base code

11 files changed, 2372 insertions, 0 deletions

diff --git a/src/assets.zig b/src/assets.zig
new file mode 100644
index 0000000..229466e
--- /dev/null
+++ b/src/assets.zig
@@ -0,0 +1,2 @@
+const c = @import("../c.zig");
+const std = @import("std");
diff --git a/src/c.zig b/src/c.zig
new file mode 100644
index 0000000..d91f49f
--- /dev/null
+++ b/src/c.zig
@@ -0,0 +1,5 @@
+pub usingnamespace @cImport({
+    @cDefine("GLFW_INCLUDE_NONE", {});
+    @cInclude("vulkan/vulkan.h");
+    @cInclude("GLFW/glfw3.h");
+});
diff --git a/src/main.zig b/src/main.zig
new file mode 100644
index 0000000..5cbeef2
--- /dev/null
+++ b/src/main.zig
@@ -0,0 +1,45 @@
+const std = @import("std");
+const c = @import("c.zig");
+const window = @import("render/window.zig");
+
+const config = @import("config");
+const Renderer = @import("render/renderer_vulkan.zig");
+const math = @import("math.zig");
+const Parser = @import("vm/parse.zig");
+const vm = @import("vm/vm.zig");
+const wasm = @import("vm/wasm.zig");
+
+pub fn main() !void {
+    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
+    const allocator = gpa.allocator();
+    {
+        var global_runtime = wasm.GlobalRuntime.init(allocator);
+        defer global_runtime.deinit();
+        try global_runtime.addFunction("debug", wasm.debug);
+
+        const file = try std.fs.cwd().openFile("assets/core.wasm", .{});
+        const module = try Parser.parseWasm(allocator, file.reader());
+        var runtime = try vm.Runtime.init(allocator, module, &global_runtime);
+        defer runtime.deinit(allocator);
+
+        var parameters = [_]usize{};
+        try runtime.callExternal(allocator, "fibonacci", &parameters);
+
+        const w = try window.Window.create(800, 600, "sideros");
+        defer w.destroy();
+
+        // TODO: Renderer.destroy should not return an error?
+        var r = try Renderer.create(allocator, w);
+        defer r.destroy() catch {};
+
+        while (!w.shouldClose()) {
+            c.glfwPollEvents();
+            try r.tick();
+        }
+        try r.device.waitIdle();
+    }
+
+    if (gpa.detectLeaks()) {
+        return error.leaked_memory;
+    }
+}
diff --git a/src/math.zig b/src/math.zig
new file mode 100644
index 0000000..a2cbcd2
--- /dev/null
+++ b/src/math.zig
@@ -0,0 +1,64 @@
+const std = @import("std");
+pub const tan = std.math.tan;
+pub const cos = std.math.cos;
+pub const sin = std.math.sin;
+pub const rad = std.math.degreesToRadians;
+
+pub const Matrix = struct {
+    rows: [4]@Vector(4, f32),
+
+    pub fn lookAt(eye: @Vector(3, f32), target: @Vector(3, f32), arbitrary_up: @Vector(3, f32)) Matrix {
+        const forward = normalize(eye - target);
+        const right = normalize(cross(arbitrary_up, forward));
+        const up = cross(forward, right);
+
+        const view = [_]@Vector(4, f32){
+            @Vector(4, f32){ right[0], right[1], right[2], 0.0 },
+            @Vector(4, f32){ up[0], up[1], up[2], 0.0 },
+            @Vector(4, f32){ forward[0], forward[1], forward[2], 0.0 },
+            @Vector(4, f32){ 0.0, 0.0, 1.0, eye[2] },
+        };
+
+        return Matrix{
+            .rows = view,
+        };
+    }
+
+    pub fn perspective(fov: f32, aspect: f32, near: f32, far: f32) Matrix {
+        const projection = [_]@Vector(4, f32){
+            @Vector(4, f32){ 1.0 / (aspect * tan(fov / 2.0)), 0.0, 0.0, 0.0 },
+            @Vector(4, f32){ 0.0, 1.0 / tan(fov / 2.0), 0.0, 0.0 },
+            @Vector(4, f32){ 0.0, 0.0, -((far + near) / (far - near)), -((2 * far * near) / (far - near)) },
+            @Vector(4, f32){ 0.0, 0.0, -1.0, 1.0 },
+        };
+
+        return Matrix{
+            .rows = projection,
+        };
+    }
+
+    pub fn identity() Matrix {
+        const view = [_]@Vector(4, f32){
+            @Vector(4, f32){ 1.0, 0.0, 0.0, 0.0 },
+            @Vector(4, f32){ 0.0, 1.0, 0.0, 0.0 },
+            @Vector(4, f32){ 0.0, 0.0, 1.0, 0.0 },
+            @Vector(4, f32){ 0.0, 0.0, 0.0, 1.0 },
+        };
+
+        return Matrix{
+            .rows = view,
+        };
+    }
+};
+
+pub fn dot(a: @Vector(3, f32), b: @Vector(3, f32)) f32 {
+    return @reduce(.Add, a * b);
+}
+
+pub fn cross(a: @Vector(3, f32), b: @Vector(3, f32)) @Vector(3, f32) {
+    return @Vector(3, f32){ a[1] * b[2] - a[2] * b[1], a[2] * b[0] - a[0] * b[2], a[0] * b[1] - a[1] * b[0] };
+}
+
+pub fn normalize(a: @Vector(3, f32)) @Vector(3, f32) {
+    return a / @as(@Vector(3, f32), @splat(@sqrt(dot(a, a))));
+}
diff --git a/src/render/mesh.zig b/src/render/mesh.zig
new file mode 100644
index 0000000..92fa303
--- /dev/null
+++ b/src/render/mesh.zig
@@ -0,0 +1,119 @@
+const c = @import("../c.zig");
+const std = @import("std");
+const vk = @import("vulkan.zig");
+const Allocator = std.mem.Allocator;
+
+pub const Vertex = struct {
+    position: [3]f32,
+
+    pub fn create(x: f32, y: f32, z: f32) Vertex {
+        return Vertex{
+            .position = .{ x, y, z },
+        };
+    }
+
+    pub fn bindingDescription() c.VkVertexInputBindingDescription {
+        const binding_description: c.VkVertexInputBindingDescription = .{
+            .binding = 0,
+            .stride = @sizeOf(Vertex),
+            .inputRate = c.VK_VERTEX_INPUT_RATE_VERTEX,
+        };
+
+        return binding_description;
+    }
+
+    pub fn attributeDescription() c.VkVertexInputAttributeDescription {
+        const attribute_description: c.VkVertexInputAttributeDescription = .{
+            .location = 0,
+            .binding = 0,
+            .format = c.VK_FORMAT_R32G32B32_SFLOAT,
+            .offset = 0,
+        };
+
+        return attribute_description;
+    }
+};
+
+pub const Mesh = struct {
+    vertex_buffer: vk.Buffer,
+    index_buffer: vk.Buffer,
+
+    pub fn createVertexBuffer(device: anytype) !vk.Buffer {
+        const vertices = [_]Vertex{
+            Vertex.create(0.5, -0.5, 0.0),
+            Vertex.create(0.5, 0.5, 0.0),
+            Vertex.create(-0.5, 0.5, 0.0),
+            Vertex.create(-0.5, -0.5, 0.0),
+        };
+
+        var data: [*c]?*anyopaque = null;
+
+        const buffer = try device.createBuffer(vk.BufferUsage{ .transfer_src = true }, vk.BufferFlags{ .host_visible = true, .host_coherent = true }, @sizeOf(Vertex) * vertices.len);
+
+        try vk.mapError(c.vkMapMemory(
+            device.handle,
+            buffer.memory,
+            0,
+            buffer.size,
+            0,
+            @ptrCast(&data),
+        ));
+
+        if (data) |ptr| {
+            const gpu_vertices: [*]Vertex = @ptrCast(@alignCast(ptr));
+
+            @memcpy(gpu_vertices, vertices[0..]);
+        }
+
+        c.vkUnmapMemory(device.handle, buffer.memory);
+
+        const vertex_buffer = try device.createBuffer(vk.BufferUsage{ .vertex_buffer = true, .transfer_dst = true }, vk.BufferFlags{ .device_local = true }, @sizeOf(Vertex) * vertices.len);
+
+        try buffer.copyTo(device, vertex_buffer);
+        buffer.destroy(device.handle);
+
+        return vertex_buffer;
+    }
+
+    pub fn createIndexBuffer(device: anytype) !vk.Buffer {
+        const indices = [_]u16{ 0, 1, 2, 3, 0, 2 };
+
+        var data: [*c]?*anyopaque = null;
+
+        const buffer = try device.createBuffer(vk.BufferUsage{ .transfer_src = true }, vk.BufferFlags{ .host_visible = true, .host_coherent = true }, @sizeOf(u16) * indices.len);
+
+        try vk.mapError(c.vkMapMemory(
+            device.handle,
+            buffer.memory,
+            0,
+            buffer.size,
+            0,
+            @ptrCast(&data),
+        ));
+
+        if (data) |ptr| {
+            const gpu_indices: [*]u16 = @ptrCast(@alignCast(ptr));
+
+            @memcpy(gpu_indices, indices[0..]);
+        }
+
+        c.vkUnmapMemory(device.handle, buffer.memory);
+
+        const index_buffer = try device.createBuffer(vk.BufferUsage{ .index_buffer = true, .transfer_dst = true }, vk.BufferFlags{ .device_local = true }, @sizeOf(u16) * indices.len);
+
+        try buffer.copyTo(device, index_buffer);
+        buffer.destroy(device.handle);
+
+        return index_buffer;
+    }
+
+    pub fn create(device: anytype) !Mesh {
+        const vertex_buffer = try Mesh.createVertexBuffer(device);
+        const index_buffer = try Mesh.createIndexBuffer(device);
+
+        return Mesh{
+            .vertex_buffer = vertex_buffer,
+            .index_buffer = index_buffer,
+        };
+    }
+};
diff --git a/src/render/renderer_vulkan.zig b/src/render/renderer_vulkan.zig
new file mode 100644
index 0000000..93088c9
--- /dev/null
+++ b/src/render/renderer_vulkan.zig
@@ -0,0 +1,94 @@
+const c = @import("../c.zig");
+const std = @import("std");
+const vk = @import("vulkan.zig");
+const window = @import("window.zig");
+const mesh = @import("mesh.zig");
+const Allocator = std.mem.Allocator;
+
+const Renderer = @This();
+
+instance: vk.Instance,
+surface: vk.Surface,
+physical_device: vk.PhysicalDevice,
+device: vk.Device(2),
+render_pass: vk.RenderPass(2),
+swapchain: vk.Swapchain(2),
+graphics_pipeline: vk.GraphicsPipeline(2),
+current_frame: u32,
+vertex_buffer: vk.Buffer,
+index_buffer: vk.Buffer,
+
+pub fn create(allocator: Allocator, w: window.Window) !Renderer {
+    const instance = try vk.Instance.create();
+
+    const surface = try vk.Surface.create(instance, w);
+
+    var physical_device = try vk.PhysicalDevice.pick(allocator, instance);
+    const device = try physical_device.create_device(surface, allocator, 2);
+
+    const vertex_shader = try device.createShader("shader_vert");
+    defer device.destroyShader(vertex_shader);
+    const fragment_shader = try device.createShader("shader_frag");
+    defer device.destroyShader(fragment_shader);
+
+    const render_pass = try vk.RenderPass(2).create(allocator, device, surface, physical_device);
+
+    const swapchain = try vk.Swapchain(2).create(allocator, surface, device, physical_device, w, render_pass);
+
+    const graphics_pipeline = try vk.GraphicsPipeline(2).create(device, swapchain, render_pass, vertex_shader, fragment_shader);
+
+    // TODO: I think the renderer shouldn't have to interact with buffers. I think the API should change to
+    // something along the lines of
+    //    renderer.begin()
+    //    renderer.render(triangle);
+    //    renderer.render(some_other_thing);
+    //    ...
+    //    renderer.submit()
+    const triangle = try mesh.Mesh.create(device);
+
+    return Renderer{
+        .instance = instance,
+        .surface = surface,
+        .physical_device = physical_device,
+        .device = device,
+        .render_pass = render_pass,
+        .swapchain = swapchain,
+        .graphics_pipeline = graphics_pipeline,
+        .current_frame = 0,
+        // TODO: Why are we storing the buffer and not the Mesh?
+        .vertex_buffer = triangle.vertex_buffer,
+        .index_buffer = triangle.index_buffer,
+    };
+}
+
+pub fn destroy(self: Renderer) !void {
+    try self.device.waitIdle();
+    self.index_buffer.destroy(self.device.handle);
+    self.vertex_buffer.destroy(self.device.handle);
+    self.graphics_pipeline.destroy(self.device);
+    self.swapchain.destroy(self.device);
+    self.render_pass.destroy(self.device);
+    self.device.destroy();
+    self.surface.destroy(self.instance);
+    self.instance.destroy();
+}
+
+// TODO: tick is maybe a bad name? something like present() or submit() is better?
+pub fn tick(self: *Renderer) !void {
+    try self.device.waitFence(self.current_frame);
+    const image = try self.swapchain.nextImage(self.device, self.current_frame);
+    try self.device.resetCommand(self.current_frame);
+    try self.device.beginCommand(self.current_frame);
+    self.render_pass.begin(self.swapchain, self.device, image, self.current_frame);
+    self.graphics_pipeline.bind(self.device, self.current_frame);
+    self.device.bindVertexBuffer(self.vertex_buffer, self.current_frame);
+    self.device.bindIndexBuffer(self.index_buffer, self.current_frame);
+    self.device.bindDescriptorSets(self.graphics_pipeline, self.current_frame);
+    self.device.draw(@intCast(self.index_buffer.size / @sizeOf(u16)), self.current_frame);
+    self.render_pass.end(self.device, self.current_frame);
+    try self.device.endCommand(self.current_frame);
+
+    try self.device.submit(self.swapchain, image, self.current_frame);
+
+    self.current_frame = (self.current_frame + 1) % 2;
+}
diff --git a/src/render/vulkan.zig b/src/render/vulkan.zig
new file mode 100644
index 0000000..e9993a5
--- /dev/null
+++ b/src/render/vulkan.zig
@@ -0,0 +1,1097 @@
+const std = @import("std");
+const c = @import("../c.zig");
+const window = @import("./window.zig");
+const mesh = @import("./mesh.zig");
+const math = @import("../math.zig");
+const Allocator = std.mem.Allocator;
+
+const builtin = @import("builtin");
+const debug = (builtin.mode == .Debug);
+
+const Uniform = struct {
+    proj: math.Matrix,
+    view: math.Matrix,
+    model: math.Matrix,
+};
+
+const validation_layers: []const [*c]const u8 = if (!debug) &[0][*c]const u8{} else &[_][*c]const u8{
+    "VK_LAYER_KHRONOS_validation",
+};
+
+const device_extensions: []const [*c]const u8 = &[_][*c]const u8{
+    c.VK_KHR_SWAPCHAIN_EXTENSION_NAME,
+};
+
+pub const Error = error{
+    out_of_host_memory,
+    out_of_device_memory,
+    initialization_failed,
+    layer_not_present,
+    extension_not_present,
+    incompatible_driver,
+    unknown_error,
+};
+
+pub fn mapError(result: c_int) !void {
+    return switch (result) {
+        c.VK_SUCCESS => {},
+        c.VK_ERROR_OUT_OF_HOST_MEMORY => Error.out_of_host_memory,
+        c.VK_ERROR_OUT_OF_DEVICE_MEMORY => Error.out_of_device_memory,
+        c.VK_ERROR_INITIALIZATION_FAILED => Error.initialization_failed,
+        c.VK_ERROR_LAYER_NOT_PRESENT => Error.layer_not_present,
+        c.VK_ERROR_EXTENSION_NOT_PRESENT => Error.extension_not_present,
+        c.VK_ERROR_INCOMPATIBLE_DRIVER => Error.incompatible_driver,
+        else => Error.unknown_error,
+    };
+}
+
+pub const BufferUsage = packed struct(u32) {
+    transfer_src: bool = false,
+    transfer_dst: bool = false,
+    uniform_texel_buffer: bool = false,
+    storage_texel_buffer: bool = false,
+    uniform_buffer: bool = false,
+    storage_buffer: bool = false,
+    index_buffer: bool = false,
+    vertex_buffer: bool = false,
+    indirect_buffer: bool = false,
+    _padding: enum(u23) { unset } = .unset,
+};
+
+pub const BufferFlags = packed struct(u32) {
+    device_local: bool = false,
+    host_visible: bool = false,
+    host_coherent: bool = false,
+    host_cached: bool = false,
+    lazily_allocated: bool = false,
+    _padding: enum(u27) { unset } = .unset,
+};
+
+pub const Instance = struct {
+    handle: c.VkInstance,
+
+    pub fn create() !Instance {
+        const extensions = window.getExtensions();
+
+        const app_info: c.VkApplicationInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_APPLICATION_INFO,
+            .pApplicationName = "sideros",
+            .applicationVersion = c.VK_MAKE_VERSION(1, 0, 0),
+            .engineVersion = c.VK_MAKE_VERSION(1, 0, 0),
+            .pEngineName = "sideros",
+            .apiVersion = c.VK_MAKE_VERSION(1, 3, 0),
+        };
+
+        const instance_info: c.VkInstanceCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
+            .pApplicationInfo = &app_info,
+            .enabledExtensionCount = @intCast(extensions.len),
+            .ppEnabledExtensionNames = extensions.ptr,
+            .enabledLayerCount = @intCast(validation_layers.len),
+            .ppEnabledLayerNames = validation_layers.ptr,
+        };
+
+        var instance: c.VkInstance = undefined;
+
+        try mapError(c.vkCreateInstance(&instance_info, null, &instance));
+
+        return Instance{
+            .handle = instance,
+        };
+    }
+
+    pub fn destroy(self: Instance) void {
+        c.vkDestroyInstance(self.handle, null);
+    }
+};
+
+pub const Buffer = struct {
+    handle: c.VkBuffer,
+    memory: c.VkDeviceMemory,
+    size: usize,
+
+    pub fn copyTo(self: Buffer, device: anytype, dest: Buffer) !void {
+        const command_buffer_info: c.VkCommandBufferAllocateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
+            .commandPool = device.command_pool,
+            .level = c.VK_COMMAND_BUFFER_LEVEL_PRIMARY,
+            .commandBufferCount = 1,
+        };
+
+        var command_buffer: c.VkCommandBuffer = undefined;
+        try mapError(c.vkAllocateCommandBuffers(device.handle, &command_buffer_info, @ptrCast(&command_buffer)));
+
+        const begin_info: c.VkCommandBufferBeginInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+            .flags = c.VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+        };
+
+        try mapError(c.vkBeginCommandBuffer(command_buffer, &begin_info));
+
+        const copy_region: c.VkBufferCopy = .{
+            .srcOffset = 0,
+            .dstOffset = 0,
+            .size = self.size,
+        };
+
+        c.vkCmdCopyBuffer(command_buffer, self.handle, dest.handle, 1, &copy_region);
+        try mapError(c.vkEndCommandBuffer(command_buffer));
+
+        const submit_info: c.VkSubmitInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_SUBMIT_INFO,
+            .commandBufferCount = 1,
+            .pCommandBuffers = &command_buffer,
+        };
+
+        try mapError(c.vkQueueSubmit(device.graphics_queue, 1, &submit_info, null));
+        try mapError(c.vkQueueWaitIdle(device.graphics_queue));
+        c.vkFreeCommandBuffers(device.handle, device.command_pool, 1, &command_buffer);
+    }
+
+    pub fn destroy(self: Buffer, device_handle: c.VkDevice) void {
+        c.vkDestroyBuffer(device_handle, self.handle, null);
+        c.vkFreeMemory(device_handle, self.memory, null);
+    }
+};
+
+pub fn RenderPass(comptime n: usize) type {
+    return struct {
+        handle: c.VkRenderPass,
+
+        const Self = @This();
+
+        pub fn create(allocator: Allocator, device: Device(n), surface: Surface, physical_device: PhysicalDevice) !Self {
+            const color_attachment: c.VkAttachmentDescription = .{
+                .format = (try Swapchain(n).pickFormat(allocator, surface, physical_device)).format,
+                .samples = c.VK_SAMPLE_COUNT_1_BIT,
+                .loadOp = c.VK_ATTACHMENT_LOAD_OP_CLEAR,
+                .storeOp = c.VK_ATTACHMENT_STORE_OP_STORE,
+                .stencilLoadOp = c.VK_ATTACHMENT_LOAD_OP_DONT_CARE,
+                .stencilStoreOp = c.VK_ATTACHMENT_STORE_OP_DONT_CARE,
+                .initialLayout = c.VK_IMAGE_LAYOUT_UNDEFINED,
+                .finalLayout = c.VK_IMAGE_LAYOUT_PRESENT_SRC_KHR,
+            };
+
+            const color_attachment_reference: c.VkAttachmentReference = .{
+                .attachment = 0,
+                .layout = c.VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
+            };
+
+            const subpass: c.VkSubpassDescription = .{
+                .pipelineBindPoint = c.VK_PIPELINE_BIND_POINT_GRAPHICS,
+                .colorAttachmentCount = 1,
+                .pColorAttachments = &color_attachment_reference,
+            };
+
+            const dependency: c.VkSubpassDependency = .{
+                .srcSubpass = c.VK_SUBPASS_EXTERNAL,
+                .dstSubpass = 0,
+                .srcStageMask = c.VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+                .srcAccessMask = 0,
+                .dstStageMask = c.VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+                .dstAccessMask = c.VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+            };
+
+            const render_pass_info: c.VkRenderPassCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
+                .attachmentCount = 1,
+                .pAttachments = &color_attachment,
+                .subpassCount = 1,
+                .pSubpasses = &subpass,
+                .dependencyCount = 1,
+                .pDependencies = &dependency,
+            };
+
+            var render_pass: c.VkRenderPass = undefined;
+
+            try mapError(c.vkCreateRenderPass(device.handle, &render_pass_info, null, &render_pass));
+
+            return Self{
+                .handle = render_pass,
+            };
+        }
+
+        pub fn begin(self: Self, swapchain: Swapchain(n), device: Device(n), image: usize, frame: usize) void {
+            std.debug.assert(frame < n);
+            const clear_color: c.VkClearValue = .{ .color = .{ .float32 = .{ 0.0, 0.0, 0.0, 1.0 } } };
+
+            const begin_info: c.VkRenderPassBeginInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+                .renderPass = self.handle,
+                .framebuffer = swapchain.framebuffers[image],
+                .renderArea = .{
+                    .offset = .{ .x = 0, .y = 0 },
+                    .extent = swapchain.extent,
+                },
+                .clearValueCount = 1,
+                .pClearValues = &clear_color,
+            };
+
+            c.vkCmdBeginRenderPass(device.command_buffers[frame], &begin_info, c.VK_SUBPASS_CONTENTS_INLINE);
+        }
+
+        pub fn end(self: Self, device: Device(n), frame: usize) void {
+            _ = self;
+            std.debug.assert(frame < n);
+            c.vkCmdEndRenderPass(device.command_buffers[frame]);
+        }
+
+        pub fn destroy(self: Self, device: Device(n)) void {
+            c.vkDestroyRenderPass(device.handle, self.handle, null);
+        }
+    };
+}
+
+pub fn GraphicsPipeline(comptime n: usize) type {
+    return struct {
+        layout: c.VkPipelineLayout,
+        handle: c.VkPipeline,
+        descriptor_pool: c.VkDescriptorPool,
+        descriptor_set: c.VkDescriptorSet,
+        descriptor_set_layout: c.VkDescriptorSetLayout,
+        uniform_buffer: Buffer,
+
+        const Self = @This();
+
+        pub fn create(device: Device(n), swapchain: Swapchain(n), render_pass: RenderPass(n), vertex_shader: c.VkShaderModule, fragment_shader: c.VkShaderModule) !Self {
+            const vertex_shader_stage_info: c.VkPipelineShaderStageCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+                .stage = c.VK_SHADER_STAGE_VERTEX_BIT,
+                .module = vertex_shader,
+                .pName = "main",
+            };
+
+            const fragment_shader_stage_info: c.VkPipelineShaderStageCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+                .stage = c.VK_SHADER_STAGE_FRAGMENT_BIT,
+                .module = fragment_shader,
+                .pName = "main",
+            };
+
+            // TODO: shouldn't this be closer to usage?
+            const shader_stage_infos: []const c.VkPipelineShaderStageCreateInfo = &.{ vertex_shader_stage_info, fragment_shader_stage_info };
+
+            const vertex_attributes: []const c.VkVertexInputAttributeDescription = &.{mesh.Vertex.attributeDescription()};
+            const vertex_bindings: []const c.VkVertexInputBindingDescription = &.{mesh.Vertex.bindingDescription()};
+
+            const vertex_input_info: c.VkPipelineVertexInputStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+                .vertexBindingDescriptionCount = 1,
+                .pVertexBindingDescriptions = vertex_bindings.ptr,
+                .vertexAttributeDescriptionCount = 1,
+                .pVertexAttributeDescriptions = vertex_attributes.ptr,
+            };
+
+            const input_assembly_info: c.VkPipelineInputAssemblyStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+                .topology = c.VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST,
+                .primitiveRestartEnable = c.VK_FALSE,
+            };
+
+            const viewport: c.VkViewport = .{
+                .x = 0.0,
+                .y = 0.0,
+                .width = @floatFromInt(swapchain.extent.width),
+                .height = @floatFromInt(swapchain.extent.height),
+                .minDepth = 0.0,
+                .maxDepth = 1.0,
+            };
+
+            const scissor: c.VkRect2D = .{
+                .offset = .{
+                    .x = 0.0,
+                    .y = 0.0,
+                },
+                .extent = swapchain.extent,
+            };
+
+            const viewport_state_info: c.VkPipelineViewportStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+                .viewportCount = 1,
+                .pViewports = &viewport,
+                .scissorCount = 1,
+                .pScissors = &scissor,
+            };
+
+            const rasterizer_info: c.VkPipelineRasterizationStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+                .depthClampEnable = c.VK_FALSE,
+                .rasterizerDiscardEnable = c.VK_FALSE,
+                .polygonMode = c.VK_POLYGON_MODE_FILL,
+                .lineWidth = 1.0,
+                .cullMode = c.VK_CULL_MODE_BACK_BIT,
+                .frontFace = c.VK_FRONT_FACE_CLOCKWISE,
+                .depthBiasEnable = c.VK_FALSE,
+            };
+
+            const multisampling_info: c.VkPipelineMultisampleStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+                .sampleShadingEnable = c.VK_FALSE,
+                .rasterizationSamples = c.VK_SAMPLE_COUNT_1_BIT,
+            };
+
+            const color_blend_attachment: c.VkPipelineColorBlendAttachmentState = .{
+                .colorWriteMask = c.VK_COLOR_COMPONENT_R_BIT | c.VK_COLOR_COMPONENT_G_BIT | c.VK_COLOR_COMPONENT_B_BIT | c.VK_COLOR_COMPONENT_A_BIT,
+                .blendEnable = c.VK_TRUE,
+                .srcColorBlendFactor = c.VK_BLEND_FACTOR_SRC_ALPHA,
+                .dstColorBlendFactor = c.VK_BLEND_FACTOR_ONE_MINUS_SRC_ALPHA,
+                .colorBlendOp = c.VK_BLEND_OP_ADD,
+                .srcAlphaBlendFactor = c.VK_BLEND_FACTOR_ONE,
+                .dstAlphaBlendFactor = c.VK_BLEND_FACTOR_ZERO,
+                .alphaBlendOp = c.VK_BLEND_OP_ADD,
+            };
+
+            const color_blend_info: c.VkPipelineColorBlendStateCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+                .logicOpEnable = c.VK_FALSE,
+                .logicOp = c.VK_LOGIC_OP_COPY,
+                .attachmentCount = 1,
+                .pAttachments = &color_blend_attachment,
+                .blendConstants = .{ 0.0, 0.0, 0.0, 0.0 },
+            };
+
+            const set_binding = c.VkDescriptorSetLayoutBinding{
+                .binding = 0,
+                .descriptorType = c.VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER,
+                .descriptorCount = 1,
+                .stageFlags = c.VK_SHADER_STAGE_VERTEX_BIT,
+            };
+
+            const bindings = [_]c.VkDescriptorSetLayoutBinding{set_binding};
+
+            var descriptor_set_layout: c.VkDescriptorSetLayout = undefined;
+
+            const descriptor_set_layout_info = c.VkDescriptorSetLayoutCreateInfo{
+                .sType = c.VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+                .bindingCount = 1,
+                .pBindings = bindings[0..].ptr,
+            };
+
+            try mapError(c.vkCreateDescriptorSetLayout(device.handle, &descriptor_set_layout_info, null, &descriptor_set_layout));
+
+            const set_layouts = [_]c.VkDescriptorSetLayout{descriptor_set_layout};
+
+            const layout_info: c.VkPipelineLayoutCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+                .setLayoutCount = 1,
+                .pSetLayouts = set_layouts[0..].ptr,
+                .pushConstantRangeCount = 0,
+                .pPushConstantRanges = null,
+            };
+
+            var layout: c.VkPipelineLayout = undefined;
+
+            try mapError(c.vkCreatePipelineLayout(device.handle, &layout_info, null, @ptrCast(&layout)));
+
+            const pipeline_info: c.VkGraphicsPipelineCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+                .stageCount = 2,
+                .pStages = shader_stage_infos.ptr,
+                .pVertexInputState = &vertex_input_info,
+                .pInputAssemblyState = &input_assembly_info,
+                .pViewportState = &viewport_state_info,
+                .pRasterizationState = &rasterizer_info,
+                .pMultisampleState = &multisampling_info,
+                .pDepthStencilState = null,
+                .pColorBlendState = &color_blend_info,
+                .pDynamicState = null,
+                .layout = layout,
+                .renderPass = render_pass.handle,
+                .subpass = 0,
+                .basePipelineHandle = null,
+                .basePipelineIndex = -1,
+            };
+
+            var pipeline: c.VkPipeline = undefined;
+
+            try mapError(c.vkCreateGraphicsPipelines(device.handle, null, 1, &pipeline_info, null, @ptrCast(&pipeline)));
+
+            var size = c.VkDescriptorPoolSize{
+                .type = c.VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER,
+                .descriptorCount = 1,
+            };
+
+            const descriptor_pool_info = c.VkDescriptorPoolCreateInfo{
+                .sType = c.VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
+                .maxSets = 1,
+                .poolSizeCount = 1,
+                .pPoolSizes = &size,
+            };
+
+            var descriptor_pool: c.VkDescriptorPool = undefined;
+
+            try mapError(c.vkCreateDescriptorPool(device.handle, &descriptor_pool_info, null, &descriptor_pool));
+
+            const descriptor_allocate_info = c.VkDescriptorSetAllocateInfo{
+                .sType = c.VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
+                .descriptorPool = descriptor_pool,
+                .descriptorSetCount = 1,
+                .pSetLayouts = set_layouts[0..].ptr,
+            };
+
+            var descriptor_set: c.VkDescriptorSet = undefined;
+
+            try mapError(c.vkAllocateDescriptorSets(device.handle, &descriptor_allocate_info, &descriptor_set));
+
+            const proj = math.Matrix.perspective(math.rad(45.0), (@as(f32, @floatFromInt(swapchain.extent.width)) / @as(f32, @floatFromInt(swapchain.extent.height))), 0.1, 10.0);
+
+            const view = math.Matrix.lookAt(@Vector(3, f32){ 0.0, 0.0, 10.0 }, @Vector(3, f32){ 0.0, 0.0, 0.0 }, @Vector(3, f32){ 0.0, 1.0, 0.0 });
+            const model = math.Matrix.identity();
+
+            const uniform = Uniform{
+                .proj = proj,
+                .view = view,
+                .model = model,
+            };
+
+            const uniform_buffer = try device.createBuffer(BufferUsage{ .uniform_buffer = true, .transfer_dst = true }, BufferFlags{ .device_local = true }, @sizeOf(Uniform));
+
+            var data: [*c]u8 = undefined;
+
+            try mapError(c.vkMapMemory(
+                device.handle,
+                uniform_buffer.memory,
+                0,
+                uniform_buffer.size,
+                0,
+                @ptrCast(&data),
+            ));
+
+            @memcpy(data[0..(@sizeOf(math.Matrix) * 3)], std.mem.asBytes(&uniform));
+
+            const descriptor_buffer_info = c.VkDescriptorBufferInfo{
+                .buffer = uniform_buffer.handle,
+                .offset = 0,
+                .range = uniform_buffer.size,
+            };
+
+            const write_descriptor_set = c.VkWriteDescriptorSet{
+                .sType = c.VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+                .dstSet = descriptor_set,
+                .dstBinding = 0,
+                .dstArrayElement = 0,
+                .descriptorCount = 1,
+                .descriptorType = c.VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER,
+                .pBufferInfo = &descriptor_buffer_info,
+            };
+
+            c.vkUpdateDescriptorSets(device.handle, 1, &write_descriptor_set, 0, null);
+
+            return Self{
+                .layout = layout,
+                .handle = pipeline,
+                .descriptor_pool = descriptor_pool,
+                .descriptor_set = descriptor_set,
+                .descriptor_set_layout = descriptor_set_layout,
+                .uniform_buffer = uniform_buffer,
+            };
+        }
+
+        pub fn bind(self: Self, device: Device(n), frame: usize) void {
+            std.debug.assert(frame < n);
+            c.vkCmdBindPipeline(device.command_buffers[frame], c.VK_PIPELINE_BIND_POINT_GRAPHICS, self.handle);
+        }
+
+        pub fn destroy(self: Self, device: Device(n)) void {
+            self.uniform_buffer.destroy(device.handle);
+            c.vkDestroyDescriptorSetLayout(device.handle, self.descriptor_set_layout, null);
+            c.vkDestroyDescriptorPool(device.handle, self.descriptor_pool, null);
+            c.vkDestroyPipeline(device.handle, self.handle, null);
+            c.vkDestroyPipelineLayout(device.handle, self.layout, null);
+        }
+    };
+}
+
+//pub const Shader = struct {
+//    handle: c.VkShaderModule,
+//
+//    pub fn create(comptime name: []const u8, device: Device) !Shader {
+//        const code = @embedFile(name);
+//
+//        const create_info: c.VkShaderModuleCreateInfo = .{
+//            .sType = c.VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
+//            .codeSize = code.len,
+//            .pCode = @ptrCast(@alignCast(code)),
+//        };
+//
+//        var shader_module: c.VkShaderModule = undefined;
+//
+//        try mapError(c.vkCreateShaderModule(device.handle, &create_info, null, @ptrCast(&shader_module)));
+//
+//        return Shader{
+//            .handle = shader_module,
+//        };
+//    }
+//
+//    pub fn destroy(self: Shader, device: Device) void {
+//        c.vkDestroyShaderModule(device.handle, self.handle, null);
+//    }
+//};
+
+pub fn Swapchain(comptime n: usize) type {
+    return struct {
+        handle: c.VkSwapchainKHR,
+        images: []c.VkImage,
+        image_views: []c.VkImageView,
+        format: c.VkSurfaceFormatKHR,
+        extent: c.VkExtent2D,
+        framebuffers: []c.VkFramebuffer,
+
+        allocator: Allocator,
+
+        const Self = @This();
+
+        // TODO: This should not be part of the Swapchain?
