Refactor to put SkXfermode_opts inside SK_OPTS_NS.

src/opts/Sk4px_NEON.h

Index: src/opts/Sk4px_NEON.h
diff --git a/src/opts/Sk4px_NEON.h b/src/opts/Sk4px_NEON.h
index 89841d927ee5efd927012ce25f8c110be071f713..e88771d2a42e09567e0a615186a1e4b028a383d9 100644
--- a/src/opts/Sk4px_NEON.h
+++ b/src/opts/Sk4px_NEON.h
@@ -5,64 +5,64 @@
  * found in the LICENSE file.
  */
 
-namespace { // See Sk4px.h
-
-inline Sk4px Sk4px::DupPMColor(SkPMColor px) { return Sk16b((uint8x16_t)vdupq_n_u32(px)); }
+SK_ALWAYS_INLINE Sk4px Sk4px::DupPMColor(SkPMColor px) {
+    return Sk16b((uint8x16_t)vdupq_n_u32(px));
+}
 
-inline Sk4px Sk4px::Load4(const SkPMColor px[4]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load4(const SkPMColor px[4]) {
     return Sk16b((uint8x16_t)vld1q_u32(px));
 }
-inline Sk4px Sk4px::Load2(const SkPMColor px[2]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load2(const SkPMColor px[2]) {
     uint32x2_t px2 = vld1_u32(px);
     return Sk16b((uint8x16_t)vcombine_u32(px2, px2));
 }
-inline Sk4px Sk4px::Load1(const SkPMColor px[1]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load1(const SkPMColor px[1]) {
     return Sk16b((uint8x16_t)vdupq_n_u32(*px));
 }
 
-inline void Sk4px::store4(SkPMColor px[4]) const {
+SK_ALWAYS_INLINE void Sk4px::store4(SkPMColor px[4]) const {
     vst1q_u32(px, (uint32x4_t)this->fVec);
 }
-inline void Sk4px::store2(SkPMColor px[2]) const {
+SK_ALWAYS_INLINE void Sk4px::store2(SkPMColor px[2]) const {
     vst1_u32(px, (uint32x2_t)vget_low_u8(this->fVec));
 }
-inline void Sk4px::store1(SkPMColor px[1]) const {
+SK_ALWAYS_INLINE void Sk4px::store1(SkPMColor px[1]) const {
     vst1q_lane_u32(px, (uint32x4_t)this->fVec, 0);
 }
 
-inline Sk4px::Wide Sk4px::widenLo() const {
+SK_ALWAYS_INLINE Sk4px::Wide Sk4px::widenLo() const {
     return Sk16h(vmovl_u8(vget_low_u8 (this->fVec)),
                  vmovl_u8(vget_high_u8(this->fVec)));
 }
 
-inline Sk4px::Wide Sk4px::widenHi() const {
+SK_ALWAYS_INLINE Sk4px::Wide Sk4px::widenHi() const {
     return Sk16h(vshll_n_u8(vget_low_u8 (this->fVec), 8),
                  vshll_n_u8(vget_high_u8(this->fVec), 8));
 }
 
-inline Sk4px::Wide Sk4px::widenLoHi() const {
+SK_ALWAYS_INLINE Sk4px::Wide Sk4px::widenLoHi() const {
     auto zipped = vzipq_u8(this->fVec, this->fVec);
     return Sk16h((uint16x8_t)zipped.val[0],
                  (uint16x8_t)zipped.val[1]);
 }
 
-inline Sk4px::Wide Sk4px::mulWiden(const Sk16b& other) const {
+SK_ALWAYS_INLINE Sk4px::Wide Sk4px::mulWiden(const Sk16b& other) const {
     return Sk16h(vmull_u8(vget_low_u8 (this->fVec), vget_low_u8 (other.fVec)),
                  vmull_u8(vget_high_u8(this->fVec), vget_high_u8(other.fVec)));
 }
 
-inline Sk4px Sk4px::Wide::addNarrowHi(const Sk16h& other) const {
+SK_ALWAYS_INLINE Sk4px Sk4px::Wide::addNarrowHi(const Sk16h& other) const {
     const Sk4px::Wide o(other);  // Should be no code, but allows us to access fLo, fHi.
     return Sk16b(vcombine_u8(vaddhn_u16(this->fLo.fVec, o.fLo.fVec),
                              vaddhn_u16(this->fHi.fVec, o.fHi.fVec)));
 }
 
-inline Sk4px Sk4px::alphas() const {
+SK_ALWAYS_INLINE Sk4px Sk4px::alphas() const {
     auto as = vshrq_n_u32((uint32x4_t)fVec, SK_A32_SHIFT);  // ___3 ___2 ___1 ___0
     return Sk16b((uint8x16_t)vmulq_n_u32(as, 0x01010101));  // 3333 2222 1111 0000
 }
 
