diff --git a/Common/Math/CrossSIMD.h b/Common/Math/CrossSIMD.h
index ce42d9dcd4..c2851b5775 100644
--- a/Common/Math/CrossSIMD.h
+++ b/Common/Math/CrossSIMD.h
@@ -38,7 +38,7 @@ inline float32x4_t vmulq_laneq_f32(float32x4_t a, float32x4_t b, int lane) {
 	case 0: return vmulq_lane_f32(a, vget_low_f32(b), 0);
 	case 1: return vmulq_lane_f32(a, vget_low_f32(b), 1);
 	case 2: return vmulq_lane_f32(a, vget_high_f32(b), 0);
-	case 3: return vmulq_lane_f32(a, vget_high_f32(b), 1);
+	default: return vmulq_lane_f32(a, vget_high_f32(b), 1);
 	}
 }
 
@@ -47,8 +47,12 @@ inline float32x4_t vmlaq_laneq_f32(float32x4_t a, float32x4_t b, float32x4_t c,
 	case 0: return vmlaq_lane_f32(a, b, vget_low_f32(c), 0);
 	case 1: return vmlaq_lane_f32(a, b, vget_low_f32(c), 1);
 	case 2: return vmlaq_lane_f32(a, b, vget_high_f32(c), 0);
-	case 3: return vmlaq_lane_f32(a, b, vget_high_f32(c), 1);
+	default: return vmlaq_lane_f32(a, b, vget_high_f32(c), 1);
 	}
 }
 
+inline uint32x4_t vcgezq_f32(float32x4_t v) {
+	return vcgeq_f32(v, vdupq_n_f32(0.0f));
+}
+
 #endif
diff --git a/GPU/Common/DrawEngineCommon.cpp b/GPU/Common/DrawEngineCommon.cpp
index 52b48defb3..71ead60a02 100644
--- a/GPU/Common/DrawEngineCommon.cpp
+++ b/GPU/Common/DrawEngineCommon.cpp
@@ -444,11 +444,10 @@ bool DrawEngineCommon::TestBoundingBoxFast(const void *vdata, int vertexCount, u
 			_mm_storeu_ps(verts + i * 3, pos);  // TODO: use stride 4 to avoid clashing writes?
 		}
 #elif PPSSPP_ARCH(ARM_NEON)
-		float32x4_t scaleFactor = vdupq_n_f32(1.0f / 32768.0f);
 		for (int i = 0; i < vertexCount; i++) {
 			const s16 *dataPtr = ((const s16 *)((const s8 *)vdata + i * stride + offset));
 			int32x4_t data = vmovl_s16(vld1_s16(dataPtr));
-			float32x4_t pos = vmulq_f32(vcvtq_f32_s32(data), scaleFactor);
+			float32x4_t pos = vcvtq_n_f32_s32(data, 15);  // >> 15 = division by 32768.0f
 			vst1q_f32(verts + i * 3, pos);
 		}
 #else