dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2024-11-20 12:28:26 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

dec_neon: add SaturateAndStore4x4

Browse Source 
converts 2 s16 vectors to 2 u8 and store to uint8_t destination;
TransformAC3 can reuse this after a rework

Change-Id: Ia9370283ee3d9bfbc8c008fa883412100ff483d0
This commit is contained in:
James Zern 2014-02-18 19:42:35 -08:00
parent e02f16ef45
commit b7b60ca16c
1 changed files with 16 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
src/dsp/dec_neon.c
View File

@ -89,6 +89,21 @@
"vst2.8 {" #c1"[6], " #c2"[6]}," #p "," #stride " \n" \ "vst2.8 {" #c1"[6], " #c2"[6]}," #p "," #stride " \n" \
"vst2.8 {" #c1"[7], " #c2"[7]}," #p "," #stride " \n" "vst2.8 {" #c1"[7], " #c2"[7]}," #p "," #stride " \n"
// Performs unsigned 8b saturation on 'dst01' and 'dst23' storing the result
// to the corresponding rows of 'dst'.
static WEBP_INLINE void SaturateAndStore4x4(uint8_t* const dst,
int16x8_t dst01, int16x8_t dst23) {
// Unsigned saturate to 8b.
const uint8x8_t dst01_u8 = vqmovun_s16(dst01);
const uint8x8_t dst23_u8 = vqmovun_s16(dst23);
// Store the results.
*(int*)(dst + 0 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst01_u8), 0);
*(int*)(dst + 1 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst01_u8), 1);
*(int*)(dst + 2 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst23_u8), 0);
*(int*)(dst + 3 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst23_u8), 1);
}
//----------------------------------------------------------------------------- //-----------------------------------------------------------------------------
// Simple In-loop filtering (Paragraph 15.2) // Simple In-loop filtering (Paragraph 15.2)
@ -335,17 +350,8 @@ static void TransformDC(const int16_t* in, uint8_t* dst) {
// Add the inverse transform. // Add the inverse transform.
dst01_s16 = vaddq_s16(dst01_s16, DC); dst01_s16 = vaddq_s16(dst01_s16, DC);
dst23_s16 = vaddq_s16(dst23_s16, DC); dst23_s16 = vaddq_s16(dst23_s16, DC);
{
// Unsigned saturate to 8b.
const uint8x8_t dst01_u8 = vqmovun_s16(dst01_s16);
const uint8x8_t dst23_u8 = vqmovun_s16(dst23_s16);
// Store the results. SaturateAndStore4x4(dst, dst01_s16, dst23_s16);
*(int*)(dst + 0 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst01_u8), 0);
*(int*)(dst + 1 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst01_u8), 1);
*(int*)(dst + 2 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst23_u8), 0);
*(int*)(dst + 3 * BPS) = vget_lane_s32(vreinterpret_s32_u8(dst23_u8), 1);
}
} }
} }