Simplify benchmark runner and pipelines (#494)

[only benchmarks]

.buildkite/pipeline.yml

@@ -156,6 +156,6 @@ steps:
         if: |
           !build.pull_request.repository.fork &&
           (build.message =~ /\[only benchmarks\]/ ||
-          build.message !~ /\[only/ && !build.pull_request.draft &&
-            build.message !~ /\[skip benchmarks\]/)
+           build.message !~ /\[only/ && !build.pull_request.draft &&
+             build.message !~ /\[skip benchmarks\]/)
         timeout_in_minutes: 30

.github/workflows/Benchmark.yml

@@ -31,7 +31,13 @@ on:
 
 jobs:
   benchmark:
-    if: ${{ (github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository) && (contains(github.event.head_commit.message, '[only benchmarks]') || !contains(github.event.head_commit.message, '[only') && !contains(github.event.head_commit.message, '[skip benchmarks]') && github.event.pull_request.draft == false) }}
+    if: |
+      (github.event_name != 'pull_request' ||
+       github.event.pull_request.head.repo.full_name == github.repository) &&
+      (contains(github.event.head_commit.message, '[only benchmarks]') ||
+       !contains(github.event.head_commit.message, '[only') &&
+       !contains(github.event.head_commit.message, '[skip benchmarks]') &&
+       github.event.pull_request.draft == false)
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4

perf/runbenchmarks.jl

@@ -27,8 +27,6 @@ latency_results = include("latency.jl")
 
 SUITE = BenchmarkGroup()
 
-# NOTE: don't use spaces in benchmark names (tobami/codespeed#256)
-
 include("metal.jl")
 include("kernel.jl")
 include("array.jl")
@@ -58,24 +56,5 @@ integration_results["metaldevrt"] = include("metaldevrt.jl")
 results["latency"] = latency_results
 results["integration"] = integration_results
 
-println(results)
-
-
-## comparison
-
 # write out the results
 BenchmarkTools.save("benchmarkresults.json", median(results))
-
-# compare against previous results
-# TODO: store these results so that we can compare when benchmarking PRs
-reference_path = joinpath(@__DIR__, "reference.json")
-if ispath(reference_path)
-    reference = BenchmarkTools.load(reference_path)[1]
-    comparison = judge(minimum(results), minimum(reference))
-
-    println("Improvements:")
-    println(improvements(comparison))
-
-    println("Regressions:")
-    println(regressions(comparison))
-end