Skip to content

GitLab

Commit 9f21776b authored by Alex Converse's avatar Alex Converse
Browse files

Cleanup get_energy_distribution_fine 

- normalize casts
- remove unused array
- add consts
- declare variables late

BUG=aomedia:450

Change-Id: I2d3bc6ba2a9b05c779d3b451907abb3fc747d79a
parent ed29ea2b
Hide whitespace changes
Inline Side-by-side
Showing with 51 additions and 83 deletions
av1/encoder/rdopt.c
View file @ 9f21776b
......@@ -626,27 +626,24 @@ static int64_t av1_daala_dist(const uint8_t *src, int src_stride,
#endif // CONFIG_DAALA_DIST
static void get_energy_distribution_fine(const AV1_COMP *cpi, BLOCK_SIZE bsize,
uint8_t *src, int src_stride,
uint8_t *dst, int dst_stride,
const uint8_t *src, int src_stride,
const uint8_t *dst, int dst_stride,
double *hordist, double *verdist) {
int bw = block_size_wide[bsize];
int bh = block_size_high[bsize];
const int bw = block_size_wide[bsize];
const int bh = block_size_high[bsize];
unsigned int esq[16] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
unsigned int var[16];
double total = 0;
const int f_index = bsize - BLOCK_16X16;
if (f_index < 0) {
int i, j, index;
int w_shift = bw == 8 ? 1 : 2;
int h_shift = bh == 8 ? 1 : 2;
const int w_shift = bw == 8 ? 1 : 2;
const int h_shift = bh == 8 ? 1 : 2;
#if CONFIG_HIGHBITDEPTH
if (cpi->common.use_highbitdepth) {
uint16_t *src16 = CONVERT_TO_SHORTPTR(src);
uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
for (i = 0; i < bh; ++i)
for (j = 0; j < bw; ++j) {
index = (j >> w_shift) + ((i >> h_shift) << 2);
const uint16_t *src16 = CONVERT_TO_SHORTPTR(src);
const uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
for (int i = 0; i < bh; ++i)
for (int j = 0; j < bw; ++j) {
const int index = (j >> w_shift) + ((i >> h_shift) << 2);
esq[index] +=
(src16[j + i * src_stride] - dst16[j + i * dst_stride]) *
(src16[j + i * src_stride] - dst16[j + i * dst_stride]);
......@@ -654,9 +651,9 @@ static void get_energy_distribution_fine(const AV1_COMP *cpi, BLOCK_SIZE bsize,
} else {
#endif // CONFIG_HIGHBITDEPTH
for (i = 0; i < bh; ++i)
for (j = 0; j < bw; ++j) {
index = (j >> w_shift) + ((i >> h_shift) << 2);
for (int i = 0; i < bh; ++i)
for (int j = 0; j < bw; ++j) {
const int index = (j >> w_shift) + ((i >> h_shift) << 2);
esq[index] += (src[j + i * src_stride] - dst[j + i * dst_stride]) *
(src[j + i * src_stride] - dst[j + i * dst_stride]);
}
......@@ -664,90 +661,61 @@ static void get_energy_distribution_fine(const AV1_COMP *cpi, BLOCK_SIZE bsize,
}
#endif // CONFIG_HIGHBITDEPTH
} else {
var[0] = cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[0]);
var[1] = cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4,
dst_stride, &esq[1]);
var[2] = cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2,
dst_stride, &esq[2]);
var[3] = cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride,
dst + 3 * bw / 4, dst_stride, &esq[3]);
cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[0]);
cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4, dst_stride,
&esq[1]);
cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2, dst_stride,
&esq[2]);
cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride, dst + 3 * bw / 4,
dst_stride, &esq[3]);
src += bh / 4 * src_stride;
dst += bh / 4 * dst_stride;
var[4] = cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[4]);
var[5] = cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4,
dst_stride, &esq[5]);
var[6] = cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2,
dst_stride, &esq[6]);
var[7] = cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride,
dst + 3 * bw / 4, dst_stride, &esq[7]);
cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[4]);
cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4, dst_stride,
&esq[5]);
cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2, dst_stride,
&esq[6]);
cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride, dst + 3 * bw / 4,
dst_stride, &esq[7]);
src += bh / 4 * src_stride;
dst += bh / 4 * dst_stride;
var[8] = cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[8]);
var[9] = cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4,
dst_stride, &esq[9]);
var[10] = cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2,
dst_stride, &esq[10]);
var[11] = cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride,
dst + 3 * bw / 4, dst_stride, &esq[11]);
cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[8]);
cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4, dst_stride,
&esq[9]);
cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2, dst_stride,
&esq[10]);
cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride, dst + 3 * bw / 4,
dst_stride, &esq[11]);
src += bh / 4 * src_stride;
dst += bh / 4 * dst_stride;
var[12] =
cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[12]);
var[13] = cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4,
dst_stride, &esq[13]);
var[14] = cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2,
dst_stride, &esq[14]);
var[15] = cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride,
dst + 3 * bw / 4, dst_stride, &esq[15]);
cpi->fn_ptr[f_index].vf(src, src_stride, dst, dst_stride, &esq[12]);
cpi->fn_ptr[f_index].vf(src + bw / 4, src_stride, dst + bw / 4, dst_stride,
&esq[13]);
cpi->fn_ptr[f_index].vf(src + bw / 2, src_stride, dst + bw / 2, dst_stride,
&esq[14]);
cpi->fn_ptr[f_index].vf(src + 3 * bw / 4, src_stride, dst + 3 * bw / 4,
dst_stride, &esq[15]);
}
total = esq[0] + esq[1] + esq[2] + esq[3] + esq[4] + esq[5] + esq[6] +
esq[7] + esq[8] + esq[9] + esq[10] + esq[11] + esq[12] + esq[13] +
esq[14] + esq[15];
double total = (double)esq[0] + esq[1] + esq[2] + esq[3] + esq[4] + esq[5] +
esq[6] + esq[7] + esq[8] + esq[9] + esq[10] + esq[11] +
esq[12] + esq[13] + esq[14] + esq[15];
if (total > 0) {
const double e_recip = 1.0 / total;
hordist[0] =
((double)esq[0] + (double)esq[4] + (double)esq[8] + (double)esq[12]) *
e_recip;
hordist[1] =
((double)esq[1] + (double)esq[5] + (double)esq[9] + (double)esq[13]) *
e_recip;
hordist[2] =
((double)esq[2] + (double)esq[6] + (double)esq[10] + (double)esq[14]) *
e_recip;
verdist[0] =
((double)esq[0] + (double)esq[1] + (double)esq[2] + (double)esq[3]) *
e_recip;
verdist[1] =
((double)esq[4] + (double)esq[5] + (double)esq[6] + (double)esq[7]) *
e_recip;
verdist[2] =
((double)esq[8] + (double)esq[9] + (double)esq[10] + (double)esq[11]) *
e_recip;
hordist[0] = ((double)esq[0] + esq[4] + esq[8] + esq[12]) * e_recip;
hordist[1] = ((double)esq[1] + esq[5] + esq[9] + esq[13]) * e_recip;
hordist[2] = ((double)esq[2] + esq[6] + esq[10] + esq[14]) * e_recip;
verdist[0] = ((double)esq[0] + esq[1] + esq[2] + esq[3]) * e_recip;
verdist[1] = ((double)esq[4] + esq[5] + esq[6] + esq[7]) * e_recip;
verdist[2] = ((double)esq[8] + esq[9] + esq[10] + esq[11]) * e_recip;
} else {
hordist[0] = verdist[0] = 0.25;
hordist[1] = verdist[1] = 0.25;
hordist[2] = verdist[2] = 0.25;
}
(void)var[0];
(void)var[1];
(void)var[2];
(void)var[3];
(void)var[4];
(void)var[5];
(void)var[6];
(void)var[7];
(void)var[8];
(void)var[9];
(void)var[10];
(void)var[11];
(void)var[12];
(void)var[13];
(void)var[14];
(void)var[15];
}
static int adst_vs_flipadst(const AV1_COMP *cpi, BLOCK_SIZE bsize, uint8_t *src,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment