diff --git a/CLAUDE.md b/CLAUDE.md new file mode 100644 index 0000000..eecfeb9 --- /dev/null +++ b/CLAUDE.md @@ -0,0 +1,236 @@ +# CLAUDE.md + +This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. + +## Project Overview + +This is a cross-platform Shadertoy-like fragment shader viewer built with SDL3 + OpenGL 3.3. The application loads and displays GLSL shaders from the `shaders/` directory with runtime switching capabilities. + +## Build and Development Commands + +### Building (CMake - Development) +```bash +mkdir build && cd build +cmake .. +cmake --build . --config Release +``` + +### Building (Makefile - Release Packages) +```bash +make windows_release # Creates .zip with DLLs +make macos_release # Creates .dmg with app bundle +make linux_release # Creates .tar.gz +make show_version # Display build version (YYYY-MM-DD format) +``` + +Platform-specific debug builds: +```bash +make windows_debug +make macos_debug +make linux_debug +``` + +### Running the Application +```bash +./shadertoy [SHADER_PATH] [-F|--fullscreen] + +# Examples: +./shadertoy shaders/test.frag.glsl +./shadertoy -F shaders/fractal_pyramid.frag.glsl +``` + +**Runtime Controls:** +- `ESC` - Exit +- `F3` - Toggle fullscreen +- `LEFT/RIGHT ARROW` - Cycle through shaders in directory + +## Architecture + +### Core Design +All application logic resides in `src/main.cpp` (~469 lines) - a monolithic design that's straightforward to understand. Key components: + +1. **Shader Loading System** - Automatic directory scanning of `.glsl` files, sorted alphabetically +2. **OpenGL Rendering** - Single fullscreen quad with fragment shader using GL_TRIANGLE_STRIP +3. **Event Loop** - SDL3-based with vsync (SDL_GL_SwapWindow + 1ms delay) +4. **Resource Path Resolution** - Multi-path fallback system for executable, relative, and macOS bundle paths + +### Global State (main.cpp) +```cpp +shader_list_ // Vector of discovered .glsl shader paths +current_shader_index_ // Active shader in rotation +current_program_ // OpenGL shader program handle +shader_start_ticks_ // Base time for iTime uniform calculation +window_ // SDL3 window pointer +shaders_directory_ // Shader directory path (resolved at startup) +``` + +### Dependencies +- **SDL3** - Window/input management, OpenGL context +- **GLAD** - OpenGL 3.3 loader (statically linked via `third_party/glad/`) +- **C++17 stdlib** - filesystem, fstream, vector, algorithms + +### Platform-Specific Code +Uses preprocessor defines (`WINDOWS_BUILD`, `MACOS_BUILD`, `LINUX_BUILD`) for: +- `getExecutableDirectory()` - Windows API, mach-o dyld, or /proc/self/exe +- `getResourcesDirectory()` - Special macOS app bundle handling (Contents/Resources) +- Linking flags - Windows uses static linking, macOS links OpenGL framework + +## Shader System + +### Shader Format (GLSL 3.3 Core) +All shaders must follow this structure: + +```glsl +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; // Normalized [0,1] coordinates from vertex shader +uniform vec2 iResolution; // Window resolution in pixels +uniform float iTime; // Time since shader loaded (seconds) + +// Shadertoy-style entry point +void mainImage(out vec4 fragColor, in vec2 fragCoord) { + // fragCoord is in pixel coordinates + // Your shader code here +} + +// Wrapper converts vUV to pixel coordinates +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} +``` + +### Adding New Shaders +1. Create `.glsl` file in `shaders/` directory (use `.frag.glsl` convention) +2. Follow required format above +3. File automatically appears in runtime shader rotation (arrow keys to navigate) +4. **No code changes required** - directory is scanned at startup + +### Shader Loading Pipeline +1. Directory scan on startup (`scanShaderDirectory()`) +2. Sort alphabetically +3. Load fragment shader source from disk +4. Compile vertex shader (fixed, embedded in main.cpp) +5. Compile fragment shader with error logging +6. Link program with error handling +7. Update uniforms each frame (iResolution, iTime) + +### Supported Shadertoy Features +- ✅ `iTime` - Time uniform +- ✅ `iResolution` - Window resolution (vec2, not vec3) +- ✅ `mainImage()` function signature +- ❌ `iMouse` - Not implemented +- ❌ `iChannel0-3` - No texture channels (multi-pass not supported) +- ❌ `iFrame`, `iTimeDelta`, `iDate` - Not implemented + +### Converting from Shadertoy +When porting Shadertoy shaders: +1. Copy the `mainImage()` function as-is +2. Add standard header (see format above) +3. Add wrapper `main()` function +4. Remove texture channel references (iChannel0-3) +5. Change `iResolution.xy` to `iResolution` (this project uses vec2) + +### Common Conversion Issues and Solutions + +Based on experience converting complex shaders like ddla_light_tunnel: + +#### iResolution vec3 vs vec2 +- **Shadertoy:** `iResolution` is `vec3(width, height, width/height)` +- **This project:** `iResolution` is `vec2(width, height)` +- **Solution:** Create vec3 manually: `vec3 r = vec3(iResolution.xy, iResolution.x/iResolution.y);` +- **Then use:** `r.xy` for resolution, `r.y` for height (as in original) + +#### Uninitialized Variables +- **Problem:** Shadertoy code may have `vec3 rgb;` without initialization +- **Shadertoy behavior:** Likely initializes to `vec3(0.0)` (black) +- **This project:** Uninitialized variables contain undefined values (often causes black screen or wrong colors) +- **Solution:** Always initialize: `vec3 rgb = vec3(0.0);` +- **Wrong approach:** Don't use random noise unless shader explicitly uses iChannel texture for noise + +#### Low mix() Factors Are Intentional +- **Example:** `rgb = mix(rgb, calculated_color, 0.01);` means 1% new color, 99% existing +- **Don't change these factors** - they create subtle effects intentionally +- **If output is black:** Problem is likely the base value (rgb), not the mix factor + +#### iChannel Textures +- **Shadertoy shows iChannel0-3** in UI, but shader may not use them +- **If iChannel3 has RGB noise but code doesn't reference it:** The noise is not actually used +- **Check code for `texture(iChannelN, ...)` calls** - if none exist, ignore the iChannel setup +- **Procedural noise replacement:** Only if shader explicitly samples the texture + +#### Color Swapping Issues +- **If colors are completely wrong:** Don't randomly swap color variables +- **First verify:** Code matches original Shadertoy exactly (except required GLSL changes) +- **Common mistake:** Changing color applications that were correct in original +- **Debug approach:** Revert to exact original code, only modify for GLSL 3.3 compatibility + +#### Division by Small Values - Color Overflow Artifacts +- **Problem:** Division by values near zero causes extreme values → color overflow artifacts (green points, strange colors) +- **Example:** `.01*vec4(6,2,1,0)/length(u*sin(iTime))` can divide by ~0.0 when near center and sin≈0 +- **Symptoms:** Bright white center that pulses to green, or random color artifacts in specific areas +- **Root cause:** Even with tonemap (tanh/clamp), extreme intermediate values cause precision issues or saturation +- **Solution - Double Protection:** + 1. **Add epsilon to denominator:** `max(denominator, 0.001)` prevents division by exact zero + 2. **Clamp the result:** `min(result, vec4(50.0))` prevents extreme accumulation + 3. **Only clamp problematic terms:** Don't clamp everything or scene becomes dim +- **Example fix:** + ```glsl + // Original (causes overflow): + o += .01*vec4(6,2,1,0)/length(u*sin(t+t+t)) + 1./s * length(u); + + // Fixed (prevents overflow while maintaining brightness): + vec4 brightTerm = min(.01*vec4(6,2,1,0)/max(length(u*sin(t+t+t)), 0.001), vec4(50.0)); + o += brightTerm + 1./s * length(u); + ``` +- **Key values:** epsilon ~0.001 (not too small, not too large), clamp ~50.0 (allows brightness without explosion) +- **Why this works in Shadertoy:** WebGL/browsers may handle float overflow differently than native OpenGL drivers + +#### Debugging Black/Wrong Output +1. **Check compilation errors first** - shader must compile without errors +2. **Initialize all variables** - especially vec3 colors +3. **Verify vec3 iResolution handling** - create vec3 from vec2 if needed +4. **Don't modify mix factors** - keep original values +5. **Compare with original code** - ensure logic is identical +6. **Test progressive changes** - add header, test; add wrapper, test; etc. +7. **Check for division by small values** - if you see color artifacts (especially green), look for divisions that can approach zero + +## Build System Details + +### Version Numbering +Releases use build date format: `shadertoy-YYYY-MM-DD-{platform}` (auto-generated by Makefile) + +### Release Artifacts +- **Windows:** `.zip` with `shadertoy.exe` + `SDL3.dll` + shaders +- **macOS:** `.dmg` with app bundle containing embedded SDL3.framework (arm64 only) +- **Linux:** `.tar.gz` with binary + shaders + +### Resource Bundling +- Shaders copied to release packages +- LICENSE and README.md included +- Platform-specific dependencies bundled (DLLs on Windows, frameworks on macOS) + +## Important Notes + +### Vertex Shader +The vertex shader is hardcoded in `main.cpp` and creates a fullscreen quad. It: +- Takes `vec2 aPos` (location 0) in NDC space [-1, 1] +- Outputs `vec2 vUV` normalized to [0, 1] +- No transformation matrices needed + +### Shader Hot-Reloading +Currently **not implemented**. Shader changes require application restart. The architecture would support adding this via file watching. + +### Multi-Pass Rendering +Single-pass only. To add multi-pass (BufferA/B/C like Shadertoy): +- Create FBOs and textures per buffer +- Render buffers in dependency order +- Pass textures as `iChannel0-3` uniforms +- Use ping-pong for feedback loops + +### OpenGL Context +Created via SDL3 with core profile (no deprecated functions). Context version: 3.3 core. \ No newline at end of file diff --git a/CMakeLists.txt b/CMakeLists.txt index 38f91a8..1b02ba8 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -22,6 +22,7 @@ set(APP_SOURCES # Fuentes de librerías de terceros set(EXTERNAL_SOURCES third_party/glad/src/glad.c + third_party/jail_audio.cpp ) # Configuración de SDL3 @@ -35,6 +36,7 @@ add_executable(${PROJECT_NAME} ${APP_SOURCES} ${EXTERNAL_SOURCES}) target_include_directories(${PROJECT_NAME} PUBLIC "${CMAKE_SOURCE_DIR}/src" "${CMAKE_SOURCE_DIR}/third_party/glad/include" + "${CMAKE_SOURCE_DIR}/third_party" ) # Enlazar la librería SDL3 diff --git a/data/music/485077__mat397__confused-voices.ogg b/data/music/485077__mat397__confused-voices.ogg new file mode 100644 index 0000000..22d25e8 Binary files /dev/null and b/data/music/485077__mat397__confused-voices.ogg differ diff --git a/data/music/485078__mat397__polyflute-pad.ogg b/data/music/485078__mat397__polyflute-pad.ogg new file mode 100644 index 0000000..bae2eaf Binary files /dev/null and b/data/music/485078__mat397__polyflute-pad.ogg differ diff --git a/data/music/485079__mat397__melancholic-flutes-pad.ogg b/data/music/485079__mat397__melancholic-flutes-pad.ogg new file mode 100644 index 0000000..a00f2f4 Binary files /dev/null and b/data/music/485079__mat397__melancholic-flutes-pad.ogg differ diff --git a/data/music/486083__mat397__world-of-ants-pad.ogg b/data/music/486083__mat397__world-of-ants-pad.ogg new file mode 100644 index 0000000..f697a07 Binary files /dev/null and b/data/music/486083__mat397__world-of-ants-pad.ogg differ diff --git a/data/music/618041__mat397__mangle-dark-pad.ogg b/data/music/618041__mat397__mangle-dark-pad.ogg new file mode 100644 index 0000000..7df122e Binary files /dev/null and b/data/music/618041__mat397__mangle-dark-pad.ogg differ diff --git a/data/music/618042__mat397__bad-electric-dark-pad.ogg b/data/music/618042__mat397__bad-electric-dark-pad.ogg new file mode 100644 index 0000000..fc817ae Binary files /dev/null and b/data/music/618042__mat397__bad-electric-dark-pad.ogg differ diff --git a/data/music/_readme_and_license.txt b/data/music/_readme_and_license.txt new file mode 100644 index 0000000..09686a3 --- /dev/null +++ b/data/music/_readme_and_license.txt @@ -0,0 +1,41 @@ +Pack downloaded from Freesound +---------------------------------------- + +"Ambient pads" + +This Pack of sounds contains sounds by the following user: + - Mat397 ( https://freesound.org/people/Mat397/ ) + +You can find this pack online at: https://freesound.org/people/Mat397/packs/27416/ + + +Licenses in this Pack (see below for individual sound licenses) +--------------------------------------------------------------- + +Creative Commons 0: http://creativecommons.org/publicdomain/zero/1.0/ +Attribution 3.0: http://creativecommons.org/licenses/by/3.0/ + + +Sounds in this Pack +------------------- + + * 618042__mat397__bad-electric-dark-pad.wav.wav + * url: https://freesound.org/s/618042/ + * license: Creative Commons 0 + * 618041__mat397__mangle-dark-pad.wav.wav + * url: https://freesound.org/s/618041/ + * license: Creative Commons 0 + * 486083__mat397__world-of-ants-pad.wav.wav + * url: https://freesound.org/s/486083/ + * license: Attribution 3.0 + * 485079__mat397__melancholic-flutes-pad.wav.wav + * url: https://freesound.org/s/485079/ + * license: Attribution 3.0 + * 485078__mat397__polyflute-pad.wav.wav + * url: https://freesound.org/s/485078/ + * license: Attribution 3.0 + * 485077__mat397__confused-voices.wav.wav + * url: https://freesound.org/s/485077/ + * license: Attribution 3.0 + + diff --git a/release/icon.afdesign b/release/icon/icon.afdesign similarity index 100% rename from release/icon.afdesign rename to release/icon/icon.afdesign diff --git a/release/icon/icon.png b/release/icon/icon.png new file mode 100644 index 0000000..8525269 Binary files /dev/null and b/release/icon/icon.png differ diff --git a/shaders/cineshader_lava.frag.glsl b/shaders/cineshader_lava.frag.glsl new file mode 100644 index 0000000..ecf6bb6 --- /dev/null +++ b/shaders/cineshader_lava.frag.glsl @@ -0,0 +1,82 @@ +// Name: Cineshader Lava +// Author: [TO_BE_COMPLETED] +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +float opSmoothUnion( float d1, float d2, float k ) +{ + float h = clamp( 0.5 + 0.5*(d2-d1)/k, 0.0, 1.0 ); + return mix( d2, d1, h ) - k*h*(1.0-h); +} + +float sdSphere( vec3 p, float s ) +{ + return length(p)-s; +} + +float map(vec3 p) +{ + float d = 2.0; + for (int i = 0; i < 16; i++) { + float fi = float(i); + float time = iTime * (fract(fi * 412.531 + 0.513) - 0.5) * 2.0; + d = opSmoothUnion( + sdSphere(p + sin(time + fi * vec3(52.5126, 64.62744, 632.25)) * vec3(2.0, 2.0, 0.8), mix(0.5, 1.0, fract(fi * 412.531 + 0.5124))), + d, + 0.4 + ); + } + return d; +} + +vec3 calcNormal( in vec3 p ) +{ + const float h = 1e-5; // or some other value + const vec2 k = vec2(1,-1); + return normalize( k.xyy*map( p + k.xyy*h ) + + k.yyx*map( p + k.yyx*h ) + + k.yxy*map( p + k.yxy*h ) + + k.xxx*map( p + k.xxx*h ) ); +} + +void mainImage( out vec4 fragColor, in vec2 fragCoord ) +{ + vec2 uv = fragCoord/iResolution.xy; + + // screen size is 6m x 6m + vec3 rayOri = vec3((uv - 0.5) * vec2(iResolution.x/iResolution.y, 1.0) * 6.0, 3.0); + vec3 rayDir = vec3(0.0, 0.0, -1.0); + + float depth = 0.0; + vec3 p; + + for(int i = 0; i < 64; i++) { + p = rayOri + rayDir * depth; + float dist = map(p); + depth += dist; + if (dist < 1e-6) { + break; + } + } + + depth = min(6.0, depth); + vec3 n = calcNormal(p); + float b = max(0.0, dot(n, vec3(0.577))); + vec3 col = (0.5 + 0.5 * cos((b + iTime * 3.0) + uv.xyx * 2.0 + vec3(0,2,4))) * (0.85 + b * 0.35); + col *= exp( -depth * 0.15 ); + + // maximum thickness is 2m in alpha channel + fragColor = vec4(col, 1.0 - (depth - 0.5) / 2.0); +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/creation.frac.glsl b/shaders/creation.frac.glsl index be3e6f9..7f16f6f 100644 --- a/shaders/creation.frac.glsl +++ b/shaders/creation.frac.glsl @@ -1,3 +1,5 @@ +// Name: Creation +// Author: [TO_BE_COMPLETED] #version 330 core precision highp float; diff --git a/shaders/dbz.frag.glsl b/shaders/dbz.frag.glsl new file mode 100644 index 0000000..54e2673 --- /dev/null +++ b/shaders/dbz.frag.glsl @@ -0,0 +1,189 @@ +// Name: DBZ +// Author: [TO_BE_COMPLETED] +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +// A small improv/fanart from yesterday. + +#define s(a,b,c) smoothstep(a,b,c) +#define PI 3.14159 +#define NBCaps 3. + +mat2 r2d( float a ){ float c = cos(a), s = sin(a); return mat2( c, s, -s, c ); } +float metaDiamond(vec2 p, vec2 pixel, float r){ vec2 d = abs(p-pixel); return r / (d.x + d.y); } + +// Dave Hoskins's hash ! Noone can hash hashes like he hashes ! +float hash12(vec2 p) +{ + vec3 p3 = fract(vec3(p.xyx) * .1031); + p3 += dot(p3, p3.yzx + 33.33); + return fract((p3.x + p3.y) * p3.z); +} +// Smoothed 1D-noise. Just like Zoltraak : stupid simple. Powerful. +float fbm(in vec2 v_p) +{ + float pvpx = v_p.x; + vec2 V1 = vec2(floor(pvpx )); + vec2 V2 = vec2(floor(pvpx + 1.0)); + return mix(hash12(V1),hash12(V2),smoothstep(0.0,1.0,fract(pvpx))); +} + +void mainImage( out vec4 fragColor, in vec2 fragCoord ) +{ + // Get Base Coordinates + vec2 p = vec2( (1.0/iResolution.y)*(fragCoord.x - iResolution.x/2.0),fragCoord.y / iResolution.y - 0.5); + + // reversed, for accurate Martian perspective... :) + p.x=-p.x; + + // Zoom out + p*=150.0; + + // Init the Accumulator + vec4 col = vec4(0.05,0.05,0.15,1.0); + + // Make up other boxes and save base in them. + vec2 save1 = p; + vec2 save2 = p; + + // Faint Nebula Background + + // Tilt the camera + p*= r2d(-0.05); + // Space Background Gradient + col = mix(col,vec4(0.2,0.3,0.5,1.0),smoothstep(75.0,0.0,abs(p.y - 5.0*fbm(vec2(0.01*(p.x - 33.333*iTime))) + 3.5))); + // Untilt the camera + p*= r2d( 0.05); + + // BG Starfield + + // Rotate + p*= r2d(-0.05); + // Zoom In + p*= 0.35; + // Scroll Left + p+= vec2(-5.0*iTime,0.0); + + // Hack the coords... + vec2 b = fract(5.0*p); + p = floor(5.0*p); + // Draw the stars + if( fbm(vec2(p.x*p.y)) > 0.996)col += clamp(1.0-pow(3.0*length(b+vec2(-0.5)),0.5),0.0,1.0); + + // Reload because the coords are all f.. up now ! + p = save1; + // Another Box... + vec2 save3; + + // We're going to draw max 4 capsules max. + // Yes we could draw more but Earth must survive, man. Have mercy ! + float Nb_Capsules = clamp(NBCaps,0.0,4.0); + for( float i = 0.0;i 0.0 ) + { + // Green Jet + col += vec4(0.0,1.0,0.5,1.0)*smoothstep(0.2,0.0,abs(p.y - 0.05*fbm(vec2(1.5*p.x - 40.0*iTime)))-0.05)*smoothstep(29.0,0.0,abs(p.x)); + // White Jet + col += vec4(1.0,1.0,1.0,1.0)*smoothstep(0.1,0.0,abs(p.y - 0.05*fbm(vec2(1.5*p.x - 40.0*iTime)))-0.05)*smoothstep(29.0,0.0,abs(p.x)); + }; + + // Reload ! + p = save3; + // (Ox)-axis Symetry for the flames + p.y = abs(p.y); + // Fine-tuning Flames position + p+= vec2(-10.0,0.0); + // Fine-tuning Flames Shape + p *= vec2(0.75,1.0); + + // Green Flames + col += 0.8*vec4(0.0,1.0,0.5,1.0)*s(20.0,0.0,length(p)-25.0+7.0*sin(0.30*length(p)*atan(p.y,p.x) + 55.0*iTime)); + // White flames + col += 0.8*vec4(1.0,1.0,1.0,1.0)*s(20.0,0.0,length(p)-20.0+7.0*sin(0.30*length(p)*atan(p.y,p.x) + 55.0*iTime)); + + p = save3; + + // Fat Aura + col = mix(col,vec4(1.0),0.5*s(10.0,0.0,length(p + vec2(5.0,0.0))-20.0)*abs(sin(50.0*iTime))); + // Less-Fat Aura + col = mix(col,vec4(1.0),0.5*s(20.0,0.0,length(p + vec2(5.0,0.0))-20.0)); + // Frieren : "Aura ? Shader yourself !" + + // The Pod + + // White Disk + col = mix(col,vec4(1.0),s(0.01,0.0,length(p)-20.0)); + + if( length(p) - 20.0 < 0.0 ) // Basic Masking + { + // 2D Shading : bluish large shadow + col = mix(col,vec4(0.65,0.68,0.68 + 0.1*(3.0-i),1.0),s(0.5,0.0,length(p - vec2(2.0,0.0))-17.0)); + // 2D Shading : dark small shadow + // If Outside Porthole Zone + if(s(0.0,1.0,length(vec2(3.0,2.0)*p + vec2(33.5,0.0))-23.0)>0.0) + col = mix(col,vec4(0.45,0.55,0.55 + 0.1*(3.0-i),1.0),0.75*s(0.5,0.0,length(p - vec2(2.0,0.0)+ 0.5*fbm(vec2(4.5*atan(p.y,p.x))))-9.0)); + + // Small 2D Indentation Details On The Spheres Using A Procedural Texture + // NOTE: Original used texture(iChannel0, ...) which is not supported + // Texture detail removed - not essential for the effect + // vec4 colorCapsule = vec4(hash12(0.0003*p*dot(p,p) + 0.789*i)); + // if(colorCapsule.x>0.75)if(s(0.0,1.0,length(vec2(3.0,2.0)*p + vec2(33.5,0.0))-23.0)>0.0)col *= vec4(0.25,0.25,0.25,1.0); + + // Bigger Dark Line All Around The Pod + col = mix(col,vec4(0.0),s(0.2,0.0,abs(length(p)-19.9)-0.20)); + // Draw The Porthole : + col = mix(col,vec4(0.5,0.2,0.3,1.0) // Base Color + -s(5.0,0.0,length(p + vec2(-6.0,15.0))-20.0) // Main Shadow + -s(0.25,0.0,abs(length(p + vec2(0.0,3.0))-15.0)-0.4)// Vertical Shadow + -s(0.0,1.5,p.y-8.5) // top Shadow + +0.25*vec4(1.0,0.5,0.0,1.0)*s(10.0,0.0,abs(p.y)) // Fake Glass Gradient + , + s(0.5,0.0,length(vec2(3.0,2.0)*p + vec2(35.0,0.0))-19.9)); + + // Porthole Black Rings + // Internal + col = mix(col,vec4(0.0,0.0,0.0,1.0),s(1.0,0.0,abs(length(vec2(3.0,2.0)*p + vec2(35.0,0.0))-19.9)-0.1)); + // External + col = mix(col,vec4(0.0,0.0,0.0,1.0),s(1.0,0.0,abs(length(vec2(3.0,2.0)*p + vec2(33.5,0.0))-23.0)-0.1)); + + // Pod Tennis-Ball Door Line... + if(p.y>0.0)col = mix(col,vec4(0.0,0.0,0.0,1.0),s(1.0,0.0,abs(length(vec2(3.0,2.0)*p + vec2( 29.0,0.0))-30.0)-0.1)); + if(p.y<0.0)col = mix(col,vec4(0.0,0.0,0.0,1.0),s(1.0,0.0,abs(length(vec2(3.0,2.0)*p + vec2(-31.0,0.0))-30.0)-0.1)); + }; + }; + // WAKE UP SHEEPLE ! + fragColor = clamp(col,0.0,1.0); +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/fractal_pyramid.frag.glsl b/shaders/fractal_pyramid.frag.glsl index 85b324d..e64cb00 100644 --- a/shaders/fractal_pyramid.frag.glsl +++ b/shaders/fractal_pyramid.frag.glsl @@ -1,3 +1,5 @@ +// Name: Fractal Pyramid +// Author: [TO_BE_COMPLETED] #version 330 core precision highp float; diff --git a/shaders/just_another_cube.frag.glsl b/shaders/just_another_cube.frag.glsl new file mode 100644 index 0000000..6a30699 --- /dev/null +++ b/shaders/just_another_cube.frag.glsl @@ -0,0 +1,158 @@ +// Name: Just Another Cube +// Author: [TO_BE_COMPLETED] +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +// CC0: Just another cube +// Glowtracers are great for compact coding, but I wanted to see how much +// I could squeeze a more normal raymarcher in terms of characters used. + +// Twigl: https://twigl.app?ol=true&ss=-OW-y9xgRgWubwKcn0Nd + +// == Globals == +// Single-letter variable names are used to save characters (code golfing). +mat2 R; // A 2D rotation matrix, calculated once per frame in mainImage and used by D. +float d=1. // Stores the most recent distance to the scene from the ray's position. + , z=0. // Stores the total distance traveled along the ray (initialized to avoid undefined behavior) + , G=9. // "Glow" variable. Tracks the closest the ray comes to the object (for volumetric glow effect). + , M=1e-3 + ; +// == Distance Function (SDF - Signed Distance Field) == +// This function calculates the shortest distance from a given point 'p' to the scene geometry. +// A positive result means the point is outside an object, negative is inside, and zero is on the surface. +// This is the core of "raymarching", as it tells us the largest safe step we can take along a ray. +float D(vec3 p) { + // Apply two rotations to the point's coordinates. This twists the space the object + // exists in, making the simple cube shape appear more complex and animated. + p.xy *= R; + p.xz *= R; + + // Create a higher-frequency version of the coordinate for detailed surface patterns. + vec3 S = sin(123.*p); + + // This creates a volumetric glow effect by tracking the minimum distance + // to either the existing glow value or a glowing shell around the object. + G = min( + G + // The glowing shell + , max( + abs(length(p)-.6) + // The main object distance calculation: + // 1. A superquadric (rounded cube shape) is created using an L8-norm. + // The expression `pow(dot(p=p*p*p*p,p),.125)` is a golfed version of + // `pow(pow(p.x,8)+pow(p.y,8)+pow(p.z,8), 1./8.)`. + // The `- .5` defines the object's size. + , d = pow(dot(p*=p*p*p,p),.125) - .5 + // 2. Surface detail subtraction. This creates small surface variations + // using high-frequency sine waves for more appealing reflections. + - pow(1.+S.x*S.y*S.z,8.)/1e5 + ) + ); + + return d; +} + +// == Main Render Function == +// This function is called for every pixel on the screen to determine its color. +// 'o' is the final output color (rgba). 'C' is the input pixel coordinate (xy). +void mainImage(out vec4 o, vec2 C) { + // Single-letter variable names are used to save characters (code golfing). + vec3 p // The current point in 3D space along the ray. + , O // Multi-purpose vector: color accumulator, then normal vector, then final color. + , r=vec3(iResolution.xy, iResolution.y) // 'r' holds screen resolution, later re-used for the epsilon vector and reflection. + // 'I' is the Ray Direction vector. It's calculated once per pixel. + // This converts the 2D screen coordinate 'C' into a 3D direction, creating the camera perspective. + , I=normalize(vec3(C-.5*r.xy, r.y)) + // Base glow color (dark bluish tint). + , B=vec3(1,2,9)*M + ; + + // == Raymarching Loop == + // This loop "marches" a ray from the camera out into the scene to find what it hits. + // It uses a golfed structure where the body of the loop updates the ray position 'p', + // and the "advancement" step moves the ray forward. + for( + // -- Initializer (runs once before the loop) -- + // Calculate the rotation matrix for this frame based on time. + R = mat2(cos(.3*iTime+vec4(0,11,33,0))) + // -- Condition -- + // Loop while total distance 'z' is less than 9 and we are not yet touching a surface (d > 1e-3). + ; z<9. && d > M + // -- Advancement -- + // The ray advances by the safe distance 'd' returned by D(p). + // The result of D(p) is also assigned to the global 'd' inside the function. + ; z += D(p) + ) + // -- Loop Body -- + // Calculate the current position 'p' in world space. + // The camera starts at (0,0,-2) and points forward. + p = z*I + , p.z -= 2. + ; + + // -- Hit Condition -- + // If the loop finished because z exceeded the max distance, we hit nothing. Otherwise, we hit the surface. + if (z < 9.) { + // -- Calculate Surface Normal -- + // Estimate the gradient ∇D at the hit point 'p' via central differences on the SDF D. + // We use ε = 1e-3 and loop over each axis (x, y, z): + // • Zero r, then set r[i] = ε. + // • Compute O[i] = D(p + r) – D(p – r). + // After the loop, O holds the unnormalized normal vector. + for ( + int i=0 // axis index: 0→x, 1→y, 2→z (initialized to avoid warnings) + ; i < 3 + ; O[i++] = D(p+r) - D(p-r) + ) + r -= r // clear r to vec3(0) + , r[i] = M // set only the i-th component + ; + + // -- Lighting and Shading -- + // 'z' is re-purposed to store a fresnel factor (1 - cos(angle)) for edge brightness. + // `dot(O, I)` calculates how much the surface faces away from the camera. + // O is also normalized here to become a proper normal vector. + z = 1.+dot(O = normalize(O),I); + + // 'r' is re-purposed to store the reflection vector. + r = reflect(I,O); + + // Calculate a point 'C' along the reflection vector 'r' to sample a background color. + // For upward reflections (r.y > 0), this finds the intersection with the plane y=5. + C = (p+r*(5.-p.y)/abs(r.y)).xz; + + // Calculate the final color 'O' of the hit point. + O = + // Multiply by the fresnel factor squared for stronger edge reflections. + z*z * + // Use a ternary operator to decide the color based on where the reflection ray goes. + ( + // If the reflection vector points upward... + r.y>0. + // ...sample a procedural "sky" with a radial gradient and blue tint. + ? 5e2*smoothstep(5., 4., d = sqrt(length(C*C))+1.)*d*B + // ...otherwise, sample a "floor" with a deep blue exponential falloff. + : exp(-2.*length(C))*(B/M-1.) + ) + // Add rim lighting (brighter on upward-facing surfaces). + + pow(1.+O.y,5.)*B + ; + } + + // == Tonemapping & Output == + // Apply final effects and map the High Dynamic Range (HDR) color to a displayable range. + // Add glow contribution: smaller G values (closer ray passes) create a brighter blue glow. + o = sqrt(O+B/G).xyzx; +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/kishimisu.frag.glsl b/shaders/kishimisu.frag.glsl index 7c8ac3c..d10ca76 100644 --- a/shaders/kishimisu.frag.glsl +++ b/shaders/kishimisu.frag.glsl @@ -1,3 +1,5 @@ +// Name: Kishimisu +// Author: [TO_BE_COMPLETED] #version 330 core precision highp float; diff --git a/shaders/octograms.frag.glsl b/shaders/octograms.frag.glsl index 4a601d6..0e9c2ca 100644 --- a/shaders/octograms.frag.glsl +++ b/shaders/octograms.frag.glsl @@ -1,3 +1,5 @@ +// Name: Octograms +// Author: [TO_BE_COMPLETED] #version 330 core precision highp float; diff --git a/shaders/remember.frag.glsl b/shaders/remember.frag.glsl new file mode 100644 index 0000000..3d32043 --- /dev/null +++ b/shaders/remember.frag.glsl @@ -0,0 +1,73 @@ +// Name: Remember +// Author: diatribes +// URL: https://www.shadertoy.com/view/tXSBDK +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +// fuzzy brain + +// Hash function to replace iChannel0 texture noise +float hash12(vec2 p) { + vec3 p3 = fract(vec3(p.xyx) * .1031); + p3 += dot(p3, p3.yzx + 33.33); + return fract((p3.x + p3.y) * p3.z); +} + +void mainImage(out vec4 o, vec2 u) { + + vec3 q,p = vec3(iResolution.xy, iResolution.x / iResolution.y); + + float i = 0.0, s, + // start the ray at a small random distance, + // this will reduce banding + // Replaced texelFetch(iChannel0, ...) with hash function + d = .125 * hash12(u), + t = iTime * .1; + + // scale coords + u = (u+u-p.xy)/p.y; + if (abs(u.y) > .8) { o = vec4(0); return; } + + // Initialize output color (out parameter must be initialized before use) + o = vec4(0.0); + + for(; i<64.; i++) { + + // shorthand for standard raymarch sample, then move forward: + // p = ro + rd * d, p.z + t + q = p = vec3(u * d, d + t*5.); + p.xy *= mat2(cos(.1*p.z+.1*t+vec4(0,33,11,0))); + + q.xz = cos(q.xz); + p.z = cos(p.z) ; + // turbulence + for (s = 1.; s++ <6.; + q += sin(.6*t+p.zxy*.6), + p += sin(t+t+p.yzx*s)*.6); + + // distance to spheres + d += s = .02 + abs(min(length(p+3.