Add experiment CONFIG_CDEF_SINGLEPASS: Make CDEF single pass
Low latency, cpu-used=0:
PSNR | PSNR Cb | PSNR Cr | PSNR HVS | SSIM | MS SSIM | CIEDE 2000
-0.3162 | -0.6719 | -0.6535 | 0.0089 | -0.3890 | -0.1515 | -0.6682
High latency, cpu-used=0:
PSNR | PSNR Cb | PSNR Cr | PSNR HVS | SSIM | MS SSIM | CIEDE 2000
-0.0293 | -0.3556 | -0.5505 | 0.0684 | -0.0862 | 0.0513 | -0.2765
Low latency, cpu-used=4:
PSNR | PSNR Cb | PSNR Cr | PSNR HVS | SSIM | MS SSIM | CIEDE 2000
-0.2248 | -0.7764 | -0.6630 | -0.2109 | -0.3240 | -0.2532 | -0.6980
High latency, cpu-used=4:
PSNR | PSNR Cb | PSNR Cr | PSNR HVS | SSIM | MS SSIM | CIEDE 2000
-0.1118 | -0.5841 | -0.7406 | -0.0463 | -0.2442 | -0.1064 | -0.4187
Change-Id: I9ca8399c8f45489541a66f535fb3d771eb1d59ab
diff --git a/av1/encoder/bitstream.c b/av1/encoder/bitstream.c
index 4a6fbd8..c86ef3a 100644
--- a/av1/encoder/bitstream.c
+++ b/av1/encoder/bitstream.c
@@ -3454,8 +3454,13 @@
#if CONFIG_CDEF
static void encode_cdef(const AV1_COMMON *cm, struct aom_write_bit_buffer *wb) {
int i;
+#if CONFIG_CDEF_SINGLEPASS
+ aom_wb_write_literal(wb, cm->cdef_pri_damping - 3, 2);
+ assert(cm->cdef_pri_damping == cm->cdef_sec_damping);
+#else
aom_wb_write_literal(wb, cm->cdef_pri_damping - 5, 1);
aom_wb_write_literal(wb, cm->cdef_sec_damping - 3, 2);
+#endif
aom_wb_write_literal(wb, cm->cdef_bits, 2);
for (i = 0; i < cm->nb_cdef_strengths; i++) {
aom_wb_write_literal(wb, cm->cdef_strengths[i], CDEF_STRENGTH_BITS);
diff --git a/av1/encoder/encoder.c b/av1/encoder/encoder.c
index da2370b..f3f7799 100644
--- a/av1/encoder/encoder.c
+++ b/av1/encoder/encoder.c
@@ -4206,7 +4206,7 @@
cm->cdef_strengths[0] = 0;
cm->nb_cdef_strengths = 1;
} else {
- // Find cm->dering_level, cm->clpf_strength_u and cm->clpf_strength_v
+ // Find CDEF parameters
av1_cdef_search(cm->frame_to_show, cpi->source, cm, xd,
cpi->oxcf.speed > 0);
diff --git a/av1/encoder/pickcdef.c b/av1/encoder/pickcdef.c
index 443e9e5..accc97e 100644
--- a/av1/encoder/pickcdef.c
+++ b/av1/encoder/pickcdef.c
@@ -68,11 +68,16 @@
uint64_t (**mse)[TOTAL_STRENGTHS], int sb_count,
int fast) {
uint64_t tot_mse[TOTAL_STRENGTHS][TOTAL_STRENGTHS];
+#if !CONFIG_CDEF_SINGLEPASS
const int total_strengths = fast ? REDUCED_TOTAL_STRENGTHS : TOTAL_STRENGTHS;
+#endif
int i, j;
uint64_t best_tot_mse = (uint64_t)1 << 63;
int best_id0 = 0;
int best_id1 = 0;
+#if CONFIG_CDEF_SINGLEPASS
+ const int total_strengths = fast ? REDUCED_TOTAL_STRENGTHS : TOTAL_STRENGTHS;
+#endif
memset(tot_mse, 0, sizeof(tot_mse));
for (i = 0; i < sb_count; i++) {
int gi;
@@ -305,7 +310,11 @@
int *sb_index = aom_malloc(nvfb * nhfb * sizeof(*sb_index));
int *selected_strength = aom_malloc(nvfb * nhfb * sizeof(*sb_index));
uint64_t(*mse[2])[TOTAL_STRENGTHS];
+#if CONFIG_CDEF_SINGLEPASS
+ int pri_damping = 3 + (cm->base_qindex >> 6);
+#else
int pri_damping = 6;
+#endif
int sec_damping = 3 + (cm->base_qindex >> 6);
int i;
int nb_strengths;
@@ -414,6 +423,17 @@
int xsize = (nhb << mi_wide_l2[pli]) +
CDEF_HBORDER * (fbc != nhfb - 1) + xoff;
sec_strength = gi % CDEF_SEC_STRENGTHS;
+#if CONFIG_CDEF_SINGLEPASS
+ copy_sb16_16(&in[(-yoff * CDEF_BSTRIDE - xoff)], CDEF_BSTRIDE,
+ src[pli],
+ (fbr * MI_SIZE_64X64 << mi_high_l2[pli]) - yoff,
+ (fbc * MI_SIZE_64X64 << mi_wide_l2[pli]) - xoff,
+ stride[pli], ysize, xsize);
+ cdef_filter_fb(NULL, tmp_dst, CDEF_BSTRIDE, in, xdec[pli], ydec[pli],
+ dir, &dirinit, var, pli, dlist, cdef_count, threshold,
+ sec_strength + (sec_strength == 3), pri_damping,
+ sec_damping, coeff_shift);
+#else
if (sec_strength == 0)
copy_sb16_16(&in[(-yoff * CDEF_BSTRIDE - xoff)], CDEF_BSTRIDE,
src[pli],
@@ -425,6 +445,7 @@
pli, dlist, cdef_count, threshold,
sec_strength + (sec_strength == 3), sec_damping,
pri_damping, coeff_shift, sec_strength != 0, 1);
+#endif
curr_mse = compute_cdef_dist(
ref_coeff[pli] +
(fbr * MI_SIZE_64X64 << mi_high_l2[pli]) * stride[pli] +
