58, 59, 52, 45, 38, 31, 39, 46,
53, 60, 61, 54, 47, 55, 62, 63
};
+const uint stream_mapping[64] = {
+ 0, 0, 1, 1, 2, 2, 3, 3,
+ 0, 0, 1, 2, 2, 2, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 2, 2, 2, 2, 3, 3, 3,
+ 2, 2, 2, 2, 3, 3, 3, 3,
+ 2, 2, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+};
layout(std430, binding = 9) buffer layoutName
{
//const uint coeff_num = ff_zigzag_direct[thread_num];
const uint coeff_num = thread_num;
const uint stream_num = coeff_num * num_blocks + block_row;
- const uint model_num = min((coeff_num % 8) + (coeff_num / 8), 7);
+ const uint model_num = stream_mapping[coeff_num];
const uint sign_bias = sign_bias_per_model[model_num];
// Initialize rANS decoder.
const unsigned prob_bits = 12;
const unsigned prob_scale = 1 << prob_bits;
const unsigned NUM_SYMS = 256;
-const unsigned NUM_TABLES = 16;
+const unsigned NUM_TABLES = 8;
struct RansDecSymbol {
unsigned sym_start;
// of coefficients to rANS probability distributions. This is randomized,
// so you might want to run it a few times.
#define FIND_OPTIMAL_STREAM_ASSIGNMENT 0
-#define NUM_CLUSTERS 8
+#define NUM_CLUSTERS 4
static constexpr uint32_t prob_bits = 12;
static constexpr uint32_t prob_scale = 1 << prob_bits;
float kl_dist[64][64];
#endif
+const int luma_mapping[64] = {
+ 0, 0, 1, 1, 2, 2, 3, 3,
+ 0, 0, 1, 2, 2, 2, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 2, 2, 2, 2, 3, 3, 3,
+ 2, 2, 2, 2, 3, 3, 3, 3,
+ 2, 2, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+};
+const int chroma_mapping[64] = {
+ 0, 1, 1, 2, 2, 2, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 2, 2, 2, 2, 3, 3, 3, 3,
+ 2, 2, 2, 3, 3, 3, 3, 3,
+ 2, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+};
+
int pick_stats_for(int x, int y, bool is_chroma)
{
#if FIND_OPTIMAL_STREAM_ASSIGNMENT
return y * 8 + x + is_chroma * 64;
#else
- return std::min<int>(x + y, 7) + is_chroma * 8;
+ if (is_chroma) {
+ return chroma_mapping[y * 8 + x] + 4;
+ } else {
+ return luma_mapping[y * 8 + x];
+ }
#endif
}
int cum2sym[prob_scale];
RansDecSymbol dsyms[NUM_SYMS];
};
-RansDecodeTable decode_tables[16];
+#define NUM_TABLES 8
+RansDecodeTable decode_tables[NUM_TABLES];
static const unsigned char std_luminance_quant_tbl[64] = {
#if 0
};
-int pick_stats_for(int y, int x)
+const int luma_mapping[64] = {
+ 0, 0, 1, 1, 2, 2, 3, 3,
+ 0, 0, 1, 2, 2, 2, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 1, 2, 2, 2, 2, 3, 3, 3,
+ 2, 2, 2, 2, 3, 3, 3, 3,
+ 2, 2, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+};
+const int chroma_mapping[64] = {
+ 0, 1, 1, 2, 2, 2, 3, 3,
+ 1, 1, 2, 2, 2, 3, 3, 3,
+ 2, 2, 2, 2, 3, 3, 3, 3,
+ 2, 2, 2, 3, 3, 3, 3, 3,
+ 2, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+};
+
+int pick_stats_for(int x, int y, bool is_chroma)
{
- if (x + y >= 7) return 7;
- return x + y;
+ if (is_chroma) {
+ return chroma_mapping[y * 8 + x] + 4;
+ } else {
+ return luma_mapping[y * 8 + x];
+ }
}
uint32_t read_varint(FILE *fp)
exit(1);
}
- uint32_t sign_bias[16];
- for (unsigned table = 0; table < 16; ++table) {
+ uint32_t sign_bias[NUM_TABLES];
+ for (unsigned table = 0; table < NUM_TABLES; ++table) {
uint32_t cum_freq = 0;
for (unsigned sym = 0; sym < NUM_SYMS; ++sym) {
uint32_t freq = read_varint(fp);
// loop over all coefficients
for (unsigned y = 0; y < 8; ++y) {
for (unsigned x = 0; x < 8; ++x) {
- unsigned tbl = pick_stats_for(x, y);
+ unsigned tbl = pick_stats_for(x, y, false);
RansState rans = 0;