fix: makefile

feat(macos): suport Metal per GpuShaderPreset via spirv-cross
- CMakeLists: bloc if(APPLE) que transpila .spv → .spv.msl amb spirv-cross - gpu_shader_preset: carrega MSL (main0) a macOS, SPIR-V (main) a la resta - Afegeix null-terminator als buffers MSL (SDL3 els tracta com C-string) - README: secció de dependències de shaders per plataforma (Vulkan SDK, spirv-cross) - Inclou .spv.msl generats per ntsc-md-rainbows Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-20 16:44:11 +01:00 · 2026-03-20 13:54:05 +01:00
8 changed files with 612 additions and 27 deletions
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -94,6 +94,33 @@ if(NOT APPLE)
    endif()
 endif()

+# macOS: transpile .spv → .msl for GpuShaderPreset
+if(APPLE)
+    find_program(SPIRV_CROSS spirv-cross)
+    if(SPIRV_CROSS)
+        file(GLOB_RECURSE DATA_SHADER_SPVS
+            "${CMAKE_SOURCE_DIR}/data/shaders/*.vert.spv"
+            "${CMAKE_SOURCE_DIR}/data/shaders/*.frag.spv"
+        )
+        set(DATA_SHADER_MSLS)
+        foreach(SPV_FILE ${DATA_SHADER_SPVS})
+            set(MSL_FILE "${SPV_FILE}.msl")
+            add_custom_command(
+                OUTPUT "${MSL_FILE}"
+                COMMAND "${SPIRV_CROSS}" --msl --output "${MSL_FILE}" "${SPV_FILE}"
+                DEPENDS "${SPV_FILE}"
+                COMMENT "Transpiling ${SPV_FILE} to MSL"
+            )
+            list(APPEND DATA_SHADER_MSLS "${MSL_FILE}")
+        endforeach()
+        if(DATA_SHADER_MSLS)
+            add_custom_target(data_shader_msls ALL DEPENDS ${DATA_SHADER_MSLS})
+        endif()
+    else()
+        message(STATUS "spirv-cross not found — GpuShaderPreset will not work on macOS (run: brew install spirv-cross)")
+    endif()
+endif()
+
 # Archivos fuente (excluir main_old.cpp)
 file(GLOB SOURCE_FILES source/*.cpp source/external/*.cpp source/boids_mgr/*.cpp source/gpu/*.cpp source/input/*.cpp source/scene/*.cpp source/shapes/*.cpp source/shapes_mgr/*.cpp source/state/*.cpp source/themes/*.cpp source/text/*.cpp source/ui/*.cpp)
 list(REMOVE_ITEM SOURCE_FILES "${CMAKE_SOURCE_DIR}/source/main_old.cpp")
@@ -138,6 +165,10 @@ if(NOT APPLE)
    endif()
 endif()

+if(APPLE AND TARGET data_shader_msls)
+    add_dependencies(${PROJECT_NAME} data_shader_msls)
+endif()
+
 # Tool: pack_resources
 add_executable(pack_resources tools/pack_resources.cpp source/resource_pack.cpp)
 target_include_directories(pack_resources PRIVATE ${CMAKE_SOURCE_DIR}/source)
--- a/1
+++ b/1
@@ -51,6 +51,7 @@ RASPI_RELEASE               := $(DIST_DIR)/$(TARGET_NAME)-$(VERSION)-raspberry.t
 # Lista completa de archivos fuente (detección automática con wildcards, como CMakeLists.txt)
 APP_SOURCES := $(wildcard source/*.cpp) \
               $(wildcard source/external/*.cpp) \
+               $(wildcard source/gpu/*.cpp) \
               $(wildcard source/shapes/*.cpp) \
               $(wildcard source/themes/*.cpp) \
               $(wildcard source/state/*.cpp) \
--- a/README.md
+++ b/README.md
@@ -531,49 +531,92 @@ vibe3_physics/
 ## 🔧 Requisitos del Sistema

 - **SDL3** (Simple DirectMedia Layer 3)
+- **SDL3_ttf** (fuentes TTF para SDL3)
 - **C++20** compatible compiler (GCC 10+, Clang 12+, MSVC 2019+)
- **CMake 3.20+** o **Make**
+- **CMake 3.20+**
 - **Plataforma**: Windows, Linux, macOS

-### Instalación de SDL3
+### Dependencias de shaders
+
+El proyecto compila shaders GLSL a SPIR-V en todas las plataformas usando `glslc` (incluido en el Vulkan SDK). En macOS, adicionalmente transpila los `.spv` a MSL (Metal) con `spirv-cross`.