*sin(p.z*.5))-4., length(q-2.*sin(p.z*.4))-6.))*.2; + + // color: 1.+cos so we don't go negative, cos(d+vec4(6,4,2,0)) samples from the palette + // divide by s for form and distance + // Clamp only the first term to prevent extreme overflow, leave second term free + vec4 brightTerm = min(.01*vec4(6,2,1,0)/max(length(u*sin(t+t+t)), 0.001), vec4(50.0)); + o += brightTerm + 1. / s * length(u); + + } + + // tonemap and divide brightness + o = tanh(max(o /6e2 + dot(u,u)*.35, 0.)); +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/seascape.frag.glsl b/shaders/seascape.frag.glsl new file mode 100644 index 0000000..3584bb6 --- /dev/null +++ b/shaders/seascape.frag.glsl @@ -0,0 +1,222 @@ +// Name: Seascape +// Author: Alexander Alekseev (TDM) +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +/* + * "Seascape" by Alexander Alekseev aka TDM - 2014 + * License Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. + * Contact: tdmaav@gmail.com + */ + +const int NUM_STEPS = 32; +const float PI = 3.141592; +const float EPSILON = 1e-3; +#define EPSILON_NRM (0.1 / iResolution.x) +//#define AA + +// sea +const int ITER_GEOMETRY = 3; +const int ITER_FRAGMENT = 5; +const float SEA_HEIGHT = 0.6; +const float SEA_CHOPPY = 4.0; +const float SEA_SPEED = 0.8; +const float SEA_FREQ = 0.16; +const vec3 SEA_BASE = vec3(0.0,0.09,0.18); +const vec3 SEA_WATER_COLOR = vec3(0.8,0.9,0.6)*0.6; +#define SEA_TIME (1.0 + iTime * SEA_SPEED) +const mat2 octave_m = mat2(1.6,1.2,-1.2,1.6); + +// math +mat3 fromEuler(vec3 ang) { + vec2 a1 = vec2(sin(ang.x),cos(ang.x)); + vec2 a2 = vec2(sin(ang.y),cos(ang.y)); + vec2 a3 = vec2(sin(ang.z),cos(ang.z)); + mat3 m; + m[0] = vec3(a1.y*a3.y+a1.x*a2.x*a3.x,a1.y*a2.x*a3.x+a3.y*a1.x,-a2.y*a3.x); + m[1] = vec3(-a2.y*a1.x,a1.y*a2.y,a2.x); + m[2] = vec3(a3.y*a1.x*a2.x+a1.y*a3.x,a1.x*a3.x-a1.y*a3.y*a2.x,a2.y*a3.y); + return m; +} +float hash( vec2 p ) { + float h = dot(p,vec2(127.1,311.7)); + return fract(sin(h)*43758.5453123); +} +float noise( in vec2 p ) { + vec2 i = floor( p ); + vec2 f = fract( p ); + vec2 u = f*f*(3.0-2.0*f); + return -1.0+2.0*mix( mix( hash( i + vec2(0.0,0.0) ), + hash( i + vec2(1.0,0.0) ), u.x), + mix( hash( i + vec2(0.0,1.0) ), + hash( i + vec2(1.0,1.0) ), u.x), u.y); +} + +// lighting +float diffuse(vec3 n,vec3 l,float p) { + return pow(dot(n,l) * 0.4 + 0.6,p); +} +float specular(vec3 n,vec3 l,vec3 e,float s) { + float nrm = (s + 8.0) / (PI * 8.0); + return pow(max(dot(reflect(e,n),l),0.0),s) * nrm; +} + +// sky +vec3 getSkyColor(vec3 e) { + e.y = (max(e.y,0.0)*0.8+0.2)*0.8; + return vec3(pow(1.0-e.y,2.0), 1.0-e.y, 0.6+(1.0-e.y)*0.4) * 1.1; +} + +// sea +float sea_octave(vec2 uv, float choppy) { + uv += noise(uv); + vec2 wv = 1.0-abs(sin(uv)); + vec2 swv = abs(cos(uv)); + wv = mix(wv,swv,wv); + return pow(1.0-pow(wv.x * wv.y,0.65),choppy); +} + +float map(vec3 p) { + float freq = SEA_FREQ; + float amp = SEA_HEIGHT; + float choppy = SEA_CHOPPY; + vec2 uv = p.xz; uv.x *= 0.75; + + float d, h = 0.0; + for(int i = 0; i < ITER_GEOMETRY; i++) { + d = sea_octave((uv+SEA_TIME)*freq,choppy); + d += sea_octave((uv-SEA_TIME)*freq,choppy); + h += d * amp; + uv *= octave_m; freq *= 1.9; amp *= 0.22; + choppy = mix(choppy,1.0,0.2); + } + return p.y - h; +} + +float map_detailed(vec3 p) { + float freq = SEA_FREQ; + float amp = SEA_HEIGHT; + float choppy = SEA_CHOPPY; + vec2 uv = p.xz; uv.x *= 0.75; + + float d, h = 0.0; + for(int i = 0; i < ITER_FRAGMENT; i++) { + d = sea_octave((uv+SEA_TIME)*freq,choppy); + d += sea_octave((uv-SEA_TIME)*freq,choppy); + h += d * amp; + uv *= octave_m; freq *= 1.9; amp *= 0.22; + choppy = mix(choppy,1.0,0.2); + } + return p.y - h; +} + +vec3 getSeaColor(vec3 p, vec3 n, vec3 l, vec3 eye, vec3 dist) { + float fresnel = clamp(1.0 - dot(n, -eye), 0.0, 1.0); + fresnel = min(fresnel * fresnel * fresnel, 0.5); + + vec3 reflected = getSkyColor(reflect(eye, n)); + vec3 refracted = SEA_BASE + diffuse(n, l, 80.0) * SEA_WATER_COLOR * 0.12; + + vec3 color = mix(refracted, reflected, fresnel); + + float atten = max(1.0 - dot(dist, dist) * 0.001, 0.0); + color += SEA_WATER_COLOR * (p.y - SEA_HEIGHT) * 0.18 * atten; + + color += specular(n, l, eye, 600.0 * inversesqrt(dot(dist,dist))); + + return color; +} + +// tracing +vec3 getNormal(vec3 p, float eps) { + vec3 n; + n.y = map_detailed(p); + n.x = map_detailed(vec3(p.x+eps,p.y,p.z)) - n.y; + n.z = map_detailed(vec3(p.x,p.y,p.z+eps)) - n.y; + n.y = eps; + return normalize(n); +} + +float heightMapTracing(vec3 ori, vec3 dir, out vec3 p) { + float tm = 0.0; + float tx = 1000.0; + float hx = map(ori + dir * tx); + if(hx > 0.0) { + p = ori + dir * tx; + return tx; + } + float hm = map(ori); + for(int i = 0; i < NUM_STEPS; i++) { + float tmid = mix(tm, tx, hm / (hm - hx)); + p = ori + dir * tmid; + float hmid = map(p); + if(hmid < 0.0) { + tx = tmid; + hx = hmid; + } else { + tm = tmid; + hm = hmid; + } + if(abs(hmid) < EPSILON) break; + } + return mix(tm, tx, hm / (hm - hx)); +} + +vec3 getPixel(in vec2 coord, float time) { + vec2 uv = coord / iResolution.xy; + uv = uv * 2.0 - 1.0; + uv.x *= iResolution.x / iResolution.y; + + // ray + vec3 ang = vec3(sin(time*3.0)*0.1,sin(time)*0.2+0.3,time); + vec3 ori = vec3(0.0,3.5,time*5.0); + vec3 dir = normalize(vec3(uv.xy,-2.0)); dir.z += length(uv) * 0.14; + dir = normalize(dir) * fromEuler(ang); + + // tracing + vec3 p; + heightMapTracing(ori,dir,p); + vec3 dist = p - ori; + vec3 n = getNormal(p, dot(dist,dist) * EPSILON_NRM); + vec3 light = normalize(vec3(0.0,1.0,0.8)); + + // color + return mix( + getSkyColor(dir), + getSeaColor(p,n,light,dir,dist), + pow(smoothstep(0.0,-0.02,dir.y),0.2)); +} + +// main +void mainImage( out vec4 fragColor, in vec2 fragCoord ) { + // Removed mouse interaction (iMouse not available) + float time = iTime * 0.3; + +#ifdef AA + vec3 color = vec3(0.0); + for(int i = -1; i <= 1; i++) { + for(int j = -1; j <= 1; j++) { + vec2 uv = fragCoord+vec2(i,j)/3.0; + color += getPixel(uv, time); + } + } + color /= 9.0; +#else + vec3 color = getPixel(fragCoord, time); +#endif + + // post + fragColor = vec4(pow(color,vec3(0.65)), 1.0); +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/test.frag.glsl b/shaders/test.frag.glsl index ff836f5..193961b 100644 --- a/shaders/test.frag.glsl +++ b/shaders/test.frag.glsl @@ -1,3 +1,5 @@ +// Name: Test +// Author: [TO_BE_COMPLETED] #version 330 core out vec4 FragColor; in vec2 vUV; diff --git a/shaders/voxel_descent.frag.glsl b/shaders/voxel_descent.frag.glsl new file mode 100644 index 0000000..53a3ee3 --- /dev/null +++ b/shaders/voxel_descent.frag.glsl @@ -0,0 +1,52 @@ +// Name: Voxel Descent +// Author: Jaenman +// URL: https://www.shadertoy.com/view/Wc3cRr +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +/*================================ += Voxel Descent = += Author: Jaenam = +================================*/ +// Date: 2025-11-14 +// License: Creative Commons (CC BY-NC-SA 4.0) + +// Thanks to @diatribes for the plasma :D + +void mainImage( out vec4 O, vec2 I ) +{ + float i=0.,d=0.,s=0.,t = iTime*.8; + vec3 p,q,z,k; + vec4 c = vec4(1,2,3,1); + mat2 R = mat2(cos(t/3.+vec4(0,33,11,0))); + + // Initialize O before loop (was O*=i where i=0, reading uninitialized out parameter) + O = vec4(0.0); + + for(; i++<1e2;O+=sin(.5*c+i*.2)/s) + + z = normalize(vec3(I+I, 0) - vec3(iResolution.xy, iResolution.y)), + z.xy*=R, + p = d*z+sqrt(i)*i, + k=z/=length(z.xy), + z.z-=t, p.z+=t, + + d+=s=.012+.07*abs(mix(cos(dot(sin(floor(p/8.)).yzx,cos(ceil(z/.1)))),length(k),.2)-i/1e2); + + O*= exp(-.02*d*d); + // Add epsilon protection to prevent division overflow + O=tanh(O*O*mix(length(abs(c/max(dot(cos(t+p),vec3(6)), 0.001))),length(k),.5)/6e4); + +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/shaders/water.frag.glsl b/shaders/water.frag.glsl new file mode 100644 index 0000000..285c29b --- /dev/null +++ b/shaders/water.frag.glsl @@ -0,0 +1,37 @@ +// Name: Water +// Author: diatribes +#version 330 core +precision highp float; + +out vec4 FragColor; +in vec2 vUV; +uniform vec2 iResolution; +uniform float iTime; + +/* + -2 by @FabriceNeyret2 + + thanks!! :D + + If it doesn't display correctly, change line 17 "r/r" to "vec3(1)" + +*/ + +void mainImage( out vec4 o, vec2 u ) { + float s=.002,i,n; + vec3 r = vec3(iResolution.xy, iResolution.x/iResolution.y); + vec3 p = vec3(0); + u = (u-r.xy/2.)/r.y-.3; + for(o *= i; i++ < 32. && s > .001;o += vec4(5,2,1,0)/max(length(u-.1), 0.001)) + for (p += vec3(u*s,s),s = 1. + p.y, + n =.01; n < 1.;n+=n) + s += abs(dot(sin(p.z+iTime+p / n), vec3(1))) * n*.1; + o = tanh(o/5e2); +} + +void main() { + vec2 fragCoordPixels = vUV * iResolution; + vec4 outColor; + mainImage(outColor, fragCoordPixels); + FragColor = outColor; +} diff --git a/src/main.cpp b/src/main.cpp index d32479d..f265839 100644 --- a/src/main.cpp +++ b/src/main.cpp @@ -7,8 +7,10 @@ #include #include #include +#include #include #include +#include "jail_audio.h" #include "defines.hpp" @@ -21,6 +23,7 @@ struct Logger { // Opciones mínimas parecidas a las tuyas struct VideoOptions { bool fullscreen = false; + bool vsync = true; } Options_video; // Estructura para guardar info del display @@ -35,13 +38,30 @@ struct DisplayMonitor { static DisplayMonitor display_monitor_; static SDL_Window* window_ = nullptr; -// Sistema de shaders +// Sistema de shaders (legacy - kept for backward compatibility with single-pass shaders) static std::vector shader_list_; +static std::vector shader_names_; // Custom names from "// Name: XXX" comments +static std::vector shader_authors_; // Custom authors from "// Author: XXX" comments static size_t current_shader_index_ = 0; static std::filesystem::path shaders_directory_; static GLuint current_program_ = 0; static Uint32 shader_start_ticks_ = 0; +// Multi-pass shader system +static std::vector shader_passes_; +static int current_window_width_ = 0; +static int current_window_height_ = 0; + +// FPS tracking +static Uint32 fps_frame_count_ = 0; +static Uint32 fps_last_update_ticks_ = 0; +static float current_fps_ = 0.0f; + +// Sistema de música +static std::vector music_list_; +static size_t current_music_index_ = 0; +static JA_Music_t* current_music_ = nullptr; + // Vertex shader embebido static const char* vertexShaderSrc = R"glsl( #version 330 core @@ -63,6 +83,89 @@ static bool loadFileToString(const std::filesystem::path& path, std::string& out return true; } +struct ShaderMetadata { + std::string name; + std::string author; + std::string iChannel0; // "BufferA", "BufferB", "none", etc. + std::string iChannel1; + std::string iChannel2; + std::string iChannel3; +}; + +struct ShaderBuffer { + GLuint program = 0; // Shader program for this buffer + GLuint fbo = 0; // Framebuffer object + GLuint texture = 0; // Output texture + std::string name; // "BufferA", "BufferB", etc. +}; + +struct ShaderPass { + std::string shaderName; // Base name (e.g., "water") + std::string displayName; // Custom name from metadata + std::string author; // Author from metadata + GLuint imageProgram = 0; // Main image shader program + std::vector buffers; // BufferA, BufferB, etc. + ShaderMetadata metadata; // iChannel configuration +}; + +static std::string trimString(const std::string& str) { + size_t start = str.find_first_not_of(" \t\r\n"); + size_t end = str.find_last_not_of(" \t\r\n"); + if (start != std::string::npos && end != std::string::npos) { + return str.substr(start, end - start + 1); + } + return ""; +} + +static ShaderMetadata extractShaderMetadata(const std::string& shaderSource) { + ShaderMetadata metadata; + metadata.iChannel0 = "none"; + metadata.iChannel1 = "none"; + metadata.iChannel2 = "none"; + metadata.iChannel3 = "none"; + + std::istringstream stream(shaderSource); + std::string line; + int lineCount = 0; + const int maxLinesToCheck = 30; + + while (std::getline(stream, line) && lineCount < maxLinesToCheck) { + lineCount++; + + // Look for "// XXX: YYY" patterns (case-insensitive) + size_t pos = line.find("//"); + if (pos != std::string::npos) { + std::string comment = line.substr(pos + 2); + std::string commentLower = comment; + std::transform(commentLower.begin(), commentLower.end(), commentLower.begin(), ::tolower); + + // Check for Name: + if (commentLower.find("name:") != std::string::npos) { + metadata.name = trimString(comment.substr(comment.find(":") + 1)); + } + // Check for Author: + else if (commentLower.find("author:") != std::string::npos) { + metadata.author = trimString(comment.substr(comment.find(":") + 1)); + } + // Check for iChannel0-3: + else if (commentLower.find("ichannel0:") != std::string::npos) { + metadata.iChannel0 = trimString(comment.substr(comment.find(":") + 1)); + } + else if (commentLower.find("ichannel1:") != std::string::npos) { + metadata.iChannel1 = trimString(comment.substr(comment.find(":") + 1)); + } + else if (commentLower.find("ichannel2:") != std::string::npos) { + metadata.iChannel2 = trimString(comment.substr(comment.find(":") + 1)); + } + else if (commentLower.find("ichannel3:") != std::string::npos) { + metadata.iChannel3 = trimString(comment.substr(comment.find(":") + 1)); + } + } + } + + return metadata; +} + static std::vector scanShaderDirectory(const std::filesystem::path& directory) { std::vector shaders; @@ -84,14 +187,168 @@ static std::vector scanShaderDirectory(const std::filesys std::sort(shaders.begin(), shaders.end()); Logger::info("Found " + std::to_string(shaders.size()) + " shader(s) in " + directory.string()); + + // Initialize shader metadata vectors with empty strings (will be filled when shaders are loaded) + shader_names_.resize(shaders.size(), ""); + shader_authors_.resize(shaders.size(), ""); + return shaders; } +static std::vector scanMusicDirectory(const std::filesystem::path& directory) { + std::vector music_files; + + if (!std::filesystem::exists(directory) || !std::filesystem::is_directory(directory)) { + Logger::info("Music directory does not exist: " + directory.string()); + return music_files; + } + + for (const auto& entry : std::filesystem::directory_iterator(directory)) { + if (entry.is_regular_file()) { + auto ext = entry.path().extension().string(); + if (ext == ".ogg") { + music_files.push_back(entry.path()); + } + } + } + + // Ordenar alfabéticamente + std::sort(music_files.begin(), music_files.end()); + + Logger::info("Found " + std::to_string(music_files.size()) + " music file(s) in " + directory.string()); + return music_files; +} + +static void playRandomMusic() { + if (music_list_.empty()) return; + + // Liberar música anterior si existe + if (current_music_) { + JA_DeleteMusic(current_music_); + current_music_ = nullptr; + } + + // Elegir índice aleatorio + current_music_index_ = rand() % music_list_.size(); + + // Cargar y reproducir música (sin loop, loop=0) + const auto& music_path = music_list_[current_music_index_]; + current_music_ = JA_LoadMusic(music_path.string().c_str()); + + if (current_music_) { + JA_PlayMusic(current_music_, 0); // 0 = no loop, se reproduce una vez + Logger::info("Now playing: " + music_path.filename().string()); + } else { + Logger::error("Failed to load music: " + music_path.string()); + } +} + +// ===== Multi-pass FBO/Texture Management ===== + +static bool createBufferFBO(ShaderBuffer& buffer, int width, int height) { + // Create texture + glGenTextures(1, &buffer.texture); + glBindTexture(GL_TEXTURE_2D, buffer.texture); + glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA32F, width, height, 0, GL_RGBA, GL_FLOAT, nullptr); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + glBindTexture(GL_TEXTURE_2D, 0); + + // Create FBO + glGenFramebuffers(1, &buffer.fbo); + glBindFramebuffer(GL_FRAMEBUFFER, buffer.fbo); + glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, buffer.texture, 0); + + // Check FBO completeness + GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER); + glBindFramebuffer(GL_FRAMEBUFFER, 0); + + if (status != GL_FRAMEBUFFER_COMPLETE) { + Logger::error("FBO creation failed for " + buffer.name + ": " + std::to_string(status)); + return false; + } + + Logger::info("Created FBO for " + buffer.name + " (" + std::to_string(width) + "x" + std::to_string(height) + ")"); + return true; +} + +static void destroyBuffer(ShaderBuffer& buffer) { + if (buffer.fbo != 0) { + glDeleteFramebuffers(1, &buffer.fbo); + buffer.fbo = 0; + } + if (buffer.texture != 0) { + glDeleteTextures(1, &buffer.texture); + buffer.texture = 0; + } + if (buffer.program != 0) { + glDeleteProgram(buffer.program); + buffer.program = 0; + } +} + +static void destroyShaderPass(ShaderPass& pass) { + if (pass.imageProgram != 0) { + glDeleteProgram(pass.imageProgram); + pass.imageProgram = 0; + } + for (auto& buffer : pass.buffers) { + destroyBuffer(buffer); + } + pass.buffers.clear(); +} + +static bool resizeBuffersIfNeeded(ShaderPass& pass, int width, int height) { + if (current_window_width_ == width && current_window_height_ == height) { + return false; // No resize needed + } + + Logger::info("Resizing buffers: " + std::to_string(width) + "x" + std::to_string(height)); + + // Destroy and recreate all buffers with new size + for (auto& buffer : pass.buffers) { + // Keep program, destroy FBO/texture only + if (buffer.fbo != 0) glDeleteFramebuffers(1, &buffer.fbo); + if (buffer.texture != 0) glDeleteTextures(1, &buffer.texture); + buffer.fbo = 0; + buffer.texture = 0; + + if (!createBufferFBO(buffer, width, height)) { + return false; + } + } + + current_window_width_ = width; + current_window_height_ = height; + return true; +} + static void updateWindowTitle() { if (!window_ || shader_list_.empty()) return; - std::string filename = shader_list_[current_shader_index_].filename().string(); - std::string title = std::string(APP_NAME) + " (" + filename + ")"; + // Use custom shader name if available, otherwise fallback to filename + std::string shaderName; + if (!shader_names_.empty() && !shader_names_[current_shader_index_].empty()) { + shaderName = shader_names_[current_shader_index_]; + } else { + shaderName = shader_list_[current_shader_index_].filename().string(); + } + + // Add author if available + if (!shader_authors_.empty() && !shader_authors_[current_shader_index_].empty()) { + shaderName += " by " + shader_authors_[current_shader_index_]; + } + + std::string title = std::string(APP_NAME) + " (" + shaderName + ")"; + + if (current_fps_ > 0.0f) { + title += " - " + std::to_string(static_cast(current_fps_ + 0.5f)) + " FPS"; + } + + title += Options_video.vsync ? " [VSync ON]" : " [VSync OFF]"; + SDL_SetWindowTitle(window_, title.c_str()); } @@ -147,6 +404,17 @@ static GLuint loadAndCompileShader(size_t index) { return 0; } + // Extract custom shader metadata (name and author) from source code + ShaderMetadata metadata = extractShaderMetadata(fragSrc); + if (!metadata.name.empty()) { + shader_names_[index] = metadata.name; + Logger::info("Shader name: " + metadata.name); + } + if (!metadata.author.empty()) { + shader_authors_[index] = metadata.author; + Logger::info("Shader author: " + metadata.author); + } + GLuint vs = compileShader(GL_VERTEX_SHADER, vertexShaderSrc); GLuint fs = compileShader(GL_FRAGMENT_SHADER, fragSrc.c_str()); @@ -208,11 +476,15 @@ void setFullscreenMode() { SDL_SetWindowFullscreen(window_, false); SDL_SetWindowSize(window_, WINDOW_WIDTH, WINDOW_HEIGHT); Options_video.fullscreen = false; + SDL_ShowCursor(); // Show cursor on fallback to windowed + } else { + SDL_HideCursor(); // Hide cursor in fullscreen } } else { // Volver a modo ventana 800x800 SDL_SetWindowFullscreen(window_, false); SDL_SetWindowSize(window_, WINDOW_WIDTH, WINDOW_HEIGHT); + SDL_ShowCursor(); // Show cursor in windowed mode } } @@ -221,6 +493,17 @@ void toggleFullscreen() { setFullscreenMode(); } +void toggleVSync() { + Options_video.vsync = !Options_video.vsync; + int result = SDL_GL_SetSwapInterval(Options_video.vsync ? 1 : 0); + + if (result == 0) { + Logger::info(Options_video.vsync ? "VSync enabled" : "VSync disabled"); + } else { + Logger::error(std::string("Failed to set VSync: ") + SDL_GetError()); + } +} + void switchShader(int direction) { if (shader_list_.empty()) return; @@ -260,6 +543,10 @@ void handleDebugEvents(const SDL_Event& event) { toggleFullscreen(); break; } + case SDLK_F4: { + toggleVSync(); + break; + } case SDLK_LEFT: { switchShader(-1); break; @@ -287,7 +574,7 @@ int main(int argc, char** argv) { Options_video.fullscreen = fullscreenFlag; // Inicializar SDL3 - auto initResult = SDL_Init(SDL_INIT_VIDEO); + auto initResult = SDL_Init(SDL_INIT_VIDEO | SDL_INIT_AUDIO); if constexpr (std::is_same_v) { if (!initResult) { Logger::error(SDL_GetError()); return -1; } } else { @@ -326,9 +613,34 @@ int main(int argc, char** argv) { return -1; } + // Set initial vsync state + int vsync_result = SDL_GL_SetSwapInterval(Options_video.vsync ? 1 : 0); + if (vsync_result == 0) { + Logger::info(Options_video.vsync ? "VSync enabled" : "VSync disabled"); + } else { + Logger::error(std::string("Failed to set initial VSync: ") + SDL_GetError()); + } + + // Inicializar jail_audio + JA_Init(48000, SDL_AUDIO_S16LE, 2); + // Obtener directorio de recursos std::string resources_dir = getResourcesDirectory(); + // Inicializar generador de números aleatorios + srand(static_cast(time(nullptr))); + + // Escanear directorio de música + std::filesystem::path music_directory = std::filesystem::path(resources_dir) / "data" / "music"; + music_list_ = scanMusicDirectory(music_directory); + + // Reproducir primera canción aleatoria + if (!music_list_.empty()) { + playRandomMusic(); + } else { + Logger::info("No music files found in " + music_directory.string()); + } + // Determinar carpeta de shaders std::filesystem::path shaderFile(shaderPath); if (shaderFile.has_parent_path()) { @@ -391,6 +703,7 @@ int main(int argc, char** argv) { } shader_start_ticks_ = SDL_GetTicks(); + fps_last_update_ticks_ = SDL_GetTicks(); updateWindowTitle(); // Quad setup @@ -415,6 +728,27 @@ int main(int argc, char** argv) { bool running = true; while (running) { + // Update FPS counter + fps_frame_count_++; + Uint32 current_ticks = SDL_GetTicks(); + + // Update FPS display every 500ms + if (current_ticks - fps_last_update_ticks_ >= 500) { + float elapsed_seconds = (current_ticks - fps_last_update_ticks_) / 1000.0f; + current_fps_ = fps_frame_count_ / elapsed_seconds; + fps_frame_count_ = 0; + fps_last_update_ticks_ = current_ticks; + updateWindowTitle(); + } + + // Actualizar audio (necesario para streaming y loops) + JA_Update(); + + // Verificar si la música actual terminó y reproducir siguiente aleatoria + if (!music_list_.empty() && JA_GetMusicState() == JA_MUSIC_STOPPED) { + playRandomMusic(); + } + SDL_Event e; while (SDL_PollEvent(&e)) { if (e.type == SDL_EVENT_QUIT) running = false; @@ -452,7 +786,9 @@ int main(int argc, char** argv) { glBindVertexArray(0); SDL_GL_SwapWindow(window_); - SDL_Delay(1); + if (!Options_video.vsync) { + SDL_Delay(1); // Prevent CPU spinning when vsync is off + } } // Cleanup @@ -462,6 +798,13 @@ int main(int argc, char** argv) { glDeleteProgram(current_program_); } + // Cleanup audio + if (current_music_) { + JA_DeleteMusic(current_music_); + current_music_ = nullptr; + } + JA_Quit(); + SDL_GL_DestroyContext(glContext); SDL_DestroyWindow(window_); SDL_Quit(); diff --git a/third_party/jail_audio.cpp b/third_party/jail_audio.cpp new file mode 100644 index 0000000..e8c2c3d --- /dev/null +++ b/third_party/jail_audio.cpp @@ -0,0 +1,477 @@ +#ifndef JA_USESDLMIXER +#include "jail_audio.h" + +#include // Para SDL_AudioFormat, SDL_BindAudioStream, SDL_SetAudioStreamGain, SDL_PutAudioStreamData, SDL_DestroyAudioStream, SDL_GetAudioStreamAvailable, Uint8, SDL_CreateAudioStream, SDL_UnbindAudioStream, Uint32, SDL_CloseAudioDevice, SDL_GetTicks, SDL_Log, SDL_free, SDL_AudioSpec, SDL_AudioStream, SDL_IOFromMem, SDL_LoadWAV, SDL_LoadWAV_IO, SDL_OpenAudioDevice, SDL_clamp, SDL_malloc, SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK, SDL_AudioDeviceID, SDL_memcpy +#include // Para uint32_t, uint8_t +#include // Para NULL, fseek, printf, fclose, fopen, fread, ftell, FILE, SEEK_END, SEEK_SET +#include // Para free, malloc +#include // Para strcpy, strlen + +#include "stb_vorbis.h" // Para stb_vorbis_decode_memory + +#define JA_MAX_SIMULTANEOUS_CHANNELS 20 +#define JA_MAX_GROUPS 2 + +struct JA_Sound_t +{ + SDL_AudioSpec spec { SDL_AUDIO_S16, 2, 48000 }; + Uint32 length { 0 }; + Uint8 *buffer { NULL }; +}; + +struct JA_Channel_t +{ + JA_Sound_t *sound { nullptr }; + int pos { 0 }; + int times { 0 }; + int group { 0 }; + SDL_AudioStream *stream { nullptr }; + JA_Channel_state state { JA_CHANNEL_FREE }; +}; + +struct JA_Music_t +{ + SDL_AudioSpec spec { SDL_AUDIO_S16, 2, 48000 }; + Uint32 length { 0 }; + Uint8 *buffer { nullptr }; + char *filename { nullptr }; + + int pos { 0 }; + int times { 0 }; + SDL_AudioStream *stream { nullptr }; + JA_Music_state state { JA_MUSIC_INVALID }; +}; + +JA_Music_t *current_music { nullptr }; +JA_Channel_t channels[JA_MAX_SIMULTANEOUS_CHANNELS]; + +SDL_AudioSpec JA_audioSpec { SDL_AUDIO_S16, 2, 48000 }; +float JA_musicVolume { 1.0f }; +float JA_soundVolume[JA_MAX_GROUPS]; +bool JA_musicEnabled { true }; +bool JA_soundEnabled { true }; +SDL_AudioDeviceID sdlAudioDevice { 0 }; +//SDL_TimerID JA_timerID { 0 }; + +bool fading = false; +int fade_start_time; +int fade_duration; +int fade_initial_volume; + + +void JA_Update() +{ + if (JA_musicEnabled && current_music && current_music->state == JA_MUSIC_PLAYING) + { + if (fading) { + int time = SDL_GetTicks(); + if (time > (fade_start_time+fade_duration)) { + fading = false; + JA_StopMusic(); + return; + } else { + const int time_passed = time - fade_start_time; + const float percent = (float)time_passed / (float)fade_duration; + SDL_SetAudioStreamGain(current_music->stream, JA_musicVolume*(1.0 - percent)); + } + } + + if (current_music->times != 0) + { + if ((Uint32)SDL_GetAudioStreamAvailable(current_music->stream) < (current_music->length/2)) { + SDL_PutAudioStreamData(current_music->stream, current_music->buffer, current_music->length); + } + if (current_music->times>0) current_music->times--; + } + else + { + if (SDL_GetAudioStreamAvailable(current_music->stream) == 0) JA_StopMusic(); + } + } + + if (JA_soundEnabled) + { + for (int i=0; i < JA_MAX_SIMULTANEOUS_CHANNELS; ++i) + if (channels[i].state == JA_CHANNEL_PLAYING) + { + if (channels[i].times != 0) + { + if ((Uint32)SDL_GetAudioStreamAvailable(channels[i].stream) < (channels[i].sound->length/2)) { + SDL_PutAudioStreamData(channels[i].stream, channels[i].sound->buffer, channels[i].sound->length); + if (channels[i].times>0) channels[i].times--; + } + } + else + { + if (SDL_GetAudioStreamAvailable(channels[i].stream) == 0) JA_StopChannel(i); + } + } + + } + + return; +} + +void JA_Init(const int freq, const SDL_AudioFormat format, const int num_channels) +{ + #ifdef DEBUG + SDL_SetLogPriority(SDL_LOG_CATEGORY_APPLICATION, SDL_LOG_PRIORITY_DEBUG); + #endif + + JA_audioSpec = {format, num_channels, freq }; + if (!sdlAudioDevice) SDL_CloseAudioDevice(sdlAudioDevice); + sdlAudioDevice = SDL_OpenAudioDevice(SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK, &JA_audioSpec); + if (sdlAudioDevice==0) SDL_Log("Failed to initialize SDL audio!"); + for (int i=0; ilength = stb_vorbis_decode_memory(buffer, length, &chan, &samplerate, &output) * chan * 2; + + music->spec.channels = chan; + music->spec.freq = samplerate; + music->spec.format = SDL_AUDIO_S16; + music->buffer = (Uint8*)SDL_malloc(music->length); + SDL_memcpy(music->buffer, output, music->length); + free(output); + music->pos = 0; + music->state = JA_MUSIC_STOPPED; + + return music; +} + +JA_Music_t *JA_LoadMusic(const char* filename) +{ + // [RZC 28/08/22] Carreguem primer el arxiu en memòria i després el descomprimim. Es algo més rapid. + FILE *f = fopen(filename, "rb"); + fseek(f, 0, SEEK_END); + long fsize = ftell(f); + fseek(f, 0, SEEK_SET); + Uint8 *buffer = (Uint8*)malloc(fsize + 1); + if (fread(buffer, fsize, 1, f)!=1) return NULL; + fclose(f); + + JA_Music_t *music = JA_LoadMusic(buffer, fsize); + music->filename = (char*)malloc(strlen(filename)+1); + strcpy(music->filename, filename); + + free(buffer); + + return music; +} + +void JA_PlayMusic(JA_Music_t *music, const int loop) +{ + if (!JA_musicEnabled) return; + + JA_StopMusic(); + + current_music = music; + current_music->pos = 0; + current_music->state = JA_MUSIC_PLAYING; + current_music->times = loop; + + current_music->stream = SDL_CreateAudioStream(¤t_music->spec, &JA_audioSpec); + if (!SDL_PutAudioStreamData(current_music->stream, current_music->buffer, current_music->length)) printf("[ERROR] SDL_PutAudioStreamData failed!\n"); + SDL_SetAudioStreamGain(current_music->stream, JA_musicVolume); + if (!SDL_BindAudioStream(sdlAudioDevice, current_music->stream)) printf("[ERROR] SDL_BindAudioStream failed!