+        pub fn pickFormat(allocator: Allocator, surface: Surface, physical_device: PhysicalDevice) !c.VkSurfaceFormatKHR {
+            const formats = try surface.formats(allocator, physical_device);
+            defer allocator.free(formats);
+            var format: ?c.VkSurfaceFormatKHR = null;
+
+            for (formats) |fmt| {
+                if (fmt.format == c.VK_FORMAT_B8G8R8A8_SRGB and fmt.colorSpace == c.VK_COLOR_SPACE_SRGB_NONLINEAR_KHR) {
+                    format = fmt;
+                }
+            }
+
+            if (format == null) {
+                format = formats[0];
+            }
+
+            return format.?;
+        }
+
+        // TODO: Allow to recreate so Window can be resized
+        pub fn create(allocator: Allocator, surface: Surface, device: Device(n), physical_device: PhysicalDevice, w: window.Window, render_pass: RenderPass(n)) !Self {
+            const present_modes = try surface.presentModes(allocator, physical_device);
+            defer allocator.free(present_modes);
+            const capabilities = try surface.capabilities(physical_device);
+            var present_mode: ?c.VkPresentModeKHR = null;
+            var extent: c.VkExtent2D = undefined;
+            const format = try Self.pickFormat(allocator, surface, physical_device);
+
+            for (present_modes) |mode| {
+                if (mode == c.VK_PRESENT_MODE_MAILBOX_KHR) {
+                    present_mode = mode;
+                }
+            }
+
+            if (present_mode == null) {
+                present_mode = c.VK_PRESENT_MODE_FIFO_KHR;
+            }
+
+            if (capabilities.currentExtent.width != std.math.maxInt(u32)) {
+                extent = capabilities.currentExtent;
+            } else {
+                const width, const height = w.size();
+
+                extent = .{
+                    .width = @intCast(width),
+                    .height = @intCast(height),
+                };
+
+                extent.width = std.math.clamp(extent.width, capabilities.minImageExtent.width, capabilities.maxImageExtent.width);
+                extent.height = std.math.clamp(extent.height, capabilities.minImageExtent.height, capabilities.maxImageExtent.height);
+            }
+
+            var create_info: c.VkSwapchainCreateInfoKHR = .{
+                .sType = c.VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR,
+                .surface = surface.handle,
+                .minImageCount = capabilities.minImageCount + 1,
+                .imageFormat = format.format,
+                .imageColorSpace = format.colorSpace,
+                .imageExtent = extent,
+                .imageArrayLayers = 1,
+                .imageUsage = c.VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
+                .preTransform = capabilities.currentTransform,
+                .compositeAlpha = c.VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR,
+                .presentMode = present_mode.?,
+                .clipped = c.VK_TRUE,
+                .oldSwapchain = null,
+            };
+
+            const graphics_family = try physical_device.graphicsQueue(allocator);
+            const present_family = try physical_device.presentQueue(surface, allocator);
+            const family_indices: []const u32 = &.{ graphics_family, present_family };
+
+            if (graphics_family != present_family) {
+                create_info.imageSharingMode = c.VK_SHARING_MODE_CONCURRENT;
+                create_info.queueFamilyIndexCount = @intCast(family_indices.len);
+                create_info.pQueueFamilyIndices = family_indices.ptr;
+            } else {
+                create_info.imageSharingMode = c.VK_SHARING_MODE_EXCLUSIVE;
+                create_info.queueFamilyIndexCount = 0;
+                create_info.pQueueFamilyIndices = null;
+            }
+
+            var swapchain: c.VkSwapchainKHR = undefined;
+
+            try mapError(c.vkCreateSwapchainKHR(device.handle, &create_info, null, &swapchain));
+
+            var image_count: u32 = 0;
+            try mapError(c.vkGetSwapchainImagesKHR(device.handle, swapchain, &image_count, null));
+            const images = try allocator.alloc(c.VkImage, image_count);
+
+            try mapError(c.vkGetSwapchainImagesKHR(device.handle, swapchain, &image_count, @ptrCast(images)));
+
+            const image_views = try allocator.alloc(c.VkImageView, image_count);
+            for (images, 0..) |image, index| {
+                const view_create_info: c.VkImageViewCreateInfo = .{
+                    .sType = c.VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+                    .image = image,
+                    .viewType = c.VK_IMAGE_VIEW_TYPE_2D,
+                    .format = format.format,
+                    .components = .{
+                        .r = c.VK_COMPONENT_SWIZZLE_IDENTITY,
+                        .g = c.VK_COMPONENT_SWIZZLE_IDENTITY,
+                        .b = c.VK_COMPONENT_SWIZZLE_IDENTITY,
+                        .a = c.VK_COMPONENT_SWIZZLE_IDENTITY,
+                    },
+                    .subresourceRange = .{
+                        .aspectMask = c.VK_IMAGE_ASPECT_COLOR_BIT,
+                        .baseMipLevel = 0,
+                        .levelCount = 1,
+                        .baseArrayLayer = 0,
+                        .layerCount = 1,
+                    },
+                };
+
+                try mapError(c.vkCreateImageView(device.handle, &view_create_info, null, &(image_views[index])));
+            }
+
+            const framebuffers = try allocator.alloc(c.VkFramebuffer, image_count);
+            for (image_views, 0..) |view, index| {
+                const framebuffer_info: c.VkFramebufferCreateInfo = .{
+                    .sType = c.VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+                    .renderPass = render_pass.handle,
+                    .attachmentCount = 1,
+                    .pAttachments = &view,
+                    .width = extent.width,
+                    .height = extent.height,
+                    .layers = 1,
+                };
+
+                try mapError(c.vkCreateFramebuffer(device.handle, &framebuffer_info, null, &(framebuffers[index])));
+            }
+
+            return Self{
+                .handle = swapchain,
+                .format = format,
+                .extent = extent,
+                .images = images[0..image_count],
+                .image_views = image_views[0..image_count],
+                .framebuffers = framebuffers,
+                .allocator = allocator,
+            };
+        }
+
+        pub fn nextImage(self: Self, device: Device(n), frame: usize) !usize {
+            std.debug.assert(frame < n);
+            var index: u32 = undefined;
+            try mapError(c.vkAcquireNextImageKHR(device.handle, self.handle, std.math.maxInt(u64), device.image_available[frame], null, &index));
+
+            return @intCast(index);
+        }
+
+        pub fn destroy(self: Self, device: Device(n)) void {
+            for (self.image_views) |view| {
+                c.vkDestroyImageView(device.handle, view, null);
+            }
+
+            for (self.framebuffers) |framebuffer| {
+                c.vkDestroyFramebuffer(device.handle, framebuffer, null);
+            }
+
+            c.vkDestroySwapchainKHR(device.handle, self.handle, null);
+
+            self.allocator.free(self.images);
+            self.allocator.free(self.image_views);
+            self.allocator.free(self.framebuffers);
+        }
+    };
+}
+
+pub const Surface = struct {
+    handle: c.VkSurfaceKHR,
+
+    pub fn create(instance: Instance, w: window.Window) !Surface {
+        var handle: c.VkSurfaceKHR = undefined;
+        try mapError(c.glfwCreateWindowSurface(instance.handle, w.raw, null, &handle));
+        return Surface{
+            .handle = handle,
+        };
+    }
+
+    pub fn presentModes(self: Surface, allocator: Allocator, device: PhysicalDevice) ![]c.VkPresentModeKHR {
+        var mode_count: u32 = 0;
+        try mapError(c.vkGetPhysicalDeviceSurfacePresentModesKHR(device.handle, self.handle, &mode_count, null));
+        const modes = try allocator.alloc(c.VkPresentModeKHR, mode_count);
+        try mapError(c.vkGetPhysicalDeviceSurfacePresentModesKHR(device.handle, self.handle, &mode_count, @ptrCast(modes)));
+
+        return modes[0..mode_count];
+    }
+
+    pub fn formats(self: Surface, allocator: Allocator, device: PhysicalDevice) ![]c.VkSurfaceFormatKHR {
+        var format_count: u32 = 0;
+        try mapError(c.vkGetPhysicalDeviceSurfaceFormatsKHR(device.handle, self.handle, &format_count, null));
+        const fmts = try allocator.alloc(c.VkSurfaceFormatKHR, format_count);
+        try mapError(c.vkGetPhysicalDeviceSurfaceFormatsKHR(device.handle, self.handle, &format_count, @ptrCast(fmts)));
+
+        return fmts[0..format_count];
+    }
+
+    pub fn capabilities(self: Surface, device: PhysicalDevice) !c.VkSurfaceCapabilitiesKHR {
+        var caps: c.VkSurfaceCapabilitiesKHR = undefined;
+        try mapError(c.vkGetPhysicalDeviceSurfaceCapabilitiesKHR(device.handle, self.handle, &caps));
+        return caps;
+    }
+
+    pub fn destroy(self: Surface, instance: Instance) void {
+        c.vkDestroySurfaceKHR(instance.handle, self.handle, null);
+    }
+};
+
+pub fn Device(comptime n: usize) type {
+    return struct {
+        handle: c.VkDevice,
+        graphics_queue: c.VkQueue,
+        present_queue: c.VkQueue,
+        command_pool: c.VkCommandPool,
+        command_buffers: [n]c.VkCommandBuffer,
+        image_available: [n]c.VkSemaphore,
+        render_finished: [n]c.VkSemaphore,
+        in_flight_fence: [n]c.VkFence,
+        graphics_family: u32,
+        present_family: u32,
+        memory_properties: c.VkPhysicalDeviceMemoryProperties,
+
+        const Self = @This();
+
+        pub fn resetCommand(self: Self, frame: usize) !void {
+            std.debug.assert(frame < n);
+            try mapError(c.vkResetCommandBuffer(self.command_buffers[frame], 0));
+        }
+
+        pub fn beginCommand(self: Self, frame: usize) !void {
+            std.debug.assert(frame < n);
+            const begin_info: c.VkCommandBufferBeginInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+                .flags = c.VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+            };
