Jmh bench suite bench sinewave end to end

modules/termflow-bench/src/main/scala/termflow/bench/SineWaveFrameBench.scala

@@ -0,0 +1,56 @@
+package termflow.bench
+
+import java.util.concurrent.TimeUnit
+
+import org.openjdk.jmh.annotations.*
+
+import termflow.apps.stress.SineWaveApp
+import termflow.testkit.TuiTestDriver
+import termflow.tui.AnsiRenderer
+import termflow.tui.AnsiRenderer.{ DiffResult, RenderFrame }
+
+/**
+ * End-to-end per-frame benchmark for [[termflow.apps.stress.SineWaveApp]]:
+ * each `@Benchmark` invocation simulates one animation tick the way the
+ * runtime would — advance the wave phase, rebuild the `RootNode` via
+ * `App.view`, `buildFrame` it, then `diff` against the previous frame.
+ *
+ * The terminal size is a `@Param` axis so the sweep covers small and large
+ * grids. The starting [[SineWaveApp.Model]] carries `Sub` fields and so cannot
+ * be hand-constructed; it is obtained from a real `App.init` via
+ * [[TuiTestDriver]] in [[setup]].
+ *
+ * Thread safety is self-enforcing: the mutable per-invocation state (current
+ * model + previous frame) lives in a `@State(Scope.Thread)` object and the
+ * class is `@Threads(1)`, so each worker thread owns its own copy regardless
+ * of any CLI flag.
+ */
+@BenchmarkMode(Array(Mode.AverageTime))
+@OutputTimeUnit(TimeUnit.MICROSECONDS)
+@Threads(1)
+@State(Scope.Thread)
+class SineWaveFrameBench:
+
+  @Param(Array("80x24", "200x50"))
+  var size: String = ""
+
+  private var model: SineWaveApp.Model = null
+  private var prev: RenderFrame        = null
+
+  @Setup
+  def setup(): Unit =
+    val Array(w, h) = size.split("x").map(_.toInt): @unchecked
+    val driver      = TuiTestDriver(SineWaveApp.App, w, h)
+    driver.init()
+    model = driver.model
+    prev = AnsiRenderer.buildFrame(SineWaveApp.App.view(model))
+
+  @Benchmark
+  def frame: DiffResult =
+    val next = model.copy(phase = model.phase + model.step)
+    val root = SineWaveApp.App.view(next)
+    val cur  = AnsiRenderer.buildFrame(root)
+    val result = AnsiRenderer.diff(Some(prev), cur)
+    model = next
+    prev = cur
+    result