avcodec/rpzaenc: replace float-point calculations with integer ones

This commit is contained in:
Paul B Mahol 2023-09-07 11:00:38 +02:00
parent b97ac6b3df
commit 03ed48aeca
1 changed files with 16 additions and 15 deletions

View File

@ -30,6 +30,7 @@
#include "avcodec.h" #include "avcodec.h"
#include "codec_internal.h" #include "codec_internal.h"
#include "encode.h" #include "encode.h"
#include "mathops.h"
#include "put_bits.h" #include "put_bits.h"
typedef struct RpzaContext { typedef struct RpzaContext {
@ -65,7 +66,7 @@ typedef struct rgb {
#define SQR(x) ((x) * (x)) #define SQR(x) ((x) * (x))
/* 15 bit components */ /* 15 bit components */
#define GET_CHAN(color, chan) (((color) >> ((chan) * 5) & 0x1F) * 8) #define GET_CHAN(color, chan) (((color) >> ((chan) * 5) & 0x1F))
#define R(color) GET_CHAN(color, RED) #define R(color) GET_CHAN(color, RED)
#define G(color) GET_CHAN(color, GREEN) #define G(color) GET_CHAN(color, GREEN)
#define B(color) GET_CHAN(color, BLUE) #define B(color) GET_CHAN(color, BLUE)
@ -141,9 +142,9 @@ static uint16_t rgb24_to_rgb555(const uint8_t *rgb24)
uint16_t rgb555 = 0; uint16_t rgb555 = 0;
uint32_t r, g, b; uint32_t r, g, b;
r = rgb24[0] >> 3; r = rgb24[0];
g = rgb24[1] >> 3; g = rgb24[1];
b = rgb24[2] >> 3; b = rgb24[2];
rgb555 |= (r << 10); rgb555 |= (r << 10);
rgb555 |= (g << 5); rgb555 |= (g << 5);
@ -185,7 +186,7 @@ static int max_component_diff(const uint16_t *colorA, const uint16_t *colorB)
if (diff > max) { if (diff > max) {
max = diff; max = diff;
} }
return max * 8; return max;
} }
/* /*
@ -266,9 +267,9 @@ static int compare_blocks(const uint16_t *block1, const uint16_t *block2,
*/ */
static int leastsquares(const uint16_t *block_ptr, const BlockInfo *bi, static int leastsquares(const uint16_t *block_ptr, const BlockInfo *bi,
channel_offset xchannel, channel_offset ychannel, channel_offset xchannel, channel_offset ychannel,
double *slope, double *y_intercept, double *correlation_coef) int *slope, int *y_intercept, int *correlation_coef)
{ {
double sumx = 0, sumy = 0, sumx2 = 0, sumy2 = 0, sumxy = 0, int sumx = 0, sumy = 0, sumx2 = 0, sumy2 = 0, sumxy = 0,
sumx_sq = 0, sumy_sq = 0, tmp, tmp2; sumx_sq = 0, sumy_sq = 0, tmp, tmp2;
int i, j, count; int i, j, count;
uint8_t x, y; uint8_t x, y;
@ -305,10 +306,10 @@ static int leastsquares(const uint16_t *block_ptr, const BlockInfo *bi,
tmp2 = count * sumy2 - sumy_sq; tmp2 = count * sumy2 - sumy_sq;
if (tmp2 == 0) { if (tmp2 == 0) {
*correlation_coef = 0.0; *correlation_coef = 0;
} else { } else {
*correlation_coef = (count * sumxy - sumx * sumy) / *correlation_coef = (count * sumxy - sumx * sumy) /
sqrt(tmp * tmp2); ff_sqrt((unsigned)tmp * tmp2);
} }
return 0; // success return 0; // success
@ -332,18 +333,18 @@ static int calc_lsq_max_fit_error(const uint16_t *block_ptr, const BlockInfo *bi
y = GET_CHAN(block_ptr[j], ychannel); y = GET_CHAN(block_ptr[j], ychannel);
/* calculate x_inc as the 4-color index (0..3) */ /* calculate x_inc as the 4-color index (0..3) */
x_inc = floor( (x - min) * 3.0 / (max - min) + 0.5); x_inc = (x - min) * 3 / (max - min) + 1;
x_inc = FFMAX(FFMIN(3, x_inc), 0); x_inc = FFMAX(FFMIN(3, x_inc), 0);
/* calculate lin_y corresponding to x_inc */ /* calculate lin_y corresponding to x_inc */
lin_y = (int)(tmp_min + (tmp_max - tmp_min) * x_inc / 3.0 + 0.5); lin_y = tmp_min + (tmp_max - tmp_min) * x_inc / 3 + 1;
err = FFABS(lin_y - y); err = FFABS(lin_y - y);
if (err > max_err) if (err > max_err)
max_err = err; max_err = err;
/* calculate lin_x corresponding to x_inc */ /* calculate lin_x corresponding to x_inc */
lin_x = (int)(min + (max - min) * x_inc / 3.0 + 0.5); lin_x = min + (max - min) * x_inc / 3 + 1;
err = FFABS(lin_x - x); err = FFABS(lin_x - x);
if (err > max_err) if (err > max_err)
@ -577,7 +578,7 @@ static void rpza_encode_stream(RpzaContext *s, const AVFrame *pict)
uint8_t avg_color[3]; uint8_t avg_color[3];
int pixel_count; int pixel_count;
uint8_t min_color[3], max_color[3]; uint8_t min_color[3], max_color[3];
double slope, y_intercept, correlation_coef; int slope, y_intercept, correlation_coef;
const uint16_t *src_pixels = (const uint16_t *)pict->data[0]; const uint16_t *src_pixels = (const uint16_t *)pict->data[0];
uint16_t *prev_pixels = (uint16_t *)s->prev_frame->data[0]; uint16_t *prev_pixels = (uint16_t *)s->prev_frame->data[0];
@ -730,8 +731,8 @@ post_skip :
min_color[i] = GET_CHAN(src_pixels[block_offset], i); min_color[i] = GET_CHAN(src_pixels[block_offset], i);
max_color[i] = GET_CHAN(src_pixels[block_offset], i); max_color[i] = GET_CHAN(src_pixels[block_offset], i);
} else { } else {
tmp_min = (int)(0.5 + min * slope + y_intercept); tmp_min = 1 + min * slope + y_intercept;
tmp_max = (int)(0.5 + max * slope + y_intercept); tmp_max = 1 + max * slope + y_intercept;
av_assert0(tmp_min <= tmp_max); av_assert0(tmp_min <= tmp_max);
// clamp min and max color values // clamp min and max color values