+            try mapError(c.vkBeginCommandBuffer(self.command_buffers[frame], &begin_info));
+        }
+
+        pub fn endCommand(self: Self, frame: usize) !void {
+            std.debug.assert(frame < n);
+            try mapError(c.vkEndCommandBuffer(self.command_buffers[frame]));
+        }
+
+        pub fn draw(self: Self, indices: u32, frame: usize) void {
+            std.debug.assert(frame < n);
+            c.vkCmdDrawIndexed(self.command_buffers[frame], indices, 1, 0, 0, 0);
+        }
+
+        pub fn waitFence(self: Self, frame: usize) !void {
+            std.debug.assert(frame < n);
+            try mapError(c.vkWaitForFences(self.handle, 1, &self.in_flight_fence[frame], c.VK_TRUE, std.math.maxInt(u64)));
+            try mapError(c.vkResetFences(self.handle, 1, &self.in_flight_fence[frame]));
+        }
+
+        pub fn waitIdle(self: Self) !void {
+            try mapError(c.vkDeviceWaitIdle(self.handle));
+        }
+
+        pub fn bindIndexBuffer(self: Self, buffer: Buffer, frame: usize) void {
+            std.debug.assert(frame < n);
+            c.vkCmdBindIndexBuffer(self.command_buffers[frame], buffer.handle, 0, c.VK_INDEX_TYPE_UINT16);
+        }
+
+        pub fn bindVertexBuffer(self: Self, buffer: Buffer, frame: usize) void {
+            std.debug.assert(frame < n);
+            const offset: u64 = 0;
+            c.vkCmdBindVertexBuffers(self.command_buffers[frame], 0, 1, &buffer.handle, &offset);
+        }
+
+        pub fn bindDescriptorSets(self: Self, pipeline: GraphicsPipeline(n), frame: usize) void {
+            c.vkCmdBindDescriptorSets(self.command_buffers[frame], c.VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline.layout, 0, 1, &pipeline.descriptor_set, 0, null);
+        }
+
+        pub fn pick_memory_type(self: Self, type_bits: u32, flags: u32) u32 {
+            var memory_type_index: u32 = 0;
+            for (0..self.memory_properties.memoryTypeCount) |index| {
+                const memory_type = self.memory_properties.memoryTypes[index];
+
+                if (((type_bits & (@as(u64, 1) << @intCast(index))) != 0) and (memory_type.propertyFlags & flags) != 0 and (memory_type.propertyFlags & c.VK_MEMORY_PROPERTY_DEVICE_COHERENT_BIT_AMD) == 0) {
+                    memory_type_index = @intCast(index);
+                }
+            }
+
+            return memory_type_index;
+        }
+
+        pub fn createBuffer(self: Self, usage: BufferUsage, flags: BufferFlags, size: usize) !Buffer {
+            const family_indices: []const u32 = &.{self.graphics_family};
+
+            const create_info: c.VkBufferCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
+                .size = size,
+                .sharingMode = c.VK_SHARING_MODE_EXCLUSIVE,
+                .usage = @bitCast(usage),
+                .queueFamilyIndexCount = 1,
+                .pQueueFamilyIndices = family_indices.ptr,
+            };
+
+            var buffer: c.VkBuffer = undefined;
+            try mapError(c.vkCreateBuffer(self.handle, &create_info, null, &buffer));
+
+            var memory_requirements: c.VkMemoryRequirements = undefined;
+            c.vkGetBufferMemoryRequirements(self.handle, buffer, &memory_requirements);
+
+            const alloc_info: c.VkMemoryAllocateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,
+                .allocationSize = memory_requirements.size,
+                .memoryTypeIndex = self.pick_memory_type(memory_requirements.memoryTypeBits, @bitCast(flags)),
+            };
+
+            var device_memory: c.VkDeviceMemory = undefined;
+
+            try mapError(c.vkAllocateMemory(self.handle, &alloc_info, null, &device_memory));
+
+            try mapError(c.vkBindBufferMemory(self.handle, buffer, device_memory, 0));
+
+            return Buffer{
+                .handle = buffer,
+                .size = size,
+                .memory = device_memory,
+            };
+        }
+
+        pub fn submit(self: Self, swapchain: Swapchain(n), image: usize, frame: usize) !void {
+            std.debug.assert(frame < n);
+            //const wait_semaphores: [1]c.VkSemaphore = .{self.image_available[frame]};
+            //const signal_semaphores: [1]c.VkSemaphore = .{self.render_finished[frame]};
+            //const swapchains: [1]c.VkSwapchainKHR = .{swapchain.handle};
+            const stages: []const u32 = &[_]u32{c.VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT};
+
+            const submit_info: c.VkSubmitInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_SUBMIT_INFO,
+                .waitSemaphoreCount = 1,
+                .pWaitSemaphores = &self.image_available[frame],
+                .pWaitDstStageMask = stages.ptr,
+                .commandBufferCount = 1,
+                .pCommandBuffers = &self.command_buffers[frame],
+                .signalSemaphoreCount = 1,
+                .pSignalSemaphores = &self.render_finished[frame],
+            };
+
+            try mapError(c.vkQueueSubmit(self.graphics_queue, 1, &submit_info, self.in_flight_fence[frame]));
+
+            const present_info: c.VkPresentInfoKHR = .{
+                .sType = c.VK_STRUCTURE_TYPE_PRESENT_INFO_KHR,
+                .waitSemaphoreCount = 1,
+                .pWaitSemaphores = &self.render_finished[frame],
+                .swapchainCount = 1,
+                .pSwapchains = &swapchain.handle,
+                .pImageIndices = @ptrCast(&image),
+                .pResults = null,
+            };
+
+            try mapError(c.vkQueuePresentKHR(self.present_queue, &present_info));
+        }
+
+        pub fn createShader(self: Self, comptime name: []const u8) !c.VkShaderModule {
+            const code = @embedFile(name);
+
+            const create_info: c.VkShaderModuleCreateInfo = .{
+                .sType = c.VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
+                .codeSize = code.len,
+                .pCode = @ptrCast(@alignCast(code)),
+            };
+
+            var shader_module: c.VkShaderModule = undefined;
+
+            try mapError(c.vkCreateShaderModule(self.handle, &create_info, null, @ptrCast(&shader_module)));
+
+            return shader_module;
+        }
+
+        pub fn destroyShader(self: Self, shader: c.VkShaderModule) void {
+            c.vkDestroyShaderModule(self.handle, shader, null);
+        }
+
+        pub fn destroy(self: Self) void {
+            inline for (0..n) |index| {
+                c.vkDestroySemaphore(self.handle, self.image_available[index], null);
+                c.vkDestroySemaphore(self.handle, self.render_finished[index], null);
+                c.vkDestroyFence(self.handle, self.in_flight_fence[index], null);
+            }
+
+            c.vkDestroyCommandPool(self.handle, self.command_pool, null);
+            c.vkDestroyDevice(self.handle, null);
+        }
+    };
+}
+
+pub const PhysicalDevice = struct {
+    handle: c.VkPhysicalDevice,
+
+    pub fn pick(allocator: Allocator, instance: Instance) !PhysicalDevice {
+        var device_count: u32 = 0;
+        try mapError(c.vkEnumeratePhysicalDevices(instance.handle, &device_count, null));
+        const devices = try allocator.alloc(c.VkPhysicalDevice, device_count);
+        defer allocator.free(devices);
+        try mapError(c.vkEnumeratePhysicalDevices(instance.handle, &device_count, @ptrCast(devices)));
+
+        return PhysicalDevice{ .handle = devices[0] };
+    }
+
+    pub fn queueFamilyProperties(self: PhysicalDevice, allocator: Allocator) ![]const c.VkQueueFamilyProperties {
+        var count: u32 = 0;
+        c.vkGetPhysicalDeviceQueueFamilyProperties(self.handle, &count, null);
+        const family_properties = try allocator.alloc(c.VkQueueFamilyProperties, count);
+        c.vkGetPhysicalDeviceQueueFamilyProperties(self.handle, &count, @ptrCast(family_properties));
+
+        return family_properties;
+    }
+
+    pub fn graphicsQueue(self: PhysicalDevice, allocator: Allocator) !u32 {
+        const queue_families = try self.queueFamilyProperties(allocator);
+        defer allocator.free(queue_families);
+        var graphics_queue: ?u32 = null;
+
+        for (queue_families, 0..) |family, index| {
+            if (graphics_queue) |_| {
+                break;
+            }
+
+            if ((family.queueFlags & c.VK_QUEUE_GRAPHICS_BIT) != 0x0) {
+                graphics_queue = @intCast(index);
+            }
+        }
+
+        return graphics_queue.?;
+    }
+
+    pub fn presentQueue(self: PhysicalDevice, surface: Surface, allocator: Allocator) !u32 {
+        const queue_families = try self.queueFamilyProperties(allocator);
+        defer allocator.free(queue_families);
+        var present_queue: ?u32 = null;
+
+        for (queue_families, 0..) |_, index| {
+            if (present_queue) |_| {
+                break;
+            }
+
+            var support: u32 = undefined;
+            try mapError(c.vkGetPhysicalDeviceSurfaceSupportKHR(self.handle, @intCast(index), surface.handle, &support));
+
+            if (support == c.VK_TRUE) {
+                present_queue = @intCast(index);
+            }
+        }
+
+        return present_queue.?;
+    }
+
+    pub fn create_device(self: *PhysicalDevice, surface: Surface, allocator: Allocator, comptime n: usize) !Device(n) {
+        const graphics_queue_index = try self.graphicsQueue(allocator);
+        const present_queue_index = try self.presentQueue(surface, allocator);
+
+        const priorities: f32 = 1.0;
+
+        const graphics_queue_info: c.VkDeviceQueueCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO,
+            .queueFamilyIndex = graphics_queue_index,
+            .queueCount = 1,
+            .pQueuePriorities = &priorities,
+        };
+
+        const present_queue_info: c.VkDeviceQueueCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO,
+            .queueFamilyIndex = present_queue_index,
+            .queueCount = 1,
+            .pQueuePriorities = &priorities,
+        };
+
+        const queues: []const c.VkDeviceQueueCreateInfo = &.{ graphics_queue_info, present_queue_info };
+
+        var device_features: c.VkPhysicalDeviceFeatures2 = .{
+            .sType = c.VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2,
+        };
+
+        c.vkGetPhysicalDeviceFeatures2(self.handle, &device_features);
+
+        const device_info: c.VkDeviceCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO,
+            .pNext = &device_features,
+            .queueCreateInfoCount = 1,
+            .pQueueCreateInfos = queues.ptr,
+            .enabledLayerCount = 0,
+            .enabledExtensionCount = @intCast(device_extensions.len),
+            .ppEnabledExtensionNames = device_extensions.ptr,
+        };
+
+        var device: c.VkDevice = undefined;
+        try mapError(c.vkCreateDevice(self.handle, &device_info, null, &device));
+
+        var graphics_queue: c.VkQueue = undefined;
+        var present_queue: c.VkQueue = undefined;
+
+        c.vkGetDeviceQueue(device, graphics_queue_index, 0, &graphics_queue);
+        c.vkGetDeviceQueue(device, present_queue_index, 0, &present_queue);
+
+        const command_pool_info: c.VkCommandPoolCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
+            .flags = c.VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
+            .queueFamilyIndex = graphics_queue_index,
+        };
+
+        var command_pool: c.VkCommandPool = undefined;
+        try mapError(c.vkCreateCommandPool(device, &command_pool_info, null, &command_pool));
+
+        const command_buffer_info: c.VkCommandBufferAllocateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
+            .commandPool = command_pool,
+            .level = c.VK_COMMAND_BUFFER_LEVEL_PRIMARY,
+            .commandBufferCount = n,
+        };
+
+        var command_buffers: [n]c.VkCommandBuffer = undefined;
+        try mapError(c.vkAllocateCommandBuffers(device, &command_buffer_info, command_buffers[0..n].ptr));
+
+        const semaphore_info: c.VkSemaphoreCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO,
+        };
+
+        const fence_info: c.VkFenceCreateInfo = .{
+            .sType = c.VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
+            .flags = c.VK_FENCE_CREATE_SIGNALED_BIT,
+        };
+
+        var image_available: [n]c.VkSemaphore = undefined;
+        var render_finished: [n]c.VkSemaphore = undefined;
+        var in_flight_fence: [n]c.VkFence = undefined;
+
+        inline for (0..n) |index| {
+            try mapError(c.vkCreateSemaphore(device, &semaphore_info, null, &image_available[index]));
+            try mapError(c.vkCreateSemaphore(device, &semaphore_info, null, &render_finished[index]));
+            try mapError(c.vkCreateFence(device, &fence_info, null, &in_flight_fence[index]));
+        }
+
+        var memory_properties: c.VkPhysicalDeviceMemoryProperties = undefined;
+        c.vkGetPhysicalDeviceMemoryProperties(self.handle, &memory_properties);
+
+        return Device(n){
+            .handle = device,
+            .graphics_queue = graphics_queue,
+            .present_queue = present_queue,
+            .command_pool = command_pool,
+            .command_buffers = command_buffers,
+            .image_available = image_available,
+            .render_finished = render_finished,
+            .in_flight_fence = in_flight_fence,
+            .graphics_family = graphics_queue_index,
+            .present_family = present_queue_index,
+            .memory_properties = memory_properties,
+        };
+    }
+};
diff --git a/src/render/window.zig b/src/render/window.zig
new file mode 100644
index 0000000..28c22d1
--- /dev/null
+++ b/src/render/window.zig
@@ -0,0 +1,64 @@
+const c = @import("../c.zig");
+const std = @import("std");
+
+pub const Error = error{
+    platform_unavailable,
+    platform_error,
+};
+
+pub fn getExtensions() [][*c]const u8 {
+    var extension_count: u32 = undefined;
+    const raw: [*c][*c]const u8 = c.glfwGetRequiredInstanceExtensions(&extension_count);
+    const extensions = raw[0..extension_count];
+
+    return extensions;
+}
+
+pub const Window = struct {
+    title: []const u8,
+    width: usize,
+    height: usize,
+    raw: *c.GLFWwindow,
+
+    pub fn create(width: usize, height: usize, title: []const u8) !Window {
+        if (c.glfwInit() != c.GLFW_TRUE) {
+            const status = c.glfwGetError(null);
+
+            return switch (status) {
+                c.GLFW_PLATFORM_UNAVAILABLE => Error.platform_unavailable,
+                c.GLFW_PLATFORM_ERROR => Error.platform_error,
+                else => unreachable,
+            };
+        }
+
+        c.glfwWindowHint(c.GLFW_RESIZABLE, c.GLFW_FALSE);
+        c.glfwWindowHint(c.GLFW_CLIENT_API, c.GLFW_NO_API);
+        const raw = c.glfwCreateWindow(@intCast(width), @intCast(height), title.ptr, null, null);
+        c.glfwShowWindow(raw);
+
+        return Window{
+            .title = title,
+            .width = width,
+            .height = height,
+            .raw = raw.?,
+        };
+    }
+
+    pub fn shouldClose(self: Window) bool {
+        return c.glfwWindowShouldClose(self.raw) == c.GLFW_TRUE;
+    }
+
+    pub fn size(self: Window) struct { usize, usize } {
+        var width: u32 = undefined;
+        var height: u32 = undefined;
+
+        c.glfwGetFramebufferSize(self.raw, @ptrCast(&width), @ptrCast(&height));
+
+        return .{ @intCast(width), @intCast(height) };
+    }
+
+    pub fn destroy(self: Window) void {
+        c.glfwDestroyWindow(self.raw);
+        c.glfwTerminate();
+    }
+};
diff --git a/src/vm/parse.zig b/src/vm/parse.zig
new file mode 100644
index 0000000..40eefc1
--- /dev/null
+++ b/src/vm/parse.zig
@@ -0,0 +1,288 @@
+const std = @import("std");
+const wasm = @import("wasm.zig");
+const Allocator = std.mem.Allocator;
+
+pub const Error = error{
+    malformed_wasm,
+    invalid_utf8,
+};
+
+pub const Module = struct {
+    types: []FunctionType,
+    imports: std.ArrayList(Import),
+    exports: std.StringHashMap(u32),
+    functions: []u32,
+    memory: Memory,
+    code: []FunctionBody,
+    funcs: std.ArrayList(Function),
+
+    pub fn deinit(self: *Module, allocator: Allocator) void {
+        for (self.types) |t| {
+            t.deinit(allocator);
+        }
+        allocator.free(self.types);
+
+        for (self.imports.items) |i| {
+            i.deinit(allocator);
+        }
+        self.imports.deinit();
+
+        var iter = self.exports.iterator();
+        while (iter.next()) |entry| {
+            allocator.free(entry.key_ptr.*);
+        }
+        self.exports.deinit();
+
+        allocator.free(self.functions);
+
+        for (self.code) |f| {
+            for (f.locals) |l| {
+                allocator.free(l.types);
+            }
+            allocator.free(f.code);
+        }
+        allocator.free(self.code);
+
+        self.funcs.deinit();
+    }
+};
+
+pub const FunctionScope = enum {
+    external,
+    internal,
+};
+
+pub const Function = union(FunctionScope) {
+    external: u8,
+    internal: u8,
+};
+
+// TODO: refactor locals
+pub const Local = struct {
+    types: []u8,
+};
+
+pub const FunctionBody = struct {
+    locals: []Local,
+    code: []u8,
+};
+
+pub const Memory = struct {
+    initial: u32,
+    max: u32,
+};
+
+pub const FunctionType = struct {
+    parameters: []u8,
+    results: []u8,
+
+    pub fn deinit(self: FunctionType, allocator: Allocator) void {
+        allocator.free(self.parameters);
+        allocator.free(self.results);
+    }
+};
+
+pub const Import = struct {
+    name: []u8,
+    module: []u8,
+    signature: u32,
+
+    pub fn deinit(self: Import, allocator: Allocator) void {
+        allocator.free(self.name);
+        allocator.free(self.module);
+    }
+};
+
+pub fn parseType(t: u8) wasm.Type {
+    return @enumFromInt(t);
+}
+
+pub fn parseName(allocator: Allocator, stream: anytype) ![]u8 {
+    const size = try std.leb.readULEB128(u32, stream);
+    const str = try allocator.alloc(u8, size);
+    if (try stream.read(str) != size) {
+        // TODO: better error
+        return Error.malformed_wasm;
+    }
+
+    if (!std.unicode.utf8ValidateSlice(str)) return Error.invalid_utf8;
+
+    return str;
+}
+
+// TODO: parse Global Section
+// TODO: Consider Arena allocator
+pub fn parseWasm(allocator: Allocator, stream: anytype) !Module {
+    var types: []FunctionType = undefined;
+    var imports = std.ArrayList(Import).init(allocator);
+    var exports = std.StringHashMap(u32).init(allocator);
+    var funcs = std.ArrayList(Function).init(allocator);
+    var functions: []u32 = undefined;
+    var memory: Memory = undefined;
+    var code: []FunctionBody = undefined;
+
+    // Parse magic
+    if (!(try stream.isBytes(&[_]u8{ 0x00, 0x61, 0x73, 0x6d }))) return Error.malformed_wasm;
+    // Parse version
+    if (!(try stream.isBytes(&[_]u8{ 0x01, 0x00, 0x00, 0x00 }))) return Error.malformed_wasm;
+
+    // NOTE: This ensures that (in this block) illegal behavior is safety-checked.
+    //     This slows down the code but since this function is only called at the start
+    //     I believe it is better to take the ``hit'' in performance (should only be @enumFromInt)
+    //     rather than  having undefined behavior when user provides an invalid wasm file.
+    @setRuntimeSafety(true);
+    loop: while (stream.readByte()) |byte| {
+        const section_size = try std.leb.readULEB128(u32, stream);
+        switch (@as(std.wasm.Section, @enumFromInt(byte))) {
+            std.wasm.Section.custom => {
+                // TODO: unimplemented
+                break :loop;
+            },
+            std.wasm.Section.type => {
+                const type_count = try std.leb.readULEB128(u32, stream);
+                types = try allocator.alloc(FunctionType, type_count);
+                for (types) |*t| {
+                    if (!(try stream.isBytes(&.{0x60}))) return Error.malformed_wasm;
+                    const params_count = try std.leb.readULEB128(u32, stream);
+                    t.parameters = try allocator.alloc(u8, params_count);
+                    if (try stream.read(t.parameters) != params_count) {
+                        // TODO: better errors
+                        return Error.malformed_wasm;
+                    }
+                    const results = try std.leb.readULEB128(u32, stream);
+                    t.results = try allocator.alloc(u8, results);
+                    if (try stream.read(t.results) != results) {
+                        // TODO: better errors
+                        return Error.malformed_wasm;
+                    }
+                }
+            },
+            std.wasm.Section.import => {
+                // Can there be more than one import section?
