Make num_blocks a uniform.

diff --git a/decoder.shader b/decoder.shader
index 09027915633af93b69a883039836bdecc16f8df2..e7b83e0ccbb52402dcb0a827119fe272eff892fc 100644 (file)
--- a/decoder.shader
+++ b/decoder.shader
@@ -10,6 +10,7 @@ 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;
 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;
 
 const uint prob_bits = 12;
 const uint prob_scale = 1 << prob_bits;


@@ -206,7 +207,6 @@ void main()
        const uint local_y = (gl_LocalInvocationID.x / 8) % 8;
        const uint local_z = gl_LocalInvocationID.x / 64;
 
        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 slice_num = local_z;
        const uint thread_num = local_y * 8 + local_x;
 

diff --git a/narabu.cpp b/narabu.cpp
index 49afa9ad9d502703281820f3ba5880d848745cf1..11117f4fd9f9c744e62b4fe15a7b06b365120812 100644 (file)
--- a/narabu.cpp
+++ b/narabu.cpp
@@ -198,14 +198,18 @@ int main(int argc, char **argv)
        GLint out_tex_pos = glGetUniformLocation(glsl_program_num, "out_tex");
        GLint coeff_tex_pos = glGetUniformLocation(glsl_program_num, "coeff_tex");
        GLint sign_bias_pos = glGetUniformLocation(glsl_program_num, "sign_bias_per_model");
        GLint out_tex_pos = glGetUniformLocation(glsl_program_num, "out_tex");
        GLint coeff_tex_pos = glGetUniformLocation(glsl_program_num, "coeff_tex");
        GLint sign_bias_pos = glGetUniformLocation(glsl_program_num, "sign_bias_per_model");

+       GLint num_blocks_pos = glGetUniformLocation(glsl_program_num, "num_blocks");

        printf("%d err=0x%x pos=%d,%d,%d,%d\n", __LINE__, glGetError(), cum2sym_tex_pos, dsyms_tex_pos, out_tex_pos, sign_bias_pos);
 
        printf("%d err=0x%x pos=%d,%d,%d,%d\n", __LINE__, glGetError(), cum2sym_tex_pos, dsyms_tex_pos, out_tex_pos, sign_bias_pos);
 

+       unsigned num_blocks = (HEIGHT / 16);
+

        // Bind the textures.
        glUniform1i(cum2sym_tex_pos, 0);
        glUniform1i(dsyms_tex_pos, 1);
        glUniform1i(out_tex_pos, 2);
        glUniform1i(coeff_tex_pos, 3);
        glUniform1uiv(sign_bias_pos, 16, sign_bias);
        // Bind the textures.
        glUniform1i(cum2sym_tex_pos, 0);
        glUniform1i(dsyms_tex_pos, 1);
        glUniform1i(out_tex_pos, 2);
        glUniform1i(coeff_tex_pos, 3);
        glUniform1uiv(sign_bias_pos, 16, sign_bias);

+       glUniform1i(num_blocks_pos, num_blocks);

         glBindImageTexture(0, cum2sym_tex, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R8UI);
         glBindImageTexture(1, dsyms_tex, 0, GL_FALSE, 0, GL_READ_ONLY, GL_RG16UI);
         glBindImageTexture(2, out_tex, 0, GL_FALSE, 0, GL_WRITE_ONLY, GL_R8);
         glBindImageTexture(0, cum2sym_tex, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R8UI);
         glBindImageTexture(1, dsyms_tex, 0, GL_FALSE, 0, GL_READ_ONLY, GL_RG16UI);
         glBindImageTexture(2, out_tex, 0, GL_FALSE, 0, GL_WRITE_ONLY, GL_R8);


@@ -213,7 +217,6 @@ int main(int argc, char **argv)
        printf("%d err=0x%x\n", __LINE__, glGetError());
 
        // Decode all luma blocks.
        printf("%d err=0x%x\n", __LINE__, glGetError());
 
        // Decode all luma blocks.

-       unsigned num_blocks = (HEIGHT / 16);

        for (unsigned y = 0; y < 8; ++y) {
                 for (unsigned x = 0; x < 8; ++x) {
                        unsigned coeff_num = y * 8 + x;
        for (unsigned y = 0; y < 8; ++y) {
                 for (unsigned x = 0; x < 8; ++x) {
                        unsigned coeff_num = y * 8 + x;