+
+| Plataforma | Backend GPU | Herramienta de shaders | Resultado |
+|------------|-------------|------------------------|-----------|
+| Linux      | Vulkan      | `glslc` (Vulkan SDK)   | `.spv`    |
+| Windows    | Vulkan      | `glslc` (Vulkan SDK)   | `.spv`    |
+| macOS      | Metal       | `glslc` + `spirv-cross`| `.spv.msl`|
+
+---
+
+### Instalación de dependencias
+
+#### macOS

-#### Windows (MinGW)
 ```bash
-# Usando vcpkg
-vcpkg install sdl3
+# SDL3 + SDL3_ttf
+brew install sdl3 sdl3_ttf

-# O compilar desde fuente
-git clone https://github.com/libsdl-org/SDL
-cd SDL
-mkdir build && cd build
-cmake ..
-make
-sudo make install
+# Vulkan SDK (incluye glslc para compilar shaders)
+# Descargar desde https://vulkan.lunarg.com/ o via Homebrew:
+brew install --cask vulkan-sdk
+
+# spirv-cross (transpilación SPIR-V → MSL para Metal)
+brew install spirv-cross
+```
+
+Tras instalar el Vulkan SDK es posible que necesites añadir `glslc` al PATH:
+```bash
+# Añadir a ~/.zshrc o ~/.bash_profile (ajusta la versión):
+export PATH="$HOME/VulkanSDK/<version>/macOS/bin:$PATH"
+source ~/.zshrc
 ```

 #### Linux (Ubuntu/Debian)
-```bash
-sudo apt update
-sudo apt install libsdl3-dev

-# Si no está disponible, compilar desde fuente
+```bash
+# SDL3 + SDL3_ttf
+sudo apt update
+sudo apt install libsdl3-dev libsdl3-ttf-dev
+
+# Vulkan SDK (incluye glslc)
+# Opción A: paquete del sistema
+sudo apt install glslc
+
+# Opción B: Vulkan SDK completo (recomendado para versión reciente)
+wget -qO- https://packages.lunarg.com/lunarg-signing-key-pub.asc | sudo apt-key add -
+sudo wget -qO /etc/apt/sources.list.d/lunarg-vulkan.list \
+    https://packages.lunarg.com/vulkan/lunarg-vulkan-focal.list
+sudo apt update
+sudo apt install vulkan-sdk
 ```

 #### Linux (Arch)
+
 ```bash
-sudo pacman -S sdl3
+sudo pacman -S sdl3 sdl3_ttf vulkan-devel glslang shaderc
 ```

-#### macOS
+#### Windows (MinGW / MSYS2)
+
 ```bash
-brew install sdl3
+# En MSYS2 shell:
+pacman -S mingw-w64-x86_64-SDL3 mingw-w64-x86_64-SDL3_ttf
+
+# Vulkan SDK: descargar el instalador desde https://vulkan.lunarg.com/
+# El instalador añade glslc al PATH automáticamente.
+```
+
+Con Visual Studio + vcpkg:
+```bash
+vcpkg install sdl3 sdl3-ttf
+# Instalar Vulkan SDK desde https://vulkan.lunarg.com/ (incluye glslc)
 ```

 ---

 ## 🚀 Compilación

-### Opción 1: CMake (Recomendado)
+### CMake (todas las plataformas)

 ```bash
 mkdir -p build && cd build
@@ -581,7 +624,9 @@ cmake ..
 cmake --build .