\n"); + //SDL_ResumeAudioStreamDevice(current_music->stream); +} + +char *JA_GetMusicFilename(JA_Music_t *music) +{ + if (!music) music = current_music; + return music->filename; +} + +void JA_PauseMusic() +{ + if (!JA_musicEnabled) return; + if (!current_music || current_music->state == JA_MUSIC_INVALID) return; + + current_music->state = JA_MUSIC_PAUSED; + //SDL_PauseAudioStreamDevice(current_music->stream); + SDL_UnbindAudioStream(current_music->stream); +} + +void JA_ResumeMusic() +{ + if (!JA_musicEnabled) return; + if (!current_music || current_music->state == JA_MUSIC_INVALID) return; + + current_music->state = JA_MUSIC_PLAYING; + //SDL_ResumeAudioStreamDevice(current_music->stream); + SDL_BindAudioStream(sdlAudioDevice, current_music->stream); +} + +void JA_StopMusic() +{ + if (!JA_musicEnabled) return; + if (!current_music || current_music->state == JA_MUSIC_INVALID) return; + + current_music->pos = 0; + current_music->state = JA_MUSIC_STOPPED; + //SDL_PauseAudioStreamDevice(current_music->stream); + SDL_DestroyAudioStream(current_music->stream); + current_music->stream = nullptr; + free(current_music->filename); + current_music->filename = nullptr; +} + +void JA_FadeOutMusic(const int milliseconds) +{ + if (!JA_musicEnabled) return; + if (current_music == NULL || current_music->state == JA_MUSIC_INVALID) return; + + fading = true; + fade_start_time = SDL_GetTicks(); + fade_duration = milliseconds; + fade_initial_volume = JA_musicVolume; +} + +JA_Music_state JA_GetMusicState() +{ + if (!JA_musicEnabled) return JA_MUSIC_DISABLED; + if (!current_music) return JA_MUSIC_INVALID; + + return current_music->state; +} + +void JA_DeleteMusic(JA_Music_t *music) +{ + if (current_music == music) current_music = nullptr; + SDL_free(music->buffer); + if (music->stream) SDL_DestroyAudioStream(music->stream); + delete music; +} + +float JA_SetMusicVolume(float volume) +{ + JA_musicVolume = SDL_clamp( volume, 0.0f, 1.0f ); + if (current_music) SDL_SetAudioStreamGain(current_music->stream, JA_musicVolume); + return JA_musicVolume; +} + +void JA_SetMusicPosition(float value) +{ + if (!current_music) return; + current_music->pos = value * current_music->spec.freq; +} + +float JA_GetMusicPosition() +{ + if (!current_music) return 0; + return float(current_music->pos)/float(current_music->spec.freq); +} + +void JA_EnableMusic(const bool value) +{ + if ( !value && current_music && (current_music->state==JA_MUSIC_PLAYING) ) JA_StopMusic(); + + JA_musicEnabled = value; +} + + + + + +JA_Sound_t *JA_NewSound(Uint8* buffer, Uint32 length) +{ + JA_Sound_t *sound = new JA_Sound_t(); + sound->buffer = buffer; + sound->length = length; + return sound; +} + +JA_Sound_t *JA_LoadSound(uint8_t* buffer, uint32_t size) +{ + JA_Sound_t *sound = new JA_Sound_t(); + SDL_LoadWAV_IO(SDL_IOFromMem(buffer, size),1, &sound->spec, &sound->buffer, &sound->length); + + return sound; +} + +JA_Sound_t *JA_LoadSound(const char* filename) +{ + JA_Sound_t *sound = new JA_Sound_t(); + SDL_LoadWAV(filename, &sound->spec, &sound->buffer, &sound->length); + + return sound; +} + +int JA_PlaySound(JA_Sound_t *sound, const int loop, const int group) +{ + if (!JA_soundEnabled) return -1; + + int channel = 0; + while (channel < JA_MAX_SIMULTANEOUS_CHANNELS && channels[channel].state != JA_CHANNEL_FREE) { channel++; } + if (channel == JA_MAX_SIMULTANEOUS_CHANNELS) channel = 0; + JA_StopChannel(channel); + + channels[channel].sound = sound; + channels[channel].times = loop; + channels[channel].pos = 0; + channels[channel].state = JA_CHANNEL_PLAYING; + channels[channel].stream = SDL_CreateAudioStream(&channels[channel].sound->spec, &JA_audioSpec); + SDL_PutAudioStreamData(channels[channel].stream, channels[channel].sound->buffer, channels[channel].sound->length); + SDL_SetAudioStreamGain(channels[channel].stream, JA_soundVolume[group]); + SDL_BindAudioStream(sdlAudioDevice, channels[channel].stream); + + return channel; +} + +int JA_PlaySoundOnChannel(JA_Sound_t *sound, const int channel, const int loop, const int group) +{ + if (!JA_soundEnabled) return -1; + + if (channel < 0 || channel >= JA_MAX_SIMULTANEOUS_CHANNELS) return -1; + JA_StopChannel(channel); + + channels[channel].sound = sound; + channels[channel].times = loop; + channels[channel].pos = 0; + channels[channel].state = JA_CHANNEL_PLAYING; + channels[channel].stream = SDL_CreateAudioStream(&channels[channel].sound->spec, &JA_audioSpec); + SDL_PutAudioStreamData(channels[channel].stream, channels[channel].sound->buffer, channels[channel].sound->length); + SDL_SetAudioStreamGain(channels[channel].stream, JA_soundVolume[group]); + SDL_BindAudioStream(sdlAudioDevice, channels[channel].stream); + + return channel; +} + +void JA_DeleteSound(JA_Sound_t *sound) +{ + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) { + if (channels[i].sound == sound) JA_StopChannel(i); + } + SDL_free(sound->buffer); + delete sound; +} + +void JA_PauseChannel(const int channel) +{ + if (!JA_soundEnabled) return; + + if (channel == -1) + { + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) + if (channels[i].state == JA_CHANNEL_PLAYING) + { + channels[i].state = JA_CHANNEL_PAUSED; + //SDL_PauseAudioStreamDevice(channels[i].stream); + SDL_UnbindAudioStream(channels[i].stream); + } + } + else if (channel >= 0 && channel < JA_MAX_SIMULTANEOUS_CHANNELS) + { + if (channels[channel].state == JA_CHANNEL_PLAYING) + { + channels[channel].state = JA_CHANNEL_PAUSED; + //SDL_PauseAudioStreamDevice(channels[channel].stream); + SDL_UnbindAudioStream(channels[channel].stream); + } + } +} + +void JA_ResumeChannel(const int channel) +{ + if (!JA_soundEnabled) return; + + if (channel == -1) + { + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) + if (channels[i].state == JA_CHANNEL_PAUSED) + { + channels[i].state = JA_CHANNEL_PLAYING; + //SDL_ResumeAudioStreamDevice(channels[i].stream); + SDL_BindAudioStream(sdlAudioDevice, channels[i].stream); + } + } + else if (channel >= 0 && channel < JA_MAX_SIMULTANEOUS_CHANNELS) + { + if (channels[channel].state == JA_CHANNEL_PAUSED) + { + channels[channel].state = JA_CHANNEL_PLAYING; + //SDL_ResumeAudioStreamDevice(channels[channel].stream); + SDL_BindAudioStream(sdlAudioDevice, channels[channel].stream); + } + } +} + +void JA_StopChannel(const int channel) +{ + if (!JA_soundEnabled) return; + + if (channel == -1) + { + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) { + if (channels[i].state != JA_CHANNEL_FREE) SDL_DestroyAudioStream(channels[i].stream); + channels[i].stream = nullptr; + channels[i].state = JA_CHANNEL_FREE; + channels[i].pos = 0; + channels[i].sound = NULL; + } + } + else if (channel >= 0 && channel < JA_MAX_SIMULTANEOUS_CHANNELS) + { + if (channels[channel].state != JA_CHANNEL_FREE) SDL_DestroyAudioStream(channels[channel].stream); + channels[channel].stream = nullptr; + channels[channel].state = JA_CHANNEL_FREE; + channels[channel].pos = 0; + channels[channel].sound = NULL; + } +} + +JA_Channel_state JA_GetChannelState(const int channel) +{ + if (!JA_soundEnabled) return JA_SOUND_DISABLED; + + if (channel < 0 || channel >= JA_MAX_SIMULTANEOUS_CHANNELS) return JA_CHANNEL_INVALID; + + return channels[channel].state; +} + +float JA_SetSoundVolume(float volume, const int group) +{ + const float v = SDL_clamp( volume, 0.0f, 1.0f ); + for (int i = 0; i < JA_MAX_GROUPS; ++i) { + if (group==-1 || group==i) JA_soundVolume[i]=v; + } + + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) + if ( ((channels[i].state == JA_CHANNEL_PLAYING) || (channels[i].state == JA_CHANNEL_PAUSED)) && + ((group==-1) || (channels[i].group==group)) ) + SDL_SetAudioStreamGain(channels[i].stream, JA_soundVolume[i]); + + return v; +} + +void JA_EnableSound(const bool value) +{ + for (int i = 0; i < JA_MAX_SIMULTANEOUS_CHANNELS; i++) + { + if (channels[i].state == JA_CHANNEL_PLAYING) JA_StopChannel(i); + } + JA_soundEnabled = value; +} + +float JA_SetVolume(float volume) +{ + JA_SetSoundVolume(JA_SetMusicVolume(volume) / 2.0f); + + return JA_musicVolume; +} + +#endif \ No newline at end of file diff --git a/third_party/jail_audio.h b/third_party/jail_audio.h new file mode 100644 index 0000000..716b7f9 --- /dev/null +++ b/third_party/jail_audio.h @@ -0,0 +1,43 @@ +#pragma once +#include + +enum JA_Channel_state { JA_CHANNEL_INVALID, JA_CHANNEL_FREE, JA_CHANNEL_PLAYING, JA_CHANNEL_PAUSED, JA_SOUND_DISABLED }; +enum JA_Music_state { JA_MUSIC_INVALID, JA_MUSIC_PLAYING, JA_MUSIC_PAUSED, JA_MUSIC_STOPPED, JA_MUSIC_DISABLED }; + +struct JA_Sound_t; +struct JA_Music_t; + +void JA_Update(); + +void JA_Init(const int freq, const SDL_AudioFormat format, const int num_channels); +void JA_Quit(); + +JA_Music_t *JA_LoadMusic(const char* filename); +JA_Music_t *JA_LoadMusic(Uint8* buffer, Uint32 length); +void JA_PlayMusic(JA_Music_t *music, const int loop = -1); +char *JA_GetMusicFilename(JA_Music_t *music = nullptr); +void JA_PauseMusic(); +void JA_ResumeMusic(); +void JA_StopMusic(); +void JA_FadeOutMusic(const int milliseconds); +JA_Music_state JA_GetMusicState(); +void JA_DeleteMusic(JA_Music_t *music); +float JA_SetMusicVolume(float volume); +void JA_SetMusicPosition(float value); +float JA_GetMusicPosition(); +void JA_EnableMusic(const bool value); + +JA_Sound_t *JA_NewSound(Uint8* buffer, Uint32 length); +JA_Sound_t *JA_LoadSound(Uint8* buffer, Uint32 length); +JA_Sound_t *JA_LoadSound(const char* filename); +int JA_PlaySound(JA_Sound_t *sound, const int loop = 0, const int group=0); +int JA_PlaySoundOnChannel(JA_Sound_t *sound, const int channel, const int loop = 0, const int group=0); +void JA_PauseChannel(const int channel); +void JA_ResumeChannel(const int channel); +void JA_StopChannel(const int channel); +JA_Channel_state JA_GetChannelState(const int channel); +void JA_DeleteSound(JA_Sound_t *sound); +float JA_SetSoundVolume(float volume, const int group=0); +void JA_EnableSound(const bool value); + +float JA_SetVolume(float volume); diff --git a/third_party/stb_vorbis.h b/third_party/stb_vorbis.h new file mode 100644 index 0000000..ee203aa --- /dev/null +++ b/third_party/stb_vorbis.h @@ -0,0 +1,5631 @@ +// Ogg Vorbis audio decoder - v1.20 - public domain +// http://nothings.org/stb_vorbis/ +// +// Original version written by Sean Barrett in 2007. +// +// Originally sponsored by RAD Game Tools. Seeking implementation +// sponsored by Phillip Bennefall, Marc Andersen, Aaron Baker, +// Elias Software, Aras Pranckevicius, and Sean Barrett. +// +// LICENSE +// +// See end of file for license information. +// +// Limitations: +// +// - floor 0 not supported (used in old ogg vorbis files pre-2004) +// - lossless sample-truncation at beginning ignored +// - cannot concatenate multiple vorbis streams +// - sample positions are 32-bit, limiting seekable 192Khz +// files to around 6 hours (Ogg supports 64-bit) +// +// Feature contributors: +// Dougall Johnson (sample-exact seeking) +// +// Bugfix/warning contributors: +// Terje Mathisen Niklas Frykholm Andy Hill +// Casey Muratori John Bolton Gargaj +// Laurent Gomila Marc LeBlanc Ronny Chevalier +// Bernhard Wodo Evan Balster github:alxprd +// Tom Beaumont Ingo Leitgeb Nicolas Guillemot +// Phillip Bennefall Rohit Thiago Goulart +// github:manxorist saga musix github:infatum +// Timur Gagiev Maxwell Koo Peter Waller +// github:audinowho Dougall Johnson David Reid +// github:Clownacy Pedro J. Estebanez Remi Verschelde +// +// Partial history: +// 1.20 - 2020-07-11 - several small fixes +// 1.19 - 2020-02-05 - warnings +// 1.18 - 2020-02-02 - fix seek bugs; parse header comments; misc warnings etc. +// 1.17 - 2019-07-08 - fix CVE-2019-13217..CVE-2019-13223 (by ForAllSecure) +// 1.16 - 2019-03-04 - fix warnings +// 1.15 - 2019-02-07 - explicit failure if Ogg Skeleton data is found +// 1.14 - 2018-02-11 - delete bogus dealloca usage +// 1.13 - 2018-01-29 - fix truncation of last frame (hopefully) +// 1.12 - 2017-11-21 - limit residue begin/end to blocksize/2 to avoid large temp allocs in bad/corrupt files +// 1.11 - 2017-07-23 - fix MinGW compilation +// 1.10 - 2017-03-03 - more robust seeking; fix negative ilog(); clear error in open_memory +// 1.09 - 2016-04-04 - back out 'truncation of last frame' fix from previous version +// 1.08 - 2016-04-02 - warnings; setup memory leaks; truncation of last frame +// 1.07 - 2015-01-16 - fixes for crashes on invalid files; warning fixes; const +// 1.06 - 2015-08-31 - full, correct support for seeking API (Dougall Johnson) +// some crash fixes when out of memory or with corrupt files +// fix some inappropriately signed shifts +// 1.05 - 2015-04-19 - don't define __forceinline if it's redundant +// 1.04 - 2014-08-27 - fix missing const-correct case in API +// 1.03 - 2014-08-07 - warning fixes +// 1.02 - 2014-07-09 - declare qsort comparison as explicitly _cdecl in Windows +// 1.01 - 2014-06-18 - fix stb_vorbis_get_samples_float (interleaved was correct) +// 1.0 - 2014-05-26 - fix memory leaks; fix warnings; fix bugs in >2-channel; +// (API change) report sample rate for decode-full-file funcs +// +// See end of file for full version history. + +////////////////////////////////////////////////////////////////////////////// +// +// HEADER BEGINS HERE +// + +#ifndef STB_VORBIS_INCLUDE_STB_VORBIS_H +#define STB_VORBIS_INCLUDE_STB_VORBIS_H + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) +#define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +/////////// THREAD SAFETY + +// Individual stb_vorbis* handles are not thread-safe; you cannot decode from +// them from multiple threads at the same time. However, you can have multiple +// stb_vorbis* handles and decode from them independently in multiple thrads. + +/////////// MEMORY ALLOCATION + +// normally stb_vorbis uses malloc() to allocate memory at startup, +// and alloca() to allocate temporary memory during a frame on the +// stack. (Memory consumption will depend on the amount of setup +// data in the file and how you set the compile flags for speed +// vs. size. In my test files the maximal-size usage is ~150KB.) +// +// You can modify the wrapper functions in the source (setup_malloc, +// setup_temp_malloc, temp_malloc) to change this behavior, or you +// can use a simpler allocation model: you pass in a buffer from +// which stb_vorbis will allocate _all_ its memory (including the +// temp memory). "open" may fail with a VORBIS_outofmem if you +// do not pass in enough data; there is no way to determine how +// much you do need except to succeed (at which point you can +// query get_info to find the exact amount required. yes I know +// this is lame). +// +// If you pass in a non-NULL buffer of the type below, allocation +// will occur from it as described above. Otherwise just pass NULL +// to use malloc()/alloca() + +typedef struct +{ + char *alloc_buffer; + int alloc_buffer_length_in_bytes; +} stb_vorbis_alloc; + +/////////// FUNCTIONS USEABLE WITH ALL INPUT MODES + +typedef struct stb_vorbis stb_vorbis; + +typedef struct +{ + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int setup_temp_memory_required; + unsigned int temp_memory_required; + + int max_frame_size; +} stb_vorbis_info; + +typedef struct +{ + char *vendor; + + int comment_list_length; + char **comment_list; +} stb_vorbis_comment; + +// get general information about the file +extern stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f); + +// get ogg comments +extern stb_vorbis_comment stb_vorbis_get_comment(stb_vorbis *f); + +// get the last error detected (clears it, too) +extern int stb_vorbis_get_error(stb_vorbis *f); + +// close an ogg vorbis file and free all memory in use +extern void stb_vorbis_close(stb_vorbis *f); + +// this function returns the offset (in samples) from the beginning of the +// file that will be returned by the next decode, if it is known, or -1 +// otherwise. after a flush_pushdata() call, this may take a while before +// it becomes valid again. +// NOT WORKING YET after a seek with PULLDATA API +extern int stb_vorbis_get_sample_offset(stb_vorbis *f); + +// returns the current seek point within the file, or offset from the beginning +// of the memory buffer. In pushdata mode it returns 0. +extern unsigned int stb_vorbis_get_file_offset(stb_vorbis *f); + +/////////// PUSHDATA API + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +// this API allows you to get blocks of data from any source and hand +// them to stb_vorbis. you have to buffer them; stb_vorbis will tell +// you how much it used, and you have to give it the rest next time; +// and stb_vorbis may not have enough data to work with and you will +// need to give it the same data again PLUS more. Note that the Vorbis +// specification does not bound the size of an individual frame. + +extern stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char *datablock, int datablock_length_in_bytes, int *datablock_memory_consumed_in_bytes, int *error, const stb_vorbis_alloc *alloc_buffer); +// create a vorbis decoder by passing in the initial data block containing +// the ogg&vorbis headers (you don't need to do parse them, just provide +// the first N bytes of the file--you're told if it's not enough, see below) +// on success, returns an stb_vorbis *, does not set error, returns the amount of +// data parsed/consumed on this call in *datablock_memory_consumed_in_bytes; +// on failure, returns NULL on error and sets *error, does not change *datablock_memory_consumed +// if returns NULL and *error is VORBIS_need_more_data, then the input block was +// incomplete and you need to pass in a larger block from the start of the file + +extern int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, + const unsigned char *datablock, + int datablock_length_in_bytes, + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples +); +// decode a frame of audio sample data if possible from the passed-in data block +// +// return value: number of bytes we used from datablock +// +// possible cases: +// 0 bytes used, 0 samples output (need more data) +// N bytes used, 0 samples output (resynching the stream, keep going) +// N bytes used, M samples output (one frame of data) +// note that after opening a file, you will ALWAYS get one N-bytes,0-sample +// frame, because Vorbis always "discards" the first frame. +// +// Note that on resynch, stb_vorbis will rarely consume all of the buffer, +// instead only datablock_length_in_bytes-3 or less. This is because it wants +// to avoid missing parts of a page header if they cross a datablock boundary, +// without writing state-machiney code to record a partial detection. +// +// The number of channels returned are stored in *channels (which can be +// NULL--it is always the same as the number of channels reported by +// get_info). *output will contain an array of float* buffers, one per +// channel. In other words, (*output)[0][0] contains the first sample from +// the first channel, and (*output)[1][0] contains the first sample from +// the second channel. + +extern void stb_vorbis_flush_pushdata(stb_vorbis *f); +// inform stb_vorbis that your next datablock will not be contiguous with +// previous ones (e.g. you've seeked in the data); future attempts to decode +// frames will cause stb_vorbis to resynchronize (as noted above), and +// once it sees a valid Ogg page (typically 4-8KB, as large as 64KB), it +// will begin decoding the _next_ frame. +// +// if you want to seek using pushdata, you need to seek in your file, then +// call stb_vorbis_flush_pushdata(), then start calling decoding, then once +// decoding is returning you data, call stb_vorbis_get_sample_offset, and +// if you don't like the result, seek your file again and repeat. +#endif + +////////// PULLING INPUT API + +#ifndef STB_VORBIS_NO_PULLDATA_API +// This API assumes stb_vorbis is allowed to pull data from a source-- +// either a block of memory containing the _entire_ vorbis stream, or a +// FILE * that you or it create, or possibly some other reading mechanism +// if you go modify the source to replace the FILE * case with some kind +// of callback to your code. (But if you don't support seeking, you may +// just want to go ahead and use pushdata.) + +#if !defined(STB_VORBIS_NO_STDIO) && !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output); +#endif +#if !defined(STB_VORBIS_NO_INTEGER_CONVERSION) +extern int stb_vorbis_decode_memory(const unsigned char *mem, int len, int *channels, int *sample_rate, short **output); +#endif +// decode an entire file and output the data interleaved into a malloc()ed +// buffer stored in *output. The return value is the number of samples +// decoded, or -1 if the file could not be opened or was not an ogg vorbis file. +// When you're done with it, just free() the pointer returned in *output. + +extern stb_vorbis *stb_vorbis_open_memory(const unsigned char *data, int len, int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an ogg vorbis stream in memory (note +// this must be the entire stream!). on failure, returns NULL and sets *error + +#ifndef STB_VORBIS_NO_STDIO +extern stb_vorbis *stb_vorbis_open_filename(const char *filename, + int *error, + const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from a filename via fopen(). on failure, +// returns NULL and sets *error (possibly to VORBIS_file_open_failure). + +extern stb_vorbis *stb_vorbis_open_file(FILE *f, int close_handle_on_close, int *error, const stb_vorbis_alloc *alloc_buffer); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell). on failure, returns NULL and sets *error. +// note that stb_vorbis must "own" this stream; if you seek it in between +// calls to stb_vorbis, it will become confused. Moreover, if you attempt to +// perform stb_vorbis_seek_*() operations on this file, it will assume it +// owns the _entire_ rest of the file after the start point. Use the next +// function, stb_vorbis_open_file_section(), to limit it. + +extern stb_vorbis *stb_vorbis_open_file_section(FILE *f, int close_handle_on_close, int *error, const stb_vorbis_alloc *alloc_buffer, unsigned int len); +// create an ogg vorbis decoder from an open FILE *, looking for a stream at +// the _current_ seek point (ftell); the stream will be of length 'len' bytes. +// on failure, returns NULL and sets *error. note that stb_vorbis must "own" +// this stream; if you seek it in between calls to stb_vorbis, it will become +// confused. +#endif + +extern int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number); +extern int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number); +// these functions seek in the Vorbis file to (approximately) 'sample_number'. +// after calling seek_frame(), the next call to get_frame_*() will include +// the specified sample. after calling stb_vorbis_seek(), the next call to +// stb_vorbis_get_samples_* will start with the specified sample. If you +// do not need to seek to EXACTLY the target sample when using get_samples_*, +// you can also use seek_frame(). + +extern int stb_vorbis_seek_start(stb_vorbis *f); +// this function is equivalent to stb_vorbis_seek(f,0) + +extern unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f); +extern float stb_vorbis_stream_length_in_seconds(stb_vorbis *f); +// these functions return the total length of the vorbis stream + +extern int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output); +// decode the next frame and return the number of samples. the number of +// channels returned are stored in *channels (which can be NULL--it is always +// the same as the number of channels reported by get_info). *output will +// contain an array of float* buffers, one per channel. These outputs will +// be overwritten on the next call to stb_vorbis_get_frame_*. +// +// You generally should not intermix calls to stb_vorbis_get_frame_*() +// and stb_vorbis_get_samples_*(), since the latter calls the former. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts); +extern int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples); +#endif +// decode the next frame and return the number of *samples* per channel. +// Note that for interleaved data, you pass in the number of shorts (the +// size of your array), but the return value is the number of samples per +// channel, not the total number of samples. +// +// The data is coerced to the number of channels you request according to the +// channel coercion rules (see below). You must pass in the size of your +// buffer(s) so that stb_vorbis will not overwrite the end of the buffer. +// The maximum buffer size needed can be gotten from get_info(); however, +// the Vorbis I specification implies an absolute maximum of 4096 samples +// per channel. + +// Channel coercion rules: +// Let M be the number of channels requested, and N the number of channels present, +// and Cn be the nth channel; let stereo L be the sum of all L and center channels, +// and stereo R be the sum of all R and center channels (channel assignment from the +// vorbis spec). +// M N output +// 1 k sum(Ck) for all k +// 2 * stereo L, stereo R +// k l k > l, the first l channels, then 0s +// k l k <= l, the first k channels +// Note that this is not _good_ surround etc. mixing at all! It's just so +// you get something useful. + +extern int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats); +extern int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples); +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. DOES NOT APPLY THE COERCION RULES. +// Returns the number of samples stored per channel; it may be less than requested +// at the end of the file. If there are no more samples in the file, returns 0. + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +extern int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts); +extern int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int num_samples); +#endif +// gets num_samples samples, not necessarily on a frame boundary--this requires +// buffering so you have to supply the buffers. Applies the coercion rules above +// to produce 'channels' channels. Returns the number of samples stored per channel; +// it may be less than requested at the end of the file. If there are no more +// samples in the file, returns 0. + +#endif + +//////// ERROR CODES + +enum STBVorbisError { + VORBIS__no_error, + + VORBIS_need_more_data = 1, // not a real error + + VORBIS_invalid_api_mixing, // can't mix API modes + VORBIS_outofmem, // not enough memory + VORBIS_feature_not_supported, // uses floor 0 + VORBIS_too_many_channels, // STB_VORBIS_MAX_CHANNELS is too small + VORBIS_file_open_failure, // fopen() failed + VORBIS_seek_without_length, // can't seek in unknown-length file + + VORBIS_unexpected_eof = 10, // file is truncated? + VORBIS_seek_invalid, // seek past EOF + + // decoding errors (corrupt/invalid stream) -- you probably + // don't care about the exact details of these + + // vorbis errors: + VORBIS_invalid_setup = 20, + VORBIS_invalid_stream, + + // ogg errors: + VORBIS_missing_capture_pattern = 30, + VORBIS_invalid_stream_structure_version, + VORBIS_continued_packet_flag_invalid, + VORBIS_incorrect_stream_serial_number, + VORBIS_invalid_first_page, + VORBIS_bad_packet_type, + VORBIS_cant_find_last_page, + VORBIS_seek_failed, + VORBIS_ogg_skeleton_not_supported +}; + +#ifdef __cplusplus +} +#endif + +#endif // STB_VORBIS_INCLUDE_STB_VORBIS_H +// +// HEADER ENDS HERE +// +////////////////////////////////////////////////////////////////////////////// + +#ifndef STB_VORBIS_HEADER_ONLY + +// global configuration settings (e.g. set these in the project/makefile), +// or just set them in this file at the top (although ideally the first few +// should be visible when the header file is compiled too, although it's not +// crucial) + +// STB_VORBIS_NO_PUSHDATA_API +// does not compile the code for the various stb_vorbis_*_pushdata() +// functions +// #define STB_VORBIS_NO_PUSHDATA_API + +// STB_VORBIS_NO_PULLDATA_API +// does not compile the code for the non-pushdata APIs +// #define STB_VORBIS_NO_PULLDATA_API + +// STB_VORBIS_NO_STDIO +// does not compile the code for the APIs that use FILE *s internally +// or externally (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_STDIO + +// STB_VORBIS_NO_INTEGER_CONVERSION +// does not compile the code for converting audio sample data from +// float to integer (implied by STB_VORBIS_NO_PULLDATA_API) +// #define STB_VORBIS_NO_INTEGER_CONVERSION + +// STB_VORBIS_NO_FAST_SCALED_FLOAT +// does not use a fast float-to-int trick to accelerate float-to-int on +// most platforms which requires endianness be defined correctly. +// #define STB_VORBIS_NO_FAST_SCALED_FLOAT + +// STB_VORBIS_MAX_CHANNELS [number] +// globally define this to the maximum number of channels you need. +// The spec does not put a restriction on channels except that +// the count is stored in a byte, so 255 is the hard limit. +// Reducing this saves about 16 bytes per value, so using 16 saves +// (255-16)*16 or around 4KB. Plus anything other memory usage +// I forgot to account for. Can probably go as low as 8 (7.1 audio), +// 6 (5.1 audio), or 2 (stereo only). +#ifndef STB_VORBIS_MAX_CHANNELS +#define STB_VORBIS_MAX_CHANNELS 16 // enough for anyone? +#endif + +// STB_VORBIS_PUSHDATA_CRC_COUNT [number] +// after a flush_pushdata(), stb_vorbis begins scanning for the +// next valid page, without backtracking. when it finds something +// that looks like a page, it streams through it and verifies its +// CRC32. Should that validation fail, it keeps scanning. But it's +// possible that _while_ streaming through to check the CRC32 of +// one candidate page, it sees another candidate page. This #define +// determines how many "overlapping" candidate pages it can search +// at once. Note that "real" pages are typically ~4KB to ~8KB, whereas +// garbage pages could be as big as 64KB, but probably average ~16KB. +// So don't hose ourselves by scanning an apparent 64KB page and +// missing a ton of real ones in the interim; so minimum of 2 +#ifndef STB_VORBIS_PUSHDATA_CRC_COUNT +#define STB_VORBIS_PUSHDATA_CRC_COUNT 4 +#endif + +// STB_VORBIS_FAST_HUFFMAN_LENGTH [number] +// sets the log size of the