static void HFilter16iSSE2(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
uint8_t* b;
__m128i mask;
__m128i t1, t2, p1, p0, q0, q1;
for (k = 3; k > 0; --k) {
b = p;
Load16x4(b, b + 8 * stride, stride, &t2, &t1, &p1, &p0); // p3, p2, p1, p0
MAX_DIFF1(t2, t1, p1, p0, mask);
b += 4; // beginning of q0
Load16x4(b, b + 8 * stride, stride, &q0, &q1, &t1, &t2); // q0, q1, q2, q3
MAX_DIFF2(t2, t1, q1, q0, mask);
COMPLEX_FL_MASK(p1, p0, q0, q1, thresh, ithresh, mask);
DoFilter4(&p1, &p0, &q0, &q1, &mask, hev_thresh);
b -= 2; // beginning of p1
Store16x4(b, b + 8 * stride, stride, &p1, &p0, &q0, &q1);
p += 4;
}
}
static void HFilter16i(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
__m128i p3, p2, p1, p0; // loop invariants
Load16x4(p, p + 8 * stride, stride, &p3, &p2, &p1, &p0); // prologue
for (k = 3; k > 0; --k) {
__m128i mask, tmp1, tmp2;
uint8_t* const b = p + 2; // beginning of p1
p += 4; // beginning of q0 (and next span)
MAX_DIFF1(p3, p2, p1, p0, mask); // compute partial mask
Load16x4(p, p + 8 * stride, stride, &p3, &p2, &tmp1, &tmp2);
MAX_DIFF2(p3, p2, tmp1, tmp2, mask);
ComplexMask(&p1, &p0, &p3, &p2, thresh, ithresh, &mask);
DoFilter4(&p1, &p0, &p3, &p2, &mask, hev_thresh);
Store16x4(&p1, &p0, &p3, &p2, b, b + 8 * stride, stride);
// rotate samples
p1 = tmp1;
p0 = tmp2;
}
}
static void HFilter16SSE2(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
__m128i mask;
__m128i p3, p2, p1, p0, q0, q1, q2, q3;
uint8_t* const b = p - 4;
Load16x4(b, b + 8 * stride, stride, &p3, &p2, &p1, &p0); // p3, p2, p1, p0
MAX_DIFF1(p3, p2, p1, p0, mask);
Load16x4(p, p + 8 * stride, stride, &q0, &q1, &q2, &q3); // q0, q1, q2, q3
MAX_DIFF2(q3, q2, q1, q0, mask);
COMPLEX_FL_MASK(p1, p0, q0, q1, thresh, ithresh, mask);
DoFilter6(&p2, &p1, &p0, &q0, &q1, &q2, &mask, hev_thresh);
Store16x4(b, b + 8 * stride, stride, &p3, &p2, &p1, &p0);
Store16x4(p, p + 8 * stride, stride, &q0, &q1, &q2, &q3);
}
static void SimpleHFilter16SSE2(uint8_t* p, int stride, int thresh) {
__m128i p1, p0, q0, q1;
p -= 2; // beginning of p1
Load16x4(p, p + 8 * stride, stride, &p1, &p0, &q0, &q1);
DoFilter2(&p1, &p0, &q0, &q1, thresh);
Store16x4(p, p + 8 * stride, stride, &p1, &p0, &q0, &q1);
}
static void HFilter8iSSE2(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
__m128i mask;
__m128i t1, t2, p1, p0, q0, q1;
Load16x4(u, v, stride, &t2, &t1, &p1, &p0); // p3, p2, p1, p0
MAX_DIFF1(t2, t1, p1, p0, mask);
u += 4; // beginning of q0
v += 4;
Load16x4(u, v, stride, &q0, &q1, &t1, &t2); // q0, q1, q2, q3
MAX_DIFF2(t2, t1, q1, q0, mask);
COMPLEX_FL_MASK(p1, p0, q0, q1, thresh, ithresh, mask);
DoFilter4(&p1, &p0, &q0, &q1, &mask, hev_thresh);
u -= 2; // beginning of p1
v -= 2;
Store16x4(u, v, stride, &p1, &p0, &q0, &q1);
}
static void HFilter8SSE2(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
__m128i mask;
__m128i p3, p2, p1, p0, q0, q1, q2, q3;
uint8_t* const tu = u - 4;
uint8_t* const tv = v - 4;
Load16x4(tu, tv, stride, &p3, &p2, &p1, &p0); // p3, p2, p1, p0
MAX_DIFF1(p3, p2, p1, p0, mask);
Load16x4(u, v, stride, &q0, &q1, &q2, &q3); // q0, q1, q2, q3
MAX_DIFF2(q3, q2, q1, q0, mask);
COMPLEX_FL_MASK(p1, p0, q0, q1, thresh, ithresh, mask);
DoFilter6(&p2, &p1, &p0, &q0, &q1, &q2, &mask, hev_thresh);
Store16x4(tu, tv, stride, &p3, &p2, &p1, &p0);
Store16x4(u, v, stride, &q0, &q1, &q2, &q3);
}