 ```

-### Opción 2: Make directo
+En macOS, CMake detecta automáticamente `spirv-cross` y genera los `.spv.msl` necesarios. Si no está instalado, los shaders de preset externos no funcionarán (el resto de la app sí).
+
+### Make directo (Linux/macOS)

 ```bash
 make
--- a/data/shaders/ntsc-md-rainbows/pass0_encode.frag.spv.msl
+++ b/data/shaders/ntsc-md-rainbows/pass0_encode.frag.spv.msl
@@ -0,0 +1,61 @@
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+struct NTSCParams
+{
+    float source_width;
+    float source_height;
+    float a_value;
+    float b_value;
+    float cc_value;
+    float scan_time;
+    float notch_width;
+    float y_freq;
+    float i_freq;
+    float q_freq;
+    float _pad0;
+    float _pad1;
+};
+
+struct main0_out
+{
+    float4 FragColor [[color(0)]];
+};
+
+struct main0_in
+{
+    float2 v_uv [[user(locn0)]];
+};
+
+fragment main0_out main0(main0_in in [[stage_in]], constant NTSCParams& u [[buffer(0)]], texture2d<float> Source [[texture(0)]], sampler SourceSmplr [[sampler(0)]])
+{
+    main0_out out = {};
+    float2 source_dims = float2(u.source_width, u.source_height);
+    float2 PValueSourceTexel = float2(1.0, 0.0) / source_dims;
+    float2 C0 = in.v_uv + (PValueSourceTexel * float2(0.0));
+    float2 C1 = in.v_uv + (PValueSourceTexel * float2(0.25, 0.0));
+    float2 C2 = in.v_uv + (PValueSourceTexel * float2(0.5, 0.0));
+    float2 C3 = in.v_uv + (PValueSourceTexel * float2(0.75, 0.0));
+    float4 Cx = float4(C0.x, C1.x, C2.x, C3.x);
+    float4 Cy = float4(C0.y, C1.y, C2.y, C3.y);
+    float4 Texel0 = Source.sample(SourceSmplr, C0);
+    float4 Texel1 = Source.sample(SourceSmplr, C1);
+    float4 Texel2 = Source.sample(SourceSmplr, C2);
+    float4 Texel3 = Source.sample(SourceSmplr, C3);
+    float4 HPosition = Cx;
+    float4 VPosition = Cy;
+    float4 Y = float4(dot(Texel0, float4(0.2989999949932098388671875, 0.58700001239776611328125, 0.114000000059604644775390625, 0.0)), dot(Texel1, float4(0.2989999949932098388671875, 0.58700001239776611328125, 0.114000000059604644775390625, 0.0)), dot(Texel2, float4(0.2989999949932098388671875, 0.58700001239776611328125, 0.114000000059604644775390625, 0.0)), dot(Texel3, float4(0.2989999949932098388671875, 0.58700001239776611328125, 0.114000000059604644775390625, 0.0)));
+    float4 I = float4(dot(Texel0, float4(0.595715999603271484375, -0.2744530141353607177734375, -0.3212629854679107666015625, 0.0)), dot(Texel1, float4(0.595715999603271484375, -0.2744530141353607177734375, -0.3212629854679107666015625, 0.0)), dot(Texel2, float4(0.595715999603271484375, -0.2744530141353607177734375, -0.3212629854679107666015625, 0.0)), dot(Texel3, float4(0.595715999603271484375, -0.2744530141353607177734375, -0.3212629854679107666015625, 0.0)));
+    float4 Q = float4(dot(Texel0, float4(0.211456000804901123046875, -0.52259099483489990234375, 0.311134994029998779296875, 0.0)), dot(Texel1, float4(0.211456000804901123046875, -0.52259099483489990234375, 0.311134994029998779296875, 0.0)), dot(Texel2, float4(0.211456000804901123046875, -0.52259099483489990234375, 0.311134994029998779296875, 0.0)), dot(Texel3, float4(0.211456000804901123046875, -0.52259099483489990234375, 0.311134994029998779296875, 0.0)));