huffman-acceleration table. Maximum +// supported value is 24. with larger numbers, more decodings are O(1), +// but the table size is larger so worse cache missing, so you'll have +// to probe (and try multiple ogg vorbis files) to find the sweet spot. +#ifndef STB_VORBIS_FAST_HUFFMAN_LENGTH +#define STB_VORBIS_FAST_HUFFMAN_LENGTH 10 +#endif + +// STB_VORBIS_FAST_BINARY_LENGTH [number] +// sets the log size of the binary-search acceleration table. this +// is used in similar fashion to the fast-huffman size to set initial +// parameters for the binary search + +// STB_VORBIS_FAST_HUFFMAN_INT +// The fast huffman tables are much more efficient if they can be +// stored as 16-bit results instead of 32-bit results. This restricts +// the codebooks to having only 65535 possible outcomes, though. +// (At least, accelerated by the huffman table.) +#ifndef STB_VORBIS_FAST_HUFFMAN_INT +#define STB_VORBIS_FAST_HUFFMAN_SHORT +#endif + +// STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH +// If the 'fast huffman' search doesn't succeed, then stb_vorbis falls +// back on binary searching for the correct one. This requires storing +// extra tables with the huffman codes in sorted order. Defining this +// symbol trades off space for speed by forcing a linear search in the +// non-fast case, except for "sparse" codebooks. +// #define STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + +// STB_VORBIS_DIVIDES_IN_RESIDUE +// stb_vorbis precomputes the result of the scalar residue decoding +// that would otherwise require a divide per chunk. you can trade off +// space for time by defining this symbol. +// #define STB_VORBIS_DIVIDES_IN_RESIDUE + +// STB_VORBIS_DIVIDES_IN_CODEBOOK +// vorbis VQ codebooks can be encoded two ways: with every case explicitly +// stored, or with all elements being chosen from a small range of values, +// and all values possible in all elements. By default, stb_vorbis expands +// this latter kind out to look like the former kind for ease of decoding, +// because otherwise an integer divide-per-vector-element is required to +// unpack the index. If you define STB_VORBIS_DIVIDES_IN_CODEBOOK, you can +// trade off storage for speed. +// #define STB_VORBIS_DIVIDES_IN_CODEBOOK + +#ifdef STB_VORBIS_CODEBOOK_SHORTS +#error "STB_VORBIS_CODEBOOK_SHORTS is no longer supported as it produced incorrect results for some input formats" +#endif + +// STB_VORBIS_DIVIDE_TABLE +// this replaces small integer divides in the floor decode loop with +// table lookups. made less than 1% difference, so disabled by default. + +// STB_VORBIS_NO_INLINE_DECODE +// disables the inlining of the scalar codebook fast-huffman decode. +// might save a little codespace; useful for debugging +// #define STB_VORBIS_NO_INLINE_DECODE + +// STB_VORBIS_NO_DEFER_FLOOR +// Normally we only decode the floor without synthesizing the actual +// full curve. We can instead synthesize the curve immediately. This +// requires more memory and is very likely slower, so I don't think +// you'd ever want to do it except for debugging. +// #define STB_VORBIS_NO_DEFER_FLOOR + +////////////////////////////////////////////////////////////////////////////// + +#ifdef STB_VORBIS_NO_PULLDATA_API +#define STB_VORBIS_NO_INTEGER_CONVERSION +#define STB_VORBIS_NO_STDIO +#endif + +#if defined(STB_VORBIS_NO_CRT) && !defined(STB_VORBIS_NO_STDIO) +#define STB_VORBIS_NO_STDIO 1 +#endif + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT + +// only need endianness for fast-float-to-int, which we don't +// use for pushdata + +#ifndef STB_VORBIS_BIG_ENDIAN +#define STB_VORBIS_ENDIAN 0 +#else +#define STB_VORBIS_ENDIAN 1 +#endif + +#endif +#endif + +#ifndef STB_VORBIS_NO_STDIO +#include +#endif + +#ifndef STB_VORBIS_NO_CRT +#include +#include +#include +#include + +// find definition of alloca if it's not in stdlib.h: +#if defined(_MSC_VER) || defined(__MINGW32__) +#include +#endif +#if defined(__linux__) || defined(__linux) || defined(__EMSCRIPTEN__) || defined(__NEWLIB__) +#include +#endif +#else // STB_VORBIS_NO_CRT +#define NULL 0 +#define malloc(s) 0 +#define free(s) ((void)0) +#define realloc(s) 0 +#endif // STB_VORBIS_NO_CRT + +#include + +#ifdef __MINGW32__ +// eff you mingw: +// "fixed": +// http://sourceforge.net/p/mingw-w64/mailman/message/32882927/ +// "no that broke the build, reverted, who cares about C": +// http://sourceforge.net/p/mingw-w64/mailman/message/32890381/ +#ifdef __forceinline +#undef __forceinline +#endif +#define __forceinline +#ifndef alloca +#define alloca __builtin_alloca +#endif +#elif !defined(_MSC_VER) +#if __GNUC__ +#define __forceinline inline +#else +#define __forceinline +#endif +#endif + +#if STB_VORBIS_MAX_CHANNELS > 256 +#error "Value of STB_VORBIS_MAX_CHANNELS outside of allowed range" +#endif + +#if STB_VORBIS_FAST_HUFFMAN_LENGTH > 24 +#error "Value of STB_VORBIS_FAST_HUFFMAN_LENGTH outside of allowed range" +#endif + +#if 0 +#include +#define CHECK(f) _CrtIsValidHeapPointer(f->channel_buffers[1]) +#else +#define CHECK(f) ((void)0) +#endif + +#define MAX_BLOCKSIZE_LOG 13 // from specification +#define MAX_BLOCKSIZE (1 << MAX_BLOCKSIZE_LOG) + +typedef unsigned char uint8; +typedef signed char int8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; + +#ifndef TRUE +#define TRUE 1 +#define FALSE 0 +#endif + +typedef float codetype; + +// @NOTE +// +// Some arrays below are tagged "//varies", which means it's actually +// a variable-sized piece of data, but rather than malloc I assume it's +// small enough it's better to just allocate it all together with the +// main thing +// +// Most of the variables are specified with the smallest size I could pack +// them into. It might give better performance to make them all full-sized +// integers. It should be safe to freely rearrange the structures or change +// the sizes larger--nothing relies on silently truncating etc., nor the +// order of variables. + +#define FAST_HUFFMAN_TABLE_SIZE (1 << STB_VORBIS_FAST_HUFFMAN_LENGTH) +#define FAST_HUFFMAN_TABLE_MASK (FAST_HUFFMAN_TABLE_SIZE - 1) + +typedef struct +{ + int dimensions, entries; + uint8 *codeword_lengths; + float minimum_value; + float delta_value; + uint8 value_bits; + uint8 lookup_type; + uint8 sequence_p; + uint8 sparse; + uint32 lookup_values; + codetype *multiplicands; + uint32 *codewords; +#ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + int16 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; +#else + int32 fast_huffman[FAST_HUFFMAN_TABLE_SIZE]; +#endif + uint32 *sorted_codewords; + int *sorted_values; + int sorted_entries; +} Codebook; + +typedef struct +{ + uint8 order; + uint16 rate; + uint16 bark_map_size; + uint8 amplitude_bits; + uint8 amplitude_offset; + uint8 number_of_books; + uint8 book_list[16]; // varies +} Floor0; + +typedef struct +{ + uint8 partitions; + uint8 partition_class_list[32]; // varies + uint8 class_dimensions[16]; // varies + uint8 class_subclasses[16]; // varies + uint8 class_masterbooks[16]; // varies + int16 subclass_books[16][8]; // varies + uint16 Xlist[31 * 8 + 2]; // varies + uint8 sorted_order[31 * 8 + 2]; + uint8 neighbors[31 * 8 + 2][2]; + uint8 floor1_multiplier; + uint8 rangebits; + int values; +} Floor1; + +typedef union { + Floor0 floor0; + Floor1 floor1; +} Floor; + +typedef struct +{ + uint32 begin, end; + uint32 part_size; + uint8 classifications; + uint8 classbook; + uint8 **classdata; + int16 (*residue_books)[8]; +} Residue; + +typedef struct +{ + uint8 magnitude; + uint8 angle; + uint8 mux; +} MappingChannel; + +typedef struct +{ + uint16 coupling_steps; + MappingChannel *chan; + uint8 submaps; + uint8 submap_floor[15]; // varies + uint8 submap_residue[15]; // varies +} Mapping; + +typedef struct +{ + uint8 blockflag; + uint8 mapping; + uint16 windowtype; + uint16 transformtype; +} Mode; + +typedef struct +{ + uint32 goal_crc; // expected crc if match + int bytes_left; // bytes left in packet + uint32 crc_so_far; // running crc + int bytes_done; // bytes processed in _current_ chunk + uint32 sample_loc; // granule pos encoded in page +} CRCscan; + +typedef struct +{ + uint32 page_start, page_end; + uint32 last_decoded_sample; +} ProbedPage; + +struct stb_vorbis { + // user-accessible info + unsigned int sample_rate; + int channels; + + unsigned int setup_memory_required; + unsigned int temp_memory_required; + unsigned int setup_temp_memory_required; + + char *vendor; + int comment_list_length; + char **comment_list; + + // input config +#ifndef STB_VORBIS_NO_STDIO + FILE *f; + uint32 f_start; + int close_on_free; +#endif + + uint8 *stream; + uint8 *stream_start; + uint8 *stream_end; + + uint32 stream_len; + + uint8 push_mode; + + // the page to seek to when seeking to start, may be zero + uint32 first_audio_page_offset; + + // p_first is the page on which the first audio packet ends + // (but not necessarily the page on which it starts) + ProbedPage p_first, p_last; + + // memory management + stb_vorbis_alloc alloc; + int setup_offset; + int temp_offset; + + // run-time results + int eof; + enum STBVorbisError error; + + // user-useful data + + // header info + int blocksize[2]; + int blocksize_0, blocksize_1; + int codebook_count; + Codebook *codebooks; + int floor_count; + uint16 floor_types[64]; // varies + Floor *floor_config; + int residue_count; + uint16 residue_types[64]; // varies + Residue *residue_config; + int mapping_count; + Mapping *mapping; + int mode_count; + Mode mode_config[64]; // varies + + uint32 total_samples; + + // decode buffer + float *channel_buffers[STB_VORBIS_MAX_CHANNELS]; + float *outputs[STB_VORBIS_MAX_CHANNELS]; + + float *previous_window[STB_VORBIS_MAX_CHANNELS]; + int previous_length; + +#ifndef STB_VORBIS_NO_DEFER_FLOOR + int16 *finalY[STB_VORBIS_MAX_CHANNELS]; +#else + float *floor_buffers[STB_VORBIS_MAX_CHANNELS]; +#endif + + uint32 current_loc; // sample location of next frame to decode + int current_loc_valid; + + // per-blocksize precomputed data + + // twiddle factors + float *A[2], *B[2], *C[2]; + float *window[2]; + uint16 *bit_reverse[2]; + + // current page/packet/segment streaming info + uint32 serial; // stream serial number for verification + int last_page; + int segment_count; + uint8 segments[255]; + uint8 page_flag; + uint8 bytes_in_seg; + uint8 first_decode; + int next_seg; + int last_seg; // flag that we're on the last segment + int last_seg_which; // what was the segment number of the last seg? + uint32 acc; + int valid_bits; + int packet_bytes; + int end_seg_with_known_loc; + uint32 known_loc_for_packet; + int discard_samples_deferred; + uint32 samples_output; + + // push mode scanning + int page_crc_tests; // only in push_mode: number of tests active; -1 if not searching +#ifndef STB_VORBIS_NO_PUSHDATA_API + CRCscan scan[STB_VORBIS_PUSHDATA_CRC_COUNT]; +#endif + + // sample-access + int channel_buffer_start; + int channel_buffer_end; +}; + +#if defined(STB_VORBIS_NO_PUSHDATA_API) +#define IS_PUSH_MODE(f) FALSE +#elif defined(STB_VORBIS_NO_PULLDATA_API) +#define IS_PUSH_MODE(f) TRUE +#else +#define IS_PUSH_MODE(f) ((f)->push_mode) +#endif + +typedef struct stb_vorbis vorb; + +static int error(vorb *f, enum STBVorbisError e) { + f->error = e; + if (!f->eof && e != VORBIS_need_more_data) { + f->error = e; // breakpoint for debugging + } + return 0; +} + +// these functions are used for allocating temporary memory +// while decoding. if you can afford the stack space, use +// alloca(); otherwise, provide a temp buffer and it will +// allocate out of those. + +#define array_size_required(count, size) (count * (sizeof(void *) + (size))) + +#define temp_alloc(f, size) (f->alloc.alloc_buffer ? setup_temp_malloc(f, size) : alloca(size)) +#define temp_free(f, p) (void)0 +#define temp_alloc_save(f) ((f)->temp_offset) +#define temp_alloc_restore(f, p) ((f)->temp_offset = (p)) + +#define temp_block_array(f, count, size) make_block_array(temp_alloc(f, array_size_required(count, size)), count, size) + +// given a sufficiently large block of memory, make an array of pointers to subblocks of it +static void *make_block_array(void *mem, int count, int size) { + int i; + void **p = (void **)mem; + char *q = (char *)(p + count); + for (i = 0; i < count; ++i) { + p[i] = q; + q += size; + } + return p; +} + +static void *setup_malloc(vorb *f, int sz) { + sz = (sz + 7) & ~7; // round up to nearest 8 for alignment of future allocs. + f->setup_memory_required += sz; + if (f->alloc.alloc_buffer) { + void *p = (char *)f->alloc.alloc_buffer + f->setup_offset; + if (f->setup_offset + sz > f->temp_offset) + return NULL; + f->setup_offset += sz; + return p; + } + return sz ? malloc(sz) : NULL; +} + +static void setup_free(vorb *f, void *p) { + if (f->alloc.alloc_buffer) + return; // do nothing; setup mem is a stack + free(p); +} + +static void *setup_temp_malloc(vorb *f, int sz) { + sz = (sz + 7) & ~7; // round up to nearest 8 for alignment of future allocs. + if (f->alloc.alloc_buffer) { + if (f->temp_offset - sz < f->setup_offset) + return NULL; + f->temp_offset -= sz; + return (char *)f->alloc.alloc_buffer + f->temp_offset; + } + return malloc(sz); +} + +static void setup_temp_free(vorb *f, void *p, int sz) { + if (f->alloc.alloc_buffer) { + f->temp_offset += (sz + 7) & ~7; + return; + } + free(p); +} + +#define CRC32_POLY 0x04c11db7 // from spec + +static uint32 crc_table[256]; +static void crc32_init(void) { + int i, j; + uint32 s; + for (i = 0; i < 256; i++) { + for (s = (uint32)i << 24, j = 0; j < 8; ++j) + s = (s << 1) ^ (s >= (1U << 31) ? CRC32_POLY : 0); + crc_table[i] = s; + } +} + +static __forceinline uint32 crc32_update(uint32 crc, uint8 byte) { + return (crc << 8) ^ crc_table[byte ^ (crc >> 24)]; +} + +// used in setup, and for huffman that doesn't go fast path +static unsigned int bit_reverse(unsigned int n) { + n = ((n & 0xAAAAAAAA) >> 1) | ((n & 0x55555555) << 1); + n = ((n & 0xCCCCCCCC) >> 2) | ((n & 0x33333333) << 2); + n = ((n & 0xF0F0F0F0) >> 4) | ((n & 0x0F0F0F0F) << 4); + n = ((n & 0xFF00FF00) >> 8) | ((n & 0x00FF00FF) << 8); + return (n >> 16) | (n << 16); +} + +static float square(float x) { + return x * x; +} + +// this is a weird definition of log2() for which log2(1) = 1, log2(2) = 2, log2(4) = 3 +// as required by the specification. fast(?) implementation from stb.h +// @OPTIMIZE: called multiple times per-packet with "constants"; move to setup +static int ilog(int32 n) { + static signed char log2_4[16] = {0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4}; + + if (n < 0) + return 0; // signed n returns 0 + + // 2 compares if n < 16, 3 compares otherwise (4 if signed or n > 1<<29) + if (n < (1 << 14)) + if (n < (1 << 4)) + return 0 + log2_4[n]; + else if (n < (1 << 9)) + return 5 + log2_4[n >> 5]; + else + return 10 + log2_4[n >> 10]; + else if (n < (1 << 24)) + if (n < (1 << 19)) + return 15 + log2_4[n >> 15]; + else + return 20 + log2_4[n >> 20]; + else if (n < (1 << 29)) + return 25 + log2_4[n >> 25]; + else + return 30 + log2_4[n >> 30]; +} + +#ifndef M_PI +#define M_PI 3.14159265358979323846264f // from CRC +#endif + +// code length assigned to a value with no huffman encoding +#define NO_CODE 255 + +/////////////////////// LEAF SETUP FUNCTIONS ////////////////////////// +// +// these functions are only called at setup, and only a few times +// per file + +static float float32_unpack(uint32 x) { + // from the specification + uint32 mantissa = x & 0x1fffff; + uint32 sign = x & 0x80000000; + uint32 exp = (x & 0x7fe00000) >> 21; + double res = sign ? -(double)mantissa : (double)mantissa; + return (float)ldexp((float)res, exp - 788); +} + +// zlib & jpeg huffman tables assume that the output symbols +// can either be arbitrarily arranged, or have monotonically +// increasing frequencies--they rely on the lengths being sorted; +// this makes for a very simple generation algorithm. +// vorbis allows a huffman table with non-sorted lengths. This +// requires a more sophisticated construction, since symbols in +// order do not map to huffman codes "in order". +static void add_entry(Codebook *c, uint32 huff_code, int symbol, int count, int len, uint32 *values) { + if (!c->sparse) { + c->codewords[symbol] = huff_code; + } else { + c->codewords[count] = huff_code; + c->codeword_lengths[count] = len; + values[count] = symbol; + } +} + +static int compute_codewords(Codebook *c, uint8 *len, int n, uint32 *values) { + int i, k, m = 0; + uint32 available[32]; + + memset(available, 0, sizeof(available)); + // find the first entry + for (k = 0; k < n; ++k) + if (len[k] < NO_CODE) + break; + if (k == n) { + assert(c->sorted_entries == 0); + return TRUE; + } + // add to the list + add_entry(c, 0, k, m++, len[k], values); + // add all available leaves + for (i = 1; i <= len[k]; ++i) + available[i] = 1U << (32 - i); + // note that the above code treats the first case specially, + // but it's really the same as the following code, so they + // could probably be combined (except the initial code is 0, + // and I use 0 in available[] to mean 'empty') + for (i = k + 1; i < n; ++i) { + uint32 res; + int z = len[i], y; + if (z == NO_CODE) + continue; + // find lowest available leaf (should always be earliest, + // which is what the specification calls for) + // note that this property, and the fact we can never have + // more than one free leaf at a given level, isn't totally + // trivial to prove, but it seems true and the assert never + // fires, so! + while (z > 0 && !available[z]) + --z; + if (z == 0) { + return FALSE; + } + res = available[z]; + assert(z >= 0 && z < 32); + available[z] = 0; + add_entry(c, bit_reverse(res), i, m++, len[i], values); + // propagate availability up the tree + if (z != len[i]) { + assert(len[i] >= 0 && len[i] < 32); + for (y = len[i]; y > z; --y) { + assert(available[y] == 0); + available[y] = res + (1 << (32 - y)); + } + } + } + return TRUE; +} + +// accelerated huffman table allows fast O(1) match of all symbols +// of length <= STB_VORBIS_FAST_HUFFMAN_LENGTH +static void compute_accelerated_huffman(Codebook *c) { + int i, len; + for (i = 0; i < FAST_HUFFMAN_TABLE_SIZE; ++i) + c->fast_huffman[i] = -1; + + len = c->sparse ? c->sorted_entries : c->entries; +#ifdef STB_VORBIS_FAST_HUFFMAN_SHORT + if (len > 32767) + len = 32767; // largest possible value we can encode! +#endif + for (i = 0; i < len; ++i) { + if (c->codeword_lengths[i] <= STB_VORBIS_FAST_HUFFMAN_LENGTH) { + uint32 z = c->sparse ? bit_reverse(c->sorted_codewords[i]) : c->codewords[i]; + // set table entries for all bit combinations in the higher bits + while (z < FAST_HUFFMAN_TABLE_SIZE) { + c->fast_huffman[z] = i; + z += 1 << c->codeword_lengths[i]; + } + } + } +} + +#ifdef _MSC_VER +#define STBV_CDECL __cdecl +#else +#define STBV_CDECL +#endif + +static int STBV_CDECL uint32_compare(const void *p, const void *q) { + uint32 x = *(uint32 *)p; + uint32 y = *(uint32 *)q; + return x < y ? -1 : x > y; +} + +static int include_in_sort(Codebook *c, uint8 len) { + if (c->sparse) { + assert(len != NO_CODE); + return TRUE; + } + if (len == NO_CODE) + return FALSE; + if (len > STB_VORBIS_FAST_HUFFMAN_LENGTH) + return TRUE; + return FALSE; +} + +// if the fast table above doesn't work, we want to binary +// search them... need to reverse the bits +static void compute_sorted_huffman(Codebook *c, uint8 *lengths, uint32 *values) { + int i, len; + // build a list of all the entries + // OPTIMIZATION: don't include the short ones, since they'll be caught by FAST_HUFFMAN. + // this is kind of a frivolous optimization--I don't see any performance improvement, + // but it's like 4 extra lines of code, so. + if (!c->sparse) { + int k = 0; + for (i = 0; i < c->entries; ++i) + if (include_in_sort(c, lengths[i])) + c->sorted_codewords[k++] = bit_reverse(c->codewords[i]); + assert(k == c->sorted_entries); + } else { + for (i = 0; i < c->sorted_entries; ++i) + c->sorted_codewords[i] = bit_reverse(c->codewords[i]); + } + + qsort(c->sorted_codewords, c->sorted_entries, sizeof(c->sorted_codewords[0]), uint32_compare); + c->sorted_codewords[c->sorted_entries] = 0xffffffff; + + len = c->sparse ? c->sorted_entries : c->entries; + // now we need to indicate how they correspond; we could either + // #1: sort a different data structure that says who they correspond to + // #2: for each sorted entry, search the original list to find who corresponds + // #3: for each original entry, find the sorted entry + // #1 requires extra storage, #2 is slow, #3 can use binary search! + for (i = 0; i < len; ++i) { + int huff_len = c->sparse ? lengths[values[i]] : lengths[i]; + if (include_in_sort(c, huff_len)) { + uint32 code = bit_reverse(c->codewords[i]); + int x = 0, n = c->sorted_entries; + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n >> 1); + } else { + n >>= 1; + } + } + assert(c->sorted_codewords[x] == code); + if (c->sparse) { + c->sorted_values[x] = values[i]; + c->codeword_lengths[x] = huff_len; + } else { + c->sorted_values[x] = i; + } + } + } +} + +// only run while parsing the header (3 times) +static int vorbis_validate(uint8 *data) { + static uint8 vorbis[6] = {'v', 'o', 'r', 'b', 'i', 's'}; + return memcmp(data, vorbis, 6) == 0; +} + +// called from setup only, once per code book +// (formula implied by specification) +static int lookup1_values(int entries, int dim) { + int r = (int)floor(exp((float)log((float)entries) / dim)); + if ((int)floor(pow((float)r + 1, dim)) <= entries) // (int) cast for MinGW warning; + ++r; // floor() to avoid _ftol() when non-CRT + if (pow((float)r + 1, dim) <= entries) + return -1; + if ((int)floor(pow((float)r, dim)) > entries) + return -1; + return r; +} + +// called twice per file +static void compute_twiddle_factors(int n, float *A, float *B, float *C) { + int n4 = n >> 2, n8 = n >> 3; + int k, k2; + + for (k = k2 = 0; k < n4; ++k, k2 += 2) { + A[k2] = (float)cos(4 * k * M_PI / n); + A[k2 + 1] = (float)-sin(4 * k * M_PI / n); + B[k2] = (float)cos((k2 + 1) * M_PI / n / 2) * 0.5f; + B[k2 + 1] = (float)sin((k2 + 1) * M_PI / n / 2) * 0.5f; + } + for (k = k2 = 0; k < n8; ++k, k2 += 2) { + C[k2] = (float)cos(2 * (k2 + 1) * M_PI / n); + C[k2 + 1] = (float)-sin(2 * (k2 + 1) * M_PI / n); + } +} + +static void compute_window(int n, float *window) { + int n2 = n >> 1, i; + for (i = 0; i < n2; ++i) + window[i] = (float)sin(0.5 * M_PI * square((float)sin((i - 0 + 0.5) / n2 * 0.5 * M_PI))); +} + +static void compute_bitreverse(int n, uint16 *rev) { + int ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + int i, n8 = n >> 3; + for (i = 0; i < n8; ++i) + rev[i] = (bit_reverse(i) >> (32 - ld + 3)) << 2; +} + +static int init_blocksize(vorb *f, int b, int n) { + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3; + f->A[b] = (float *)setup_malloc(f, sizeof(float) * n2); + f->B[b] = (float *)setup_malloc(f, sizeof(float) * n2); + f->C[b] = (float *)setup_malloc(f, sizeof(float) * n4); + if (!f->A[b] || !f->B[b] || !f->C[b]) + return error(f, VORBIS_outofmem); + compute_twiddle_factors(n, f->A[b], f->B[b], f->C[b]); + f->window[b] = (float *)setup_malloc(f, sizeof(float) * n2); + if (!f->window[b]) + return error(f, VORBIS_outofmem); + compute_window(n, f->window[b]); + f->bit_reverse[b] = (uint16 *)setup_malloc(f, sizeof(uint16) * n8); + if (!f->bit_reverse[b]) + return error(f, VORBIS_outofmem); + compute_bitreverse(n, f->bit_reverse[b]); + return TRUE; +} + +static void neighbors(uint16 *x, int n, int *plow, int *phigh) { + int low = -1; + int high = 65536; + int i; + for (i = 0; i < n; ++i) { + if (x[i] > low && x[i] < x[n]) { + *plow = i; + low = x[i]; + } + if (x[i] < high && x[i] > x[n]) { + *phigh = i; + high = x[i]; + } + } +} + +// this has been repurposed so y is now the original index instead of y +typedef struct +{ + uint16 x, id; +} stbv__floor_ordering; + +static int STBV_CDECL point_compare(const void *p, const void *q) { + stbv__floor_ordering *a = (stbv__floor_ordering *)p; + stbv__floor_ordering *b = (stbv__floor_ordering *)q; + return a->x < b->x ? -1 : a->x > b->x; +} + +// +/////////////////////// END LEAF SETUP FUNCTIONS ////////////////////////// + +#if defined(STB_VORBIS_NO_STDIO) +#define USE_MEMORY(z) TRUE +#else +#define USE_MEMORY(z) ((z)->stream) +#endif + +static uint8 get8(vorb *z) { + if (USE_MEMORY(z)) { + if (z->stream >= z->stream_end) { + z->eof = TRUE; + return 0; + } + return *z->stream++; + } + +#ifndef STB_VORBIS_NO_STDIO + { + int c = fgetc(z->f); + if (c == EOF) { + z->eof = TRUE; + return 0; + } + return c; + } +#endif +} + +static uint32 get32(vorb *f) { + uint32 x; + x = get8(f); + x += get8(f) << 8; + x += get8(f) << 16; + x += (uint32)get8(f) << 24; + return x; +} + +static int getn(vorb *z, uint8 *data, int n) { + if (USE_MEMORY(z)) { + if (z->stream + n > z->stream_end) { + z->eof = 1; + return 0; + } + memcpy(data, z->stream, n); + z->stream += n; + return 1; + } + +#ifndef STB_VORBIS_NO_STDIO + if (fread(data, n, 1, z->f) == 1) + return 1; + else { + z->eof = 1; + return 0; + } +#endif +} + +static void skip(vorb *z, int n) { + if (USE_MEMORY(z)) { + z->stream += n; + if (z->stream >= z->stream_end) + z->eof = 1; + return; + } +#ifndef STB_VORBIS_NO_STDIO + { + long x = ftell(z->f); + fseek(z->f, x + n, SEEK_SET); + } +#endif +} + +static int set_file_offset(stb_vorbis *f, unsigned int loc) { +#ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) + return 0; +#endif + f->eof = 0; + if (USE_MEMORY(f)) { + if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) { + f->stream = f->stream_end; + f->eof = 1; + return 0; + } else { + f->stream = f->stream_start + loc; + return 1; + } + } +#ifndef STB_VORBIS_NO_STDIO + if (loc + f->f_start < loc || loc >= 0x80000000) { + loc = 0x7fffffff; + f->eof = 1; + } else { + loc += f->f_start; + } + if (!fseek(f->f, loc, SEEK_SET)) + return 1; + f->eof = 1; + fseek(f->f, f->f_start, SEEK_END); + return 0; +#endif +} + +static uint8 ogg_page_header[4] = {0x4f, 0x67, 0x67, 0x53}; + +static int capture_pattern(vorb *f) { + if (0x4f != get8(f)) + return FALSE; + if (0x67 != get8(f)) + return FALSE; + if (0x67 != get8(f)) + return FALSE; + if (0x53 != get8(f)) + return FALSE; + return TRUE; +} + +#define PAGEFLAG_continued_packet 1 +#define PAGEFLAG_first_page 2 +#define PAGEFLAG_last_page 4 + +static int start_page_no_capturepattern(vorb *f) { + uint32 loc0, loc1, n; + if (f->first_decode && !IS_PUSH_MODE(f)) { + f->p_first.page_start = stb_vorbis_get_file_offset(f) - 4; + } + // stream structure version + if (0 != get8(f)) + return error(f, VORBIS_invalid_stream_structure_version); + // header flag + f->page_flag = get8(f); + // absolute granule position + loc0 = get32(f); + loc1 = get32(f); + // @TODO: validate loc0,loc1 as valid positions? + // stream serial number -- vorbis doesn't interleave, so discard + get32(f); + // if (f->serial != get32(f)) return error(f, VORBIS_incorrect_stream_serial_number); + // page sequence number + n = get32(f); + f->last_page = n; + // CRC32 + get32(f); + // page_segments + f->segment_count = get8(f); + if (!getn(f, f->segments, f->segment_count)) + return error(f, VORBIS_unexpected_eof); + // assume we _don't_ know any the sample position of any segments + f->end_seg_with_known_loc = -2; + if (loc0 != ~0U || loc1 != ~0U) { + int i; + // determine which packet is the last one that will complete + for (i = f->segment_count - 1; i >= 0; --i) + if (f->segments[i] < 255) + break; + // 'i' is now the index of the _last_ segment of a packet that ends + if (i >= 0) { + f->end_seg_with_known_loc = i; + f->known_loc_for_packet = loc0; + } + } + if (f->first_decode) { + int i, len; + len = 0; + for (i = 0; i < f->segment_count; ++i) + len += f->segments[i]; + len += 27 + f->segment_count; + f->p_first.page_end = f->p_first.page_start + len; + f->p_first.last_decoded_sample = loc0; + } + f->next_seg = 0; + return TRUE; +} + +static int start_page(vorb *f) { + if (!capture_pattern(f)) + return error(f, VORBIS_missing_capture_pattern); + return start_page_no_capturepattern(f); +} + +static int start_packet(vorb *f) { + while (f->next_seg == -1) { + if (!start_page(f)) + return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) + return error(f, VORBIS_continued_packet_flag_invalid); + } + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + // f->next_seg is now valid + return TRUE; +} + +static int maybe_start_packet(vorb *f) { + if (f->next_seg == -1) { + int x = get8(f); + if (f->eof) + return FALSE; // EOF at page boundary is not an error! + if (0x4f != x) + return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) + return error(f, VORBIS_missing_capture_pattern); + if (0x67 != get8(f)) + return error(f, VORBIS_missing_capture_pattern); + if (0x53 != get8(f)) + return error(f, VORBIS_missing_capture_pattern); + if (!start_page_no_capturepattern(f)) + return FALSE; + if (f->page_flag & PAGEFLAG_continued_packet) { + // set up enough state that we can read this packet if we want, + // e.g. during recovery + f->last_seg = FALSE; + f->bytes_in_seg = 0; + return error(f, VORBIS_continued_packet_flag_invalid); + } + } + return start_packet(f); +} + +static int next_segment(vorb *f) { + int len; + if (f->last_seg) + return 0; + if (f->next_seg == -1) { + f->last_seg_which = f->segment_count - 1; // in case start_page fails + if (!start_page(f)) { + f->last_seg = 1; + return 0; + } + if (!