+                const import_count = try std.leb.readULEB128(u32, stream);
+                for (0..import_count) |i| {
+                    const mod = try parseName(allocator, stream);
+                    const nm = try parseName(allocator, stream);
+
+                    const b = try stream.readByte();
+                    switch (@as(std.wasm.ExternalKind, @enumFromInt(b))) {
+                        std.wasm.ExternalKind.function => try funcs.append(.{ .external = @intCast(i) }),
+                        // TODO: not implemented
+                        std.wasm.ExternalKind.table => {},
+                        std.wasm.ExternalKind.memory => {},
+                        std.wasm.ExternalKind.global => {},
+                    }
+                    const idx = try std.leb.readULEB128(u32, stream);
+                    try imports.append(.{
+                        .module = mod,
+                        .name = nm,
+                        .signature = idx,
+                    });
+                }
+            },
+            std.wasm.Section.function => {
+                const function_count = try std.leb.readULEB128(u32, stream);
+                functions = try allocator.alloc(u32, function_count);
+                for (functions) |*f| {
+                    f.* = try std.leb.readULEB128(u32, stream);
+                }
+            },
+            std.wasm.Section.table => {
+                // TODO: not implemented
+                try stream.skipBytes(section_size, .{});
+            },
+            std.wasm.Section.memory => {
+                const memory_count = try std.leb.readULEB128(u32, stream);
+                for (0..memory_count) |_| {
+                    const b = try stream.readByte();
+                    const n = try std.leb.readULEB128(u32, stream);
+                    var m: u32 = 0;
+                    switch (b) {
+                        0x00 => {},
+                        0x01 => m = try std.leb.readULEB128(u32, stream),
+                        else => return Error.malformed_wasm,
+                    }
+                    // TODO: support multiple memories
+                    memory = .{
+                        .initial = n,
+                        .max = m,
+                    };
+                }
+            },
+            std.wasm.Section.global => {
+                // TODO: unimplemented
+                try stream.skipBytes(section_size, .{});
+            },
+            // TODO: Can there be more than one export section? Otherwise we can optimize allocations
+            std.wasm.Section.@"export" => {
+                const export_count = try std.leb.readULEB128(u32, stream);
+                for (0..export_count) |_| {
+                    const nm = try parseName(allocator, stream);
+                    const b = try stream.readByte();
+                    const idx = try std.leb.readULEB128(u32, stream);
+                    switch (@as(std.wasm.ExternalKind, @enumFromInt(b))) {
+                        std.wasm.ExternalKind.function => try exports.put(nm, idx),
+                        // TODO: unimplemented,
+                        std.wasm.ExternalKind.table => allocator.free(nm),
+                        std.wasm.ExternalKind.memory => allocator.free(nm),
+                        std.wasm.ExternalKind.global => allocator.free(nm),
+                    }
+                }
+            },
+            std.wasm.Section.start => {
+                // TODO: unimplemented
+                try stream.skipBytes(section_size, .{});
+            },
+            std.wasm.Section.element => {
+                // TODO: unimplemented
+                try stream.skipBytes(section_size, .{});
+            },
+            std.wasm.Section.code => {
+                const code_count = try std.leb.readULEB128(u32, stream);
+                code = try allocator.alloc(FunctionBody, code_count);
+                for (0..code_count) |i| {
+                    const code_size = try std.leb.readULEB128(u32, stream);
+                    const local_count = try std.leb.readULEB128(u32, stream);
+                    const locals = try allocator.alloc(Local, local_count);
+                    for (locals) |*l| {
+                        const n = try std.leb.readULEB128(u32, stream);
+                        l.types = try allocator.alloc(u8, n);
+                        @memset(l.types, try stream.readByte());
+                    }
+                    code[i].locals = locals;
+
+                    // TODO: maybe is better to parse code into ast here and not do it every frame?
+                    // FIXME: This calculation is plain wrong. Resolving above TODO should help
+                    code[i].code = try allocator.alloc(u8, code_size - local_count - 1);
+                    // TODO: better error reporting
+                    if (try stream.read(code[i].code) != code_size - local_count - 1) return Error.malformed_wasm;
+
+                    const f = Function{ .internal = @intCast(i) };
+                    try funcs.append(f);
+                }
+            },
+            std.wasm.Section.data => {
+                // TODO: unimplemented
+                try stream.skipBytes(section_size, .{});
+            },
+            std.wasm.Section.data_count => {
+                // TODO: unimplemented
+                try stream.skipBytes(section_size, .{});
+            },
+            else => return Error.malformed_wasm,
+        }
+    } else |err| switch (err) {
+        error.EndOfStream => {},
+        else => return err,
+    }
+
+    return Module{
+        .types = types,
+        .imports = imports,
+        .functions = functions,
+        .memory = memory,
+        .exports = exports,
+        .code = code,
+        .funcs = funcs,
+    };
+}
diff --git a/src/vm/vm.zig b/src/vm/vm.zig
new file mode 100644
index 0000000..dfa457c
--- /dev/null
+++ b/src/vm/vm.zig
@@ -0,0 +1,559 @@
+const std = @import("std");
+const wasm = @import("wasm.zig");
+const Parser = @import("parse.zig");
+const Allocator = std.mem.Allocator;
+const AllocationError = error{OutOfMemory};
+
+pub fn leb128Decode(comptime T: type, bytes: []u8) struct { len: usize, val: T } {
+    switch (@typeInfo(T)) {
+        .int => {},
+        else => @compileError("LEB128 integer decoding only support integers, but got " ++ @typeName(T)),
+    }
+    if (@typeInfo(T).int.bits != 32 and @typeInfo(T).int.bits != 64) {
+        @compileError("LEB128 integer decoding only supports 32 or 64 bits integers but got " ++ std.fmt.comptimePrint("{d} bits", .{@typeInfo(T).int.bits}));
+    }
+
+    var result: T = 0;
+    // TODO: is the type of shift important. Reading Wikipedia (not very much tho) it seems like we can use u32 and call it a day...
+    var shift: if (@typeInfo(T).int.bits == 32) u5 else u6 = 0;
+    var byte: u8 = undefined;
+    var len: usize = 0;
+    for (bytes) |b| {
+        len += 1;
+        result |= @as(T, @intCast((b & 0x7f))) << shift;
+        if ((b & (0x1 << 7)) == 0) {
+            byte = b;
+            break;
+        }
+        shift += 7;
+    }
+    if (@typeInfo(T).int.signedness == .signed) {
+        const size = @sizeOf(T) * 8;
+        if (shift < size and (byte & 0x40) != 0) {
+            result |= (~@as(T, 0) << shift);
+        }
+    }
+
+    return .{ .len = len, .val = result };
+}
+
+pub fn decodeLittleEndian(comptime T: type, bytes: []u8) T {
+    if (T != i32 and T != i64) {
+        return @as(T, 0);
+    } else {
+        var value = @as(T, 0);
+        for (0..@sizeOf(T)) |b| {
+            value |= ((bytes[b]) << @intCast((@sizeOf(T) - b - 1)));
+        }
+        return value;
+    }
+}
+
+pub fn encodeLittleEndian(comptime T: type, bytes: *[]u8, value: T) void {
+    for (0..@sizeOf(T)) |b| {
+        bytes.*[b] = @intCast(((value >> @intCast((@sizeOf(T) - b - 1))) & 0xff));
+    }
+}
+
+pub const CallFrame = struct {
+    program_counter: usize,
+    code: []u8,
+    locals: []Value,
+};
+
+const ValueType = enum {
+    i32,
+    i64,
+};
+
+pub const Value = union(ValueType) {
+    i32: i32,
+    i64: i64,
+};
+
+pub const Runtime = struct {
+    module: Parser.Module,
+    stack: std.ArrayList(Value),
+    call_stack: std.ArrayList(CallFrame),
+    memory: []u8,
+    global_runtime: *wasm.GlobalRuntime,
+    labels: std.ArrayList(usize),
+
+    pub fn init(allocator: Allocator, module: Parser.Module, global_runtime: *wasm.GlobalRuntime) !Runtime {
+        const memory = try allocator.alloc(u8, module.memory.max);
+        return Runtime{
+            .module = module,
+            .stack = try std.ArrayList(Value).initCapacity(allocator, 10),
+            .call_stack = try std.ArrayList(CallFrame).initCapacity(allocator, 5),
+            .labels = try std.ArrayList(usize).initCapacity(allocator, 2),
+            .memory = memory,
+            .global_runtime = global_runtime,
+        };
+    }
+
+    pub fn deinit(self: *Runtime, allocator: Allocator) void {
+        self.module.deinit(allocator);
+        self.stack.deinit();
+        self.labels.deinit();
+        self.call_stack.deinit();
+        allocator.free(self.memory);
+    }
+
+    pub fn executeFrame(self: *Runtime, allocator: Allocator, frame: *CallFrame) !void {
+        var for_loop = false;
+        loop: while (true) {
+            const byte: u8 = frame.code[frame.program_counter];
+            frame.program_counter += 1;
+            std.debug.print("b: {x}\n", .{byte});
+            switch (byte) {
+                0x03 => {
+                    try self.labels.append(frame.program_counter);
+                    frame.program_counter += 1;
+                    //const a = frame.code[frame.program_counter];
+                    for_loop = true;
+                },
+                0x0d => {
+                    const label = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += label.len;
+                    var address = @as(usize, 0);
+                    for (0..(label.val)) |_| {
+                        address = self.labels.pop().?;
+                    }
+
+                    if (self.stack.pop().?.i32 != 0) {
+                        frame.program_counter = address;
+                    }
+                },
+
+                0x20 => {
+                    const integer = leb128Decode(u32, frame.code[frame.program_counter..]);
+
+                    frame.program_counter += integer.len;
+                    try self.stack.append(frame.locals[integer.val]);
+                },
+                0x21 => {
+                    const integer = leb128Decode(u32, frame.code[frame.program_counter..]);
+
+                    frame.program_counter += integer.len;
+                    frame.locals[integer.val] = self.stack.pop().?;
+                },
+                0x22 => {
+                    const integer = leb128Decode(u32, frame.code[frame.program_counter..]);
+
+                    frame.program_counter += integer.len;
+                    frame.locals[integer.val] = self.stack.pop().?;
+                    try self.stack.append(Value{ .i32 = @intCast(integer.val) });
+                },
+                0x28 => {
+                    const address = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += address.len;
+                    const offset = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += offset.len;
+                    const start = (address.val + offset.val);
+                    const end = start + @sizeOf(u32);
+                    try self.stack.append(Value{ .i32 = decodeLittleEndian(i32, self.memory[start..end]) });
+                },
+                0x29 => {
+                    const address = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += address.len;
+                    const offset = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += offset.len;
+                    const start = (address.val + offset.val);
+                    const end = start + @sizeOf(u64);
+                    try self.stack.append(Value{ .i64 = decodeLittleEndian(i64, self.memory[start..end]) });
+                },
+                0x36 => {
+                    const address = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += address.len;
+                    const offset = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += offset.len;
+                    const start = (address.val + offset.val);
+                    const end = start + @sizeOf(u32);
+                    try self.stack.append(Value{ .i32 = decodeLittleEndian(i32, self.memory[start..end]) });
+                },
+                0x37 => {
+                    const address = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += address.len;
+                    const offset = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += offset.len;
+                    const start = (address.val + offset.val);
+                    const end = start + @sizeOf(u32);
+                    encodeLittleEndian(i32, @constCast(&self.memory[start..end]), self.stack.pop().?.i32);
+                },
+                0x38 => {
+                    const address = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += address.len;
+                    const offset = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += offset.len;
+                    const start = (address.val + offset.val);
+                    const end = start + @sizeOf(u64);
+                    encodeLittleEndian(i64, @constCast(&self.memory[start..end]), self.stack.pop().?.i64);
+                },
+                0x41 => {
+                    const integer = leb128Decode(i32, frame.code[frame.program_counter..]);
+
+                    frame.program_counter += integer.len;
+                    try self.stack.append(Value{ .i32 = integer.val });
+                },
+                0x42 => {
+                    const integer = leb128Decode(i64, frame.code[frame.program_counter..]);
+
+                    frame.program_counter += integer.len;
+                    try self.stack.append(Value{ .i64 = integer.val });
+                },
+                0x45 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 == 0))) });
+                },
