diff --git a/src/CodeGen_D3D12Compute_Dev.cpp b/src/CodeGen_D3D12Compute_Dev.cpp
index ad4f6451f918..e40dd1b316ef 100644
--- a/src/CodeGen_D3D12Compute_Dev.cpp
+++ b/src/CodeGen_D3D12Compute_Dev.cpp
@@ -388,10 +388,12 @@ void CodeGen_D3D12Compute_Dev::CodeGen_D3D12Compute_C::visit(const Call *op) {
     } else if (op->name == "pow_f32" && can_prove(op->args[0] > 0)) {
         // If we know pow(x, y) is called with x > 0, we can use HLSL's pow
         // directly.
-        stream << "pow(" << print_expr(op->args[0]) << ", " << print_expr(op->args[1]) << ")";
+        Expr equiv = Call::make(op->type, "pow", op->args, Call::PureExtern);
+        equiv.accept(this);
     } else if (op->is_intrinsic(Call::round)) {
         // HLSL's round intrinsic has the correct semantics for our rounding.
-        print_assignment(op->type, "round(" + print_expr(op->args[0]) + ")");
+        Expr equiv = Call::make(op->type, "round", op->args, Call::PureExtern);
+        equiv.accept(this);
     } else {
         CodeGen_GPU_C::visit(op);
     }
diff --git a/test/correctness/math.cpp b/test/correctness/math.cpp
index 68ff3c0e56e8..9d09f19c9193 100644
--- a/test/correctness/math.cpp
+++ b/test/correctness/math.cpp
@@ -118,6 +118,33 @@ struct TestArgs {
     }
 };
 
+// For D3D12Compute, we lower directly to HLSL's pow function if the base is provably positive.
+// This test ensures that lowering is correct.
+void test_pow_positive() {
+    printf("Testing pow(x, y) where x > 0\n");
+    TestArgs<float> args(256, 0.0f, 10.0f);
+    Func test_pow_positive("test_pow_positive");
+    Var x("x"), xi("xi");
+    test_pow_positive(x) = pow(1.5f, args.data(x));
+
+    Target target = get_jit_target_from_environment();
+    if (target.has_gpu_feature()) {
+        test_pow_positive.gpu_tile(x, xi, 16);
+    } else if (target.has_feature(Target::HVX)) {
+        test_pow_positive.hexagon();
+    }
+    Buffer<float> result = test_pow_positive.realize({args.data.extent(0)}, target);
+    for (int i = 0; i < args.data.extent(0); i++) {
+        float c_result = pow(1.5f, args.data(i));
+        if (!relatively_equal(c_result, result(i), target)) {
+            fprintf(stderr, "For pow(1.5f, %.20f) == %.20f from C and %.20f from %s.\n",
+                    (double)args.data(i), (double)c_result, (double)result(i),
+                    target.to_string().c_str());
+            num_errors++;
+        }
+    }
+}
+
 // Using macros to expand name as both a C function and an Expr fragment.
 // It may well be possible to do this without macros, but that is left
 // for another day.
@@ -299,6 +326,7 @@ int main(int argc, char **argv) {
     call_1_float_types(trunc, 256, -25, 25);
 
     call_2_float_types(pow, 256, -10.0, 10.0, -4.0f, 4.0f);
+    test_pow_positive();
 
     const int8_t int8_min = std::numeric_limits<int8_t>::min();
     const int16_t int16_min = std::numeric_limits<int16_t>::min();