(f->page_flag & PAGEFLAG_continued_packet)) + return error(f, VORBIS_continued_packet_flag_invalid); + } + len = f->segments[f->next_seg++]; + if (len < 255) { + f->last_seg = TRUE; + f->last_seg_which = f->next_seg - 1; + } + if (f->next_seg >= f->segment_count) + f->next_seg = -1; + assert(f->bytes_in_seg == 0); + f->bytes_in_seg = len; + return len; +} + +#define EOP (-1) +#define INVALID_BITS (-1) + +static int get8_packet_raw(vorb *f) { + if (!f->bytes_in_seg) { // CLANG! + if (f->last_seg) + return EOP; + else if (!next_segment(f)) + return EOP; + } + assert(f->bytes_in_seg > 0); + --f->bytes_in_seg; + ++f->packet_bytes; + return get8(f); +} + +static int get8_packet(vorb *f) { + int x = get8_packet_raw(f); + f->valid_bits = 0; + return x; +} + +static int get32_packet(vorb *f) { + uint32 x; + x = get8_packet(f); + x += get8_packet(f) << 8; + x += get8_packet(f) << 16; + x += (uint32)get8_packet(f) << 24; + return x; +} + +static void flush_packet(vorb *f) { + while (get8_packet_raw(f) != EOP); +} + +// @OPTIMIZE: this is the secondary bit decoder, so it's probably not as important +// as the huffman decoder? +static uint32 get_bits(vorb *f, int n) { + uint32 z; + + if (f->valid_bits < 0) + return 0; + if (f->valid_bits < n) { + if (n > 24) { + // the accumulator technique below would not work correctly in this case + z = get_bits(f, 24); + z += get_bits(f, n - 24) << 24; + return z; + } + if (f->valid_bits == 0) + f->acc = 0; + while (f->valid_bits < n) { + int z = get8_packet_raw(f); + if (z == EOP) { + f->valid_bits = INVALID_BITS; + return 0; + } + f->acc += z << f->valid_bits; + f->valid_bits += 8; + } + } + + assert(f->valid_bits >= n); + z = f->acc & ((1 << n) - 1); + f->acc >>= n; + f->valid_bits -= n; + return z; +} + +// @OPTIMIZE: primary accumulator for huffman +// expand the buffer to as many bits as possible without reading off end of packet +// it might be nice to allow f->valid_bits and f->acc to be stored in registers, +// e.g. cache them locally and decode locally +static __forceinline void prep_huffman(vorb *f) { + if (f->valid_bits <= 24) { + if (f->valid_bits == 0) + f->acc = 0; + do { + int z; + if (f->last_seg && !f->bytes_in_seg) + return; + z = get8_packet_raw(f); + if (z == EOP) + return; + f->acc += (unsigned)z << f->valid_bits; + f->valid_bits += 8; + } while (f->valid_bits <= 24); + } +} + +enum { + VORBIS_packet_id = 1, + VORBIS_packet_comment = 3, + VORBIS_packet_setup = 5 +}; + +static int codebook_decode_scalar_raw(vorb *f, Codebook *c) { + int i; + prep_huffman(f); + + if (c->codewords == NULL && c->sorted_codewords == NULL) + return -1; + + // cases to use binary search: sorted_codewords && !c->codewords + // sorted_codewords && c->entries > 8 + if (c->entries > 8 ? c->sorted_codewords != NULL : !c->codewords) { + // binary search + uint32 code = bit_reverse(f->acc); + int x = 0, n = c->sorted_entries, len; + + while (n > 1) { + // invariant: sc[x] <= code < sc[x+n] + int m = x + (n >> 1); + if (c->sorted_codewords[m] <= code) { + x = m; + n -= (n >> 1); + } else { + n >>= 1; + } + } + // x is now the sorted index + if (!c->sparse) + x = c->sorted_values[x]; + // x is now sorted index if sparse, or symbol otherwise + len = c->codeword_lengths[x]; + if (f->valid_bits >= len) { + f->acc >>= len; + f->valid_bits -= len; + return x; + } + + f->valid_bits = 0; + return -1; + } + + // if small, linear search + assert(!c->sparse); + for (i = 0; i < c->entries; ++i) { + if (c->codeword_lengths[i] == NO_CODE) + continue; + if (c->codewords[i] == (f->acc & ((1 << c->codeword_lengths[i]) - 1))) { + if (f->valid_bits >= c->codeword_lengths[i]) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + return i; + } + f->valid_bits = 0; + return -1; + } + } + + error(f, VORBIS_invalid_stream); + f->valid_bits = 0; + return -1; +} + +#ifndef STB_VORBIS_NO_INLINE_DECODE + +#define DECODE_RAW(var, f, c) \ + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) \ + prep_huffman(f); \ + var = f->acc & FAST_HUFFMAN_TABLE_MASK; \ + var = c->fast_huffman[var]; \ + if (var >= 0) { \ + int n = c->codeword_lengths[var]; \ + f->acc >>= n; \ + f->valid_bits -= n; \ + if (f->valid_bits < 0) { \ + f->valid_bits = 0; \ + var = -1; \ + } \ + } else { \ + var = codebook_decode_scalar_raw(f, c); \ + } + +#else + +static int codebook_decode_scalar(vorb *f, Codebook *c) { + int i; + if (f->valid_bits < STB_VORBIS_FAST_HUFFMAN_LENGTH) + prep_huffman(f); + // fast huffman table lookup + i = f->acc & FAST_HUFFMAN_TABLE_MASK; + i = c->fast_huffman[i]; + if (i >= 0) { + f->acc >>= c->codeword_lengths[i]; + f->valid_bits -= c->codeword_lengths[i]; + if (f->valid_bits < 0) { + f->valid_bits = 0; + return -1; + } + return i; + } + return codebook_decode_scalar_raw(f, c); +} + +#define DECODE_RAW(var, f, c) var = codebook_decode_scalar(f, c); + +#endif + +#define DECODE(var, f, c) \ + DECODE_RAW(var, f, c) \ + if (c->sparse) \ + var = c->sorted_values[var]; + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK +#define DECODE_VQ(var, f, c) DECODE_RAW(var, f, c) +#else +#define DECODE_VQ(var, f, c) DECODE(var, f, c) +#endif + +// CODEBOOK_ELEMENT_FAST is an optimization for the CODEBOOK_FLOATS case +// where we avoid one addition +#define CODEBOOK_ELEMENT(c, off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_FAST(c, off) (c->multiplicands[off]) +#define CODEBOOK_ELEMENT_BASE(c) (0) + +static int codebook_decode_start(vorb *f, Codebook *c) { + int z = -1; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) + error(f, VORBIS_invalid_stream); + else { + DECODE_VQ(z, f, c); + if (c->sparse) + assert(z < c->sorted_entries); + if (z < 0) { // check for EOP + if (!f->bytes_in_seg) + if (f->last_seg) + return z; + error(f, VORBIS_invalid_stream); + } + } + return z; +} + +static int codebook_decode(vorb *f, Codebook *c, float *output, int len) { + int i, z = codebook_decode_start(f, c); + if (z < 0) + return FALSE; + if (len > c->dimensions) + len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + float last = CODEBOOK_ELEMENT_BASE(c); + int div = 1; + for (i = 0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c, off) + last; + output[i] += val; + if (c->sequence_p) + last = val + c->minimum_value; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + if (c->sequence_p) { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i = 0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c, z + i) + last; + output[i] += val; + last = val + c->minimum_value; + } + } else { + float last = CODEBOOK_ELEMENT_BASE(c); + for (i = 0; i < len; ++i) { + output[i] += CODEBOOK_ELEMENT_FAST(c, z + i) + last; + } + } + + return TRUE; +} + +static int codebook_decode_step(vorb *f, Codebook *c, float *output, int len, int step) { + int i, z = codebook_decode_start(f, c); + float last = CODEBOOK_ELEMENT_BASE(c); + if (z < 0) + return FALSE; + if (len > c->dimensions) + len = c->dimensions; + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i = 0; i < len; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c, off) + last; + output[i * step] += val; + if (c->sequence_p) + last = val; + div *= c->lookup_values; + } + return TRUE; + } +#endif + + z *= c->dimensions; + for (i = 0; i < len; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c, z + i) + last; + output[i * step] += val; + if (c->sequence_p) + last = val; + } + + return TRUE; +} + +static int codebook_decode_deinterleave_repeat(vorb *f, Codebook *c, float **outputs, int ch, int *c_inter_p, int *p_inter_p, int len, int total_decode) { + int c_inter = *c_inter_p; + int p_inter = *p_inter_p; + int i, z, effective = c->dimensions; + + // type 0 is only legal in a scalar context + if (c->lookup_type == 0) + return error(f, VORBIS_invalid_stream); + + while (total_decode > 0) { + float last = CODEBOOK_ELEMENT_BASE(c); + DECODE_VQ(z, f, c); +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + assert(!c->sparse || z < c->sorted_entries); +#endif + if (z < 0) { + if (!f->bytes_in_seg) + if (f->last_seg) + return FALSE; + return error(f, VORBIS_invalid_stream); + } + + // if this will take us off the end of the buffers, stop short! + // we check by computing the length of the virtual interleaved + // buffer (len*ch), our current offset within it (p_inter*ch)+(c_inter), + // and the length we'll be using (effective) + if (c_inter + p_inter * ch + effective > len * ch) { + effective = len * ch - (p_inter * ch - c_inter); + } + +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int div = 1; + for (i = 0; i < effective; ++i) { + int off = (z / div) % c->lookup_values; + float val = CODEBOOK_ELEMENT_FAST(c, off) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { + c_inter = 0; + ++p_inter; + } + if (c->sequence_p) + last = val; + div *= c->lookup_values; + } + } else +#endif + { + z *= c->dimensions; + if (c->sequence_p) { + for (i = 0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c, z + i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { + c_inter = 0; + ++p_inter; + } + last = val; + } + } else { + for (i = 0; i < effective; ++i) { + float val = CODEBOOK_ELEMENT_FAST(c, z + i) + last; + if (outputs[c_inter]) + outputs[c_inter][p_inter] += val; + if (++c_inter == ch) { + c_inter = 0; + ++p_inter; + } + } + } + } + + total_decode -= effective; + } + *c_inter_p = c_inter; + *p_inter_p = p_inter; + return TRUE; +} + +static int predict_point(int x, int x0, int x1, int y0, int y1) { + int dy = y1 - y0; + int adx = x1 - x0; + // @OPTIMIZE: force int division to round in the right direction... is this necessary on x86? + int err = abs(dy) * (x - x0); + int off = err / adx; + return dy < 0 ? y0 - off : y0 + off; +} + +// the following table is block-copied from the specification +static float inverse_db_table[256] = + { + 1.0649863e-07f, 1.1341951e-07f, 1.2079015e-07f, 1.2863978e-07f, 1.3699951e-07f, 1.4590251e-07f, 1.5538408e-07f, 1.6548181e-07f, 1.7623575e-07f, 1.8768855e-07f, 1.9988561e-07f, 2.1287530e-07f, 2.2670913e-07f, 2.4144197e-07f, 2.5713223e-07f, 2.7384213e-07f, 2.9163793e-07f, 3.1059021e-07f, 3.3077411e-07f, 3.5226968e-07f, 3.7516214e-07f, 3.9954229e-07f, 4.2550680e-07f, 4.5315863e-07f, 4.8260743e-07f, 5.1396998e-07f, 5.4737065e-07f, 5.8294187e-07f, 6.2082472e-07f, 6.6116941e-07f, 7.0413592e-07f, 7.4989464e-07f, 7.9862701e-07f, 8.5052630e-07f, 9.0579828e-07f, 9.6466216e-07f, 1.0273513e-06f, 1.0941144e-06f, 1.1652161e-06f, 1.2409384e-06f, 1.3215816e-06f, 1.4074654e-06f, 1.4989305e-06f, 1.5963394e-06f, 1.7000785e-06f, 1.8105592e-06f, 1.9282195e-06f, 2.0535261e-06f, 2.1869758e-06f, 2.3290978e-06f, 2.4804557e-06f, 2.6416497e-06f, 2.8133190e-06f, 2.9961443e-06f, 3.1908506e-06f, 3.3982101e-06f, 3.6190449e-06f, 3.8542308e-06f, 4.1047004e-06f, 4.3714470e-06f, 4.6555282e-06f, 4.9580707e-06f, 5.2802740e-06f, 5.6234160e-06f, 5.9888572e-06f, 6.3780469e-06f, 6.7925283e-06f, 7.2339451e-06f, 7.7040476e-06f, 8.2047000e-06f, 8.7378876e-06f, 9.3057248e-06f, 9.9104632e-06f, 1.0554501e-05f, 1.1240392e-05f, 1.1970856e-05f, 1.2748789e-05f, 1.3577278e-05f, 1.4459606e-05f, 1.5399272e-05f, 1.6400004e-05f, 1.7465768e-05f, 1.8600792e-05f, 1.9809576e-05f, 2.1096914e-05f, 2.2467911e-05f, 2.3928002e-05f, 2.5482978e-05f, 2.7139006e-05f, 2.8902651e-05f, 3.0780908e-05f, 3.2781225e-05f, 3.4911534e-05f, 3.7180282e-05f, 3.9596466e-05f, 4.2169667e-05f, 4.4910090e-05f, 4.7828601e-05f, 5.0936773e-05f, 5.4246931e-05f, 5.7772202e-05f, 6.1526565e-05f, 6.5524908e-05f, 6.9783085e-05f, 7.4317983e-05f, 7.9147585e-05f, 8.4291040e-05f, 8.9768747e-05f, 9.5602426e-05f, 0.00010181521f, 0.00010843174f, 0.00011547824f, 0.00012298267f, 0.00013097477f, 0.00013948625f, 0.00014855085f, 0.00015820453f, 0.00016848555f, 0.00017943469f, 0.00019109536f, 0.00020351382f, 0.00021673929f, 0.00023082423f, 0.00024582449f, 0.00026179955f, 0.00027881276f, 0.00029693158f, 0.00031622787f, 0.00033677814f, 0.00035866388f, 0.00038197188f, 0.00040679456f, 0.00043323036f, 0.00046138411f, 0.00049136745f, 0.00052329927f, 0.00055730621f, 0.00059352311f, 0.00063209358f, 0.00067317058f, 0.00071691700f, 0.00076350630f, 0.00081312324f, 0.00086596457f, 0.00092223983f, 0.00098217216f, 0.0010459992f, 0.0011139742f, 0.0011863665f, 0.0012634633f, 0.0013455702f, 0.0014330129f, 0.0015261382f, 0.0016253153f, 0.0017309374f, 0.0018434235f, 0.0019632195f, 0.0020908006f, 0.0022266726f, 0.0023713743f, 0.0025254795f, 0.0026895994f, 0.0028643847f, 0.0030505286f, 0.0032487691f, 0.0034598925f, 0.0036847358f, 0.0039241906f, 0.0041792066f, 0.0044507950f, 0.0047400328f, 0.0050480668f, 0.0053761186f, 0.0057254891f, 0.0060975636f, 0.0064938176f, 0.0069158225f, 0.0073652516f, 0.0078438871f, 0.0083536271f, 0.0088964928f, 0.009474637f, 0.010090352f, 0.010746080f, 0.011444421f, 0.012188144f, 0.012980198f, 0.013823725f, 0.014722068f, 0.015678791f, 0.016697687f, 0.017782797f, 0.018938423f, 0.020169149f, 0.021479854f, 0.022875735f, 0.024362330f, 0.025945531f, 0.027631618f, 0.029427276f, 0.031339626f, 0.033376252f, 0.035545228f, 0.037855157f, 0.040315199f, 0.042935108f, 0.045725273f, 0.048696758f, 0.051861348f, 0.055231591f, 0.058820850f, 0.062643361f, 0.066714279f, 0.071049749f, 0.075666962f, 0.080584227f, 0.085821044f, 0.091398179f, 0.097337747f, 0.10366330f, 0.11039993f, 0.11757434f, 0.12521498f, 0.13335215f, 0.14201813f, 0.15124727f, 0.16107617f, 0.17154380f, 0.18269168f, 0.19456402f, 0.20720788f, 0.22067342f, 0.23501402f, 0.25028656f, 0.26655159f, 0.28387361f, 0.30232132f, 0.32196786f, 0.34289114f, 0.36517414f, 0.38890521f, 0.41417847f, 0.44109412f, 0.46975890f, 0.50028648f, 0.53279791f, 0.56742212f, 0.60429640f, 0.64356699f, 0.68538959f, 0.72993007f, 0.77736504f, 0.82788260f, 0.88168307f, 0.9389798f, 1.0f}; + +// @OPTIMIZE: if you want to replace this bresenham line-drawing routine, +// note that you must produce bit-identical output to decode correctly; +// this specific sequence of operations is specified in the spec (it's +// drawing integer-quantized frequency-space lines that the encoder +// expects to be exactly the same) +// ... also, isn't the whole point of Bresenham's algorithm to NOT +// have to divide in the setup? sigh. +#ifndef STB_VORBIS_NO_DEFER_FLOOR +#define LINE_OP(a, b) a *= b +#else +#define LINE_OP(a, b) a = b +#endif + +#ifdef STB_VORBIS_DIVIDE_TABLE +#define DIVTAB_NUMER 32 +#define DIVTAB_DENOM 64 +int8 integer_divide_table[DIVTAB_NUMER][DIVTAB_DENOM]; // 2KB +#endif + +static __forceinline void draw_line(float *output, int x0, int y0, int x1, int y1, int n) { + int dy = y1 - y0; + int adx = x1 - x0; + int ady = abs(dy); + int base; + int x = x0, y = y0; + int err = 0; + int sy; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (adx < DIVTAB_DENOM && ady < DIVTAB_NUMER) { + if (dy < 0) { + base = -integer_divide_table[ady][adx]; + sy = base - 1; + } else { + base = integer_divide_table[ady][adx]; + sy = base + 1; + } + } else { + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base + 1; + } +#else + base = dy / adx; + if (dy < 0) + sy = base - 1; + else + sy = base + 1; +#endif + ady -= abs(base) * adx; + if (x1 > n) + x1 = n; + if (x < x1) { + LINE_OP(output[x], inverse_db_table[y & 255]); + for (++x; x < x1; ++x) { + err += ady; + if (err >= adx) { + err -= adx; + y += sy; + } else + y += base; + LINE_OP(output[x], inverse_db_table[y & 255]); + } + } +} + +static int residue_decode(vorb *f, Codebook *book, float *target, int offset, int n, int rtype) { + int k; + if (rtype == 0) { + int step = n / book->dimensions; + for (k = 0; k < step; ++k) + if (!codebook_decode_step(f, book, target + offset + k, n - offset - k, step)) + return FALSE; + } else { + for (k = 0; k < n;) { + if (!codebook_decode(f, book, target + offset, n - k)) + return FALSE; + k += book->dimensions; + offset += book->dimensions; + } + } + return TRUE; +} + +// n is 1/2 of the blocksize -- +// specification: "Correct per-vector decode length is [n]/2" +static void decode_residue(vorb *f, float *residue_buffers[], int ch, int n, int rn, uint8 *do_not_decode) { + int i, j, pass; + Residue *r = f->residue_config + rn; + int rtype = f->residue_types[rn]; + int c = r->classbook; + int classwords = f->codebooks[c].dimensions; + unsigned int actual_size = rtype == 2 ? n * 2 : n; + unsigned int limit_r_begin = (r->begin < actual_size ? r->begin : actual_size); + unsigned int limit_r_end = (r->end < actual_size ? r->end : actual_size); + int n_read = limit_r_end - limit_r_begin; + int part_read = n_read / r->part_size; + int temp_alloc_point = temp_alloc_save(f); +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + uint8 ***part_classdata = (uint8 ***)temp_block_array(f, f->channels, part_read * sizeof(**part_classdata)); +#else + int **classifications = (int **)temp_block_array(f, f->channels, part_read * sizeof(**classifications)); +#endif + + CHECK(f); + + for (i = 0; i < ch; ++i) + if (!do_not_decode[i]) + memset(residue_buffers[i], 0, sizeof(float) * n); + + if (rtype == 2 && ch != 1) { + for (j = 0; j < ch; ++j) + if (!do_not_decode[j]) + break; + if (j == ch) + goto done; + + for (pass = 0; pass < 8; ++pass) { + int pcount = 0, class_set = 0; + if (ch == 2) { + while (pcount < part_read) { + int z = r->begin + pcount * r->part_size; + int c_inter = (z & 1), p_inter = z >> 1; + if (pass == 0) { + Codebook *c = f->codebooks + r->classbook; + int q; + DECODE(q, f, c); + if (q == EOP) + goto done; +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; +#else + for (i = classwords - 1; i >= 0; --i) { + classifications[0][i + pcount] = q % r->classifications; + q /= r->classifications; + } +#endif + } + for (i = 0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount * r->part_size; +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; +#else + int c = classifications[0][pcount]; +#endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; +#ifdef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; +#else + // saves 1% + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; +#endif + } else { + z += r->part_size; + c_inter = z & 1; + p_inter = z >> 1; + } + } +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; +#endif + } + } else if (ch > 2) { + while (pcount < part_read) { + int z = r->begin + pcount * r->part_size; + int c_inter = z % ch, p_inter = z / ch; + if (pass == 0) { + Codebook *c = f->codebooks + r->classbook; + int q; + DECODE(q, f, c); + if (q == EOP) + goto done; +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[0][class_set] = r->classdata[q]; +#else + for (i = classwords - 1; i >= 0; --i) { + classifications[0][i + pcount] = q % r->classifications; + q /= r->classifications; + } +#endif + } + for (i = 0; i < classwords && pcount < part_read; ++i, ++pcount) { + int z = r->begin + pcount * r->part_size; +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[0][class_set][i]; +#else + int c = classifications[0][pcount]; +#endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + Codebook *book = f->codebooks + b; + if (!codebook_decode_deinterleave_repeat(f, book, residue_buffers, ch, &c_inter, &p_inter, n, r->part_size)) + goto done; + } else { + z += r->part_size; + c_inter = z % ch; + p_inter = z / ch; + } + } +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; +#endif + } + } + } + goto done; + } + CHECK(f); + + for (pass = 0; pass < 8; ++pass) { + int pcount = 0, class_set = 0; + while (pcount < part_read) { + if (pass == 0) { + for (j = 0; j < ch; ++j) { + if (!do_not_decode[j]) { + Codebook *c = f->codebooks + r->classbook; + int temp; + DECODE(temp, f, c); + if (temp == EOP) + goto done; +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + part_classdata[j][class_set] = r->classdata[temp]; +#else + for (i = classwords - 1; i >= 0; --i) { + classifications[j][i + pcount] = temp % r->classifications; + temp /= r->classifications; + } +#endif + } + } + } + for (i = 0; i < classwords && pcount < part_read; ++i, ++pcount) { + for (j = 0; j < ch; ++j) { + if (!do_not_decode[j]) { +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + int c = part_classdata[j][class_set][i]; +#else + int c = classifications[j][pcount]; +#endif + int b = r->residue_books[c][pass]; + if (b >= 0) { + float *target = residue_buffers[j]; + int offset = r->begin + pcount * r->part_size; + int n = r->part_size; + Codebook *book = f->codebooks + b; + if (!residue_decode(f, book, target, offset, n, rtype)) + goto done; + } + } + } + } +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + ++class_set; +#endif + } + } +done: + CHECK(f); +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + temp_free(f, part_classdata); +#else + temp_free(f, classifications); +#endif + temp_alloc_restore(f, temp_alloc_point); +} + +#if 0 +// slow way for debugging +void inverse_mdct_slow(float *buffer, int n) +{ + int i,j; + int n2 = n >> 1; + float *x = (float *) malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i=0; i < n; ++i) { + float acc = 0; + for (j=0; j < n2; ++j) + // formula from paper: + //acc += n/4.0f * x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + // formula from wikipedia + //acc += 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + // these are equivalent, except the formula from the paper inverts the multiplier! + // however, what actually works is NO MULTIPLIER!?! + //acc += 64 * 2.0f / n2 * x[j] * (float) cos(M_PI/n2 * (i + 0.5 + n2/2)*(j + 0.5)); + acc += x[j] * (float) cos(M_PI / 2 / n * (2 * i + 1 + n/2.0)*(2*j+1)); + buffer[i] = acc; + } + free(x); +} +#elif 0 +// same as above, but just barely able to run in real time on modern machines +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) { + float mcos[16384]; + int i, j; + int n2 = n >> 1, nmask = (n << 2) - 1; + float *x = (float *)malloc(sizeof(*x) * n2); + memcpy(x, buffer, sizeof(*x) * n2); + for (i = 0; i < 4 * n; ++i) + mcos[i] = (float)cos(M_PI / 2 * i / n); + + for (i = 0; i < n; ++i) { + float acc = 0; + for (j = 0; j < n2; ++j) + acc += x[j] * mcos[(2 * i + 1 + n2) * (2 * j + 1) & nmask]; + buffer[i] = acc; + } + free(x); +} +#elif 0 +// transform to use a slow dct-iv; this is STILL basically trivial, +// but only requires half as many ops +void dct_iv_slow(float *buffer, int n) { + float mcos[16384]; + float x[2048]; + int i, j; + int n2 = n >> 1, nmask = (n << 3) - 1; + memcpy(x, buffer, sizeof(*x) * n); + for (i = 0; i < 8 * n; ++i) + mcos[i] = (float)cos(M_PI / 4 * i / n); + for (i = 0; i < n; ++i) { + float acc = 0; + for (j = 0; j < n; ++j) + acc += x[j] * mcos[((2 * i + 1) * (2 * j + 1)) & nmask]; + buffer[i] = acc; + } +} + +void inverse_mdct_slow(float *buffer, int n, vorb *f, int blocktype) { + int i, n4 = n >> 2, n2 = n >> 1, n3_4 = n - n4; + float temp[4096]; + + memcpy(temp, buffer, n2 * sizeof(float)); + dct_iv_slow(temp, n2); // returns -c'-d, a-b' + + for (i = 0; i < n4; ++i) + buffer[i] = temp[i + n4]; // a-b' + for (; i < n3_4; ++i) + buffer[i] = -temp[n3_4 - i - 1]; // b-a', c+d' + for (; i < n; ++i) + buffer[i] = -temp[i - n3_4]; // c'+d +} +#endif + +#ifndef LIBVORBIS_MDCT +#define LIBVORBIS_MDCT 0 +#endif + +#if LIBVORBIS_MDCT +// directly call the vorbis MDCT using an interface documented +// by Jeff Roberts... useful for performance comparison +typedef struct +{ + int n; + int log2n; + + float *trig; + int *bitrev; + + float scale; +} mdct_lookup; + +extern void mdct_init(mdct_lookup *lookup, int n); +extern void mdct_clear(mdct_lookup *l); +extern void mdct_backward(mdct_lookup *init, float *in, float *out); + +mdct_lookup M1, M2; + +void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) { + mdct_lookup *M; + if (M1.n == n) + M = &M1; + else if (M2.n == n) + M = &M2; + else if (M1.n == 0) { + mdct_init(&M1, n); + M = &M1; + } else { + if (M2.n) + __asm int 3; + mdct_init(&M2, n); + M = &M2; + } + + mdct_backward(M, buffer, buffer); +} +#endif + +// the following were split out into separate functions while optimizing; +// they could be pushed back up but eh. __forceinline showed no change; +// they're probably already being inlined. +static void imdct_step3_iter0_loop(int n, float *e, int i_off, int k_off, float *A) { + float *ee0 = e + i_off; + float *ee2 = ee0 + k_off; + int i; + + assert((n & 3) == 0); + for (i = (n >> 2); i > 0; --i) { + float k00_20, k01_21; + k00_20 = ee0[0] - ee2[0]; + k01_21 = ee0[-1] - ee2[-1]; + ee0[0] += ee2[0]; // ee0[ 0] = ee0[ 0] + ee2[ 0]; + ee0[-1] += ee2[-1]; // ee0[-1] = ee0[-1] + ee2[-1]; + ee2[0] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-1] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-2] - ee2[-2]; + k01_21 = ee0[-3] - ee2[-3]; + ee0[-2] += ee2[-2]; // ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] += ee2[-3]; // ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-3] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-4] - ee2[-4]; + k01_21 = ee0[-5] - ee2[-5]; + ee0[-4] += ee2[-4]; // ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] += ee2[-5]; // ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-5] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + + k00_20 = ee0[-6] - ee2[-6]; + k01_21 = ee0[-7] - ee2[-7]; + ee0[-6] += ee2[-6]; // ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] += ee2[-7]; // ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = k00_20 * A[0] - k01_21 * A[1]; + ee2[-7] = k01_21 * A[0] + k00_20 * A[1]; + A += 8; + ee0 -= 8; + ee2 -= 8; + } +} + +static void imdct_step3_inner_r_loop(int lim, float *e, int d0, int k_off, float *A, int k1) { + int i; + float k00_20, k01_21; + + float *e0 = e + d0; + float *e2 = e0 + k_off; + + for (i = lim >> 2; i > 0; --i) { + k00_20 = e0[-0] - e2[-0]; + k01_21 = e0[-1] - e2[-1]; + e0[-0] += e2[-0]; // e0[-0] = e0[-0] + e2[-0]; + e0[-1] += e2[-1]; // e0[-1] = e0[-1] + e2[-1]; + e2[-0] = (k00_20)*A[0] - (k01_21)*A[1]; + e2[-1] = (k01_21)*A[0] + (k00_20)*A[1]; + + A += k1; + + k00_20 = e0[-2] - e2[-2]; + k01_21 = e0[-3] - e2[-3]; + e0[-2] += e2[-2]; // e0[-2] = e0[-2] + e2[-2]; + e0[-3] += e2[-3]; // e0[-3] = e0[-3] + e2[-3]; + e2[-2] = (k00_20)*A[0] - (k01_21)*A[1]; + e2[-3] = (k01_21)*A[0] + (k00_20)*A[1]; + + A += k1; + + k00_20 = e0[-4] - e2[-4]; + k01_21 = e0[-5] - e2[-5]; + e0[-4] += e2[-4]; // e0[-4] = e0[-4] + e2[-4]; + e0[-5] += e2[-5]; // e0[-5] = e0[-5] + e2[-5]; + e2[-4] = (k00_20)*A[0] - (k01_21)*A[1]; + e2[-5] = (k01_21)*A[0] + (k00_20)*A[1]; + + A += k1; + + k00_20 = e0[-6] - e2[-6]; + k01_21 = e0[-7] - e2[-7]; + e0[-6] += e2[-6]; // e0[-6] = e0[-6] + e2[-6]; + e0[-7] += e2[-7]; // e0[-7] = e0[-7] + e2[-7]; + e2[-6] = (k00_20)*A[0] - (k01_21)*A[1]; + e2[-7] = (k01_21)*A[0] + (k00_20)*A[1]; + + e0 -= 8; + e2 -= 8; + + A += k1; + } +} + +static void imdct_step3_inner_s_loop(int n, float *e, int i_off, int k_off, float *A, int a_off, int k0) { + int i; + float A0 = A[0]; + float A1 = A[0 + 1]; + float A2 = A[0 + a_off]; + float A3 = A[0 + a_off + 1]; + float A4 = A[0 + a_off * 2 + 0]; + float A5 = A[0 + a_off * 2 + 1]; + float A6 = A[0 + a_off * 3 + 0]; + float A7 = A[0 + a_off * 3 + 1]; + + float k00, k11; + + float *ee0 = e + i_off; + float *ee2 = ee0 + k_off; + + for (i = n; i > 0; --i) { + k00 = ee0[0] - ee2[0]; + k11 = ee0[-1] - ee2[-1]; + ee0[0] = ee0[0] + ee2[0]; + ee0[-1] = ee0[-1] + ee2[-1]; + ee2[0] = (k00)*A0 - (k11)*A1; + ee2[-1] = (k11)*A0 + (k00)*A1; + + k00 = ee0[-2] - ee2[-2]; + k11 = ee0[-3] - ee2[-3]; + ee0[-2] = ee0[-2] + ee2[-2]; + ee0[-3] = ee0[-3] + ee2[-3]; + ee2[-2] = (k00)*A2 - (k11)*A3; + ee2[-3] = (k11)*A2 + (k00)*A3; + + k00 = ee0[-4] - ee2[-4]; + k11 = ee0[-5] - ee2[-5]; + ee0[-4] = ee0[-4] + ee2[-4]; + ee0[-5] = ee0[-5] + ee2[-5]; + ee2[-4] = (k00)*A4 - (k11)*A5; + ee2[-5] = (k11)*A4 + (k00)*A5; + + k00 = ee0[-6] - ee2[-6]; + k11 = ee0[-7] - ee2[-7]; + ee0[-6] = ee0[-6] + ee2[-6]; + ee0[-7] = ee0[-7] + ee2[-7]; + ee2[-6] = (k00)*A6 - (k11)*A7; + ee2[-7] = (k11)*A6 + (k00)*A7; + + ee0 -= k0; + ee2 -= k0; + } +} + +static __forceinline void iter_54(float *z) { + float k00, k11, k22, k33; + float y0, y1, y2, y3; + + k00 = z[0] - z[-4]; + y0 = z[0] + z[-4]; + y2 = z[-2] + z[-6]; + k22 = z[-2] - z[-6]; + + z[-0] = y0 + y2; // z0 + z4 + z2 + z6 + z[-2] = y0 - y2; // z0 + z4 - z2 - z6 + + // done with y0,y2 + + k33 = z[-3] - z[-7]; + + z[-4] = k00 + k33; // z0 - z4 + z3 - z7 + z[-6] = k00 - k33; // z0 - z4 - z3 + z7 + + // done with k33 + + k11 = z[-1] - z[-5]; + y1 = z[-1] + z[-5]; + y3 = z[-3] + z[-7]; + + z[-1] = y1 + y3; // z1 + z5 + z3 + z7 + z[-3] = y1 - y3; // z1 + z5 - z3 - z7 + z[-5] = k11 - k22; // z1 - z5 + z2 - z6 + z[-7] = k11 + k22; // z1 - z5 - z2 + z6 +} + +static void imdct_step3_inner_s_loop_ld654(int n, float *e, int i_off, float *A, int base_n) { + int a_off = base_n >> 3; + float A2 = A[0 + a_off]; + float *z = e + i_off; + float *base = z - 16 * n; + + while (z > base) { + float k00, k11; + + k00 = z[-0] - z[-8]; + k11 = z[-1] - z[-9]; + z[-0] = z[-0] + z[-8]; + z[-1] = z[-1] + z[-9]; + z[-8] = k00; + z[-9] = k11; + + k00 = z[-2] - z[-10]; + k11 = z[-3] - z[-11]; + z[-2] = z[-2] + z[-10]; + z[-3] = z[-3] + z[-11]; + z[-10] = (k00 + k11) * A2; + z[-11] = (k11 - k00) * A2; + + k00 = z[-12] - z[-4]; // reverse to avoid a unary negation + k11 = z[-5] - z[-13]; + z[-4] = z[-4] + z[-12]; + z[-5] = z[-5] + z[-13]; + z[-12] = k11; + z[-13] = k00; + + k00 = z[-14] - z[-6]; // reverse to avoid a unary negation + k11 = z[-7] - z[-15]; + z[-6] = z[-6] + z[-14]; + z[-7] = z[-7] + z[-15]; + z[-14] = (k00 + k11) * A2; + z[-15] = (k00 - k11) * A2; + + iter_54(z); + iter_54(z - 8); + z -= 16; + } +} + +static void inverse_mdct(float *buffer, int n, vorb *f, int blocktype) { + int n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int ld; + // @OPTIMIZE: reduce register pressure by using fewer variables? + int save_point = temp_alloc_save(f); + float *buf2 = (float *)temp_alloc(f, n2 * sizeof(*buf2)); + float *u = NULL, *v = NULL; + // twiddle factors + float *A = f->A[blocktype]; + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // See notes about bugs in that paper in less-optimal implementation 'inverse_mdct_old' after this function. + + // kernel from paper + + // merged: + // copy and reflect spectral data + // step 0 + + // note that it turns out that the items added together during + // this step are, in fact, being added to