X-Git-Url: https://git.sesse.net/?p=movit;a=blobdiff_plain;f=util.cpp;h=c2ac59301778906894410cb27971ea4956289e56;hp=b12ee80712ba83e541f28472760167e5e8fb1d22;hb=83032a75853c0fadda1157e12d119b94991961a8;hpb=185ced44b129739c9b5438da691e71d664d6443a diff --git a/util.cpp b/util.cpp index b12ee80..c2ac593 100644 --- a/util.cpp +++ b/util.cpp @@ -4,6 +4,9 @@ #include #include #include +#include +#include +#include #include #include "fp16.h" @@ -84,24 +87,46 @@ string read_file(const string &filename) { const string full_pathname = *movit_data_directory + "/" + filename; - static char buf[131072]; FILE *fp = fopen(full_pathname.c_str(), "r"); if (fp == NULL) { perror(full_pathname.c_str()); exit(1); } - int len = fread(buf, 1, sizeof(buf), fp); + int ret = fseek(fp, 0, SEEK_END); + if (ret == -1) { + perror("fseek(SEEK_END)"); + exit(1); + } + + int size = ftell(fp); + + ret = fseek(fp, 0, SEEK_SET); + if (ret == -1) { + perror("fseek(SEEK_SET)"); + exit(1); + } + + string str; + str.resize(size); + ret = fread(&str[0], size, 1, fp); + if (ret == -1) { + perror("fread"); + exit(1); + } + if (ret == 0) { + fprintf(stderr, "Short read when trying to read %d bytes from %s\n", + size, full_pathname.c_str()); + exit(1); + } fclose(fp); - return string(buf, len); + return str; } string read_version_dependent_file(const string &base, const string &extension) { - if (movit_shader_model == MOVIT_GLSL_110) { - return read_file(base + "." + extension); - } else if (movit_shader_model == MOVIT_GLSL_130) { + if (movit_shader_model == MOVIT_GLSL_130) { return read_file(base + ".130." + extension); } else if (movit_shader_model == MOVIT_ESSL_300) { return read_file(base + ".300es." + extension); @@ -129,6 +154,7 @@ GLuint compile_shader(const string &shader_src, GLenum type) GLint status; glGetShaderiv(obj, GL_COMPILE_STATUS, &status); if (status == GL_FALSE) { + fprintf(stderr, "Failed to compile shader: %s\n", shader_src.c_str()); exit(1); } @@ -145,21 +171,53 @@ void print_3x3_matrix(const Eigen::Matrix3d& m) string output_glsl_mat3(const string &name, const Eigen::Matrix3d &m) { - char buf[1024]; - sprintf(buf, - "const mat3 %s = mat3(\n" - " %.8f, %.8f, %.8f,\n" - " %.8f, %.8f, %.8f,\n" - " %.8f, %.8f, %.8f);\n\n", - name.c_str(), - m(0,0), m(1,0), m(2,0), - m(0,1), m(1,1), m(2,1), - m(0,2), m(1,2), m(2,2)); - return buf; + // Use stringstream to be independent of the current locale in a thread-safe manner. + stringstream ss; + ss.imbue(locale("C")); + ss.precision(8); + ss << scientific; + ss << "const mat3 " << name << " = mat3(\n"; + ss << " " << m(0,0) << ", " << m(1,0) << ", " << m(2,0) << ",\n"; + ss << " " << m(0,1) << ", " << m(1,1) << ", " << m(2,1) << ",\n"; + ss << " " << m(0,2) << ", " << m(1,2) << ", " << m(2,2) << ");\n\n"; + return ss.str(); +} + +string output_glsl_float(const string &name, float x) +{ + // Use stringstream to be independent of the current locale in a thread-safe manner. + stringstream ss; + ss.imbue(locale("C")); + ss.precision(8); + ss << scientific; + ss << "const float " << name << " = " << x << ";\n"; + return ss.str(); +} + +string output_glsl_vec2(const string &name, float x, float y) +{ + // Use stringstream to be independent of the current locale in a thread-safe manner. + stringstream ss; + ss.imbue(locale("C")); + ss.precision(8); + ss << scientific; + ss << "const vec2 " << name << " = vec2(" << x << ", " << y << ");\n"; + return ss.str(); +} + +string output_glsl_vec3(const string &name, float x, float y, float z) +{ + // Use stringstream to be independent of the current locale in a thread-safe manner. + stringstream ss; + ss.imbue(locale("C")); + ss.precision(8); + ss << scientific; + ss << "const vec3 " << name << " = vec3(" << x << ", " << y << ", " << z << ");\n"; + return ss.str(); } template -void combine_two_samples(float w1, float w2, float pos1, float pos2, unsigned size, +void combine_two_samples(float w1, float w2, float pos1, float pos2, float num_subtexels, float inv_num_subtexels, DestFloat *offset, DestFloat *total_weight, float *sum_sq_error) { assert(movit_initialized); @@ -178,9 +236,7 @@ void combine_two_samples(float w1, float w2, float pos1, float pos2, unsigned si // Round to the minimum number of bits we have measured earlier. // The card will do this for us anyway, but if we know what the real z // is, we can pick a better total_weight below. - z *= size; // Move to pixel coordinates, - z = lrintf(z / movit_texel_subpixel_precision) * movit_texel_subpixel_precision; // Round. - z /= size; // Move back to normalized coordinates. + z = lrintf(z * num_subtexels) * inv_num_subtexels; // Choose total weight w so that we minimize total squared error // for the effective weights: @@ -195,22 +251,22 @@ void combine_two_samples(float w1, float w2, float pos1, float pos2, unsigned si // w = (a(1-z) + bz) / ((1-z)² + z²) // // If z had infinite precision, this would simply reduce to w = w1 + w2. - *total_weight = (w1 * (1 - z) + w2 * z) / (z * z + (1 - z) * (1 - z)); + *total_weight = from_fp64((w1 + z * (w2 - w1)) / (z * z + (1 - z) * (1 - z))); if (sum_sq_error != NULL) { - float err1 = *total_weight * (1 - z) - w1; - float err2 = *total_weight * z - w2; + float err1 = to_fp64(*total_weight) * (1 - z) - w1; + float err2 = to_fp64(*total_weight) * z - w2; *sum_sq_error = err1 * err1 + err2 * err2; } } // Explicit instantiations. template -void combine_two_samples(float w1, float w2, float pos1, float pos2, unsigned size, +void combine_two_samples(float w1, float w2, float pos1, float pos2, float num_subtexels, float inv_num_subtexels, float *offset, float *total_weight, float *sum_sq_error); template -void combine_two_samples(float w1, float w2, float pos1, float pos2, unsigned size, +void combine_two_samples(float w1, float w2, float pos1, float pos2, float num_subtexels, float inv_num_subtexels, fp16_int_t *offset, fp16_int_t *total_weight, float *sum_sq_error); GLuint fill_vertex_attribute(GLuint glsl_program_num, const string &attribute_name, GLint size, GLenum type, GLsizeiptr data_size, const GLvoid *data)