-inline Sk4px Sk4px::Load4Alphas(const SkAlpha a[4]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load4Alphas(const SkAlpha a[4]) {
     uint8x16_t a8 = vdupq_n_u8(0);                           // ____ ____ ____ ____
     a8 = vld1q_lane_u8(a+0, a8,  0);                         // ____ ____ ____ ___0
     a8 = vld1q_lane_u8(a+1, a8,  4);                         // ____ ____ ___1 ___0
@@ -72,7 +72,7 @@ inline Sk4px Sk4px::Load4Alphas(const SkAlpha a[4]) {
     return Sk16b((uint8x16_t)vmulq_n_u32(a32, 0x01010101));  // 3333 2222 1111 0000
 }
 
-inline Sk4px Sk4px::Load2Alphas(const SkAlpha a[2]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load2Alphas(const SkAlpha a[2]) {
     uint8x16_t a8 = vdupq_n_u8(0);                           // ____ ____ ____ ____
     a8 = vld1q_lane_u8(a+0, a8,  0);                         // ____ ____ ____ ___0
     a8 = vld1q_lane_u8(a+1, a8,  4);                         // ____ ____ ___1 ___0
@@ -80,16 +80,16 @@ inline Sk4px Sk4px::Load2Alphas(const SkAlpha a[2]) {
     return Sk16b((uint8x16_t)vmulq_n_u32(a32, 0x01010101));  // ____ ____ 1111 0000
 }
 
-inline Sk4px Sk4px::zeroColors() const {
+SK_ALWAYS_INLINE Sk4px Sk4px::zeroColors() const {
     return Sk16b(vandq_u8(this->fVec, (uint8x16_t)vdupq_n_u32(0xFF << SK_A32_SHIFT)));
 }
 
-inline Sk4px Sk4px::zeroAlphas() const {
+SK_ALWAYS_INLINE Sk4px Sk4px::zeroAlphas() const {
     // vbic(a,b) == a & ~b
     return Sk16b(vbicq_u8(this->fVec, (uint8x16_t)vdupq_n_u32(0xFF << SK_A32_SHIFT)));
 }
 
-static inline uint8x16_t widen_to_8888(uint16x4_t v) {
+static SK_ALWAYS_INLINE uint8x16_t widen_to_8888(uint16x4_t v) {
     // RGB565 format:   |R....|G.....|B....|
     //           Bit:  16    11      5     0
 
@@ -115,7 +115,7 @@ static inline uint8x16_t widen_to_8888(uint16x4_t v) {
                                  vdupq_n_u32(0xFF << SK_A32_SHIFT))));
 }
 
-static inline uint16x4_t narrow_to_565(uint8x16_t w8x16) {
+static SK_ALWAYS_INLINE uint16x4_t narrow_to_565(uint8x16_t w8x16) {
     uint32x4_t w = (uint32x4_t)w8x16;
 
     // Extract out top RGB 565 bits of each pixel, with no rounding.
@@ -136,29 +136,27 @@ static inline uint16x4_t narrow_to_565(uint8x16_t w8x16) {
 }
 
 
-inline Sk4px Sk4px::Load4(const SkPMColor16 src[4]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load4(const SkPMColor16 src[4]) {
     return Sk16b(widen_to_8888(vld1_u16(src)));
 }
-inline Sk4px Sk4px::Load2(const SkPMColor16 src[2]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load2(const SkPMColor16 src[2]) {
     auto src2 = ((uint32_t)src[0]      )
               | ((uint32_t)src[1] << 16);
     return Sk16b(widen_to_8888(vcreate_u16(src2)));
 }
-inline Sk4px Sk4px::Load1(const SkPMColor16 src[1]) {
+SK_ALWAYS_INLINE Sk4px Sk4px::Load1(const SkPMColor16 src[1]) {
     return Sk16b(widen_to_8888(vcreate_u16(src[0])));
 }
 
-inline void Sk4px::store4(SkPMColor16 dst[4]) const {
+SK_ALWAYS_INLINE void Sk4px::store4(SkPMColor16 dst[4]) const {
     vst1_u16(dst, narrow_to_565(this->fVec));
 }
-inline void Sk4px::store2(SkPMColor16 dst[2]) const {
+SK_ALWAYS_INLINE void Sk4px::store2(SkPMColor16 dst[2]) const {
     auto v = narrow_to_565(this->fVec);
     dst[0] = vget_lane_u16(v, 0);
     dst[1] = vget_lane_u16(v, 1);
 }
-inline void Sk4px::store1(SkPMColor16 dst[1]) const {
+SK_ALWAYS_INLINE void Sk4px::store1(SkPMColor16 dst[1]) const {
     dst[0] = vget_lane_u16(narrow_to_565(this->fVec), 0);
 }
 
-} // namespace
-