cl, ssa: fix zero-sized global init, fold composite literal alloc stores, and add LTO baseline benchmarks

.github/workflows/benchmark.yml

@@ -31,13 +31,24 @@ jobs:
       LLGO_ROOT: ${{ github.workspace }}
     steps:
       - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+
+      - name: Determine pull request merge-base
+        if: github.event_name == 'pull_request'
+        id: merge-base
+        run: |
+          git fetch https://github.com/${{ github.event.pull_request.base.repo.full_name }}.git ${{ github.event.pull_request.base.ref }}
+          base_sha=$(git merge-base FETCH_HEAD ${{ github.event.pull_request.head.sha }})
+          echo "sha=$base_sha" >> "$GITHUB_OUTPUT"
+          echo "Computed pull request merge-base: $base_sha (head: ${{ github.event.pull_request.head.sha }})"
 
       - name: Check out pull request base benchmark source
         if: github.event_name == 'pull_request'
         uses: actions/checkout@v7
         with:
           repository: ${{ github.event.pull_request.base.repo.full_name }}
-          ref: ${{ github.event.pull_request.base.sha }}
+          ref: ${{ steps.merge-base.outputs.sha }}
           path: .benchmark/source
           persist-credentials: false
 

benchmark/baseline/README.md

@@ -10,9 +10,9 @@ commit, branch, or pull-request series.
 
 The program workloads reuse:
 
-- `benchmark/binary_size/cprintf`: only `lib/c.Printf`;
-- `benchmark/binary_size/println`: only the built-in `println`;
-- `benchmark/binary_size/fmtprintf`: `fmt.Printf`.
+- `benchmark/binary_size/cprintf`: only `lib/c.Printf` (default and `-lto=full`);
+- `benchmark/binary_size/println`: only the built-in `println` (default and `-lto=full`);
+- `benchmark/binary_size/fmtprintf`: `fmt.Printf` (default and `-lto=full`).
 
 For each workload, the collector performs an unmeasured warm build, then records
 median build time, median process time, file size, executable-code bytes,

benchmark/baseline/main.go

@@ -51,12 +51,16 @@ type workload struct {
 	name   string
 	source string
 	output string
+	flags  []string
 }
 
 var workloads = []workload{
 	{name: "cprintf", source: "benchmark/binary_size/cprintf/main.go", output: "Hello, world\n"},
+	{name: "cprintf-lto", source: "benchmark/binary_size/cprintf/main.go", output: "Hello, world\n", flags: []string{"-lto=full"}},
 	{name: "println", source: "benchmark/binary_size/println/main.go", output: "Hello, world\n"},
+	{name: "println-lto", source: "benchmark/binary_size/println/main.go", output: "Hello, world\n", flags: []string{"-lto=full"}},
 	{name: "fmtprintf", source: "benchmark/binary_size/fmtprintf/main.go", output: "Hello, world\n"},
+	{name: "fmtprintf-lto", source: "benchmark/binary_size/fmtprintf/main.go", output: "Hello, world\n", flags: []string{"-lto=full"}},
 }
 
 var expectedGoBenchmarks = []string{
@@ -223,15 +227,17 @@ func collect(ctx context.Context, root, llgo, out string, buildRuns, runRuns int
 	var sizes, timings []metric
 	for _, item := range workloads {
 		binary := filepath.Join(binDir, item.name)
+		buildArgs := append([]string{"build"}, item.flags...)
+		buildArgs = append(buildArgs, "-o", binary, filepath.Join(root, item.source))
 		// Keep first-use toolchain and filesystem caches out of the measured
 		// median so the first revision is not systematically disadvantaged.
-		if err := run(ctx, env, io.Discard, llgo, "build", "-o", binary, filepath.Join(root, item.source)); err != nil {
+		if err := run(ctx, env, io.Discard, llgo, buildArgs...); err != nil {
 			return fmt.Errorf("warm build %s: %w", item.name, err)
 		}
 		buildDurations := make([]time.Duration, 0, buildRuns)
 		for range buildRuns {
 			start := time.Now()
-			if err := run(ctx, env, io.Discard, llgo, "build", "-o", binary, filepath.Join(root, item.source)); err != nil {
+			if err := run(ctx, env, io.Discard, llgo, buildArgs...); err != nil {
 				return fmt.Errorf("build %s: %w", item.name, err)
 			}
 			buildDurations = append(buildDurations, time.Since(start))

benchmark/baseline/main_test.go

@@ -97,6 +97,7 @@ func TestExportBenchmarks(t *testing.T) {
 		"Unit file-bytes better=lower assume=exact",
 		"Unit build-ns better=lower",
 		"BenchmarkProgram/cprintf 1 1 file-bytes 1 text-bytes 1 data-bytes 1 bss-bytes 1 build-ns 1 run-ns",
+		"BenchmarkProgram/cprintf-lto 1 1 file-bytes 1 text-bytes 1 data-bytes 1 bss-bytes 1 build-ns 1 run-ns",
 		"BenchmarkRuntimeGetG-1 100 12.5 ns/op",
 	} {
 		if !strings.Contains(text, want) {