Use switch operator[](int) to kth<int>() so we can use vget_lane.

src/opts/SkNx_neon.h

 /*
  * Copyright 2015 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef SkNx_neon_DEFINED
 #define SkNx_neon_DEFINED
 
 #include <arm_neon.h>
 
 template <>
 class SkNi<2, int32_t> {
 public:
     SkNi(int32x2_t vec) : fVec(vec) {}
 
     SkNi() {}
-    bool allTrue() const { return fVec[0] && fVec[1]; }
-    bool anyTrue() const { return fVec[0] || fVec[1]; }
+    bool allTrue() const { return vget_lane_s32(fVec, 0) && vget_lane_s32(fVec, 1); }
+    bool anyTrue() const { return vget_lane_s32(fVec, 0) || vget_lane_s32(fVec, 1); }
 private:
     int32x2_t fVec;
 };
 
 template <>
 class SkNi<4, int32_t> {
 public:
     SkNi(int32x4_t vec) : fVec(vec) {}
 
     SkNi() {}
-    bool allTrue() const { return fVec[0] && fVec[1] && fVec[2] && fVec[3]; }
-    bool anyTrue() const { return fVec[0] || fVec[1] || fVec[2] || fVec[3]; }
+    bool allTrue() const { return vgetq_lane_s32(fVec, 0) && vgetq_lane_s32(fVec, 1)
+                               && vgetq_lane_s32(fVec, 2) && vgetq_lane_s32(fVec, 3); }
+    bool anyTrue() const { return vgetq_lane_s32(fVec, 0) || vgetq_lane_s32(fVec, 1)
+                               || vgetq_lane_s32(fVec, 2) || vgetq_lane_s32(fVec, 3); }
 private:
     int32x4_t fVec;
 };
 
 template <>
 class SkNf<2, float> {
     typedef SkNi<2, int32_t> Ni;
 public:
     SkNf(float32x2_t vec) : fVec(vec) {}
 
@@ skipping 48 matching lines @@
     #if defined(SK_CPU_ARM64)
         return vsqrt_f32(fVec);
     #else
         float32x2_t est1 = this->rsqrt().fVec,
         // An extra step of Newton's method to refine the estimate of 1/sqrt(this).
                     est2 = vmul_f32(vrsqrts_f32(fVec, vmul_f32(est1, est1)), est1);
         return vmul_f32(fVec, est2);
     #endif
     }
 
-    float operator[] (int k) const {
+    template <int k> float kth() const {
         SkASSERT(0 <= k && k < 2);
-        return fVec[k];
+        return vget_lane_f32(fVec, k&1);
     }
 
 private:
     float32x2_t fVec;
 };
 
 #if defined(SK_CPU_ARM64)
 template <>
 class SkNi<2, int64_t> {
 public:
     SkNi(int64x2_t vec) : fVec(vec) {}
 
     SkNi() {}
-    bool allTrue() const { return fVec[0] && fVec[1]; }
-    bool anyTrue() const { return fVec[0] || fVec[1]; }
+    bool allTrue() const { return vgetq_lane_s64(fVec, 0) && vgetq_lane_s64(fVec, 1); }
+    bool anyTrue() const { return vgetq_lane_s64(fVec, 0) || vgetq_lane_s64(fVec, 1); }
 private:
     int64x2_t fVec;
 };
 
 template <>
 class SkNf<2, double> {
     typedef SkNi<2, int64_t> Ni;
 public:
     SkNf(float64x2_t vec) : fVec(vec) {}
 
@@ skipping 34 matching lines @@
         return est1;
     }
 
     SkNf invert() const {
         float64x2_t est1 = this->approxInvert().fVec,
                     est2 = vmulq_f64(vrecpsq_f64(est1, fVec), est1),
                     est3 = vmulq_f64(vrecpsq_f64(est2, fVec), est2);
         return est3;
     }
 
-    double operator[] (int k) const {
+    template <int k> double kth() const {
         SkASSERT(0 <= k && k < 2);
-        return fVec[k];
+        return vgetq_lane_f64(fVec, k&1);
     }
 
 private:
     float64x2_t fVec;
 };
 #endif//defined(SK_CPU_ARM64)
 
 template <>
 class SkNf<4, float> {
     typedef SkNi<4, int32_t> Ni;
@@ skipping 52 matching lines @@
     #if defined(SK_CPU_ARM64)
         return vsqrtq_f32(fVec);
     #else
         float32x4_t est1 = this->rsqrt().fVec,
         // An extra step of Newton's method to refine the estimate of 1/sqrt(this).
                     est2 = vmulq_f32(vrsqrtsq_f32(fVec, vmulq_f32(est1, est1)), est1);
         return vmulq_f32(fVec, est2);
     #endif
     }
 
-    float operator[] (int k) const {
+    template <int k> float kth() const {
         SkASSERT(0 <= k && k < 4);
-        return fVec[k];
+        return vgetq_lane_f32(fVec, k&3);
     }
 
 private:
     float32x4_t fVec;
 };
 
 #endif//SkNx_neon_DEFINED