themselves (as reflected + // by step 0). inexplicable inefficiency! this became obvious + // once I combined the passes. + + // so there's a missing 'times 2' here (for adding X to itself). + // this propagates through linearly to the end, where the numbers + // are 1/2 too small, and need to be compensated for. + + { + float *d, *e, *AA, *e_stop; + d = &buf2[n2 - 2]; + AA = A; + e = &buffer[0]; + e_stop = &buffer[n2]; + while (e != e_stop) { + d[1] = (e[0] * AA[0] - e[2] * AA[1]); + d[0] = (e[0] * AA[1] + e[2] * AA[0]); + d -= 2; + AA += 2; + e += 4; + } + + e = &buffer[n2 - 3]; + while (d >= buf2) { + d[1] = (-e[2] * AA[0] - -e[0] * AA[1]); + d[0] = (-e[2] * AA[1] + -e[0] * AA[0]); + d -= 2; + AA += 2; + e -= 4; + } + } + + // now we use symbolic names for these, so that we can + // possibly swap their meaning as we change which operations + // are in place + + u = buffer; + v = buf2; + + // step 2 (paper output is w, now u) + // this could be in place, but the data ends up in the wrong + // place... _somebody_'s got to swap it, so this is nominated + { + float *AA = &A[n2 - 8]; + float *d0, *d1, *e0, *e1; + + e0 = &v[n4]; + e1 = &v[0]; + + d0 = &u[n4]; + d1 = &u[0]; + + while (AA >= A) { + float v40_20, v41_21; + + v41_21 = e0[1] - e1[1]; + v40_20 = e0[0] - e1[0]; + d0[1] = e0[1] + e1[1]; + d0[0] = e0[0] + e1[0]; + d1[1] = v41_21 * AA[4] - v40_20 * AA[5]; + d1[0] = v40_20 * AA[4] + v41_21 * AA[5]; + + v41_21 = e0[3] - e1[3]; + v40_20 = e0[2] - e1[2]; + d0[3] = e0[3] + e1[3]; + d0[2] = e0[2] + e1[2]; + d1[3] = v41_21 * AA[0] - v40_20 * AA[1]; + d1[2] = v40_20 * AA[0] + v41_21 * AA[1]; + + AA -= 8; + + d0 += 4; + d1 += 4; + e0 += 4; + e1 += 4; + } + } + + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + + // optimized step 3: + + // the original step3 loop can be nested r inside s or s inside r; + // it's written originally as s inside r, but this is dumb when r + // iterates many times, and s few. So I have two copies of it and + // switch between them halfway. + + // this is iteration 0 of step 3 + imdct_step3_iter0_loop(n >> 4, u, n2 - 1 - n4 * 0, -(n >> 3), A); + imdct_step3_iter0_loop(n >> 4, u, n2 - 1 - n4 * 1, -(n >> 3), A); + + // this is iteration 1 of step 3 + imdct_step3_inner_r_loop(n >> 5, u, n2 - 1 - n8 * 0, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2 - 1 - n8 * 1, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2 - 1 - n8 * 2, -(n >> 4), A, 16); + imdct_step3_inner_r_loop(n >> 5, u, n2 - 1 - n8 * 3, -(n >> 4), A, 16); + + l = 2; + for (; l < (ld - 3) >> 1; ++l) { + int k0 = n >> (l + 2), k0_2 = k0 >> 1; + int lim = 1 << (l + 1); + int i; + for (i = 0; i < lim; ++i) + imdct_step3_inner_r_loop(n >> (l + 4), u, n2 - 1 - k0 * i, -k0_2, A, 1 << (l + 3)); + } + + for (; l < ld - 6; ++l) { + int k0 = n >> (l + 2), k1 = 1 << (l + 3), k0_2 = k0 >> 1; + int rlim = n >> (l + 6), r; + int lim = 1 << (l + 1); + int i_off; + float *A0 = A; + i_off = n2 - 1; + for (r = rlim; r > 0; --r) { + imdct_step3_inner_s_loop(lim, u, i_off, -k0_2, A0, k1, k0); + A0 += k1 * 4; + i_off -= 8; + } + } + + // iterations with count: + // ld-6,-5,-4 all interleaved together + // the big win comes from getting rid of needless flops + // due to the constants on pass 5 & 4 being all 1 and 0; + // combining them to be simultaneous to improve cache made little difference + imdct_step3_inner_s_loop_ld654(n >> 5, u, n2 - 1, A, n); + + // output is u + + // step 4, 5, and 6 + // cannot be in-place because of step 5 + { + uint16 *bitrev = f->bit_reverse[blocktype]; + // weirdly, I'd have thought reading sequentially and writing + // erratically would have been better than vice-versa, but in + // fact that's not what my testing showed. (That is, with + // j = bitreverse(i), do you read i and write j, or read j and write i.) + + float *d0 = &v[n4 - 4]; + float *d1 = &v[n2 - 4]; + while (d0 >= v) { + int k4; + + k4 = bitrev[0]; + d1[3] = u[k4 + 0]; + d1[2] = u[k4 + 1]; + d0[3] = u[k4 + 2]; + d0[2] = u[k4 + 3]; + + k4 = bitrev[1]; + d1[1] = u[k4 + 0]; + d1[0] = u[k4 + 1]; + d0[1] = u[k4 + 2]; + d0[0] = u[k4 + 3]; + + d0 -= 4; + d1 -= 4; + bitrev += 2; + } + } + // (paper output is u, now v) + + // data must be in buf2 + assert(v == buf2); + + // step 7 (paper output is v, now v) + // this is now in place + { + float *C = f->C[blocktype]; + float *d, *e; + + d = v; + e = v + n2 - 4; + + while (d < e) { + float a02, a11, b0, b1, b2, b3; + + a02 = d[0] - e[2]; + a11 = d[1] + e[3]; + + b0 = C[1] * a02 + C[0] * a11; + b1 = C[1] * a11 - C[0] * a02; + + b2 = d[0] + e[2]; + b3 = d[1] - e[3]; + + d[0] = b2 + b0; + d[1] = b3 + b1; + e[2] = b2 - b0; + e[3] = b1 - b3; + + a02 = d[2] - e[0]; + a11 = d[3] + e[1]; + + b0 = C[3] * a02 + C[2] * a11; + b1 = C[3] * a11 - C[2] * a02; + + b2 = d[2] + e[0]; + b3 = d[3] - e[1]; + + d[2] = b2 + b0; + d[3] = b3 + b1; + e[0] = b2 - b0; + e[1] = b1 - b3; + + C += 4; + d += 4; + e -= 4; + } + } + + // data must be in buf2 + + // step 8+decode (paper output is X, now buffer) + // this generates pairs of data a la 8 and pushes them directly through + // the decode kernel (pushing rather than pulling) to avoid having + // to make another pass later + + // this cannot POSSIBLY be in place, so we refer to the buffers directly + + { + float *d0, *d1, *d2, *d3; + + float *B = f->B[blocktype] + n2 - 8; + float *e = buf2 + n2 - 8; + d0 = &buffer[0]; + d1 = &buffer[n2 - 4]; + d2 = &buffer[n2]; + d3 = &buffer[n - 4]; + while (e >= v) { + float p0, p1, p2, p3; + + p3 = e[6] * B[7] - e[7] * B[6]; + p2 = -e[6] * B[6] - e[7] * B[7]; + + d0[0] = p3; + d1[3] = -p3; + d2[0] = p2; + d3[3] = p2; + + p1 = e[4] * B[5] - e[5] * B[4]; + p0 = -e[4] * B[4] - e[5] * B[5]; + + d0[1] = p1; + d1[2] = -p1; + d2[1] = p0; + d3[2] = p0; + + p3 = e[2] * B[3] - e[3] * B[2]; + p2 = -e[2] * B[2] - e[3] * B[3]; + + d0[2] = p3; + d1[1] = -p3; + d2[2] = p2; + d3[1] = p2; + + p1 = e[0] * B[1] - e[1] * B[0]; + p0 = -e[0] * B[0] - e[1] * B[1]; + + d0[3] = p1; + d1[0] = -p1; + d2[3] = p0; + d3[0] = p0; + + B -= 8; + e -= 8; + d0 += 4; + d2 += 4; + d1 -= 4; + d3 -= 4; + } + } + + temp_free(f, buf2); + temp_alloc_restore(f, save_point); +} + +#if 0 +// this is the original version of the above code, if you want to optimize it from scratch +void inverse_mdct_naive(float *buffer, int n) +{ + float s; + float A[1 << 12], B[1 << 12], C[1 << 11]; + int i,k,k2,k4, n2 = n >> 1, n4 = n >> 2, n8 = n >> 3, l; + int n3_4 = n - n4, ld; + // how can they claim this only uses N words?! + // oh, because they're only used sparsely, whoops + float u[1 << 13], X[1 << 13], v[1 << 13], w[1 << 13]; + // set up twiddle factors + + for (k=k2=0; k < n4; ++k,k2+=2) { + A[k2 ] = (float) cos(4*k*M_PI/n); + A[k2+1] = (float) -sin(4*k*M_PI/n); + B[k2 ] = (float) cos((k2+1)*M_PI/n/2); + B[k2+1] = (float) sin((k2+1)*M_PI/n/2); + } + for (k=k2=0; k < n8; ++k,k2+=2) { + C[k2 ] = (float) cos(2*(k2+1)*M_PI/n); + C[k2+1] = (float) -sin(2*(k2+1)*M_PI/n); + } + + // IMDCT algorithm from "The use of multirate filter banks for coding of high quality digital audio" + // Note there are bugs in that pseudocode, presumably due to them attempting + // to rename the arrays nicely rather than representing the way their actual + // implementation bounces buffers back and forth. As a result, even in the + // "some formulars corrected" version, a direct implementation fails. These + // are noted below as "paper bug". + + // copy and reflect spectral data + for (k=0; k < n2; ++k) u[k] = buffer[k]; + for ( ; k < n ; ++k) u[k] = -buffer[n - k - 1]; + // kernel from paper + // step 1 + for (k=k2=k4=0; k < n4; k+=1, k2+=2, k4+=4) { + v[n-k4-1] = (u[k4] - u[n-k4-1]) * A[k2] - (u[k4+2] - u[n-k4-3])*A[k2+1]; + v[n-k4-3] = (u[k4] - u[n-k4-1]) * A[k2+1] + (u[k4+2] - u[n-k4-3])*A[k2]; + } + // step 2 + for (k=k4=0; k < n8; k+=1, k4+=4) { + w[n2+3+k4] = v[n2+3+k4] + v[k4+3]; + w[n2+1+k4] = v[n2+1+k4] + v[k4+1]; + w[k4+3] = (v[n2+3+k4] - v[k4+3])*A[n2-4-k4] - (v[n2+1+k4]-v[k4+1])*A[n2-3-k4]; + w[k4+1] = (v[n2+1+k4] - v[k4+1])*A[n2-4-k4] + (v[n2+3+k4]-v[k4+3])*A[n2-3-k4]; + } + // step 3 + ld = ilog(n) - 1; // ilog is off-by-one from normal definitions + for (l=0; l < ld-3; ++l) { + int k0 = n >> (l+2), k1 = 1 << (l+3); + int rlim = n >> (l+4), r4, r; + int s2lim = 1 << (l+2), s2; + for (r=r4=0; r < rlim; r4+=4,++r) { + for (s2=0; s2 < s2lim; s2+=2) { + u[n-1-k0*s2-r4] = w[n-1-k0*s2-r4] + w[n-1-k0*(s2+1)-r4]; + u[n-3-k0*s2-r4] = w[n-3-k0*s2-r4] + w[n-3-k0*(s2+1)-r4]; + u[n-1-k0*(s2+1)-r4] = (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1] + - (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1+1]; + u[n-3-k0*(s2+1)-r4] = (w[n-3-k0*s2-r4] - w[n-3-k0*(s2+1)-r4]) * A[r*k1] + + (w[n-1-k0*s2-r4] - w[n-1-k0*(s2+1)-r4]) * A[r*k1+1]; + } + } + if (l+1 < ld-3) { + // paper bug: ping-ponging of u&w here is omitted + memcpy(w, u, sizeof(u)); + } + } + + // step 4 + for (i=0; i < n8; ++i) { + int j = bit_reverse(i) >> (32-ld+3); + assert(j < n8); + if (i == j) { + // paper bug: original code probably swapped in place; if copying, + // need to directly copy in this case + int i8 = i << 3; + v[i8+1] = u[i8+1]; + v[i8+3] = u[i8+3]; + v[i8+5] = u[i8+5]; + v[i8+7] = u[i8+7]; + } else if (i < j) { + int i8 = i << 3, j8 = j << 3; + v[j8+1] = u[i8+1], v[i8+1] = u[j8 + 1]; + v[j8+3] = u[i8+3], v[i8+3] = u[j8 + 3]; + v[j8+5] = u[i8+5], v[i8+5] = u[j8 + 5]; + v[j8+7] = u[i8+7], v[i8+7] = u[j8 + 7]; + } + } + // step 5 + for (k=0; k < n2; ++k) { + w[k] = v[k*2+1]; + } + // step 6 + for (k=k2=k4=0; k < n8; ++k, k2 += 2, k4 += 4) { + u[n-1-k2] = w[k4]; + u[n-2-k2] = w[k4+1]; + u[n3_4 - 1 - k2] = w[k4+2]; + u[n3_4 - 2 - k2] = w[k4+3]; + } + // step 7 + for (k=k2=0; k < n8; ++k, k2 += 2) { + v[n2 + k2 ] = ( u[n2 + k2] + u[n-2-k2] + C[k2+1]*(u[n2+k2]-u[n-2-k2]) + C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n-2 - k2] = ( u[n2 + k2] + u[n-2-k2] - C[k2+1]*(u[n2+k2]-u[n-2-k2]) - C[k2]*(u[n2+k2+1]+u[n-2-k2+1]))/2; + v[n2+1+ k2] = ( u[n2+1+k2] - u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + v[n-1 - k2] = (-u[n2+1+k2] + u[n-1-k2] + C[k2+1]*(u[n2+1+k2]+u[n-1-k2]) - C[k2]*(u[n2+k2]-u[n-2-k2]))/2; + } + // step 8 + for (k=k2=0; k < n4; ++k,k2 += 2) { + X[k] = v[k2+n2]*B[k2 ] + v[k2+1+n2]*B[k2+1]; + X[n2-1-k] = v[k2+n2]*B[k2+1] - v[k2+1+n2]*B[k2 ]; + } + + // decode kernel to output + // determined the following value experimentally + // (by first figuring out what made inverse_mdct_slow work); then matching that here + // (probably vorbis encoder premultiplies by n or n/2, to save it on the decoder?) + s = 0.5; // theoretically would be n4 + + // [[[ note! the s value of 0.5 is compensated for by the B[] in the current code, + // so it needs to use the "old" B values to behave correctly, or else + // set s to 1.0 ]]] + for (i=0; i < n4 ; ++i) buffer[i] = s * X[i+n4]; + for ( ; i < n3_4; ++i) buffer[i] = -s * X[n3_4 - i - 1]; + for ( ; i < n ; ++i) buffer[i] = -s * X[i - n3_4]; +} +#endif + +static float *get_window(vorb *f, int len) { + len <<= 1; + if (len == f->blocksize_0) + return f->window[0]; + if (len == f->blocksize_1) + return f->window[1]; + return NULL; +} + +#ifndef STB_VORBIS_NO_DEFER_FLOOR +typedef int16 YTYPE; +#else +typedef int YTYPE; +#endif +static int do_floor(vorb *f, Mapping *map, int i, int n, float *target, YTYPE *finalY, uint8 *step2_flag) { + int n2 = n >> 1; + int s = map->chan[i].mux, floor; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + int j, q; + int lx = 0, ly = finalY[0] * g->floor1_multiplier; + for (q = 1; q < g->values; ++q) { + j = g->sorted_order[q]; +#ifndef STB_VORBIS_NO_DEFER_FLOOR + if (finalY[j] >= 0) +#else + if (step2_flag[j]) +#endif + { + int hy = finalY[j] * g->floor1_multiplier; + int hx = g->Xlist[j]; + if (lx != hx) + draw_line(target, lx, ly, hx, hy, n2); + CHECK(f); + lx = hx, ly = hy; + } + } + if (lx < n2) { + // optimization of: draw_line(target, lx,ly, n,ly, n2); + for (j = lx; j < n2; ++j) + LINE_OP(target[j], inverse_db_table[ly]); + CHECK(f); + } + } + return TRUE; +} + +// The meaning of "left" and "right" +// +// For a given frame: +// we compute samples from 0..n +// window_center is n/2 +// we'll window and mix the samples from left_start to left_end with data from the previous frame +// all of the samples from left_end to right_start can be output without mixing; however, +// this interval is 0-length except when transitioning between short and long frames +// all of the samples from right_start to right_end need to be mixed with the next frame, +// which we don't have, so those get saved in a buffer +// frame N's right_end-right_start, the number of samples to mix with the next frame, +// has to be the same as frame N+1's left_end-left_start (which they are by +// construction) + +static int vorbis_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) { + Mode *m; + int i, n, prev, next, window_center; + f->channel_buffer_start = f->channel_buffer_end = 0; + +retry: + if (f->eof) + return FALSE; + if (!maybe_start_packet(f)) + return FALSE; + // check packet type + if (get_bits(f, 1) != 0) { + if (IS_PUSH_MODE(f)) + return error(f, VORBIS_bad_packet_type); + while (EOP != get8_packet(f)); + goto retry; + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + i = get_bits(f, ilog(f->mode_count - 1)); + if (i == EOP) + return FALSE; + if (i >= f->mode_count) + return FALSE; + *mode = i; + m = f->mode_config + i; + if (m->blockflag) { + n = f->blocksize_1; + prev = get_bits(f, 1); + next = get_bits(f, 1); + } else { + prev = next = 0; + n = f->blocksize_0; + } + + // WINDOWING + + window_center = n >> 1; + if (m->blockflag && !prev) { + *p_left_start = (n - f->blocksize_0) >> 2; + *p_left_end = (n + f->blocksize_0) >> 2; + } else { + *p_left_start = 0; + *p_left_end = window_center; + } + if (m->blockflag && !next) { + *p_right_start = (n * 3 - f->blocksize_0) >> 2; + *p_right_end = (n * 3 + f->blocksize_0) >> 2; + } else { + *p_right_start = window_center; + *p_right_end = n; + } + + return TRUE; +} + +static int vorbis_decode_packet_rest(vorb *f, int *len, Mode *m, int left_start, int left_end, int right_start, int right_end, int *p_left) { + Mapping *map; + int i, j, k, n, n2; + int zero_channel[256]; + int really_zero_channel[256]; + + // WINDOWING + + n = f->blocksize[m->blockflag]; + map = &f->mapping[m->mapping]; + + // FLOORS + n2 = n >> 1; + + CHECK(f); + + for (i = 0; i < f->channels; ++i) { + int s = map->chan[i].mux, floor; + zero_channel[i] = FALSE; + floor = map->submap_floor[s]; + if (f->floor_types[floor] == 0) { + return error(f, VORBIS_invalid_stream); + } else { + Floor1 *g = &f->floor_config[floor].floor1; + if (get_bits(f, 1)) { + short *finalY; + uint8 step2_flag[256]; + static int range_list[4] = {256, 128, 86, 64}; + int range = range_list[g->floor1_multiplier - 1]; + int offset = 2; + finalY = f->finalY[i]; + finalY[0] = get_bits(f, ilog(range) - 1); + finalY[1] = get_bits(f, ilog(range) - 1); + for (j = 0; j < g->partitions; ++j) { + int pclass = g->partition_class_list[j]; + int cdim = g->class_dimensions[pclass]; + int cbits = g->class_subclasses[pclass]; + int csub = (1 << cbits) - 1; + int cval = 0; + if (cbits) { + Codebook *c = f->codebooks + g->class_masterbooks[pclass]; + DECODE(cval, f, c); + } + for (k = 0; k < cdim; ++k) { + int book = g->subclass_books[pclass][cval & csub]; + cval = cval >> cbits; + if (book >= 0) { + int temp; + Codebook *c = f->codebooks + book; + DECODE(temp, f, c); + finalY[offset++] = temp; + } else + finalY[offset++] = 0; + } + } + if (f->valid_bits == INVALID_BITS) + goto error; // behavior according to spec + step2_flag[0] = step2_flag[1] = 1; + for (j = 2; j < g->values; ++j) { + int low, high, pred, highroom, lowroom, room, val; + low = g->neighbors[j][0]; + high = g->neighbors[j][1]; + // neighbors(g->Xlist, j, &low, &high); + pred = predict_point(g->Xlist[j], g->Xlist[low], g->Xlist[high], finalY[low], finalY[high]); + val = finalY[j]; + highroom = range - pred; + lowroom = pred; + if (highroom < lowroom) + room = highroom * 2; + else + room = lowroom * 2; + if (val) { + step2_flag[low] = step2_flag[high] = 1; + step2_flag[j] = 1; + if (val >= room) + if (highroom > lowroom) + finalY[j] = val - lowroom + pred; + else + finalY[j] = pred - val + highroom - 1; + else if (val & 1) + finalY[j] = pred - ((val + 1) >> 1); + else + finalY[j] = pred + (val >> 1); + } else { + step2_flag[j] = 0; + finalY[j] = pred; + } + } + +#ifdef STB_VORBIS_NO_DEFER_FLOOR + do_floor(f, map, i, n, f->floor_buffers[i], finalY, step2_flag); +#else + // defer final floor computation until _after_ residue + for (j = 0; j < g->values; ++j) { + if (!step2_flag[j]) + finalY[j] = -1; + } +#endif + } else { + error: + zero_channel[i] = TRUE; + } + // So we just defer everything else to later + + // at this point we've decoded the floor into buffer + } + } + CHECK(f); + // at this point we've decoded all floors + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + + // re-enable coupled channels if necessary + memcpy(really_zero_channel, zero_channel, sizeof(really_zero_channel[0]) * f->channels); + for (i = 0; i < map->coupling_steps; ++i) + if (!zero_channel[map->chan[i].magnitude] || !zero_channel[map->chan[i].angle]) { + zero_channel[map->chan[i].magnitude] = zero_channel[map->chan[i].angle] = FALSE; + } + + CHECK(f); + // RESIDUE DECODE + for (i = 0; i < map->submaps; ++i) { + float *residue_buffers[STB_VORBIS_MAX_CHANNELS]; + int r; + uint8 do_not_decode[256]; + int ch = 0; + for (j = 0; j < f->channels; ++j) { + if (map->chan[j].mux == i) { + if (zero_channel[j]) { + do_not_decode[ch] = TRUE; + residue_buffers[ch] = NULL; + } else { + do_not_decode[ch] = FALSE; + residue_buffers[ch] = f->channel_buffers[j]; + } + ++ch; + } + } + r = map->submap_residue[i]; + decode_residue(f, residue_buffers, ch, n2, r, do_not_decode); + } + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + CHECK(f); + + // INVERSE COUPLING + for (i = map->coupling_steps - 1; i >= 0; --i) { + int n2 = n >> 1; + float *m = f->channel_buffers[map->chan[i].magnitude]; + float *a = f->channel_buffers[map->chan[i].angle]; + for (j = 0; j < n2; ++j) { + float a2, m2; + if (m[j] > 0) + if (a[j] > 0) + m2 = m[j], a2 = m[j] - a[j]; + else + a2 = m[j], m2 = m[j] + a[j]; + else if (a[j] > 0) + m2 = m[j], a2 = m[j] + a[j]; + else + a2 = m[j], m2 = m[j] - a[j]; + m[j] = m2; + a[j] = a2; + } + } + CHECK(f); + + // finish decoding the floors +#ifndef STB_VORBIS_NO_DEFER_FLOOR + for (i = 0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + do_floor(f, map, i, n, f->channel_buffers[i], f->finalY[i], NULL); + } + } +#else + for (i = 0; i < f->channels; ++i) { + if (really_zero_channel[i]) { + memset(f->channel_buffers[i], 0, sizeof(*f->channel_buffers[i]) * n2); + } else { + for (j = 0; j < n2; ++j) + f->channel_buffers[i][j] *= f->floor_buffers[i][j]; + } + } +#endif + + // INVERSE MDCT + CHECK(f); + for (i = 0; i < f->channels; ++i) + inverse_mdct(f->channel_buffers[i], n, f, m->blockflag); + CHECK(f); + + // this shouldn't be necessary, unless we exited on an error + // and want to flush to get to the next packet + flush_packet(f); + + if (f->first_decode) { + // assume we start so first non-discarded sample is sample 0 + // this isn't to spec, but spec would require us to read ahead + // and decode the size of all current frames--could be done, + // but presumably it's not a commonly used feature + f->current_loc = -n2; // start of first frame is positioned for discard + // we might have to discard samples "from" the next frame too, + // if we're lapping a large block then a small at the start? + f->discard_samples_deferred = n - right_end; + f->current_loc_valid = TRUE; + f->first_decode = FALSE; + } else if (f->discard_samples_deferred) { + if (f->discard_samples_deferred >= right_start - left_start) { + f->discard_samples_deferred -= (right_start - left_start); + left_start = right_start; + *p_left = left_start; + } else { + left_start += f->discard_samples_deferred; + *p_left = left_start; + f->discard_samples_deferred = 0; + } + } else if (f->previous_length == 0 && f->current_loc_valid) { + // we're recovering from a seek... that means we're going to discard + // the samples from this packet even though we know our position from + // the last page header, so we need to update the position based on + // the discarded samples here + // but wait, the code below is going to add this in itself even + // on a discard, so we don't need to do it here... + } + + // check if we have ogg information about the sample # for this packet + if (f->last_seg_which == f->end_seg_with_known_loc) { + // if we have a valid current loc, and this is final: + if (f->current_loc_valid && (f->page_flag & PAGEFLAG_last_page)) { + uint32 current_end = f->known_loc_for_packet; + // then let's infer the size of the (probably) short final frame + if (current_end < f->current_loc + (right_end - left_start)) { + if (current_end < f->current_loc) { + // negative truncation, that's impossible! + *len = 0; + } else { + *len = current_end - f->current_loc; + } + *len += left_start; // this doesn't seem right, but has no ill effect on my test files + if (*len > right_end) + *len = right_end; // this should never happen + f->current_loc += *len; + return TRUE; + } + } + // otherwise, just set our sample loc + // guess that the ogg granule pos refers to the _middle_ of the + // last frame? + // set f->current_loc to the position of left_start + f->current_loc = f->known_loc_for_packet - (n2 - left_start); + f->current_loc_valid = TRUE; + } + if (f->current_loc_valid) + f->current_loc += (right_start - left_start); + + if (f->alloc.alloc_buffer) + assert(f->alloc.alloc_buffer_length_in_bytes == f->temp_offset); + *len = right_end; // ignore samples after the window goes to 0 + CHECK(f); + + return TRUE; +} + +static int vorbis_decode_packet(vorb *f, int *len, int *p_left, int *p_right) { + int mode, left_end, right_end; + if (!vorbis_decode_initial(f, p_left, &left_end, p_right, &right_end, &mode)) + return 0; + return vorbis_decode_packet_rest(f, len, f->mode_config + mode, *p_left, left_end, *p_right, right_end, p_left); +} + +static int vorbis_finish_frame(stb_vorbis *f, int len, int left, int right) { + int prev, i, j; + // we use right&left (the start of the right- and left-window sin()-regions) + // to determine how much to return, rather than inferring from the rules + // (same result, clearer code); 'left' indicates where our sin() window + // starts, therefore where the previous window's right edge starts, and + // therefore where to start mixing from the previous buffer. 'right' + // indicates where our sin() ending-window starts, therefore that's where + // we start saving, and where our returned-data ends. + + // mixin from previous window + if (f->previous_length) { + int i, j, n = f->previous_length; + float *w = get_window(f, n); + if (w == NULL) + return 0; + for (i = 0; i < f->channels; ++i) { + for (j = 0; j < n; ++j) + f->channel_buffers[i][left + j] = + f->channel_buffers[i][left + j] * w[j] + + f->previous_window[i][j] * w[n - 1 - j]; + } + } + + prev = f->previous_length; + + // last half of this data becomes previous window + f->previous_length = len - right; + + // @OPTIMIZE: could avoid this copy by double-buffering the + // output (flipping previous_window with channel_buffers), but + // then previous_window would have to be 2x as large, and + // channel_buffers couldn't be temp mem (although they're NOT + // currently temp mem, they could be (unless we want to level + // performance by spreading out the computation)) + for (i = 0; i < f->channels; ++i) + for (j = 0; right + j < len; ++j) + f->previous_window[i][j] = f->channel_buffers[i][right + j]; + + if (!prev) + // there was no previous packet, so this data isn't valid... + // this isn't entirely true, only the would-have-overlapped data + // isn't valid, but this seems to be what the spec requires + return 0; + + // truncate a short frame + if (len < right) + right = len; + + f->samples_output += right - left; + + return right - left; +} + +static int vorbis_pump_first_frame(stb_vorbis *f) { + int len, right, left, res; + res = vorbis_decode_packet(f, &len, &left, &right); + if (res) + vorbis_finish_frame(f, len, left, right); + return res; +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API +static int is_whole_packet_present(stb_vorbis *f) { + // make sure that we have the packet available before continuing... + // this requires a full ogg parse, but we know we can fetch from f->stream + + // instead of coding this out explicitly, we could save the current read state, + // read the next packet with get8() until end-of-packet, check f->eof, then + // reset the state? but that would be slower, esp. since we'd have over 256 bytes + // of state to restore (primarily the page segment table) + + int s = f->next_seg, first = TRUE; + uint8 *p = f->stream; + + if (s != -1) { // if we're not starting the packet with a 'continue on next page' flag + for (; s < f->segment_count; ++s) { + p += f->segments[s]; + if (f->segments[s] < 255) // stop at first short segment + break; + } + // either this continues, or it ends it... + if (s == f->segment_count) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) + return error(f, VORBIS_need_more_data); + first = FALSE; + } + for (; s == -1;) { + uint8 *q; + int n; + + // check that we have the page header ready + if (p + 26 >= f->stream_end) + return error(f, VORBIS_need_more_data); + // validate the page + if (memcmp(p, ogg_page_header, 4)) + return error(f, VORBIS_invalid_stream); + if (p[4] != 0) + return error(f, VORBIS_invalid_stream); + if (first) { // the first segment must NOT have 'continued_packet', later ones MUST + if (f->previous_length) + if ((p[5] & PAGEFLAG_continued_packet)) + return error(f, VORBIS_invalid_stream); + // if no previous length, we're resynching, so we can come in on a continued-packet, + // which we'll just drop + } else { + if (!(p[5] & PAGEFLAG_continued_packet)) + return error(f, VORBIS_invalid_stream); + } + n = p[26]; // segment counts + q = p + 27; // q points to segment table + p = q + n; // advance past header + // make sure we've read the segment table + if (p > f->stream_end) + return error(f, VORBIS_need_more_data); + for (s = 0; s < n; ++s) { + p += q[s]; + if (q[s] < 255) + break; + } + if (s == n) + s = -1; // set 'crosses page' flag + if (p > f->stream_end) + return error(f, VORBIS_need_more_data); + first = FALSE; + } + return TRUE; +} +#endif // !STB_VORBIS_NO_PUSHDATA_API + +static int start_decoder(vorb *f) { + uint8 header[6], x, y; + int len, i, j, k, max_submaps = 0; + int longest_floorlist = 0; + + // first page, first packet + f->first_decode = TRUE; + + if (!start_page(f)) + return FALSE; + // validate page flag + if (!(f->page_flag & PAGEFLAG_first_page)) + return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_last_page) + return error(f, VORBIS_invalid_first_page); + if (f->page_flag & PAGEFLAG_continued_packet) + return error(f, VORBIS_invalid_first_page); + // check for expected packet length + if (f->segment_count != 1) + return error(f, VORBIS_invalid_first_page); + if (f->segments[0] != 30) { + // check for the Ogg skeleton fishead identifying header to refine our error + if (f->segments[0] == 64 && + getn(f, header, 6) && + header[0] == 'f' && + header[1] == 'i' && + header[2] == 's' && + header[3] == 'h' && + header[4] == 'e' && + header[5] == 'a' && + get8(f) == 'd' && + get8(f) == '\0') + return error(f, VORBIS_ogg_skeleton_not_supported); + else + return error(f, VORBIS_invalid_first_page); + } + + // read packet + // check packet header + if (get8(f) != VORBIS_packet_id) + return error(f, VORBIS_invalid_first_page); + if (!getn(f, header, 6)) + return error(f, VORBIS_unexpected_eof); + if (!vorbis_validate(header)) + return error(f, VORBIS_invalid_first_page); + // vorbis_version + if (get32(f) != 0) + return error(f, VORBIS_invalid_first_page); + f->channels = get8(f); + if (!f->channels) + return error(f, VORBIS_invalid_first_page); + if (f->channels > STB_VORBIS_MAX_CHANNELS) + return error(f, VORBIS_too_many_channels); + f->sample_rate = get32(f); + if (!f->sample_rate) + return error(f, VORBIS_invalid_first_page); + get32(f); // bitrate_maximum + get32(f); // bitrate_nominal + get32(f); // bitrate_minimum + x = get8(f); + { + int log0, log1; + log0 = x & 15; + log1 = x >> 4; + f->blocksize_0 = 1 << log0; + f->blocksize_1 = 1 << log1; + if (log0 < 6 || log0 > 13) + return error(f, VORBIS_invalid_setup); + if (log1 < 6 || log1 > 13) + return error(f, VORBIS_invalid_setup); + if (log0 > log1) + return error(f, VORBIS_invalid_setup); + } + + // framing_flag + x = get8(f); + if (!(x & 1)) + return error(f, VORBIS_invalid_first_page); + + // second packet! + if (!start_page(f)) + return FALSE; + + if (!start_packet(f)) + return FALSE; + + if (!next_segment(f)) + return FALSE; + + if (get8_packet(f) != VORBIS_packet_comment) + return error(f, VORBIS_invalid_setup); + for (i = 0; i < 6; ++i) + header[i] = get8_packet(f); + if (!vorbis_validate(header)) + return error(f, VORBIS_invalid_setup); + // file vendor + len = get32_packet(f); + f->vendor = (char *)setup_malloc(f, sizeof(char) * (len + 1)); + if (f->vendor == NULL) + return error(f, VORBIS_outofmem); + for (i = 0; i < len; ++i) { + f->vendor[i] = get8_packet(f); + } + f->vendor[len] = (char)'\0'; + // user comments + f->comment_list_length = get32_packet(f); + if (f->comment_list_length > 0) { + f->comment_list = (char **)setup_malloc(f, sizeof(char *) * (f->comment_list_length)); + if (f->comment_list == NULL) + return error(f, VORBIS_outofmem); + } + + for (i = 0; i < f->comment_list_length; ++i) { + len = get32_packet(f); + f->comment_list[i] = (char *)setup_malloc(f, sizeof(char) * (len + 1)); + if (f->comment_list[i] == NULL) + return error(f, VORBIS_outofmem); + + for (j = 0; j < len; ++j) { + f->comment_list[i][j] = get8_packet(f); + } + f->comment_list[i][len] = (char)'\0'; + } + + // framing_flag + x = get8_packet(f); + if (!