Prepare for more flexible slices.

[narabu] / decoder.shader

diff --git a/decoder.shader b/decoder.shader
index 77ae2869f08c54cf3a4fafbc6d101e6dba62a408..09027915633af93b69a883039836bdecc16f8df2 100644 (file)
--- a/decoder.shader
+++ b/decoder.shader
@@ -9,6 +9,7 @@ layout(local_size_x = 64*PARALLEL_SLICES) in;
 layout(r8ui) uniform restrict readonly uimage2D cum2sym_tex;
 layout(rg16ui) uniform restrict readonly uimage2D dsyms_tex;
 layout(r8) uniform restrict writeonly image2D out_tex;
+layout(r16i) uniform restrict writeonly iimage2D coeff_tex;
 
 const uint prob_bits = 12;
 const uint prob_scale = 1 << prob_bits;
@@ -39,6 +40,16 @@ const uint ff_zigzag_direct[64] = {
     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
 {
@@ -203,7 +214,7 @@ void main()
        //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.
@@ -251,6 +262,12 @@ void main()
                                last_k = k;
                        }
 
+#if 0
+                       uint y = block_row * 16 + block_y * 8 + local_y;
+                       uint x = block_x * 64 + subblock_idx * 8 + local_x;
+                       imageStore(coeff_tex, ivec2(x, y), ivec4(k, 0,0,0));
+#endif
+
                        temp[slice_num * 64 * 8 + subblock_idx * 64 + coeff_num] = k * q;
                        //temp[subblock_idx * 64 + 8 * y + x] = (2 * k * w * 4) / 32;  // 100% matching unquant
                }