sx primitives: add trig/transcendental/bit-op math helpers

Adds sin/cos/tan + inverse + hyperbolic + inverse-hyperbolic, log/log2/log10/log1p,
exp/expm1, cbrt, hypot (variadic), sign, fround/clz32/imul. All one-liners over
Float.* / Int32.*. Needed by JS-on-SX to unblock built-ins/Math tests.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

hosts/ocaml/lib/sx_primitives.ml

@@ -144,6 +144,90 @@ let () =
   register "pow" (fun args ->
     match args with [a; b] -> Number (as_number a ** as_number b)
     | _ -> raise (Eval_error "pow: 2 args"));
+  register "cbrt" (fun args ->
+    match args with [a] -> Number (Float.cbrt (as_number a)) | _ -> raise (Eval_error "cbrt: 1 arg"));
+  register "exp" (fun args ->
+    match args with [a] -> Number (Float.exp (as_number a)) | _ -> raise (Eval_error "exp: 1 arg"));
+  register "expm1" (fun args ->
+    match args with [a] -> Number (Float.expm1 (as_number a)) | _ -> raise (Eval_error "expm1: 1 arg"));
+  register "log" (fun args ->
+    match args with [a] -> Number (Float.log (as_number a)) | _ -> raise (Eval_error "log: 1 arg"));
+  register "log2" (fun args ->
+    match args with [a] -> Number (Float.log (as_number a) /. Float.log 2.0) | _ -> raise (Eval_error "log2: 1 arg"));
+  register "log10" (fun args ->
+    match args with [a] -> Number (Float.log10 (as_number a)) | _ -> raise (Eval_error "log10: 1 arg"));
+  register "log1p" (fun args ->
+    match args with [a] -> Number (Float.log1p (as_number a)) | _ -> raise (Eval_error "log1p: 1 arg"));
+  register "sin" (fun args ->
+    match args with [a] -> Number (Float.sin (as_number a)) | _ -> raise (Eval_error "sin: 1 arg"));
+  register "cos" (fun args ->
+    match args with [a] -> Number (Float.cos (as_number a)) | _ -> raise (Eval_error "cos: 1 arg"));
+  register "tan" (fun args ->
+    match args with [a] -> Number (Float.tan (as_number a)) | _ -> raise (Eval_error "tan: 1 arg"));
+  register "asin" (fun args ->
+    match args with [a] -> Number (Float.asin (as_number a)) | _ -> raise (Eval_error "asin: 1 arg"));
+  register "acos" (fun args ->
+    match args with [a] -> Number (Float.acos (as_number a)) | _ -> raise (Eval_error "acos: 1 arg"));
+  register "atan" (fun args ->
+    match args with [a] -> Number (Float.atan (as_number a)) | _ -> raise (Eval_error "atan: 1 arg"));
+  register "atan2" (fun args ->
+    match args with [a; b] -> Number (Float.atan2 (as_number a) (as_number b))
+    | _ -> raise (Eval_error "atan2: 2 args"));
+  register "sinh" (fun args ->
+    match args with [a] -> Number (Float.sinh (as_number a)) | _ -> raise (Eval_error "sinh: 1 arg"));
+  register "cosh" (fun args ->
+    match args with [a] -> Number (Float.cosh (as_number a)) | _ -> raise (Eval_error "cosh: 1 arg"));
+  register "tanh" (fun args ->
+    match args with [a] -> Number (Float.tanh (as_number a)) | _ -> raise (Eval_error "tanh: 1 arg"));
+  register "asinh" (fun args ->
+    match args with [a] -> Number (Float.asinh (as_number a)) | _ -> raise (Eval_error "asinh: 1 arg"));
+  register "acosh" (fun args ->
+    match args with [a] -> Number (Float.acosh (as_number a)) | _ -> raise (Eval_error "acosh: 1 arg"));
+  register "atanh" (fun args ->
+    match args with [a] -> Number (Float.atanh (as_number a)) | _ -> raise (Eval_error "atanh: 1 arg"));
+  register "hypot" (fun args ->
+    let square x = x *. x in
+    let sum = List.fold_left (fun acc a -> acc +. square (as_number a)) 0.0 args in
+    Number (Float.sqrt sum));
+  register "sign" (fun args ->
+    match args with
+    | [a] ->
+      let n = as_number a in
+      Number (if Float.is_nan n then Float.nan
+              else if n > 0.0 then 1.0
+              else if n < 0.0 then -1.0
+              else n)
+    | _ -> raise (Eval_error "sign: 1 arg"));
+  register "fround" (fun args ->
+    match args with [a] -> Number (Int32.float_of_bits (Int32.bits_of_float (as_number a)))
+    | _ -> raise (Eval_error "fround: 1 arg"));
+  register "clz32" (fun args ->
+    match args with
+    | [a] ->
+      let n = as_number a in
+      let i = if Float.is_nan n || Float.is_infinite n then 0l
+              else Int32.of_float (Float.rem n 4294967296.0) in
+      if i = 0l then Number 32.0
+      else
+        let high_bit = Int32.shift_left 1l 31 in
+        let count = ref 0 in
+        let x = ref i in
+        while Int32.logand !x high_bit = 0l do
+          incr count;
+          x := Int32.shift_left !x 1
+        done;
+        Number (float_of_int !count)
+    | _ -> raise (Eval_error "clz32: 1 arg"));
+  register "imul" (fun args ->
+    match args with
+    | [a; b] ->
+      let tou32 f =
+        if Float.is_nan f || Float.is_infinite f then 0l
+        else Int32.of_float (Float.rem f 4294967296.0) in
+      let ai = tou32 (as_number a) and bi = tou32 (as_number b) in
+      let r = Int32.mul ai bi in
+      Number (Int32.to_float r)
+    | _ -> raise (Eval_error "imul: 2 args"));
   register "clamp" (fun args ->
     match args with
     | [x; lo; hi] ->