(x & 1)) + return error(f, VORBIS_invalid_setup); + + skip(f, f->bytes_in_seg); + f->bytes_in_seg = 0; + + do { + len = next_segment(f); + skip(f, len); + f->bytes_in_seg = 0; + } while (len); + + // third packet! + if (!start_packet(f)) + return FALSE; + +#ifndef STB_VORBIS_NO_PUSHDATA_API + if (IS_PUSH_MODE(f)) { + if (!is_whole_packet_present(f)) { + // convert error in ogg header to write type + if (f->error == VORBIS_invalid_stream) + f->error = VORBIS_invalid_setup; + return FALSE; + } + } +#endif + + crc32_init(); // always init it, to avoid multithread race conditions + + if (get8_packet(f) != VORBIS_packet_setup) + return error(f, VORBIS_invalid_setup); + for (i = 0; i < 6; ++i) + header[i] = get8_packet(f); + if (!vorbis_validate(header)) + return error(f, VORBIS_invalid_setup); + + // codebooks + + f->codebook_count = get_bits(f, 8) + 1; + f->codebooks = (Codebook *)setup_malloc(f, sizeof(*f->codebooks) * f->codebook_count); + if (f->codebooks == NULL) + return error(f, VORBIS_outofmem); + memset(f->codebooks, 0, sizeof(*f->codebooks) * f->codebook_count); + for (i = 0; i < f->codebook_count; ++i) { + uint32 *values; + int ordered, sorted_count; + int total = 0; + uint8 *lengths; + Codebook *c = f->codebooks + i; + CHECK(f); + x = get_bits(f, 8); + if (x != 0x42) + return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); + if (x != 0x43) + return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); + if (x != 0x56) + return error(f, VORBIS_invalid_setup); + x = get_bits(f, 8); + c->dimensions = (get_bits(f, 8) << 8) + x; + x = get_bits(f, 8); + y = get_bits(f, 8); + c->entries = (get_bits(f, 8) << 16) + (y << 8) + x; + ordered = get_bits(f, 1); + c->sparse = ordered ? 0 : get_bits(f, 1); + + if (c->dimensions == 0 && c->entries != 0) + return error(f, VORBIS_invalid_setup); + + if (c->sparse) + lengths = (uint8 *)setup_temp_malloc(f, c->entries); + else + lengths = c->codeword_lengths = (uint8 *)setup_malloc(f, c->entries); + + if (!lengths) + return error(f, VORBIS_outofmem); + + if (ordered) { + int current_entry = 0; + int current_length = get_bits(f, 5) + 1; + while (current_entry < c->entries) { + int limit = c->entries - current_entry; + int n = get_bits(f, ilog(limit)); + if (current_length >= 32) + return error(f, VORBIS_invalid_setup); + if (current_entry + n > (int)c->entries) { + return error(f, VORBIS_invalid_setup); + } + memset(lengths + current_entry, current_length, n); + current_entry += n; + ++current_length; + } + } else { + for (j = 0; j < c->entries; ++j) { + int present = c->sparse ? get_bits(f, 1) : 1; + if (present) { + lengths[j] = get_bits(f, 5) + 1; + ++total; + if (lengths[j] == 32) + return error(f, VORBIS_invalid_setup); + } else { + lengths[j] = NO_CODE; + } + } + } + + if (c->sparse && total >= c->entries >> 2) { + // convert sparse items to non-sparse! + if (c->entries > (int)f->setup_temp_memory_required) + f->setup_temp_memory_required = c->entries; + + c->codeword_lengths = (uint8 *)setup_malloc(f, c->entries); + if (c->codeword_lengths == NULL) + return error(f, VORBIS_outofmem); + memcpy(c->codeword_lengths, lengths, c->entries); + setup_temp_free(f, lengths, c->entries); // note this is only safe if there have been no intervening temp mallocs! + lengths = c->codeword_lengths; + c->sparse = 0; + } + + // compute the size of the sorted tables + if (c->sparse) { + sorted_count = total; + } else { + sorted_count = 0; +#ifndef STB_VORBIS_NO_HUFFMAN_BINARY_SEARCH + for (j = 0; j < c->entries; ++j) + if (lengths[j] > STB_VORBIS_FAST_HUFFMAN_LENGTH && lengths[j] != NO_CODE) + ++sorted_count; +#endif + } + + c->sorted_entries = sorted_count; + values = NULL; + + CHECK(f); + if (!c->sparse) { + c->codewords = (uint32 *)setup_malloc(f, sizeof(c->codewords[0]) * c->entries); + if (!c->codewords) + return error(f, VORBIS_outofmem); + } else { + unsigned int size; + if (c->sorted_entries) { + c->codeword_lengths = (uint8 *)setup_malloc(f, c->sorted_entries); + if (!c->codeword_lengths) + return error(f, VORBIS_outofmem); + c->codewords = (uint32 *)setup_temp_malloc(f, sizeof(*c->codewords) * c->sorted_entries); + if (!c->codewords) + return error(f, VORBIS_outofmem); + values = (uint32 *)setup_temp_malloc(f, sizeof(*values) * c->sorted_entries); + if (!values) + return error(f, VORBIS_outofmem); + } + size = c->entries + (sizeof(*c->codewords) + sizeof(*values)) * c->sorted_entries; + if (size > f->setup_temp_memory_required) + f->setup_temp_memory_required = size; + } + + if (!compute_codewords(c, lengths, c->entries, values)) { + if (c->sparse) + setup_temp_free(f, values, 0); + return error(f, VORBIS_invalid_setup); + } + + if (c->sorted_entries) { + // allocate an extra slot for sentinels + c->sorted_codewords = (uint32 *)setup_malloc(f, sizeof(*c->sorted_codewords) * (c->sorted_entries + 1)); + if (c->sorted_codewords == NULL) + return error(f, VORBIS_outofmem); + // allocate an extra slot at the front so that c->sorted_values[-1] is defined + // so that we can catch that case without an extra if + c->sorted_values = (int *)setup_malloc(f, sizeof(*c->sorted_values) * (c->sorted_entries + 1)); + if (c->sorted_values == NULL) + return error(f, VORBIS_outofmem); + ++c->sorted_values; + c->sorted_values[-1] = -1; + compute_sorted_huffman(c, lengths, values); + } + + if (c->sparse) { + setup_temp_free(f, values, sizeof(*values) * c->sorted_entries); + setup_temp_free(f, c->codewords, sizeof(*c->codewords) * c->sorted_entries); + setup_temp_free(f, lengths, c->entries); + c->codewords = NULL; + } + + compute_accelerated_huffman(c); + + CHECK(f); + c->lookup_type = get_bits(f, 4); + if (c->lookup_type > 2) + return error(f, VORBIS_invalid_setup); + if (c->lookup_type > 0) { + uint16 *mults; + c->minimum_value = float32_unpack(get_bits(f, 32)); + c->delta_value = float32_unpack(get_bits(f, 32)); + c->value_bits = get_bits(f, 4) + 1; + c->sequence_p = get_bits(f, 1); + if (c->lookup_type == 1) { + int values = lookup1_values(c->entries, c->dimensions); + if (values < 0) + return error(f, VORBIS_invalid_setup); + c->lookup_values = (uint32)values; + } else { + c->lookup_values = c->entries * c->dimensions; + } + if (c->lookup_values == 0) + return error(f, VORBIS_invalid_setup); + mults = (uint16 *)setup_temp_malloc(f, sizeof(mults[0]) * c->lookup_values); + if (mults == NULL) + return error(f, VORBIS_outofmem); + for (j = 0; j < (int)c->lookup_values; ++j) { + int q = get_bits(f, c->value_bits); + if (q == EOP) { + setup_temp_free(f, mults, sizeof(mults[0]) * c->lookup_values); + return error(f, VORBIS_invalid_setup); + } + mults[j] = q; + } + +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + if (c->lookup_type == 1) { + int len, sparse = c->sparse; + float last = 0; + // pre-expand the lookup1-style multiplicands, to avoid a divide in the inner loop + if (sparse) { + if (c->sorted_entries == 0) + goto skip; + c->multiplicands = (codetype *)setup_malloc(f, sizeof(c->multiplicands[0]) * c->sorted_entries * c->dimensions); + } else + c->multiplicands = (codetype *)setup_malloc(f, sizeof(c->multiplicands[0]) * c->entries * c->dimensions); + if (c->multiplicands == NULL) { + setup_temp_free(f, mults, sizeof(mults[0]) * c->lookup_values); + return error(f, VORBIS_outofmem); + } + len = sparse ? c->sorted_entries : c->entries; + for (j = 0; j < len; ++j) { + unsigned int z = sparse ? c->sorted_values[j] : j; + unsigned int div = 1; + for (k = 0; k < c->dimensions; ++k) { + int off = (z / div) % c->lookup_values; + float val = mults[off]; + val = mults[off] * c->delta_value + c->minimum_value + last; + c->multiplicands[j * c->dimensions + k] = val; + if (c->sequence_p) + last = val; + if (k + 1 < c->dimensions) { + if (div > UINT_MAX / (unsigned int)c->lookup_values) { + setup_temp_free(f, mults, sizeof(mults[0]) * c->lookup_values); + return error(f, VORBIS_invalid_setup); + } + div *= c->lookup_values; + } + } + } + c->lookup_type = 2; + } else +#endif + { + float last = 0; + CHECK(f); + c->multiplicands = (codetype *)setup_malloc(f, sizeof(c->multiplicands[0]) * c->lookup_values); + if (c->multiplicands == NULL) { + setup_temp_free(f, mults, sizeof(mults[0]) * c->lookup_values); + return error(f, VORBIS_outofmem); + } + for (j = 0; j < (int)c->lookup_values; ++j) { + float val = mults[j] * c->delta_value + c->minimum_value + last; + c->multiplicands[j] = val; + if (c->sequence_p) + last = val; + } + } +#ifndef STB_VORBIS_DIVIDES_IN_CODEBOOK + skip:; +#endif + setup_temp_free(f, mults, sizeof(mults[0]) * c->lookup_values); + + CHECK(f); + } + CHECK(f); + } + + // time domain transfers (notused) + + x = get_bits(f, 6) + 1; + for (i = 0; i < x; ++i) { + uint32 z = get_bits(f, 16); + if (z != 0) + return error(f, VORBIS_invalid_setup); + } + + // Floors + f->floor_count = get_bits(f, 6) + 1; + f->floor_config = (Floor *)setup_malloc(f, f->floor_count * sizeof(*f->floor_config)); + if (f->floor_config == NULL) + return error(f, VORBIS_outofmem); + for (i = 0; i < f->floor_count; ++i) { + f->floor_types[i] = get_bits(f, 16); + if (f->floor_types[i] > 1) + return error(f, VORBIS_invalid_setup); + if (f->floor_types[i] == 0) { + Floor0 *g = &f->floor_config[i].floor0; + g->order = get_bits(f, 8); + g->rate = get_bits(f, 16); + g->bark_map_size = get_bits(f, 16); + g->amplitude_bits = get_bits(f, 6); + g->amplitude_offset = get_bits(f, 8); + g->number_of_books = get_bits(f, 4) + 1; + for (j = 0; j < g->number_of_books; ++j) + g->book_list[j] = get_bits(f, 8); + return error(f, VORBIS_feature_not_supported); + } else { + stbv__floor_ordering p[31 * 8 + 2]; + Floor1 *g = &f->floor_config[i].floor1; + int max_class = -1; + g->partitions = get_bits(f, 5); + for (j = 0; j < g->partitions; ++j) { + g->partition_class_list[j] = get_bits(f, 4); + if (g->partition_class_list[j] > max_class) + max_class = g->partition_class_list[j]; + } + for (j = 0; j <= max_class; ++j) { + g->class_dimensions[j] = get_bits(f, 3) + 1; + g->class_subclasses[j] = get_bits(f, 2); + if (g->class_subclasses[j]) { + g->class_masterbooks[j] = get_bits(f, 8); + if (g->class_masterbooks[j] >= f->codebook_count) + return error(f, VORBIS_invalid_setup); + } + for (k = 0; k < 1 << g->class_subclasses[j]; ++k) { + g->subclass_books[j][k] = get_bits(f, 8) - 1; + if (g->subclass_books[j][k] >= f->codebook_count) + return error(f, VORBIS_invalid_setup); + } + } + g->floor1_multiplier = get_bits(f, 2) + 1; + g->rangebits = get_bits(f, 4); + g->Xlist[0] = 0; + g->Xlist[1] = 1 << g->rangebits; + g->values = 2; + for (j = 0; j < g->partitions; ++j) { + int c = g->partition_class_list[j]; + for (k = 0; k < g->class_dimensions[c]; ++k) { + g->Xlist[g->values] = get_bits(f, g->rangebits); + ++g->values; + } + } + // precompute the sorting + for (j = 0; j < g->values; ++j) { + p[j].x = g->Xlist[j]; + p[j].id = j; + } + qsort(p, g->values, sizeof(p[0]), point_compare); + for (j = 0; j < g->values - 1; ++j) + if (p[j].x == p[j + 1].x) + return error(f, VORBIS_invalid_setup); + for (j = 0; j < g->values; ++j) + g->sorted_order[j] = (uint8)p[j].id; + // precompute the neighbors + for (j = 2; j < g->values; ++j) { + int low = 0, hi = 0; + neighbors(g->Xlist, j, &low, &hi); + g->neighbors[j][0] = low; + g->neighbors[j][1] = hi; + } + + if (g->values > longest_floorlist) + longest_floorlist = g->values; + } + } + + // Residue + f->residue_count = get_bits(f, 6) + 1; + f->residue_config = (Residue *)setup_malloc(f, f->residue_count * sizeof(f->residue_config[0])); + if (f->residue_config == NULL) + return error(f, VORBIS_outofmem); + memset(f->residue_config, 0, f->residue_count * sizeof(f->residue_config[0])); + for (i = 0; i < f->residue_count; ++i) { + uint8 residue_cascade[64]; + Residue *r = f->residue_config + i; + f->residue_types[i] = get_bits(f, 16); + if (f->residue_types[i] > 2) + return error(f, VORBIS_invalid_setup); + r->begin = get_bits(f, 24); + r->end = get_bits(f, 24); + if (r->end < r->begin) + return error(f, VORBIS_invalid_setup); + r->part_size = get_bits(f, 24) + 1; + r->classifications = get_bits(f, 6) + 1; + r->classbook = get_bits(f, 8); + if (r->classbook >= f->codebook_count) + return error(f, VORBIS_invalid_setup); + for (j = 0; j < r->classifications; ++j) { + uint8 high_bits = 0; + uint8 low_bits = get_bits(f, 3); + if (get_bits(f, 1)) + high_bits = get_bits(f, 5); + residue_cascade[j] = high_bits * 8 + low_bits; + } + r->residue_books = (short (*)[8])setup_malloc(f, sizeof(r->residue_books[0]) * r->classifications); + if (r->residue_books == NULL) + return error(f, VORBIS_outofmem); + for (j = 0; j < r->classifications; ++j) { + for (k = 0; k < 8; ++k) { + if (residue_cascade[j] & (1 << k)) { + r->residue_books[j][k] = get_bits(f, 8); + if (r->residue_books[j][k] >= f->codebook_count) + return error(f, VORBIS_invalid_setup); + } else { + r->residue_books[j][k] = -1; + } + } + } + // precompute the classifications[] array to avoid inner-loop mod/divide + // call it 'classdata' since we already have r->classifications + r->classdata = (uint8 **)setup_malloc(f, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + if (!r->classdata) + return error(f, VORBIS_outofmem); + memset(r->classdata, 0, sizeof(*r->classdata) * f->codebooks[r->classbook].entries); + for (j = 0; j < f->codebooks[r->classbook].entries; ++j) { + int classwords = f->codebooks[r->classbook].dimensions; + int temp = j; + r->classdata[j] = (uint8 *)setup_malloc(f, sizeof(r->classdata[j][0]) * classwords); + if (r->classdata[j] == NULL) + return error(f, VORBIS_outofmem); + for (k = classwords - 1; k >= 0; --k) { + r->classdata[j][k] = temp % r->classifications; + temp /= r->classifications; + } + } + } + + f->mapping_count = get_bits(f, 6) + 1; + f->mapping = (Mapping *)setup_malloc(f, f->mapping_count * sizeof(*f->mapping)); + if (f->mapping == NULL) + return error(f, VORBIS_outofmem); + memset(f->mapping, 0, f->mapping_count * sizeof(*f->mapping)); + for (i = 0; i < f->mapping_count; ++i) { + Mapping *m = f->mapping + i; + int mapping_type = get_bits(f, 16); + if (mapping_type != 0) + return error(f, VORBIS_invalid_setup); + m->chan = (MappingChannel *)setup_malloc(f, f->channels * sizeof(*m->chan)); + if (m->chan == NULL) + return error(f, VORBIS_outofmem); + if (get_bits(f, 1)) + m->submaps = get_bits(f, 4) + 1; + else + m->submaps = 1; + if (m->submaps > max_submaps) + max_submaps = m->submaps; + if (get_bits(f, 1)) { + m->coupling_steps = get_bits(f, 8) + 1; + if (m->coupling_steps > f->channels) + return error(f, VORBIS_invalid_setup); + for (k = 0; k < m->coupling_steps; ++k) { + m->chan[k].magnitude = get_bits(f, ilog(f->channels - 1)); + m->chan[k].angle = get_bits(f, ilog(f->channels - 1)); + if (m->chan[k].magnitude >= f->channels) + return error(f, VORBIS_invalid_setup); + if (m->chan[k].angle >= f->channels) + return error(f, VORBIS_invalid_setup); + if (m->chan[k].magnitude == m->chan[k].angle) + return error(f, VORBIS_invalid_setup); + } + } else + m->coupling_steps = 0; + + // reserved field + if (get_bits(f, 2)) + return error(f, VORBIS_invalid_setup); + if (m->submaps > 1) { + for (j = 0; j < f->channels; ++j) { + m->chan[j].mux = get_bits(f, 4); + if (m->chan[j].mux >= m->submaps) + return error(f, VORBIS_invalid_setup); + } + } else + // @SPECIFICATION: this case is missing from the spec + for (j = 0; j < f->channels; ++j) + m->chan[j].mux = 0; + + for (j = 0; j < m->submaps; ++j) { + get_bits(f, 8); // discard + m->submap_floor[j] = get_bits(f, 8); + m->submap_residue[j] = get_bits(f, 8); + if (m->submap_floor[j] >= f->floor_count) + return error(f, VORBIS_invalid_setup); + if (m->submap_residue[j] >= f->residue_count) + return error(f, VORBIS_invalid_setup); + } + } + + // Modes + f->mode_count = get_bits(f, 6) + 1; + for (i = 0; i < f->mode_count; ++i) { + Mode *m = f->mode_config + i; + m->blockflag = get_bits(f, 1); + m->windowtype = get_bits(f, 16); + m->transformtype = get_bits(f, 16); + m->mapping = get_bits(f, 8); + if (m->windowtype != 0) + return error(f, VORBIS_invalid_setup); + if (m->transformtype != 0) + return error(f, VORBIS_invalid_setup); + if (m->mapping >= f->mapping_count) + return error(f, VORBIS_invalid_setup); + } + + flush_packet(f); + + f->previous_length = 0; + + for (i = 0; i < f->channels; ++i) { + f->channel_buffers[i] = (float *)setup_malloc(f, sizeof(float) * f->blocksize_1); + f->previous_window[i] = (float *)setup_malloc(f, sizeof(float) * f->blocksize_1 / 2); + f->finalY[i] = (int16 *)setup_malloc(f, sizeof(int16) * longest_floorlist); + if (f->channel_buffers[i] == NULL || f->previous_window[i] == NULL || f->finalY[i] == NULL) + return error(f, VORBIS_outofmem); + memset(f->channel_buffers[i], 0, sizeof(float) * f->blocksize_1); +#ifdef STB_VORBIS_NO_DEFER_FLOOR + f->floor_buffers[i] = (float *)setup_malloc(f, sizeof(float) * f->blocksize_1 / 2); + if (f->floor_buffers[i] == NULL) + return error(f, VORBIS_outofmem); +#endif + } + + if (!init_blocksize(f, 0, f->blocksize_0)) + return FALSE; + if (!init_blocksize(f, 1, f->blocksize_1)) + return FALSE; + f->blocksize[0] = f->blocksize_0; + f->blocksize[1] = f->blocksize_1; + +#ifdef STB_VORBIS_DIVIDE_TABLE + if (integer_divide_table[1][1] == 0) + for (i = 0; i < DIVTAB_NUMER; ++i) + for (j = 1; j < DIVTAB_DENOM; ++j) + integer_divide_table[i][j] = i / j; +#endif + + // compute how much temporary memory is needed + + // 1. + { + uint32 imdct_mem = (f->blocksize_1 * sizeof(float) >> 1); + uint32 classify_mem; + int i, max_part_read = 0; + for (i = 0; i < f->residue_count; ++i) { + Residue *r = f->residue_config + i; + unsigned int actual_size = f->blocksize_1 / 2; + unsigned int limit_r_begin = r->begin < actual_size ? r->begin : actual_size; + unsigned int limit_r_end = r->end < actual_size ? r->end : actual_size; + int n_read = limit_r_end - limit_r_begin; + int part_read = n_read / r->part_size; + if (part_read > max_part_read) + max_part_read = part_read; + } +#ifndef STB_VORBIS_DIVIDES_IN_RESIDUE + classify_mem = f->channels * (sizeof(void *) + max_part_read * sizeof(uint8 *)); +#else + classify_mem = f->channels * (sizeof(void *) + max_part_read * sizeof(int *)); +#endif + + // maximum reasonable partition size is f->blocksize_1 + + f->temp_memory_required = classify_mem; + if (imdct_mem > f->temp_memory_required) + f->temp_memory_required = imdct_mem; + } + + if (f->alloc.alloc_buffer) { + assert(f->temp_offset == f->alloc.alloc_buffer_length_in_bytes); + // check if there's enough temp memory so we don't error later + if (f->setup_offset + sizeof(*f) + f->temp_memory_required > (unsigned)f->temp_offset) + return error(f, VORBIS_outofmem); + } + + // @TODO: stb_vorbis_seek_start expects first_audio_page_offset to point to a page + // without PAGEFLAG_continued_packet, so this either points to the first page, or + // the page after the end of the headers. It might be cleaner to point to a page + // in the middle of the headers, when that's the page where the first audio packet + // starts, but we'd have to also correctly skip the end of any continued packet in + // stb_vorbis_seek_start. + if (f->next_seg == -1) { + f->first_audio_page_offset = stb_vorbis_get_file_offset(f); + } else { + f->first_audio_page_offset = 0; + } + + return TRUE; +} + +static void vorbis_deinit(stb_vorbis *p) { + int i, j; + + setup_free(p, p->vendor); + for (i = 0; i < p->comment_list_length; ++i) { + setup_free(p, p->comment_list[i]); + } + setup_free(p, p->comment_list); + + if (p->residue_config) { + for (i = 0; i < p->residue_count; ++i) { + Residue *r = p->residue_config + i; + if (r->classdata) { + for (j = 0; j < p->codebooks[r->classbook].entries; ++j) + setup_free(p, r->classdata[j]); + setup_free(p, r->classdata); + } + setup_free(p, r->residue_books); + } + } + + if (p->codebooks) { + CHECK(p); + for (i = 0; i < p->codebook_count; ++i) { + Codebook *c = p->codebooks + i; + setup_free(p, c->codeword_lengths); + setup_free(p, c->multiplicands); + setup_free(p, c->codewords); + setup_free(p, c->sorted_codewords); + // c->sorted_values[-1] is the first entry in the array + setup_free(p, c->sorted_values ? c->sorted_values - 1 : NULL); + } + setup_free(p, p->codebooks); + } + setup_free(p, p->floor_config); + setup_free(p, p->residue_config); + if (p->mapping) { + for (i = 0; i < p->mapping_count; ++i) + setup_free(p, p->mapping[i].chan); + setup_free(p, p->mapping); + } + CHECK(p); + for (i = 0; i < p->channels && i < STB_VORBIS_MAX_CHANNELS; ++i) { + setup_free(p, p->channel_buffers[i]); + setup_free(p, p->previous_window[i]); +#ifdef STB_VORBIS_NO_DEFER_FLOOR + setup_free(p, p->floor_buffers[i]); +#endif + setup_free(p, p->finalY[i]); + } + for (i = 0; i < 2; ++i) { + setup_free(p, p->A[i]); + setup_free(p, p->B[i]); + setup_free(p, p->C[i]); + setup_free(p, p->window[i]); + setup_free(p, p->bit_reverse[i]); + } +#ifndef STB_VORBIS_NO_STDIO + if (p->close_on_free) + fclose(p->f); +#endif +} + +void stb_vorbis_close(stb_vorbis *p) { + if (p == NULL) + return; + vorbis_deinit(p); + setup_free(p, p); +} + +static void vorbis_init(stb_vorbis *p, const stb_vorbis_alloc *z) { + memset(p, 0, sizeof(*p)); // NULL out all malloc'd pointers to start + if (z) { + p->alloc = *z; + p->alloc.alloc_buffer_length_in_bytes &= ~7; + p->temp_offset = p->alloc.alloc_buffer_length_in_bytes; + } + p->eof = 0; + p->error = VORBIS__no_error; + p->stream = NULL; + p->codebooks = NULL; + p->page_crc_tests = -1; +#ifndef STB_VORBIS_NO_STDIO + p->close_on_free = FALSE; + p->f = NULL; +#endif +} + +int stb_vorbis_get_sample_offset(stb_vorbis *f) { + if (f->current_loc_valid) + return f->current_loc; + else + return -1; +} + +stb_vorbis_info stb_vorbis_get_info(stb_vorbis *f) { + stb_vorbis_info d; + d.channels = f->channels; + d.sample_rate = f->sample_rate; + d.setup_memory_required = f->setup_memory_required; + d.setup_temp_memory_required = f->setup_temp_memory_required; + d.temp_memory_required = f->temp_memory_required; + d.max_frame_size = f->blocksize_1 >> 1; + return d; +} + +stb_vorbis_comment stb_vorbis_get_comment(stb_vorbis *f) { + stb_vorbis_comment d; + d.vendor = f->vendor; + d.comment_list_length = f->comment_list_length; + d.comment_list = f->comment_list; + return d; +} + +int stb_vorbis_get_error(stb_vorbis *f) { + int e = f->error; + f->error = VORBIS__no_error; + return e; +} + +static stb_vorbis *vorbis_alloc(stb_vorbis *f) { + stb_vorbis *p = (stb_vorbis *)setup_malloc(f, sizeof(*p)); + return p; +} + +#ifndef STB_VORBIS_NO_PUSHDATA_API + +void stb_vorbis_flush_pushdata(stb_vorbis *f) { + f->previous_length = 0; + f->page_crc_tests = 0; + f->discard_samples_deferred = 0; + f->current_loc_valid = FALSE; + f->first_decode = FALSE; + f->samples_output = 0; + f->channel_buffer_start = 0; + f->channel_buffer_end = 0; +} + +static int vorbis_search_for_page_pushdata(vorb *f, uint8 *data, int data_len) { + int i, n; + for (i = 0; i < f->page_crc_tests; ++i) + f->scan[i].bytes_done = 0; + + // if we have room for more scans, search for them first, because + // they may cause us to stop early if their header is incomplete + if (f->page_crc_tests < STB_VORBIS_PUSHDATA_CRC_COUNT) { + if (data_len < 4) + return 0; + data_len -= 3; // need to look for 4-byte sequence, so don't miss + // one that straddles a boundary + for (i = 0; i < data_len; ++i) { + if (data[i] == 0x4f) { + if (0 == memcmp(data + i, ogg_page_header, 4)) { + int j, len; + uint32 crc; + // make sure we have the whole page header + if (i + 26 >= data_len || i + 27 + data[i + 26] >= data_len) { + // only read up to this page start, so hopefully we'll + // have the whole page header start next time + data_len = i; + break; + } + // ok, we have it all; compute the length of the page + len = 27 + data[i + 26]; + for (j = 0; j < data[i + 26]; ++j) + len += data[i + 27 + j]; + // scan everything up to the embedded crc (which we must 0) + crc = 0; + for (j = 0; j < 22; ++j) + crc = crc32_update(crc, data[i + j]); + // now process 4 0-bytes + for (; j < 26; ++j) + crc = crc32_update(crc, 0); + // len is the total number of bytes we need to scan + n = f->page_crc_tests++; + f->scan[n].bytes_left = len - j; + f->scan[n].crc_so_far = crc; + f->scan[n].goal_crc = data[i + 22] + (data[i + 23] << 8) + (data[i + 24] << 16) + (data[i + 25] << 24); + // if the last frame on a page is continued to the next, then + // we can't recover the sample_loc immediately + if (data[i + 27 + data[i + 26] - 1] == 255) + f->scan[n].sample_loc = ~0; + else + f->scan[n].sample_loc = data[i + 6] + (data[i + 7] << 8) + (data[i + 8] << 16) + (data[i + 9] << 24); + f->scan[n].bytes_done = i + j; + if (f->page_crc_tests == STB_VORBIS_PUSHDATA_CRC_COUNT) + break; + // keep going if we still have room for more + } + } + } + } + + for (i = 0; i < f->page_crc_tests;) { + uint32 crc; + int j; + int n = f->scan[i].bytes_done; + int m = f->scan[i].bytes_left; + if (m > data_len - n) + m = data_len - n; + // m is the bytes to scan in the current chunk + crc = f->scan[i].crc_so_far; + for (j = 0; j < m; ++j) + crc = crc32_update(crc, data[n + j]); + f->scan[i].bytes_left -= m; + f->scan[i].crc_so_far = crc; + if (f->scan[i].bytes_left == 0) { + // does it match? + if (f->scan[i].crc_so_far == f->scan[i].goal_crc) { + // Houston, we have page + data_len = n + m; // consumption amount is wherever that scan ended + f->page_crc_tests = -1; // drop out of page scan mode + f->previous_length = 0; // decode-but-don't-output one frame + f->next_seg = -1; // start a new page + f->current_loc = f->scan[i].sample_loc; // set the current sample location + // to the amount we'd have decoded had we decoded this page + f->current_loc_valid = f->current_loc != ~0U; + return data_len; + } + // delete entry + f->scan[i] = f->scan[--f->page_crc_tests]; + } else { + ++i; + } + } + + return data_len; +} + +// return value: number of bytes we used +int stb_vorbis_decode_frame_pushdata( + stb_vorbis *f, // the file we're decoding + const uint8 *data, + int data_len, // the memory available for decoding + int *channels, // place to write number of float * buffers + float ***output, // place to write float ** array of float * buffers + int *samples // place to write number of output samples +) { + int i; + int len, right, left; + + if (!IS_PUSH_MODE(f)) + return error(f, VORBIS_invalid_api_mixing); + + if (f->page_crc_tests >= 0) { + *samples = 0; + return vorbis_search_for_page_pushdata(f, (uint8 *)data, data_len); + } + + f->stream = (uint8 *)data; + f->stream_end = (uint8 *)data + data_len; + f->error = VORBIS__no_error; + + // check that we have the entire packet in memory + if (!is_whole_packet_present(f)) { + *samples = 0; + return 0; + } + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + // save the actual error we encountered + enum STBVorbisError error = f->error; + if (error == VORBIS_bad_packet_type) { + // flush and resynch + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) + break; + *samples = 0; + return (int)(f->stream - data); + } + if (error == VORBIS_continued_packet_flag_invalid) { + if (f->previous_length == 0) { + // we may be resynching, in which case it's ok to hit one + // of these; just discard the packet + f->error = VORBIS__no_error; + while (get8_packet(f) != EOP) + if (f->eof) + break; + *samples = 0; + return (int)(f->stream - data); + } + } + // if we get an error while parsing, what to do? + // well, it DEFINITELY won't work to continue from where we are! + stb_vorbis_flush_pushdata(f); + // restore the error that actually made us bail + f->error = error; + *samples = 0; + return 1; + } + + // success! + len = vorbis_finish_frame(f, len, left, right); + for (i = 0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + if (channels) + *channels = f->channels; + *samples = len; + *output = f->outputs; + return (int)(f->stream - data); +} + +stb_vorbis *stb_vorbis_open_pushdata( + const unsigned char *data, int data_len, // the memory available for decoding + int *data_used, // only defined if result is not NULL + int *error, + const stb_vorbis_alloc *alloc) { + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.stream = (uint8 *)data; + p.stream_end = (uint8 *)data + data_len; + p.push_mode = TRUE; + if (!start_decoder(&p)) { + if (p.eof) + *error = VORBIS_need_more_data; + else + *error = p.error; + return NULL; + } + f = vorbis_alloc(&p); + if (f) { + *f = p; + *data_used = (int)(f->stream - data); + *error = 0; + return f; + } else { + vorbis_deinit(&p); + return NULL; + } +} +#endif // STB_VORBIS_NO_PUSHDATA_API + +unsigned int stb_vorbis_get_file_offset(stb_vorbis *f) { +#ifndef STB_VORBIS_NO_PUSHDATA_API + if (f->push_mode) + return 0; +#endif + if (USE_MEMORY(f)) + return (unsigned int)(f->stream - f->stream_start); +#ifndef STB_VORBIS_NO_STDIO + return (unsigned int)(ftell(f->f) - f->f_start); +#endif +} + +#ifndef STB_VORBIS_NO_PULLDATA_API +// +// DATA-PULLING API +// + +static uint32 vorbis_find_page(stb_vorbis *f, uint32 *end, uint32 *last) { + for (;;) { + int n; + if (f->eof) + return 0; + n = get8(f); + if (n == 0x4f) { // page header candidate + unsigned int retry_loc = stb_vorbis_get_file_offset(f); + int i; + // check if we're off the end of a file_section stream + if (retry_loc - 25 > f->stream_len) + return 0; + // check the rest of the header + for (i = 1; i < 4; ++i) + if (get8(f) != ogg_page_header[i]) + break; + if (f->eof) + return 0; + if (i == 4) { + uint8 