+    float W = (6.283185482025146484375 * u.cc_value) * u.scan_time;
+    float WoPI = W / 3.1415927410125732421875;
+    float HOffset = u.a_value / WoPI;
+    float VScale = (u.b_value * source_dims.y) / WoPI;
+    float4 T = (HPosition + float4(HOffset)) + (VPosition * float4(VScale));
+    float4 TW = T * W;
+    out.FragColor = (Y + (I * cos(TW))) + (Q * sin(TW));
+    return out;
+}
+
--- a/data/shaders/ntsc-md-rainbows/pass0_encode.vert.spv.msl
+++ b/data/shaders/ntsc-md-rainbows/pass0_encode.vert.spv.msl
@@ -0,0 +1,63 @@
+#pragma clang diagnostic ignored "-Wmissing-prototypes"
+#pragma clang diagnostic ignored "-Wmissing-braces"
+
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+template<typename T, size_t Num>
+struct spvUnsafeArray
+{
+    T elements[Num ? Num : 1];
+    
+    thread T& operator [] (size_t pos) thread
+    {
+        return elements[pos];
+    }
+    constexpr const thread T& operator [] (size_t pos) const thread
+    {
+        return elements[pos];
+    }
+    
+    device T& operator [] (size_t pos) device
+    {
+        return elements[pos];
+    }
+    constexpr const device T& operator [] (size_t pos) const device
+    {
+        return elements[pos];
+    }
+    
+    constexpr const constant T& operator [] (size_t pos) const constant
+    {
+        return elements[pos];
+    }
+    
+    threadgroup T& operator [] (size_t pos) threadgroup
+    {
+        return elements[pos];
+    }
+    constexpr const threadgroup T& operator [] (size_t pos) const threadgroup
+    {
+        return elements[pos];
+    }
+};
+
+constant spvUnsafeArray<float2, 3> _18 = spvUnsafeArray<float2, 3>({ float2(-1.0), float2(3.0, -1.0), float2(-1.0, 3.0) });
+constant spvUnsafeArray<float2, 3> _26 = spvUnsafeArray<float2, 3>({ float2(0.0, 1.0), float2(2.0, 1.0), float2(0.0, -1.0) });
+
+struct main0_out
+{
+    float2 v_uv [[user(locn0)]];
+    float4 gl_Position [[position]];
+};
+
+vertex main0_out main0(uint gl_VertexIndex [[vertex_id]])
+{
+    main0_out out = {};
+    out.gl_Position = float4(_18[int(gl_VertexIndex)], 0.0, 1.0);
+    out.v_uv = _26[int(gl_VertexIndex)];
+    return out;
+}
+
--- a/data/shaders/ntsc-md-rainbows/pass1_decode.frag.spv.msl
+++ b/data/shaders/ntsc-md-rainbows/pass1_decode.frag.spv.msl
@@ -0,0 +1,304 @@
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+struct NTSCParams
+{
+    float source_width;
+    float source_height;
+    float a_value;
+    float b_value;
+    float cc_value;
+    float scan_time;
+    float notch_width;
+    float y_freq;
+    float i_freq;
+    float q_freq;
+    float _pad0;
+    float _pad1;
+};
+
+struct main0_out
+{
+    float4 FragColor [[color(0)]];
+};
+
+struct main0_in
+{
+    float2 v_uv [[user(locn0)]];
+};
+
+fragment main0_out main0(main0_in in [[stage_in]], constant NTSCParams& u [[buffer(0)]], texture2d<float> Source [[texture(0)]], sampler SourceSmplr [[sampler(0)]])
