         ),
         (
-            "numlist_short",
+            "numlist",
             "benches/numlist.pl",
             "run_numlist(1000000, Head).",
             Strategy::Reuse,
@@ -67,27 +60,40 @@ pub struct PrologBenchmark {
 }
 
 impl PrologBenchmark {
-    pub fn setup(&self) -> impl FnMut() {
+    pub fn make_machine(&self) -> Machine {
         let program = fs::read_to_string(self.filename).unwrap();
         let module_name = Path::new(self.filename)
             .file_stem()
             .and_then(|s| s.to_str())
             .unwrap();
-
         let mut machine = Machine::new_lib();
         machine.load_module_string(module_name, program);
+        machine
+    }
 
-        let benchmark_name = self.name;
+    pub fn setup(&self) -> impl FnMut() -> QueryResolution {
+        let mut machine = self.make_machine();
         let query = self.query;
-        let expected = QueryResolution::Matches(vec![QueryMatch::from(self.bindings.clone())]);
-
         move || {
             use criterion::black_box;
-            let result = black_box(machine.run_query(black_box(query.to_string())));
+            black_box(machine.run_query(black_box(query.to_string()))).unwrap()
-            match result {
+        }
-                Ok(r) => assert_eq!(&r, &expected),
+    }
-                Err(e) => panic!("benchmark {} failed with: {}", benchmark_name, e),
+}
-            }
+
+#[cfg(test)]
+mod test {
+    #[test]
+    fn validate_benchmarks() {
+        use super::prolog_benches;
+        use scryer_prolog::machine::parsed_results::QueryResolution;
+
+        use scryer_prolog::machine::parsed_results::QueryMatch;
+        for (_, r) in prolog_benches() {
+            let mut machine = r.make_machine();
+            let result = machine.run_query(r.query.to_string()).unwrap();
+            let expected = QueryResolution::Matches(vec![QueryMatch::from(r.bindings.clone())]);
+            assert_eq!(result, expected, "validating benchmark {}", r.name);
         }
     }
 }