feat: initial release

Assisted-by: GLM 5.3 Flash

internal/engine/engine_bench_test.go

@@ -0,0 +1,25 @@
+// Copyright (c) 2026 Petr Balvín <opensource@petrbalvin.org> (https://petrbalvin.org)
+// SPDX-License-Identifier: MIT
+
+package engine
+
+import (
+	"fmt"
+	"testing"
+)
+
+// BenchmarkPoolRoundTrip measures the full borrow/return cycle at the
+// sizes the kernels actually request, including the clear-on-borrow
+// cost the pool contract pays. Run before and after touching the pool.
+func BenchmarkPoolRoundTrip(b *testing.B) {
+	for _, n := range []int{64, 4096, 262144} {
+		b.Run(fmt.Sprintf("n=%d", n), func(b *testing.B) {
+			b.ReportAllocs()
+			for b.Loop() {
+				buf := GetFloat64Buf(n)
+				buf[0] = 1 // touch one slot: prove the buffer is writable
+				PutFloat64Buf(buf)
+			}
+		})
+	}
+}