+{
+    main0_out out = {};
+    float2 source_dims = float2(u.source_width, u.source_height);
+    float4 BaseTexel = Source.sample(SourceSmplr, in.v_uv);
+    float CCValue = u.cc_value;
+    float ScanTime = u.scan_time;
+    float NotchHalfWidth = u.notch_width / 2.0;
+    float YFreqResponse = u.y_freq;
+    float IFreqResponse = u.i_freq;
+    float QFreqResponse = u.q_freq;
+    float AValue = u.a_value;
+    float BValue = u.b_value;
+    float TimePerSample = ScanTime / (source_dims.x * 4.0);
+    float Fc_y1 = (CCValue - NotchHalfWidth) * TimePerSample;
+    float Fc_y2 = (CCValue + NotchHalfWidth) * TimePerSample;
+    float Fc_y3 = YFreqResponse * TimePerSample;
+    float Fc_i = IFreqResponse * TimePerSample;
+    float Fc_q = QFreqResponse * TimePerSample;
+    float Fc_i_2 = Fc_i * 2.0;
+    float Fc_q_2 = Fc_q * 2.0;
+    float Fc_y1_2 = Fc_y1 * 2.0;
+    float Fc_y2_2 = Fc_y2 * 2.0;
+    float Fc_y3_2 = Fc_y3 * 2.0;
+    float Fc_i_pi2 = Fc_i * 6.283185482025146484375;
+    float Fc_q_pi2 = Fc_q * 6.283185482025146484375;
+    float Fc_y1_pi2 = Fc_y1 * 6.283185482025146484375;
+    float Fc_y2_pi2 = Fc_y2 * 6.283185482025146484375;
+    float Fc_y3_pi2 = Fc_y3 * 6.283185482025146484375;
+    float PI2Length = 0.098174773156642913818359375;
+    float W = (6.283185482025146484375 * CCValue) * ScanTime;
+    float WoPI = W / 3.1415927410125732421875;
+    float HOffset = BValue / WoPI;
+    float VScale = (AValue * source_dims.y) / WoPI;
+    float4 YAccum = float4(0.0);
+    float4 IAccum = float4(0.0);
+    float4 QAccum = float4(0.0);
+    float4 Cy = float4(in.v_uv.y);
+    float4 VPosition = Cy;
+    float4 SincY1;
+    float _259;
+    float _274;
+    float _290;
+    float _306;
+    float4 SincY2;
+    float _322;
+    float _337;
+    float _352;
+    float _367;
+    float4 SincY3;
+    float _383;
+    float _398;
+    float _413;
+    float _428;
+    float4 IdealI;
+    float _457;
+    float _474;
+    float _491;
+    float _508;
+    float4 IdealQ;
+    float _526;
+    float _543;
+    float _560;
+    float _577;
+    for (float i = 0.0; i < 64.0; i += 4.0)
+    {
+        float n = i - 32.0;
+        float4 n4 = float4(n) + float4(0.0, 1.0, 2.0, 3.0);
+        float4 Cx = float4(in.v_uv.x) + ((n4 * 0.25) / float4(source_dims.x));
+        float4 HPosition = Cx;
+        float4 C = Source.sample(SourceSmplr, float2(Cx.x, Cy.x));
+        float4 T = (HPosition + float4(HOffset)) + (VPosition * float4(VScale));
+        float4 WT = T * W;
+        float4 SincKernel = float4(0.540000021457672119140625) + (cos(n4 * PI2Length) * 0.4600000083446502685546875);
+        float4 SincYIn1 = n4 * Fc_y1_pi2;
+        float4 SincYIn2 = n4 * Fc_y2_pi2;
+        float4 SincYIn3 = n4 * Fc_y3_pi2;
+        float4 SincIIn = n4 * Fc_i_pi2;
+        float4 SincQIn = n4 * Fc_q_pi2;
+        if (SincYIn1.x != 0.0)
+        {
+            _259 = sin(SincYIn1.x) / SincYIn1.x;
+        }
+        else
+        {
+            _259 = 1.0;
+        }
+        SincY1.x = _259;
+        if (SincYIn1.y != 0.0)
