Update zwasm to v0.11.0 (x86_64 JIT, tail calls, multi-memory)

Bump zwasm dependency from v0.10.0 to v0.11.0.
Record wasm benchmarks. Update docs with latest numbers.

Author: chaploud

.dev/memo.md

@@ -6,10 +6,10 @@ Session handover document. Read at session start.
 
 - **All phases through 53 COMPLETE**
 - Coverage: 835+ vars (635/706 core, 16 namespaces total)
-- Wasm engine: zwasm v0.8.0 (GitHub URL dependency, build.zig.zon)
+- Wasm engine: zwasm v0.11.0 (GitHub URL dependency, build.zig.zon)
 - Bridge: `src/wasm/types.zig` (751 lines, thin wrapper over zwasm)
 - 49 upstream test files, all passing. 6/6 e2e tests pass.
-- Benchmarks: `bench/history.yaml` (zwasm-v0.8.0 entry = latest baseline)
+- Benchmarks: `bench/history.yaml` (zwasm-v0.11.0 entry = latest baseline)
 
 ## Strategic Direction
 

.dev/roadmap.md

@@ -19,7 +19,7 @@ Goal: Babashka-competitive startup, single binary distribution, behavioral compa
 - **71 skip vars** remaining (70+ permanently JVM, ~1 implementable)
 - **49 upstream test files**, all passing (44 upstream ports + 5 SCI)
 - **31 benchmarks** (20 native + 5 wasm legacy + 4 wasm TinyGo + 2 GC)
-- **Wasm engine**: zwasm v0.7.0 (Register IR + ARM64 JIT, 0.9-2.2x of wasmtime)
+- **Wasm engine**: zwasm v0.11.0 (Register IR + ARM64/x86_64 JIT, 1.3-2.4x of wasmtime)
 
 ## Completed Phases
 

README.md

@@ -106,9 +106,9 @@ Call WebAssembly modules directly from Clojure:
 - v128 SIMD operations
 - Predecoded IR with superinstructions for optimized dispatch
 
-> **Performance note**: The Wasm runtime ([zwasm](https://github.com/clojurewasm/zwasm) v0.8.0)
-> uses Register IR with ARM64 JIT, achieving 0.9-2.2x of wasmtime performance
-> depending on workload (call-heavy: ~2.2x, compute-heavy: ~1x, sieve: CW wins).
+> **Performance note**: The Wasm runtime ([zwasm](https://github.com/clojurewasm/zwasm) v0.11.0)
+> uses Register IR with ARM64/x86_64 JIT, achieving 1.3-2.4x of wasmtime performance
+> depending on workload (call-heavy: ~2.4x, compute-heavy: ~1.3x, sieve: CW wins).
 > Module load time is faster (~4ms vs ~5ms).
 
 ### Server & Networking

bench/README.md

@@ -222,18 +222,18 @@ Note: gc_stress Zig value (462.7ms) omitted — Zig benchmark uses
 
 ## Latest Wasm Runtime Results (2026-02-12)
 
-CW's built-in Wasm runtime (zwasm v0.7.0, Register IR + ARM64 JIT)
+CW's built-in Wasm runtime (zwasm v0.11.0, Register IR + ARM64/x86_64 JIT)
 vs wasmtime JIT. Same TinyGo-compiled `.wasm` modules, startup subtracted.
 
-| Benchmark | CW (ms)  | wasmtime (ms) | Ratio (CW/wt) |
-|-----------|----------|---------------|----------------|
-| fib       | 522      | 227           | 2.3x           |
-| tak       | 2647     | 1274          | 2.1x           |
-| arith     | 0.3      | 0.1           | 3.0x           |
-| sieve     | 8.0      | 4.4           | 1.8x           |
-| fib_loop  | 13.0     | 0.5           | 26.0x          |
-| gcd       | 59.1     | 43.9          | 1.3x           |
-
-zwasm reaches 1.3-2.3x of wasmtime performance on compute-heavy benchmarks.
-fib_loop and arith show higher ratios due to loop optimization differences.
+| Benchmark | CW (ms) | wasmtime (ms) | Ratio (CW/wt) |
+|-----------|---------|---------------|----------------|
+| fib       | 546     | 226           | 2.4x           |
+| tak       | 2620    | 1257          | 2.1x           |
+| arith     | 0.2     | 0.1           | 2.0x           |
+| sieve     | 6.0     | 4.3           | 1.4x           |
+| fib_loop  | 13.6    | 0.9           | 15.1x          |
+| gcd       | 58.7    | 45.3          | 1.3x           |
+
+zwasm reaches 1.3-2.4x of wasmtime performance on compute-heavy benchmarks.
+fib_loop shows higher ratio due to loop optimization differences.
 History tracked in `wasm_history.yaml`.

bench/wasm_history.yaml

@@ -265,3 +265,46 @@ entries:
         wt_warm_ms: 41.1
         ratio: 1.4
         speedup_vs_prev: 0.97
+  - id: "zwasm-0.11.0"
+    date: "2026-02-12"
+    reason: "zwasm v0.11.0 — x86_64 JIT, tail calls, multi-memory"
+    commit: c2d5cdd
+    build: ReleaseSafe
+    cw_startup_ms: 5.1
+    wt_startup_ms: 6.3
+    results:
+      fib:
+        desc: "fib(20) x 10000"
+        cw_warm_ms: 546
+        wt_warm_ms: 226
+        ratio: 2.4
+        speedup_vs_prev: 0.93
+      tak:
+        desc: "tak(18,12,6) x 10000"
+        cw_warm_ms: 2620
+        wt_warm_ms: 1257
+        ratio: 2.1
+        speedup_vs_prev: 0.99
+      arith:
+        desc: "arith(1M) x 10"
+        cw_warm_ms: 0.2
+        wt_warm_ms: 0.1
+        ratio: 2.0
+      sieve:
+        desc: "sieve(64K) x 100"
+        cw_warm_ms: 6.0
+        wt_warm_ms: 4.3
+        ratio: 1.4
+        speedup_vs_prev: 1.05
+      fib_loop:
+        desc: "fib_loop(25) x 1000000"
+        cw_warm_ms: 13.6
+        wt_warm_ms: 0.9
+        ratio: 15.1
+        speedup_vs_prev: 0.78
+      gcd:
+        desc: "gcd(1M,700K) x 1000000"
+        cw_warm_ms: 58.7
+        wt_warm_ms: 45.3
+        ratio: 1.3
+        speedup_vs_prev: 0.95

build.zig.zon

@@ -3,8 +3,8 @@
     .version = "0.0.1",
     .dependencies = .{
         .zwasm = .{
-            .url = "https://github.com/clojurewasm/zwasm/archive/v0.8.0.tar.gz",
-            .hash = "zwasm-0.1.0-IBbzF5euDwCAlK2wSrInwc_Y2jwe5gY4-tPwR-6sNVaT",
+            .url = "https://github.com/clojurewasm/zwasm/archive/v0.11.0.tar.gz",
+            .hash = "zwasm-0.1.0-IBbzFywwEgDOKsu6xvDSvLOrCvdEgV3LWf39rmzu-9Dj",
         },
     },
     .fingerprint = 0x62a7be489d633543,