layout(rg16ui) uniform restrict readonly uimage2D dsyms_tex;
layout(r8) uniform restrict writeonly image2D out_tex;
layout(r16i) uniform restrict writeonly iimage2D coeff_tex;
+uniform int num_blocks;
const uint prob_bits = 12;
const uint prob_scale = 1 << prob_bits;
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 local_y = (gl_LocalInvocationID.x / 8) % 8;
const uint local_z = gl_LocalInvocationID.x / 64;
- const uint num_blocks = 720 / 16; // FIXME: make a uniform
const uint slice_num = local_z;
const uint thread_num = local_y * 8 + local_x;
//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.
float q = (coeff_num == 0) ? 1.0 : (quant_matrix[coeff_num] * quant_scalefac / 128.0 / sqrt(2.0)); // FIXME: fold
q *= (1.0 / 255.0);
//int w = (coeff_num == 0) ? 32 : int(quant_matrix[coeff_num]);
- int last_k = 0;
+ int last_k = 128;
pick_timer(start, local_timing[0]);
for (uint block_idx = 40; block_idx --> 0; ) {
- uint block_x = block_idx % 20;
- uint block_y = block_idx / 20;
- if (block_x == 19) last_k = 0;
-
pick_timer(start, local_timing[1]);
// rANS decode one coefficient across eight blocks (so 64x8 coefficients).
pick_timer(start, local_timing[6]);
- uint y = block_row * 16 + block_y * 8;
- uint x = block_x * 64 + local_y * 8 + local_x;
+ uint global_block_idx = (block_row * 40 + block_idx) * 8 + local_y;
+ uint block_x = global_block_idx % 160;
+ uint block_y = global_block_idx / 160;
+
+ uint y = block_y * 8;
+ uint x = block_x * 8 + local_x;
for (uint yl = 0; yl < 8; ++yl) {
imageStore(out_tex, ivec2(x, yl + y), vec4(temp[row_offset + yl * 8], 0.0, 0.0, 1.0));
}