[FFmpeg-devel] [PATCH 2/2] avfilter/vf_v360: stop using floats in interpolation
Paul B Mahol
onemda at gmail.com
Tue Sep 3 15:11:29 EEST 2019
Signed-off-by: Paul B Mahol <onemda at gmail.com>
---
libavfilter/vf_v360.c | 40 ++++++++++++++++++++--------------------
1 file changed, 20 insertions(+), 20 deletions(-)
diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c
index 2fee0016b9..dd41bb293c 100644
--- a/libavfilter/vf_v360.c
+++ b/libavfilter/vf_v360.c
@@ -118,7 +118,7 @@ typedef struct V360Context {
int nb_planes;
uint16_t *u[4], *v[4];
- float *ker[4];
+ int16_t *ker[4];
int (*remap_slice)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs);
} V360Context;
@@ -330,12 +330,12 @@ static int remap##ws##_##bits##bit_slice(AVFilterContext *ctx, void *arg, int jo
uint##bits##_t *d = dst + y * out_linesize; \
const uint16_t *u = s->u[plane] + y * width * ws * ws; \
const uint16_t *v = s->v[plane] + y * width * ws * ws; \
- const float *ker = s->ker[plane] + y * width * ws * ws; \
+ const int16_t *ker = s->ker[plane] + y * width * ws * ws; \
for (x = 0; x < width; x++) { \
const uint16_t *uu = u + x * ws * ws; \
const uint16_t *vv = v + x * ws * ws; \
- const float *kker = ker + x * ws * ws; \
- float tmp = 0.f; \
+ const int16_t *kker = ker + x * ws * ws; \
+ int tmp = 0; \
\
for (i = 0; i < ws; i++) { \
for (j = 0; j < ws; j++) { \
@@ -343,7 +343,7 @@ static int remap##ws##_##bits##bit_slice(AVFilterContext *ctx, void *arg, int jo
} \
} \
\
- *d++ = av_clip_uint##bits(roundf(tmp)); \
+ *d++ = av_clip_uint##bits(tmp >> (15 - ws)); \
} \
} \
} \
@@ -367,7 +367,7 @@ DEFINE_REMAP(4, 16, 2)
* @param ker ker remap data
*/
static void nearest_kernel(float du, float dv, const XYRemap *r_tmp,
- uint16_t *u, uint16_t *v, float *ker)
+ uint16_t *u, uint16_t *v, int16_t *ker)
{
const int i = roundf(dv) + 1;
const int j = roundf(du) + 1;
@@ -387,7 +387,7 @@ static void nearest_kernel(float du, float dv, const XYRemap *r_tmp,
* @param ker ker remap data
*/
static void bilinear_kernel(float du, float dv, const XYRemap *r_tmp,
- uint16_t *u, uint16_t *v, float *ker)
+ uint16_t *u, uint16_t *v, int16_t *ker)
{
int i, j;
@@ -398,10 +398,10 @@ static void bilinear_kernel(float du, float dv, const XYRemap *r_tmp,
}
}
- ker[0] = (1.f - du) * (1.f - dv);
- ker[1] = du * (1.f - dv);
- ker[2] = (1.f - du) * dv;
- ker[3] = du * dv;
+ ker[0] = (1.f - du) * (1.f - dv) * 8192;
+ ker[1] = du * (1.f - dv) * 8192;
+ ker[2] = (1.f - du) * dv * 8192;
+ ker[3] = du * dv * 8192;
}
/**
@@ -432,7 +432,7 @@ static inline void calculate_bicubic_coeffs(float t, float *coeffs)
* @param ker ker remap data
*/
static void bicubic_kernel(float du, float dv, const XYRemap *r_tmp,
- uint16_t *u, uint16_t *v, float *ker)
+ uint16_t *u, uint16_t *v, int16_t *ker)
{
int i, j;
float du_coeffs[4];
@@ -445,7 +445,7 @@ static void bicubic_kernel(float du, float dv, const XYRemap *r_tmp,
for (j = 0; j < 4; j++) {
u[i * 4 + j] = r_tmp->u[i][j];
v[i * 4 + j] = r_tmp->v[i][j];
- ker[i * 4 + j] = du_coeffs[j] * dv_coeffs[i];
+ ker[i * 4 + j] = du_coeffs[j] * dv_coeffs[i] * 2048;
}
}
}
@@ -487,7 +487,7 @@ static inline void calculate_lanczos_coeffs(float t, float *coeffs)
* @param ker ker remap data
*/
static void lanczos_kernel(float du, float dv, const XYRemap *r_tmp,
- uint16_t *u, uint16_t *v, float *ker)
+ uint16_t *u, uint16_t *v, int16_t *ker)
{
int i, j;
float du_coeffs[4];
@@ -500,7 +500,7 @@ static void lanczos_kernel(float du, float dv, const XYRemap *r_tmp,
for (j = 0; j < 4; j++) {
u[i * 4 + j] = r_tmp->u[i][j];
v[i * 4 + j] = r_tmp->v[i][j];
- ker[i * 4 + j] = du_coeffs[j] * dv_coeffs[i];
+ ker[i * 4 + j] = du_coeffs[j] * dv_coeffs[i] * 2048;
}
}
}
@@ -1985,7 +1985,7 @@ static int config_output(AVFilterLink *outlink)
int i, int j, int width, int height,
float *vec);
void (*calculate_kernel)(float du, float dv, const XYRemap *r_tmp,
- uint16_t *u, uint16_t *v, float *ker);
+ uint16_t *u, uint16_t *v, int16_t *ker);
float rot_mat[3][3];
switch (s->interp) {
@@ -2001,21 +2001,21 @@ static int config_output(AVFilterLink *outlink)
s->remap_slice = depth <= 8 ? remap2_8bit_slice : remap2_16bit_slice;
elements = 2 * 2;
sizeof_uv = sizeof(uint16_t) * elements;
- sizeof_ker = sizeof(float) * elements;
+ sizeof_ker = sizeof(uint16_t) * elements;
break;
case BICUBIC:
calculate_kernel = bicubic_kernel;
s->remap_slice = depth <= 8 ? remap4_8bit_slice : remap4_16bit_slice;
elements = 4 * 4;
sizeof_uv = sizeof(uint16_t) * elements;
- sizeof_ker = sizeof(float) * elements;
+ sizeof_ker = sizeof(uint16_t) * elements;
break;
case LANCZOS:
calculate_kernel = lanczos_kernel;
s->remap_slice = depth <= 8 ? remap4_8bit_slice : remap4_16bit_slice;
elements = 4 * 4;
sizeof_uv = sizeof(uint16_t) * elements;
- sizeof_ker = sizeof(float) * elements;
+ sizeof_ker = sizeof(uint16_t) * elements;
break;
}
@@ -2182,7 +2182,7 @@ static int config_output(AVFilterLink *outlink)
for (j = 0; j < height; j++) {
uint16_t *u = s->u[p] + (j * width + i) * elements;
uint16_t *v = s->v[p] + (j * width + i) * elements;
- float *ker = s->ker[p] + (j * width + i) * elements;
+ int16_t *ker = s->ker[p] + (j * width + i) * elements;
out_transform(s, i, j, width, height, vec);
rotate(rot_mat, vec);
--
2.17.1
More information about the ffmpeg-devel
mailing list