static void YuvToBgrRowSSE2(const uint8_t* y,
const uint8_t* u, const uint8_t* v,
uint8_t* dst, int len) {
int n;
for (n = 0; n + 2 < len; ++n) { // we directly stomp the *dst memory
YuvToBgrSSE2(y[0], u[0], v[0], dst); // stomps 8 bytes
dst += 3;
++y;
u += (n & 1);
v += (n & 1);
}
VP8YuvToBgr(y[0], u[0], v[0], dst + 0);
if (len > 1) {
VP8YuvToBgr(y[1], u[n & 1], v[n & 1], dst + 3);
}
}
static void YuvToBgrRow_SSE41(const uint8_t* y,
const uint8_t* u, const uint8_t* v,
uint8_t* dst, int len) {
int n;
for (n = 0; n + 32 <= len; n += 32, dst += 32 * 3) {
__m128i R0, R1, R2, R3, G0, G1, G2, G3, B0, B1, B2, B3;
__m128i bgr0, bgr1, bgr2, bgr3, bgr4, bgr5;
YUV420ToRGB_SSE41(y + 0, u + 0, v + 0, &R0, &G0, &B0);
YUV420ToRGB_SSE41(y + 8, u + 4, v + 4, &R1, &G1, &B1);
YUV420ToRGB_SSE41(y + 16, u + 8, v + 8, &R2, &G2, &B2);
YUV420ToRGB_SSE41(y + 24, u + 12, v + 12, &R3, &G3, &B3);
// Cast to 8b and store as BBBBGGGGRRRR.
bgr0 = _mm_packus_epi16(B0, B1);
bgr1 = _mm_packus_epi16(B2, B3);
bgr2 = _mm_packus_epi16(G0, G1);
bgr3 = _mm_packus_epi16(G2, G3);
bgr4 = _mm_packus_epi16(R0, R1);
bgr5 = _mm_packus_epi16(R2, R3);
// Pack as BGRBGRBGRBGR.
PlanarTo24b_SSE41(&bgr0, &bgr1, &bgr2, &bgr3, &bgr4, &bgr5, dst);
y += 32;
u += 16;
v += 16;
}
for (; n < len; ++n) { // Finish off
VP8YuvToBgr(y[0], u[0], v[0], dst);
dst += 3;
y += 1;
u += (n & 1);
v += (n & 1);
}
}
