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

DC16_NEON,aarch64: use vaddlv

Browse Source 
saves 3 instructions, neutral to mildly faster on a pixel 3a

Change-Id: I6ae57e8e38d4149167ea14e27cd2b32113b4f8e7
This commit is contained in:
James Zern 2019-12-05 21:00:45 -08:00
parent 53f3d8cf7e
commit 1b92fe75a1
1 changed files with 5 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
src/dsp/dec_neon.c
View File

@ -1511,11 +1511,16 @@ static WEBP_INLINE void DC16_NEON(uint8_t* dst, int do_top, int do_left) {
if (do_top) {
const uint8x16_t A = vld1q_u8(dst - BPS); // top row
#if defined(__aarch64__)
const uint16_t p3 = vaddlvq_u8(A);
sum_top = vdupq_n_u16(p3);
#else
const uint16x8_t p0 = vpaddlq_u8(A); // cascading summation of the top
const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
const uint16x4_t p2 = vpadd_u16(p1, p1);
const uint16x4_t p3 = vpadd_u16(p2, p2);
sum_top = vcombine_u16(p3, p3);
#endif
}
if (do_left) {