+                0x46 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 == self.stack.pop().?.i32))) });
+                },
+                0x47 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 != self.stack.pop().?.i32))) });
+                },
+                0x48 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 < self.stack.pop().?.i32))) });
+                },
+                0x49 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(@as(u32, @bitCast(self.stack.pop().?.i32)) < @as(u32, @bitCast(self.stack.pop().?.i32))))) });
+                },
+                0x4a => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 > self.stack.pop().?.i32))) });
+                },
+                0x4b => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(@as(u32, @bitCast(self.stack.pop().?.i32)) > @as(u32, @bitCast(self.stack.pop().?.i32))))) });
+                },
+                0x4c => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 <= self.stack.pop().?.i32))) });
+                },
+                0x4d => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(@as(u32, @bitCast(self.stack.pop().?.i32)) <= @as(u32, @bitCast(self.stack.pop().?.i32))))) });
+                },
+                0x4e => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i32 >= self.stack.pop().?.i32))) });
+                },
+                0x4f => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(@as(u32, @bitCast(self.stack.pop().?.i32)) >= @as(u32, @bitCast(self.stack.pop().?.i32))))) });
+                },
+
+                0x50 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 == 0))) });
+                },
+                0x51 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 == self.stack.pop().?.i64))) });
+                },
+                0x52 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 != self.stack.pop().?.i64))) });
+                },
+                0x53 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 < self.stack.pop().?.i64))) });
+                },
+                0x54 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(@as(u64, @bitCast(self.stack.pop().?.i64)) < @as(u64, @bitCast(self.stack.pop().?.i64))))) });
+                },
+                0x55 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 > self.stack.pop().?.i64))) });
+                },
+                0x56 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(@as(u64, @bitCast(self.stack.pop().?.i64)) > @as(u64, @bitCast(self.stack.pop().?.i64))))) });
+                },
+                0x57 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 <= self.stack.pop().?.i64))) });
+                },
+                0x58 => {
+                    try self.stack.append(Value{ .i64 = @intCast(@as(u1, @bitCast(@as(u64, @bitCast(self.stack.pop().?.i64)) <= @as(u64, @bitCast(self.stack.pop().?.i64))))) });
+                },
+                0x59 => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(self.stack.pop().?.i64 >= self.stack.pop().?.i64))) });
+                },
+                0x5a => {
+                    try self.stack.append(Value{ .i32 = @intCast(@as(u1, @bitCast(@as(u64, @bitCast(self.stack.pop().?.i64)) >= @as(u64, @bitCast(self.stack.pop().?.i64))))) });
+                },
+
+                0x67 => {
+                    var i = @as(i32, 0);
+                    const number = self.stack.pop().?.i32;
+                    for (0..@sizeOf(i32)) |b| {
+                        if (number & (@as(i32, 0x1) << @intCast((@sizeOf(i32) - b - 1))) == 1) {
+                            break;
+                        }
+                        i += 1;
+                    }
+                    try self.stack.append(Value{ .i32 = i });
+                },
+                0x68 => {
+                    var i = @as(i32, 0);
+                    const number = self.stack.pop().?.i32;
+                    for (0..@sizeOf(i32)) |b| {
+                        if (number & (@as(i32, 0x1) << @intCast(b)) == 1) {
+                            break;
+                        }
+                        i += 1;
+                    }
+                    try self.stack.append(Value{ .i32 = i });
+                },
+                0x69 => {
+                    var i = @as(i32, 0);
+                    const number = self.stack.pop().?.i32;
+                    for (0..@sizeOf(i32)) |b| {
+                        if (number & (@as(i32, 0x1) << @intCast(b)) == 1) {
+                            i += 1;
+                        }
+                    }
+                    try self.stack.append(Value{ .i32 = i });
+                },
+                0x6a => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 + b.i32 });
+                },
+                0x6b => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 - b.i32 });
+                },
+                0x6c => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 * b.i32 });
+                },
+                0x6d => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = @divTrunc(a.i32, b.i32) });
+                },
+                0x6e => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = @as(i32, @bitCast(@as(u32, @bitCast(a.i32)) / @as(u32, @bitCast(b.i32)))) });
+                },
+                0x6f => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = @rem(a.i32, b.i32) });
+                },
+                0x70 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = @as(i32, @bitCast(@as(u32, @bitCast(a.i32)) % @as(u32, @bitCast(b.i32)))) });
+                },
+                0x71 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 & b.i32 });
+                },
+                0x72 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 | b.i32 });
+                },
+                0x73 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 ^ b.i32 });
+                },
+                0x74 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 << @intCast(b.i32) });
+                },
+                0x75 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = a.i32 >> @intCast(b.i32) });
+                },
+                0x76 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = @as(i32, @bitCast(@as(u32, @bitCast(a.i32)) >> @intCast(@as(u32, @bitCast(b.i32))))) });
+                },
+                0x77 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = (a.i32 << @intCast(@as(u32, @bitCast(b.i32)))) | (a.i32 >> @intCast((@sizeOf(u32) * 8 - b.i32))) });
+                },
+                0x78 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i32 = (a.i32 >> @intCast(@as(u32, @bitCast(b.i32)))) | (a.i32 << @intCast((@sizeOf(u32) * 8 - b.i32))) });
+                },
+
+                0x79 => {
+                    var i = @as(i64, 0);
+                    const number = self.stack.pop().?.i64;
+                    for (0..@sizeOf(i64)) |b| {
+                        if (number & (@as(i64, 0x1) << @intCast((@sizeOf(i64) - b - 1))) == 1) {
+                            break;
+                        }
+                        i += 1;
+                    }
+                    try self.stack.append(Value{ .i64 = i });
+                },
+                0x7a => {
+                    var i = @as(i64, 0);
+                    const number = self.stack.pop().?.i64;
+                    for (0..@sizeOf(i64)) |b| {
+                        if (number & (@as(i64, 0x1) << @intCast(b)) == 1) {
+                            break;
+                        }
+                        i += 1;
+                    }
+                    try self.stack.append(Value{ .i64 = i });
+                },
+                0x7b => {
+                    var i = @as(i64, 0);
+                    const number = self.stack.pop().?.i64;
+                    for (0..@sizeOf(i64)) |b| {
+                        if (number & (@as(i64, 0x1) << @intCast(b)) == 1) {
+                            i += 1;
+                        }
+                    }
+                    try self.stack.append(Value{ .i64 = i });
+                },
+                0x7c => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 + b.i64 });
+                },
+                0x7d => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 - b.i64 });
+                },
+                0x7e => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 * b.i64 });
+                },
+                0x7f => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = @divTrunc(a.i64, b.i64) });
+                },
+                0x80 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = @as(i64, @bitCast(@as(u64, @bitCast(a.i64)) / @as(u64, @bitCast(b.i64)))) });
+                },
+                0x81 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = @rem(a.i64, b.i64) });
+                },
+                0x82 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = @as(i64, @bitCast(@as(u64, @bitCast(a.i64)) % @as(u64, @bitCast(b.i64)))) });
+                },
+                0x83 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 & b.i64 });
+                },
+                0x84 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 | b.i64 });
+                },
+                0x85 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 ^ b.i64 });
+                },
+                0x86 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 << @intCast(b.i64) });
+                },
+                0x87 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = a.i64 >> @intCast(b.i64) });
+                },
+                0x88 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = @as(i64, @bitCast(@as(u64, @bitCast(a.i64)) >> @intCast(@as(u64, @bitCast(b.i64))))) });
+                },
+                0x89 => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = (a.i64 << @intCast(@as(u64, @bitCast(b.i64)))) | (a.i64 >> @intCast((@sizeOf(u64) * 8 - b.i64))) });
+                },
+                0x8a => {
+                    const a = self.stack.pop().?;
+                    const b = self.stack.pop().?;
+                    try self.stack.append(.{ .i64 = (a.i64 >> @intCast(@as(u64, @bitCast(b.i64)))) | (a.i64 << @intCast((@sizeOf(u64) * 8 - b.i64))) });
+                },
+
+                0x10 => {
+                    const integer = leb128Decode(u32, frame.code[frame.program_counter..]);
+                    frame.program_counter += integer.len;
+
+                    self.call(allocator, integer.val, &[_]usize{}) catch {};
+                },
+                0xb => {
+                    if (for_loop) {
+                        frame.program_counter += 1;
+                    } else {
+                        break :loop;
+                    }
+                },
+                else => std.debug.print("instruction {} not implemented\n", .{byte}),
+            }
+        }
+    }
+
+    pub fn callExternal(self: *Runtime, allocator: Allocator, name: []const u8, parameters: []usize) !void {
+        if (self.module.exports.get(name)) |function| {
+            try self.call(allocator, function, parameters);
+        }
+    }
+
+    pub fn call(self: *Runtime, allocator: Allocator, function: usize, parameters: []usize) AllocationError!void {
+        const f = self.module.funcs.items[function];
+        switch (f) {
+            .internal => {
+                const function_type = self.module.types[self.module.functions[f.internal]];
+                var frame = CallFrame{
+                    .code = self.module.code[f.internal].code,
+                    .program_counter = 0x0,
+                    .locals = try allocator.alloc(Value, self.module.code[f.internal].locals.len + function_type.parameters.len),
+                };
+
+                for (parameters, 0..) |p, i| {
+                    switch (Parser.parseType(function_type.parameters[i])) {
+                        .i32 => {
+                            frame.locals[i] = .{ .i32 = @intCast(p) };
+                        },
+                        .i64 => {
+                            frame.locals[i] = .{ .i64 = @intCast(p) };
+                        },
+                        else => unreachable,
+                    }
+                }
+
+                for (self.module.code[f.internal].locals, function_type.parameters.len..) |local, i| {
+                    switch (Parser.parseType(local.types[0])) {
+                        .i32 => {
+                            frame.locals[i] = .{ .i32 = 0 };
+                        },
+                        .i64 => {
+                            frame.locals[i] = .{ .i64 = 0 };
+                        },
+                        else => unreachable,
+                    }
+                }
+
+                try self.executeFrame(allocator, &frame);
+
+                allocator.free(frame.locals);
+            },
+            .external => {
+                const name = self.module.imports.items[f.external].name;
+                if (self.global_runtime.functions.get(name)) |external| {
+                    external(&self.stack);
+                }
+            },
+        }
+    }
+};
diff --git a/src/vm/wasm.zig b/src/vm/wasm.zig
new file mode 100644
index 0000000..579309b
--- /dev/null
+++ b/src/vm/wasm.zig
@@ -0,0 +1,35 @@
+const vm = @import("vm.zig");
+const std = @import("std");
+const Allocator = std.mem.Allocator;
+
+pub const Type = enum(u8) {
+    i32 = 0x7f,
+    i64 = 0x7e,
+    f32 = 0x7d,
+    f64 = 0x7c,
+    v128 = 0x7b,
+};
+
+pub const GlobalRuntime = struct {
+    functions: std.StringHashMap(*const fn (stack: *std.ArrayList(vm.Value)) void),
+
+    pub fn init(allocator: Allocator) GlobalRuntime {
+        return GlobalRuntime{
+            .functions = std.StringHashMap(*const fn (stack: *std.ArrayList(vm.Value)) void).init(allocator),
+        };
+    }
+
+    pub fn deinit(self: *GlobalRuntime) void {
+        self.functions.deinit();
+    }
+
+    pub fn addFunction(self: *GlobalRuntime, name: []const u8, function: *const fn (stack: *std.ArrayList(vm.Value)) void) !void {
+        try self.functions.put(name, function);
+    }
+};
+
+pub fn debug(stack: *std.ArrayList(vm.Value)) void {
+    const a = stack.pop().?;
+
+    std.debug.print("{}\n", .{a.i32});
+}