+        {
+            _274 = sin(SincYIn1.y) / SincYIn1.y;
+        }
+        else
+        {
+            _274 = 1.0;
+        }
+        SincY1.y = _274;
+        if (SincYIn1.z != 0.0)
+        {
+            _290 = sin(SincYIn1.z) / SincYIn1.z;
+        }
+        else
+        {
+            _290 = 1.0;
+        }
+        SincY1.z = _290;
+        if (SincYIn1.w != 0.0)
+        {
+            _306 = sin(SincYIn1.w) / SincYIn1.w;
+        }
+        else
+        {
+            _306 = 1.0;
+        }
+        SincY1.w = _306;
+        if (SincYIn2.x != 0.0)
+        {
+            _322 = sin(SincYIn2.x) / SincYIn2.x;
+        }
+        else
+        {
+            _322 = 1.0;
+        }
+        SincY2.x = _322;
+        if (SincYIn2.y != 0.0)
+        {
+            _337 = sin(SincYIn2.y) / SincYIn2.y;
+        }
+        else
+        {
+            _337 = 1.0;
+        }
+        SincY2.y = _337;
+        if (SincYIn2.z != 0.0)
+        {
+            _352 = sin(SincYIn2.z) / SincYIn2.z;
+        }
+        else
+        {
+            _352 = 1.0;
+        }
+        SincY2.z = _352;
+        if (SincYIn2.w != 0.0)
+        {
+            _367 = sin(SincYIn2.w) / SincYIn2.w;
+        }
+        else
+        {
+            _367 = 1.0;
+        }
+        SincY2.w = _367;
+        if (SincYIn3.x != 0.0)
+        {
+            _383 = sin(SincYIn3.x) / SincYIn3.x;
+        }
+        else
+        {
+            _383 = 1.0;
+        }
+        SincY3.x = _383;
+        if (SincYIn3.y != 0.0)
+        {
+            _398 = sin(SincYIn3.y) / SincYIn3.y;
+        }
+        else
+        {
+            _398 = 1.0;
+        }
+        SincY3.y = _398;
+        if (SincYIn3.z != 0.0)
+        {
+            _413 = sin(SincYIn3.z) / SincYIn3.z;
+        }
+        else
+        {
+            _413 = 1.0;
+        }
+        SincY3.z = _413;
+        if (SincYIn3.w != 0.0)
+        {
+            _428 = sin(SincYIn3.w) / SincYIn3.w;
+        }
+        else
+        {
+            _428 = 1.0;
+        }
+        SincY3.w = _428;
+        float4 IdealY = ((SincY1 * Fc_y1_2) - (SincY2 * Fc_y2_2)) + (SincY3 * Fc_y3_2);
+        if (SincIIn.x != 0.0)
+        {
+            _457 = sin(SincIIn.x) / SincIIn.x;
+        }
+        else
+        {
+            _457 = 1.0;
+        }
+        IdealI.x = Fc_i_2 * _457;
+        if (SincIIn.y != 0.0)
+        {
+            _474 = sin(SincIIn.y) / SincIIn.y;
+        }
+        else
+        {
+            _474 = 1.0;
+        }
+        IdealI.y = Fc_i_2 * _474;
+        if (SincIIn.z != 0.0)
+        {
+            _491 = sin(SincIIn.z) / SincIIn.z;
+        }
+        else
+        {
+            _491 = 1.0;
+        }
+        IdealI.z = Fc_i_2 * _491;
+        if (SincIIn.w != 0.0)
+        {
+            _508 = sin(SincIIn.w) / SincIIn.w;
+        }
+        else
+        {
+            _508 = 1.0;
+        }
+        IdealI.w = Fc_i_2 * _508;
+        if (SincQIn.x != 0.0)
+        {
+            _526 = sin(SincQIn.x) / SincQIn.x;
+        }
+        else
