Phase 21.7 normalization: optimization pre-work + bench harness expansion

- Add opt-in optimizations (defaults OFF)
  - Ret purity verifier: NYASH_VERIFY_RET_PURITY=1
  - strlen FAST enhancement for const handles
  - FAST_INT gate for same-BB SSA optimization
  - length cache for string literals in llvmlite
- Expand bench harness (tools/perf/microbench.sh)
  - Add branch/call/stringchain/arraymap/chip8/kilo cases
  - Auto-calculate ratio vs C reference
  - Document in benchmarks/README.md
- Compiler health improvements
  - Unify PHI insertion to insert_phi_at_head()
  - Add NYASH_LLVM_SKIP_BUILD=1 for build reuse
- Runtime & safety enhancements
  - Clarify Rust/Hako ownership boundaries
  - Strengthen receiver localization (LocalSSA/pin/after-PHIs)
  - Stop excessive PluginInvoke→BoxCall rewrites
- Update CURRENT_TASK.md, docs, and canaries

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

crates/nyash-llvm-compiler/src/main.rs

@@ -131,6 +131,12 @@ fn main() -> Result<()> {
         bail!("input JSON not found: {}", input_path.display());
     }
 
+    // Optional: dump incoming MIR JSON for diagnostics (AotPrep 後の入力を観測)
+    if let Ok(dump_path) = env::var("NYASH_LLVM_DUMP_MIR_IN") {
+        let _ = std::fs::copy(&input_path, &dump_path);
+        eprintln!("[ny-llvmc] dumped MIR input to {}", dump_path);
+    }
+
     // Optional: preflight shape/hints (best-effort; no behavior change)
     if let Ok(s) = std::fs::read_to_string(&input_path) {
         if let Ok(val) = serde_json::from_str::<JsonValue>(&s) {
@@ -379,9 +385,26 @@ fn link_executable(
             cmd.arg(tok);
         }
     }
-    let status = cmd.status().context("failed to invoke system linker")?;
-    if !status.success() {
-        bail!("linker exited with status: {:?}", status.code());
+    // Run linker and capture diagnostics for better error reporting
+    let output = cmd
+        .output()
+        .with_context(|| format!("failed to invoke system linker: {}", linker))?;
+    if !output.status.success() {
+        eprintln!("[ny-llvmc/link] command: {}", linker);
+        // Show args (for debugging)
+        // Note: std::process::Command doesn't expose argv back; re-emit essential parts
+        eprintln!(
+            "[ny-llvmc/link] args: -o {} {} -Wl,--whole-archive {} -Wl,--no-whole-archive -ldl -lpthread -lm {}",
+            out_exe.display(),
+            obj.display(),
+            libnyrt.display(),
+            extra_libs.unwrap_or("")
+        );
+        let stderr = String::from_utf8_lossy(&output.stderr);
+        let stdout = String::from_utf8_lossy(&output.stdout);
+        eprintln!("[ny-llvmc/link:stdout]\n{}", stdout);
+        eprintln!("[ny-llvmc/link:stderr]\n{}", stderr);
+        bail!("linker exited with status: {:?}", output.status.code());
     }
     Ok(())
 }

crates/nyash_kernel/src/lib.rs

@@ -38,6 +38,7 @@ pub extern "C" fn nyash_string_len_h(handle: i64) -> i64 {
 
 // FAST-path helper: compute string length from raw pointer (i8*) with mode (reserved)
 // Exported as both legacy name (nyash.string.length_si) and neutral name (nyrt_string_length)
+#[inline(always)]
 #[export_name = "nyrt_string_length"]
 pub extern "C" fn nyrt_string_length(ptr: *const i8, _mode: i64) -> i64 {
     use std::ffi::CStr;