dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-12-27 06:08:21 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

dec_neon: add DC4 intra predictor

Browse Source 
~70% faster

Change-Id: I2e06907b8d69be71a8c5581832c931923c24bab0
This commit is contained in:
James Zern 2014-10-22 17:02:28 +02:00
parent 79abfbd9df
commit eba6ce06c3
1 changed files with 25 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/dsp/dec_neon.c
View File

@ -1261,6 +1261,30 @@ static void TransformAC3(const int16_t* in, uint8_t* dst) {
//------------------------------------------------------------------------------
// 4x4
static void DC4(uint8_t* dst) { // DC
const uint16x8_t A0 = vmovl_u8(vld1_u8(dst - BPS + 0));
const uint16x8_t A1 = vmovl_u8(vld1_u8(dst - BPS + 1));
const uint16x8_t A2 = vmovl_u8(vld1_u8(dst - BPS + 2));
const uint16x8_t A3 = vmovl_u8(vld1_u8(dst - BPS + 3));
const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
const uint16x8_t s0 = vaddq_u16(A0, L0);
const uint16x8_t s1 = vaddq_u16(A1, L1);
const uint16x8_t s2 = vaddq_u16(A2, L2);
const uint16x8_t s3 = vaddq_u16(A3, L3);
const uint16x8_t s01 = vaddq_u16(s0, s1);
const uint16x8_t s23 = vaddq_u16(s2, s3);
const uint16x8_t sum = vaddq_u16(s01, s23);
const uint8x8_t dc0 = vrshrn_n_u16(sum, 3); // (sum + 4) >> 3
const uint8x8_t dc = vdup_lane_u8(dc0, 0);
int i;
for (i = 0; i < 4; ++i) {
vst1_lane_u32((uint32_t*)(dst + i * BPS), vreinterpret_u32_u8(dc), 0);
}
}
static void TM4(uint8_t* dst) { // TrueMotion
const uint8x8_t TL = vdup_n_u8(dst[-BPS - 1]); // top-left pixel 'A[-1]'
const uint8x8_t T = vld1_u8(dst - BPS); // top row 'A[0..3]'
@ -1354,6 +1378,7 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitNEON(void) {
VP8SimpleVFilter16i = SimpleVFilter16i;
VP8SimpleHFilter16i = SimpleHFilter16i;
VP8PredLuma4[0] = DC4;
VP8PredLuma4[1] = TM4;
VP8PredLuma4[2] = VE4;
VP8PredLuma4[6] = LD4;