dany/libwebp
1
0
Fork 0
mirror of https://github.com/webmproject/libwebp.git synced 2025-07-17 06:19:54 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

faster rgb565/rgb4444/argb output

Browse Source 
SSE2 and NEON implementation.

Change-Id: I342a1c3d84937b8497f0aaecb7ce9bdb7f50296b
This commit is contained in:
Pascal Massimino
2015-12-18 08:11:33 +01:00
committed by James Zern
parent 4c7f565ff0
commit 7e7b6ccc7f
4 changed files with 131 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

76
src/dsp/yuv_sse2.c
View File

@ -110,6 +110,50 @@ static WEBP_INLINE void PackAndStore4(const __m128i* const R,
_mm_storeu_si128((__m128i*)(dst + 16), RGBA_hi);
}
// Pack R/G/B/A results into 16b output.
static WEBP_INLINE void PackAndStore4444(const __m128i* const R,
const __m128i* const G,
const __m128i* const B,
const __m128i* const A,
uint8_t* const dst) {
#if !defined(WEBP_SWAP_16BIT_CSP)
const __m128i rg0 = _mm_packus_epi16(*R, *G);
const __m128i ba0 = _mm_packus_epi16(*B, *A);
#else
const __m128i rg0 = _mm_packus_epi16(*B, *A);
const __m128i ba0 = _mm_packus_epi16(*R, *G);
#endif
const __m128i mask_0xf0 = _mm_set1_epi8(0xf0);
const __m128i rb1 = _mm_unpacklo_epi8(rg0, ba0); // rbrbrbrbrb...
const __m128i ga1 = _mm_unpackhi_epi8(rg0, ba0); // gagagagaga...
const __m128i rb2 = _mm_and_si128(rb1, mask_0xf0);
const __m128i ga2 = _mm_srli_epi16(_mm_and_si128(ga1, mask_0xf0), 4);
const __m128i rgba4444 = _mm_or_si128(rb2, ga2);
_mm_storeu_si128((__m128i*)dst, rgba4444);
}
// Pack R/G/B results into 16b output.
static WEBP_INLINE void PackAndStore565(const __m128i* const R,
const __m128i* const G,
const __m128i* const B,
uint8_t* const dst) {
const __m128i r0 = _mm_packus_epi16(*R, *R);
const __m128i g0 = _mm_packus_epi16(*G, *G);
const __m128i b0 = _mm_packus_epi16(*B, *B);
const __m128i r1 = _mm_and_si128(r0, _mm_set1_epi8(0xf8));
const __m128i b1 = _mm_and_si128(_mm_srli_epi16(b0, 3), _mm_set1_epi8(0x1f));
const __m128i g1 = _mm_srli_epi16(_mm_and_si128(g0, _mm_set1_epi8(0xe0)), 5);
const __m128i g2 = _mm_slli_epi16(_mm_and_si128(g0, _mm_set1_epi8(0x1c)), 3);
const __m128i rg = _mm_or_si128(r1, g1);
const __m128i gb = _mm_or_si128(g2, b1);
#if !defined(WEBP_SWAP_16BIT_CSP)
const __m128i rgb565 = _mm_unpacklo_epi8(rg, gb);
#else
const __m128i rgb565 = _mm_unpacklo_epi8(gb, rg);
#endif
_mm_storeu_si128((__m128i*)dst, rgb565);
}
// Function used several times in PlanarTo24b.
// It samples the in buffer as follows: one every two unsigned char is stored
// at the beginning of the buffer, while the other half is stored at the end.
@ -185,6 +229,38 @@ void VP8YuvToBgra32(const uint8_t* y, const uint8_t* u, const uint8_t* v,
}
}
void VP8YuvToArgb32(const uint8_t* y, const uint8_t* u, const uint8_t* v,
uint8_t* dst) {
const __m128i kAlpha = _mm_set1_epi16(255);
int n;
for (n = 0; n < 32; n += 8, dst += 32) {
__m128i R, G, B;
YUV444ToRGB(y + n, u + n, v + n, &R, &G, &B);
PackAndStore4(&kAlpha, &R, &G, &B, dst);
}
}
void VP8YuvToRgba444432(const uint8_t* y, const uint8_t* u, const uint8_t* v,
uint8_t* dst) {
const __m128i kAlpha = _mm_set1_epi16(255);
int n;
for (n = 0; n < 32; n += 8, dst += 16) {
__m128i R, G, B;
YUV444ToRGB(y + n, u + n, v + n, &R, &G, &B);
PackAndStore4444(&R, &G, &B, &kAlpha, dst);
}
}
void VP8YuvToRgb56532(const uint8_t* y, const uint8_t* u, const uint8_t* v,
uint8_t* dst) {
int n;
for (n = 0; n < 32; n += 8, dst += 16) {
__m128i R, G, B;
YUV444ToRGB(y + n, u + n, v + n, &R, &G, &B);
PackAndStore565(&R, &G, &B, dst);
}
}
void VP8YuvToRgb32(const uint8_t* y, const uint8_t* u, const uint8_t* v,
uint8_t* dst) {
__m128i R0, R1, R2, R3, G0, G1, G2, G3, B0, B1, B2, B3;