+        {
+            _526 = 1.0;
+        }
+        IdealQ.x = Fc_q_2 * _526;
+        if (SincQIn.y != 0.0)
+        {
+            _543 = sin(SincQIn.y) / SincQIn.y;
+        }
+        else
+        {
+            _543 = 1.0;
+        }
+        IdealQ.y = Fc_q_2 * _543;
+        if (SincQIn.z != 0.0)
+        {
+            _560 = sin(SincQIn.z) / SincQIn.z;
+        }
+        else
+        {
+            _560 = 1.0;
+        }
+        IdealQ.z = Fc_q_2 * _560;
+        if (SincQIn.w != 0.0)
+        {
+            _577 = sin(SincQIn.w) / SincQIn.w;
+        }
+        else
+        {
+            _577 = 1.0;
+        }
+        IdealQ.w = Fc_q_2 * _577;
+        float4 FilterY = SincKernel * IdealY;
+        float4 FilterI = SincKernel * IdealI;
+        float4 FilterQ = SincKernel * IdealQ;
+        YAccum += (C * FilterY);
+        IAccum += ((C * cos(WT)) * FilterI);
+        QAccum += ((C * sin(WT)) * FilterQ);
+    }
+    float3 YIQ = float3(((YAccum.x + YAccum.y) + YAccum.z) + YAccum.w, (((IAccum.x + IAccum.y) + IAccum.z) + IAccum.w) * 2.0, (((QAccum.x + QAccum.y) + QAccum.z) + QAccum.w) * 2.0);
+    float3 RGB = float3(dot(YIQ, float3(1.0, 0.95599997043609619140625, 0.620999991893768310546875)), dot(YIQ, float3(1.0, -0.272000014781951904296875, -0.647000014781951904296875)), dot(YIQ, float3(1.0, -1.10599994659423828125, 1.70299994945526123046875)));
+    out.FragColor = float4(RGB, BaseTexel.w);
+    return out;
+}
+
--- a/data/shaders/ntsc-md-rainbows/pass1_decode.vert.spv.msl
+++ b/data/shaders/ntsc-md-rainbows/pass1_decode.vert.spv.msl
@@ -0,0 +1,63 @@
+#pragma clang diagnostic ignored "-Wmissing-prototypes"
+#pragma clang diagnostic ignored "-Wmissing-braces"
+
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+template<typename T, size_t Num>
+struct spvUnsafeArray
+{
+    T elements[Num ? Num : 1];
+    
+    thread T& operator [] (size_t pos) thread
+    {
+        return elements[pos];
+    }
+    constexpr const thread T& operator [] (size_t pos) const thread
+    {
+        return elements[pos];
+    }
+    
+    device T& operator [] (size_t pos) device
+    {
+        return elements[pos];
+    }
+    constexpr const device T& operator [] (size_t pos) const device
+    {
+        return elements[pos];
+    }
+    
+    constexpr const constant T& operator [] (size_t pos) const constant
+    {
+        return elements[pos];
+    }
+    
+    threadgroup T& operator [] (size_t pos) threadgroup
+    {
+        return elements[pos];
+    }
+    constexpr const threadgroup T& operator [] (size_t pos) const threadgroup
+    {
+        return elements[pos];
+    }
+};
+
+constant spvUnsafeArray<float2, 3> _18 = spvUnsafeArray<float2, 3>({ float2(-1.0), float2(3.0, -1.0), float2(-1.0, 3.0) });
+constant spvUnsafeArray<float2, 3> _26 = spvUnsafeArray<float2, 3>({ float2(0.0, 1.0), float2(2.0, 1.0), float2(0.0, -1.0) });
+
+struct main0_out
+{
+    float2 v_uv [[user(locn0)]];
+    float4 gl_Position [[position]];
+};
+
+vertex main0_out main0(uint gl_VertexIndex [[vertex_id]])
+{