header[27]; + uint32 i, crc, goal, len; + for (i = 0; i < 4; ++i) + header[i] = ogg_page_header[i]; + for (; i < 27; ++i) + header[i] = get8(f); + if (f->eof) + return 0; + if (header[4] != 0) + goto invalid; + goal = header[22] + (header[23] << 8) + (header[24] << 16) + (header[25] << 24); + for (i = 22; i < 26; ++i) + header[i] = 0; + crc = 0; + for (i = 0; i < 27; ++i) + crc = crc32_update(crc, header[i]); + len = 0; + for (i = 0; i < header[26]; ++i) { + int s = get8(f); + crc = crc32_update(crc, s); + len += s; + } + if (len && f->eof) + return 0; + for (i = 0; i < len; ++i) + crc = crc32_update(crc, get8(f)); + // finished parsing probable page + if (crc == goal) { + // we could now check that it's either got the last + // page flag set, OR it's followed by the capture + // pattern, but I guess TECHNICALLY you could have + // a file with garbage between each ogg page and recover + // from it automatically? So even though that paranoia + // might decrease the chance of an invalid decode by + // another 2^32, not worth it since it would hose those + // invalid-but-useful files? + if (end) + *end = stb_vorbis_get_file_offset(f); + if (last) { + if (header[5] & 0x04) + *last = 1; + else + *last = 0; + } + set_file_offset(f, retry_loc - 1); + return 1; + } + } + invalid: + // not a valid page, so rewind and look for next one + set_file_offset(f, retry_loc); + } + } +} + +#define SAMPLE_unknown 0xffffffff + +// seeking is implemented with a binary search, which narrows down the range to +// 64K, before using a linear search (because finding the synchronization +// pattern can be expensive, and the chance we'd find the end page again is +// relatively high for small ranges) +// +// two initial interpolation-style probes are used at the start of the search +// to try to bound either side of the binary search sensibly, while still +// working in O(log n) time if they fail. + +static int get_seek_page_info(stb_vorbis *f, ProbedPage *z) { + uint8 header[27], lacing[255]; + int i, len; + + // record where the page starts + z->page_start = stb_vorbis_get_file_offset(f); + + // parse the header + getn(f, header, 27); + if (header[0] != 'O' || header[1] != 'g' || header[2] != 'g' || header[3] != 'S') + return 0; + getn(f, lacing, header[26]); + + // determine the length of the payload + len = 0; + for (i = 0; i < header[26]; ++i) + len += lacing[i]; + + // this implies where the page ends + z->page_end = z->page_start + 27 + header[26] + len; + + // read the last-decoded sample out of the data + z->last_decoded_sample = header[6] + (header[7] << 8) + (header[8] << 16) + (header[9] << 24); + + // restore file state to where we were + set_file_offset(f, z->page_start); + return 1; +} + +// rarely used function to seek back to the preceding page while finding the +// start of a packet +static int go_to_page_before(stb_vorbis *f, unsigned int limit_offset) { + unsigned int previous_safe, end; + + // now we want to seek back 64K from the limit + if (limit_offset >= 65536 && limit_offset - 65536 >= f->first_audio_page_offset) + previous_safe = limit_offset - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + + while (vorbis_find_page(f, &end, NULL)) { + if (end >= limit_offset && stb_vorbis_get_file_offset(f) < limit_offset) + return 1; + set_file_offset(f, end); + } + + return 0; +} + +// implements the search logic for finding a page and starting decoding. if +// the function succeeds, current_loc_valid will be true and current_loc will +// be less than or equal to the provided sample number (the closer the +// better). +static int seek_to_sample_coarse(stb_vorbis *f, uint32 sample_number) { + ProbedPage left, right, mid; + int i, start_seg_with_known_loc, end_pos, page_start; + uint32 delta, stream_length, padding, last_sample_limit; + double offset = 0.0, bytes_per_sample = 0.0; + int probe = 0; + + // find the last page and validate the target sample + stream_length = stb_vorbis_stream_length_in_samples(f); + if (stream_length == 0) + return error(f, VORBIS_seek_without_length); + if (sample_number > stream_length) + return error(f, VORBIS_seek_invalid); + + // this is the maximum difference between the window-center (which is the + // actual granule position value), and the right-start (which the spec + // indicates should be the granule position (give or take one)). + padding = ((f->blocksize_1 - f->blocksize_0) >> 2); + if (sample_number < padding) + last_sample_limit = 0; + else + last_sample_limit = sample_number - padding; + + left = f->p_first; + while (left.last_decoded_sample == ~0U) { + // (untested) the first page does not have a 'last_decoded_sample' + set_file_offset(f, left.page_end); + if (!get_seek_page_info(f, &left)) + goto error; + } + + right = f->p_last; + assert(right.last_decoded_sample != ~0U); + + // starting from the start is handled differently + if (last_sample_limit <= left.last_decoded_sample) { + if (stb_vorbis_seek_start(f)) { + if (f->current_loc > sample_number) + return error(f, VORBIS_seek_failed); + return 1; + } + return 0; + } + + while (left.page_end != right.page_start) { + assert(left.page_end < right.page_start); + // search range in bytes + delta = right.page_start - left.page_end; + if (delta <= 65536) { + // there's only 64K left to search - handle it linearly + set_file_offset(f, left.page_end); + } else { + if (probe < 2) { + if (probe == 0) { + // first probe (interpolate) + double data_bytes = right.page_end - left.page_start; + bytes_per_sample = data_bytes / right.last_decoded_sample; + offset = left.page_start + bytes_per_sample * (last_sample_limit - left.last_decoded_sample); + } else { + // second probe (try to bound the other side) + double error = ((double)last_sample_limit - mid.last_decoded_sample) * bytes_per_sample; + if (error >= 0 && error < 8000) + error = 8000; + if (error < 0 && error > -8000) + error = -8000; + offset += error * 2; + } + + // ensure the offset is valid + if (offset < left.page_end) + offset = left.page_end; + if (offset > right.page_start - 65536) + offset = right.page_start - 65536; + + set_file_offset(f, (unsigned int)offset); + } else { + // binary search for large ranges (offset by 32K to ensure + // we don't hit the right page) + set_file_offset(f, left.page_end + (delta / 2) - 32768); + } + + if (!vorbis_find_page(f, NULL, NULL)) + goto error; + } + + for (;;) { + if (!get_seek_page_info(f, &mid)) + goto error; + if (mid.last_decoded_sample != ~0U) + break; + // (untested) no frames end on this page + set_file_offset(f, mid.page_end); + assert(mid.page_start < right.page_start); + } + + // if we've just found the last page again then we're in a tricky file, + // and we're close enough (if it wasn't an interpolation probe). + if (mid.page_start == right.page_start) { + if (probe >= 2 || delta <= 65536) + break; + } else { + if (last_sample_limit < mid.last_decoded_sample) + right = mid; + else + left = mid; + } + + ++probe; + } + + // seek back to start of the last packet + page_start = left.page_start; + set_file_offset(f, page_start); + if (!start_page(f)) + return error(f, VORBIS_seek_failed); + end_pos = f->end_seg_with_known_loc; + assert(end_pos >= 0); + + for (;;) { + for (i = end_pos; i > 0; --i) + if (f->segments[i - 1] != 255) + break; + + start_seg_with_known_loc = i; + + if (start_seg_with_known_loc > 0 || !(f->page_flag & PAGEFLAG_continued_packet)) + break; + + // (untested) the final packet begins on an earlier page + if (!go_to_page_before(f, page_start)) + goto error; + + page_start = stb_vorbis_get_file_offset(f); + if (!start_page(f)) + goto error; + end_pos = f->segment_count - 1; + } + + // prepare to start decoding + f->current_loc_valid = FALSE; + f->last_seg = FALSE; + f->valid_bits = 0; + f->packet_bytes = 0; + f->bytes_in_seg = 0; + f->previous_length = 0; + f->next_seg = start_seg_with_known_loc; + + for (i = 0; i < start_seg_with_known_loc; i++) + skip(f, f->segments[i]); + + // start decoding (optimizable - this frame is generally discarded) + if (!vorbis_pump_first_frame(f)) + return 0; + if (f->current_loc > sample_number) + return error(f, VORBIS_seek_failed); + return 1; + +error: + // try to restore the file to a valid state + stb_vorbis_seek_start(f); + return error(f, VORBIS_seek_failed); +} + +// the same as vorbis_decode_initial, but without advancing +static int peek_decode_initial(vorb *f, int *p_left_start, int *p_left_end, int *p_right_start, int *p_right_end, int *mode) { + int bits_read, bytes_read; + + if (!vorbis_decode_initial(f, p_left_start, p_left_end, p_right_start, p_right_end, mode)) + return 0; + + // either 1 or 2 bytes were read, figure out which so we can rewind + bits_read = 1 + ilog(f->mode_count - 1); + if (f->mode_config[*mode].blockflag) + bits_read += 2; + bytes_read = (bits_read + 7) / 8; + + f->bytes_in_seg += bytes_read; + f->packet_bytes -= bytes_read; + skip(f, -bytes_read); + if (f->next_seg == -1) + f->next_seg = f->segment_count - 1; + else + f->next_seg--; + f->valid_bits = 0; + + return 1; +} + +int stb_vorbis_seek_frame(stb_vorbis *f, unsigned int sample_number) { + uint32 max_frame_samples; + + if (IS_PUSH_MODE(f)) + return error(f, VORBIS_invalid_api_mixing); + + // fast page-level search + if (!seek_to_sample_coarse(f, sample_number)) + return 0; + + assert(f->current_loc_valid); + assert(f->current_loc <= sample_number); + + // linear search for the relevant packet + max_frame_samples = (f->blocksize_1 * 3 - f->blocksize_0) >> 2; + while (f->current_loc < sample_number) { + int left_start, left_end, right_start, right_end, mode, frame_samples; + if (!peek_decode_initial(f, &left_start, &left_end, &right_start, &right_end, &mode)) + return error(f, VORBIS_seek_failed); + // calculate the number of samples returned by the next frame + frame_samples = right_start - left_start; + if (f->current_loc + frame_samples > sample_number) { + return 1; // the next frame will contain the sample + } else if (f->current_loc + frame_samples + max_frame_samples > sample_number) { + // there's a chance the frame after this could contain the sample + vorbis_pump_first_frame(f); + } else { + // this frame is too early to be relevant + f->current_loc += frame_samples; + f->previous_length = 0; + maybe_start_packet(f); + flush_packet(f); + } + } + // the next frame should start with the sample + if (f->current_loc != sample_number) + return error(f, VORBIS_seek_failed); + return 1; +} + +int stb_vorbis_seek(stb_vorbis *f, unsigned int sample_number) { + if (!stb_vorbis_seek_frame(f, sample_number)) + return 0; + + if (sample_number != f->current_loc) { + int n; + uint32 frame_start = f->current_loc; + stb_vorbis_get_frame_float(f, &n, NULL); + assert(sample_number > frame_start); + assert(f->channel_buffer_start + (int)(sample_number - frame_start) <= f->channel_buffer_end); + f->channel_buffer_start += (sample_number - frame_start); + } + + return 1; +} + +int stb_vorbis_seek_start(stb_vorbis *f) { + if (IS_PUSH_MODE(f)) { + return error(f, VORBIS_invalid_api_mixing); + } + set_file_offset(f, f->first_audio_page_offset); + f->previous_length = 0; + f->first_decode = TRUE; + f->next_seg = -1; + return vorbis_pump_first_frame(f); +} + +unsigned int stb_vorbis_stream_length_in_samples(stb_vorbis *f) { + unsigned int restore_offset, previous_safe; + unsigned int end, last_page_loc; + + if (IS_PUSH_MODE(f)) + return error(f, VORBIS_invalid_api_mixing); + if (!f->total_samples) { + unsigned int last; + uint32 lo, hi; + char header[6]; + + // first, store the current decode position so we can restore it + restore_offset = stb_vorbis_get_file_offset(f); + + // now we want to seek back 64K from the end (the last page must + // be at most a little less than 64K, but let's allow a little slop) + if (f->stream_len >= 65536 && f->stream_len - 65536 >= f->first_audio_page_offset) + previous_safe = f->stream_len - 65536; + else + previous_safe = f->first_audio_page_offset; + + set_file_offset(f, previous_safe); + // previous_safe is now our candidate 'earliest known place that seeking + // to will lead to the final page' + + if (!vorbis_find_page(f, &end, &last)) { + // if we can't find a page, we're hosed! + f->error = VORBIS_cant_find_last_page; + f->total_samples = 0xffffffff; + goto done; + } + + // check if there are more pages + last_page_loc = stb_vorbis_get_file_offset(f); + + // stop when the last_page flag is set, not when we reach eof; + // this allows us to stop short of a 'file_section' end without + // explicitly checking the length of the section + while (!last) { + set_file_offset(f, end); + if (!vorbis_find_page(f, &end, &last)) { + // the last page we found didn't have the 'last page' flag + // set. whoops! + break; + } + previous_safe = last_page_loc + 1; + last_page_loc = stb_vorbis_get_file_offset(f); + } + + set_file_offset(f, last_page_loc); + + // parse the header + getn(f, (unsigned char *)header, 6); + // extract the absolute granule position + lo = get32(f); + hi = get32(f); + if (lo == 0xffffffff && hi == 0xffffffff) { + f->error = VORBIS_cant_find_last_page; + f->total_samples = SAMPLE_unknown; + goto done; + } + if (hi) + lo = 0xfffffffe; // saturate + f->total_samples = lo; + + f->p_last.page_start = last_page_loc; + f->p_last.page_end = end; + f->p_last.last_decoded_sample = lo; + + done: + set_file_offset(f, restore_offset); + } + return f->total_samples == SAMPLE_unknown ? 0 : f->total_samples; +} + +float stb_vorbis_stream_length_in_seconds(stb_vorbis *f) { + return stb_vorbis_stream_length_in_samples(f) / (float)f->sample_rate; +} + +int stb_vorbis_get_frame_float(stb_vorbis *f, int *channels, float ***output) { + int len, right, left, i; + if (IS_PUSH_MODE(f)) + return error(f, VORBIS_invalid_api_mixing); + + if (!vorbis_decode_packet(f, &len, &left, &right)) { + f->channel_buffer_start = f->channel_buffer_end = 0; + return 0; + } + + len = vorbis_finish_frame(f, len, left, right); + for (i = 0; i < f->channels; ++i) + f->outputs[i] = f->channel_buffers[i] + left; + + f->channel_buffer_start = left; + f->channel_buffer_end = left + len; + + if (channels) + *channels = f->channels; + if (output) + *output = f->outputs; + return len; +} + +#ifndef STB_VORBIS_NO_STDIO + +stb_vorbis *stb_vorbis_open_file_section(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc, unsigned int length) { + stb_vorbis *f, p; + vorbis_init(&p, alloc); + p.f = file; + p.f_start = (uint32)ftell(file); + p.stream_len = length; + p.close_on_free = close_on_free; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + return f; + } + } + if (error) + *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +stb_vorbis *stb_vorbis_open_file(FILE *file, int close_on_free, int *error, const stb_vorbis_alloc *alloc) { + unsigned int len, start; + start = (unsigned int)ftell(file); + fseek(file, 0, SEEK_END); + len = (unsigned int)(ftell(file) - start); + fseek(file, start, SEEK_SET); + return stb_vorbis_open_file_section(file, close_on_free, error, alloc, len); +} + +stb_vorbis *stb_vorbis_open_filename(const char *filename, int *error, const stb_vorbis_alloc *alloc) { + FILE *f; +#if defined(_WIN32) && defined(__STDC_WANT_SECURE_LIB__) + if (0 != fopen_s(&f, filename, "rb")) + f = NULL; +#else + f = fopen(filename, "rb"); +#endif + if (f) + return stb_vorbis_open_file(f, TRUE, error, alloc); + if (error) + *error = VORBIS_file_open_failure; + return NULL; +} +#endif // STB_VORBIS_NO_STDIO + +stb_vorbis *stb_vorbis_open_memory(const unsigned char *data, int len, int *error, const stb_vorbis_alloc *alloc) { + stb_vorbis *f, p; + if (data == NULL) + return NULL; + vorbis_init(&p, alloc); + p.stream = (uint8 *)data; + p.stream_end = (uint8 *)data + len; + p.stream_start = (uint8 *)p.stream; + p.stream_len = len; + p.push_mode = FALSE; + if (start_decoder(&p)) { + f = vorbis_alloc(&p); + if (f) { + *f = p; + vorbis_pump_first_frame(f); + if (error) + *error = VORBIS__no_error; + return f; + } + } + if (error) + *error = p.error; + vorbis_deinit(&p); + return NULL; +} + +#ifndef STB_VORBIS_NO_INTEGER_CONVERSION +#define PLAYBACK_MONO 1 +#define PLAYBACK_LEFT 2 +#define PLAYBACK_RIGHT 4 + +#define L (PLAYBACK_LEFT | PLAYBACK_MONO) +#define C (PLAYBACK_LEFT | PLAYBACK_RIGHT | PLAYBACK_MONO) +#define R (PLAYBACK_RIGHT | PLAYBACK_MONO) + +static int8 channel_position[7][6] = + { + {0}, + {C}, + {L, R}, + {L, C, R}, + {L, R, L, R}, + {L, C, R, L, R}, + {L, C, R, L, R, C}, +}; + +#ifndef STB_VORBIS_NO_FAST_SCALED_FLOAT +typedef union { + float f; + int i; +} float_conv; +typedef char stb_vorbis_float_size_test[sizeof(float) == 4 && sizeof(int) == 4]; +#define FASTDEF(x) float_conv x +// add (1<<23) to convert to int, then divide by 2^SHIFT, then add 0.5/2^SHIFT to round +#define MAGIC(SHIFT) (1.5f * (1 << (23 - SHIFT)) + 0.5f / (1 << SHIFT)) +#define ADDEND(SHIFT) (((150 - SHIFT) << 23) + (1 << 22)) +#define FAST_SCALED_FLOAT_TO_INT(temp, x, s) (temp.f = (x) + MAGIC(s), temp.i - ADDEND(s)) +#define check_endianness() +#else +#define FAST_SCALED_FLOAT_TO_INT(temp, x, s) ((int)((x) * (1 << (s)))) +#define check_endianness() +#define FASTDEF(x) +#endif + +static void copy_samples(short *dest, float *src, int len) { + int i; + check_endianness(); + for (i = 0; i < len; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp, src[i], 15); + if ((unsigned int)(v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + dest[i] = v; + } +} + +static void compute_samples(int mask, short *output, int num_c, float **data, int d_offset, int len) { +#define BUFFER_SIZE 32 + float buffer[BUFFER_SIZE]; + int i, j, o, n = BUFFER_SIZE; + check_endianness(); + for (o = 0; o < len; o += BUFFER_SIZE) { + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) + n = len - o; + for (j = 0; j < num_c; ++j) { + if (channel_position[num_c][j] & mask) { + for (i = 0; i < n; ++i) + buffer[i] += data[j][d_offset + o + i]; + } + } + for (i = 0; i < n; ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp, buffer[i], 15); + if ((unsigned int)(v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o + i] = v; + } + } +} + +static void compute_stereo_samples(short *output, int num_c, float **data, int d_offset, int len) { +#define BUFFER_SIZE 32 + float buffer[BUFFER_SIZE]; + int i, j, o, n = BUFFER_SIZE >> 1; + // o is the offset in the source data + check_endianness(); + for (o = 0; o < len; o += BUFFER_SIZE >> 1) { + // o2 is the offset in the output data + int o2 = o << 1; + memset(buffer, 0, sizeof(buffer)); + if (o + n > len) + n = len - o; + for (j = 0; j < num_c; ++j) { + int m = channel_position[num_c][j] & (PLAYBACK_LEFT | PLAYBACK_RIGHT); + if (m == (PLAYBACK_LEFT | PLAYBACK_RIGHT)) { + for (i = 0; i < n; ++i) { + buffer[i * 2 + 0] += data[j][d_offset + o + i]; + buffer[i * 2 + 1] += data[j][d_offset + o + i]; + } + } else if (m == PLAYBACK_LEFT) { + for (i = 0; i < n; ++i) { + buffer[i * 2 + 0] += data[j][d_offset + o + i]; + } + } else if (m == PLAYBACK_RIGHT) { + for (i = 0; i < n; ++i) { + buffer[i * 2 + 1] += data[j][d_offset + o + i]; + } + } + } + for (i = 0; i < (n << 1); ++i) { + FASTDEF(temp); + int v = FAST_SCALED_FLOAT_TO_INT(temp, buffer[i], 15); + if ((unsigned int)(v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + output[o2 + i] = v; + } + } +} + +static void convert_samples_short(int buf_c, short **buffer, int b_offset, int data_c, float **data, int d_offset, int samples) { + int i; + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + static int channel_selector[3][2] = {{0}, {PLAYBACK_MONO}, {PLAYBACK_LEFT, PLAYBACK_RIGHT}}; + for (i = 0; i < buf_c; ++i) + compute_samples(channel_selector[buf_c][i], buffer[i] + b_offset, data_c, data, d_offset, samples); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + for (i = 0; i < limit; ++i) + copy_samples(buffer[i] + b_offset, data[i] + d_offset, samples); + for (; i < buf_c; ++i) + memset(buffer[i] + b_offset, 0, sizeof(short) * samples); + } +} + +int stb_vorbis_get_frame_short(stb_vorbis *f, int num_c, short **buffer, int num_samples) { + float **output = NULL; + int len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len > num_samples) + len = num_samples; + if (len) + convert_samples_short(num_c, buffer, 0, f->channels, output, 0, len); + return len; +} + +static void convert_channels_short_interleaved(int buf_c, short *buffer, int data_c, float **data, int d_offset, int len) { + int i; + check_endianness(); + if (buf_c != data_c && buf_c <= 2 && data_c <= 6) { + assert(buf_c == 2); + for (i = 0; i < buf_c; ++i) + compute_stereo_samples(buffer, data_c, data, d_offset, len); + } else { + int limit = buf_c < data_c ? buf_c : data_c; + int j; + for (j = 0; j < len; ++j) { + for (i = 0; i < limit; ++i) { + FASTDEF(temp); + float f = data[i][d_offset + j]; + int v = FAST_SCALED_FLOAT_TO_INT(temp, f, 15); // data[i][d_offset+j],15); + if ((unsigned int)(v + 32768) > 65535) + v = v < 0 ? -32768 : 32767; + *buffer++ = v; + } + for (; i < buf_c; ++i) + *buffer++ = 0; + } + } +} + +int stb_vorbis_get_frame_short_interleaved(stb_vorbis *f, int num_c, short *buffer, int num_shorts) { + float **output; + int len; + if (num_c == 1) + return stb_vorbis_get_frame_short(f, num_c, &buffer, num_shorts); + len = stb_vorbis_get_frame_float(f, NULL, &output); + if (len) { + if (len * num_c > num_shorts) + len = num_shorts / num_c; + convert_channels_short_interleaved(num_c, buffer, f->channels, output, 0, len); + } + return len; +} + +int stb_vorbis_get_samples_short_interleaved(stb_vorbis *f, int channels, short *buffer, int num_shorts) { + float **outputs; + int len = num_shorts / channels; + int n = 0; + int z = f->channels; + if (z > channels) + z = channels; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n + k >= len) + k = len - n; + if (k) + convert_channels_short_interleaved(channels, buffer, f->channels, f->channel_buffers, f->channel_buffer_start, k); + buffer += k * channels; + n += k; + f->channel_buffer_start += k; + if (n == len) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} + +int stb_vorbis_get_samples_short(stb_vorbis *f, int channels, short **buffer, int len) { + float **outputs; + int n = 0; + int z = f->channels; + if (z > channels) + z = channels; + while (n < len) { + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n + k >= len) + k = len - n; + if (k) + convert_samples_short(channels, buffer, n, f->channels, f->channel_buffers, f->channel_buffer_start, k); + n += k; + f->channel_buffer_start += k; + if (n == len) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} + +#ifndef STB_VORBIS_NO_STDIO +int stb_vorbis_decode_filename(const char *filename, int *channels, int *sample_rate, short **output) { + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_filename(filename, &error, NULL); + if (v == NULL) + return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *)malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data + offset, total - offset); + if (n == 0) + break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *)realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // NO_STDIO + +int stb_vorbis_decode_memory(const uint8 *mem, int len, int *channels, int *sample_rate, short **output) { + int data_len, offset, total, limit, error; + short *data; + stb_vorbis *v = stb_vorbis_open_memory(mem, len, &error, NULL); + if (v == NULL) + return -1; + limit = v->channels * 4096; + *channels = v->channels; + if (sample_rate) + *sample_rate = v->sample_rate; + offset = data_len = 0; + total = limit; + data = (short *)malloc(total * sizeof(*data)); + if (data == NULL) { + stb_vorbis_close(v); + return -2; + } + for (;;) { + int n = stb_vorbis_get_frame_short_interleaved(v, v->channels, data + offset, total - offset); + if (n == 0) + break; + data_len += n; + offset += n * v->channels; + if (offset + limit > total) { + short *data2; + total *= 2; + data2 = (short *)realloc(data, total * sizeof(*data)); + if (data2 == NULL) { + free(data); + stb_vorbis_close(v); + return -2; + } + data = data2; + } + } + *output = data; + stb_vorbis_close(v); + return data_len; +} +#endif // STB_VORBIS_NO_INTEGER_CONVERSION + +int stb_vorbis_get_samples_float_interleaved(stb_vorbis *f, int channels, float *buffer, int num_floats) { + float **outputs; + int len = num_floats / channels; + int n = 0; + int z = f->channels; + if (z > channels) + z = channels; + while (n < len) { + int i, j; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n + k >= len) + k = len - n; + for (j = 0; j < k; ++j) { + for (i = 0; i < z; ++i) + *buffer++ = f->channel_buffers[i][f->channel_buffer_start + j]; + for (; i < channels; ++i) + *buffer++ = 0; + } + n += k; + f->channel_buffer_start += k; + if (n == len) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} + +int stb_vorbis_get_samples_float(stb_vorbis *f, int channels, float **buffer, int num_samples) { + float **outputs; + int n = 0; + int z = f->channels; + if (z > channels) + z = channels; + while (n < num_samples) { + int i; + int k = f->channel_buffer_end - f->channel_buffer_start; + if (n + k >= num_samples) + k = num_samples - n; + if (k) { + for (i = 0; i < z; ++i) + memcpy(buffer[i] + n, f->channel_buffers[i] + f->channel_buffer_start, sizeof(float) * k); + for (; i < channels; ++i) + memset(buffer[i] + n, 0, sizeof(float) * k); + } + n += k; + f->channel_buffer_start += k; + if (n == num_samples) + break; + if (!stb_vorbis_get_frame_float(f, NULL, &outputs)) + break; + } + return n; +} +#endif // STB_VORBIS_NO_PULLDATA_API + +/* Version history + 1.17 - 2019-07-08 - fix CVE-2019-13217, -13218, -13219, -13220, -13221, -13222, -13223 + found with Mayhem by ForAllSecure + 1.16 - 2019-03-04 - fix warnings + 1.15 - 2019-02-07 - explicit failure if Ogg Skeleton data is found + 1.14 - 2018-02-11 - delete bogus dealloca usage + 1.13 - 2018-01-29 - fix truncation of last frame (hopefully) + 1.12 - 2017-11-21 - limit residue begin/end to blocksize/2 to avoid large temp allocs in bad/corrupt files + 1.11 - 2017-07-23 - fix MinGW compilation + 1.10 - 2017-03-03 - more robust seeking; fix negative ilog(); clear error in open_memory + 1.09 - 2016-04-04 - back out 'avoid discarding last frame' fix from previous version + 1.08 - 2016-04-02 - fixed multiple warnings; fix setup memory leaks; + avoid discarding last frame of audio data + 1.07 - 2015-01-16 - fixed some warnings, fix mingw, const-correct API + some more crash fixes when out of memory or with corrupt files + 1.06 - 2015-08-31 - full, correct support for seeking API (Dougall Johnson) + some crash fixes when out of memory or with corrupt files + 1.05 - 2015-04-19 - don't define __forceinline if it's redundant + 1.04 - 2014-08-27 - fix missing const-correct case in API + 1.03 - 2014-08-07 - Warning fixes + 1.02 - 2014-07-09 - Declare qsort compare function _cdecl on windows + 1.01 - 2014-06-18 - fix stb_vorbis_get_samples_float + 1.0 - 2014-05-26 - fix memory leaks; fix warnings; fix bugs in multichannel + (API change) report sample rate for decode-full-file funcs + 0.99996 - bracket #include for macintosh compilation by Laurent Gomila + 0.99995 - use union instead of pointer-cast for fast-float-to-int to avoid alias-optimization problem + 0.99994 - change fast-float-to-int to work in single-precision FPU mode, remove endian-dependence + 0.99993 - remove assert that fired on legal files with empty tables + 0.99992 - rewind-to-start + 0.99991 - bugfix to stb_vorbis_get_samples_short by Bernhard Wodo + 0.9999 - (should have been 0.99990) fix no-CRT support, compiling as C++ + 0.9998 - add a full-decode function with a memory source + 0.9997 - fix a bug in the read-from-FILE case in 0.9996 addition + 0.9996 - query length of vorbis stream in samples/seconds + 0.9995 - bugfix to another optimization that only happened in certain files + 0.9994 - bugfix to one of the optimizations that caused significant (but inaudible?) errors + 0.9993 - performance improvements; runs in 99% to 104% of time of reference implementation + 0.9992 - performance improvement of IMDCT; now performs close to reference implementation + 0.9991 - performance improvement of IMDCT + 0.999 - (should have been 0.9990) performance improvement of IMDCT + 0.998 - no-CRT support from Casey Muratori + 0.997 - bugfixes for bugs found by Terje Mathisen + 0.996 - bugfix: fast-huffman decode initialized incorrectly for sparse codebooks; fixing gives 10% speedup - found by Terje Mathisen + 0.995 - bugfix: fix to 'effective' overrun detection - found by Terje Mathisen + 0.994 - bugfix: garbage decode on final VQ symbol of a non-multiple - found by Terje Mathisen + 0.993 - bugfix: pushdata API required 1 extra byte for empty page (failed to consume final page if empty) - found by Terje Mathisen + 0.992 - fixes for MinGW warning + 0.991 - turn fast-float-conversion on by default + 0.990 - fix push-mode seek recovery if you seek into the headers + 0.98b - fix to bad release of 0.98 + 0.98 - fix push-mode seek recovery; robustify float-to-int and support non-fast mode + 0.97 - builds under c++ (typecasting, don't use 'class' keyword) + 0.96 - somehow MY 0.95 was right, but the web one was wrong, so here's my 0.95 rereleased as 0.96, fixes a typo in the clamping code + 0.95 - clamping code for 16-bit functions + 0.94 - not publically released + 0.93 - fixed all-zero-floor case (was decoding garbage) + 0.92 - fixed a memory leak + 0.91 - conditional compiles to omit parts of the API and the infrastructure to support them: STB_VORBIS_NO_PULLDATA_API, STB_VORBIS_NO_PUSHDATA_API, STB_VORBIS_NO_STDIO, STB_VORBIS_NO_INTEGER_CONVERSION + 0.90 - first public release +*/ + +#endif // STB_VORBIS_HEADER_ONLY + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/ \ No newline at end of file