1 /*
2 * Copyright 2015 Google Inc.
3 *
4 * Use of this source code is governed by a BSD-style license that can be
5 * found in the LICENSE file.
6 */
7
8 namespace { // NOLINT(google-build-namespaces)
9
widen()10 inline Sk4px::Wide Sk4px::widen() const {
11 return Sk16h(vmovl_u8(vget_low_u8 (this->fVec)),
12 vmovl_u8(vget_high_u8(this->fVec)));
13 }
14
mulWiden(const Sk16b & other)15 inline Sk4px::Wide Sk4px::mulWiden(const Sk16b& other) const {
16 return Sk16h(vmull_u8(vget_low_u8 (this->fVec), vget_low_u8 (other.fVec)),
17 vmull_u8(vget_high_u8(this->fVec), vget_high_u8(other.fVec)));
18 }
19
addNarrowHi(const Sk16h & other)20 inline Sk4px Sk4px::Wide::addNarrowHi(const Sk16h& other) const {
21 const Sk4px::Wide o(other); // Should be no code, but allows us to access fLo, fHi.
22 return Sk16b(vcombine_u8(vaddhn_u16(this->fLo.fVec, o.fLo.fVec),
23 vaddhn_u16(this->fHi.fVec, o.fHi.fVec)));
24 }
25
div255()26 inline Sk4px Sk4px::Wide::div255() const {
27 // Calculated as (x + (x+128)>>8 +128) >> 8. The 'r' in each instruction provides each +128.
28 return Sk16b(vcombine_u8(vraddhn_u16(this->fLo.fVec, vrshrq_n_u16(this->fLo.fVec, 8)),
29 vraddhn_u16(this->fHi.fVec, vrshrq_n_u16(this->fHi.fVec, 8))));
30 }
31
alphas()32 inline Sk4px Sk4px::alphas() const {
33 auto as = vshrq_n_u32((uint32x4_t)fVec, SK_A32_SHIFT); // ___3 ___2 ___1 ___0
34 return Sk16b((uint8x16_t)vmulq_n_u32(as, 0x01010101)); // 3333 2222 1111 0000
35 }
36
Load4Alphas(const SkAlpha a[4])37 inline Sk4px Sk4px::Load4Alphas(const SkAlpha a[4]) {
38 uint8x16_t a8 = vdupq_n_u8(0); // ____ ____ ____ ____
39 a8 = vld1q_lane_u8(a+0, a8, 0); // ____ ____ ____ ___0
40 a8 = vld1q_lane_u8(a+1, a8, 4); // ____ ____ ___1 ___0
41 a8 = vld1q_lane_u8(a+2, a8, 8); // ____ ___2 ___1 ___0
42 a8 = vld1q_lane_u8(a+3, a8, 12); // ___3 ___2 ___1 ___0
43 auto a32 = (uint32x4_t)a8; //
44 return Sk16b((uint8x16_t)vmulq_n_u32(a32, 0x01010101)); // 3333 2222 1111 0000
45 }
46
Load2Alphas(const SkAlpha a[2])47 inline Sk4px Sk4px::Load2Alphas(const SkAlpha a[2]) {
48 uint8x16_t a8 = vdupq_n_u8(0); // ____ ____ ____ ____
49 a8 = vld1q_lane_u8(a+0, a8, 0); // ____ ____ ____ ___0
50 a8 = vld1q_lane_u8(a+1, a8, 4); // ____ ____ ___1 ___0
51 auto a32 = (uint32x4_t)a8; //
52 return Sk16b((uint8x16_t)vmulq_n_u32(a32, 0x01010101)); // ____ ____ 1111 0000
53 }
54
55 } // namespace
56
57