+    main0_out out = {};
+    out.gl_Position = float4(_18[int(gl_VertexIndex)], 0.0, 1.0);
+    out.v_uv = _26[int(gl_VertexIndex)];
+    return out;
+}
+
--- a/source/gpu/gpu_shader_preset.cpp
+++ b/source/gpu/gpu_shader_preset.cpp
@@ -110,6 +110,14 @@ SDL_GPUGraphicsPipeline* GpuShaderPreset::buildPassPipeline(SDL_GPUDevice* devic
                                                              const std::string& vert_spv_path,
                                                              const std::string& frag_spv_path,
                                                              SDL_GPUTextureFormat target_fmt) {
+#ifdef __APPLE__
+    constexpr SDL_GPUShaderFormat kFmt = SDL_GPU_SHADERFORMAT_MSL;
+    constexpr const char* kEntry = "main0";
+#else
+    constexpr SDL_GPUShaderFormat kFmt = SDL_GPU_SHADERFORMAT_SPIRV;
+    constexpr const char* kEntry = "main";
+#endif
+
    auto vert_spv = readFile(vert_spv_path);
    auto frag_spv = readFile(frag_spv_path);
    if (vert_spv.empty()) {
@@ -120,12 +128,16 @@ SDL_GPUGraphicsPipeline* GpuShaderPreset::buildPassPipeline(SDL_GPUDevice* devic
        SDL_Log("GpuShaderPreset: cannot read %s", frag_spv_path.c_str());
        return nullptr;
    }
+#ifdef __APPLE__
+    vert_spv.push_back(0);
+    frag_spv.push_back(0);
+#endif

    SDL_GPUShaderCreateInfo vert_info = {};
    vert_info.code               = vert_spv.data();
    vert_info.code_size          = vert_spv.size();
-    vert_info.entrypoint         = "main";
-    vert_info.format             = SDL_GPU_SHADERFORMAT_SPIRV;
+    vert_info.entrypoint         = kEntry;
+    vert_info.format             = kFmt;
    vert_info.stage              = SDL_GPU_SHADERSTAGE_VERTEX;
    vert_info.num_samplers       = 0;
    vert_info.num_uniform_buffers = 0;
@@ -133,8 +145,8 @@ SDL_GPUGraphicsPipeline* GpuShaderPreset::buildPassPipeline(SDL_GPUDevice* devic
    SDL_GPUShaderCreateInfo frag_info = {};
    frag_info.code               = frag_spv.data();
    frag_info.code_size          = frag_spv.size();
-    frag_info.entrypoint         = "main";
-    frag_info.format             = SDL_GPU_SHADERFORMAT_SPIRV;
+    frag_info.entrypoint         = kEntry;
+    frag_info.format             = kFmt;
    frag_info.stage              = SDL_GPU_SHADERSTAGE_FRAGMENT;
    frag_info.num_samplers       = 1;
    frag_info.num_uniform_buffers = 1;
@@ -201,8 +213,13 @@ bool GpuShaderPreset::load(SDL_GPUDevice* device,
        bool is_last = (i == n - 1);
        SDL_GPUTextureFormat target_fmt = is_last ? swapchain_fmt : inter_fmt;

-        std::string vert_spv = dir + "/" + descs_[i].vert_name + ".spv";
-        std::string frag_spv = dir + "/" + descs_[i].frag_name + ".spv";
+#ifdef __APPLE__
+        const char* ext = ".spv.msl";
+#else
+        const char* ext = ".spv";
+#endif
+        std::string vert_spv = dir + "/" + descs_[i].vert_name + ext;
+        std::string frag_spv = dir + "/" + descs_[i].frag_name + ext;

        passes_[i].pipeline = buildPassPipeline(device, vert_spv, frag_spv, target_fmt);
        if (!passes_[i].pipeline) {