add new win32 platform implementation (#1224)

* Buffer resources from swap chain were not being released.
* sysgpu Texture.getWidth() should return width not height.
* d3d12: Fixing issues with d3d12 on sysgpu.
* Initial win32 platform

build.zig

@@ -276,6 +276,7 @@ pub const Platform = enum {
     x11,
     wayland,
     web,
+    win32,
 
     pub fn fromTarget(target: std.Target) Platform {
         if (target.cpu.arch == .wasm32) return .web;

src/Core.zig

@@ -16,6 +16,7 @@ const Platform = switch (build_options.core_platform) {
     .x11 => @import("core/X11.zig"),
     .wayland => @import("core/Wayland.zig"),
     .web => @panic("TODO: revive wasm backend"),
+    .win32 => @import("core/win32.zig"),
 };
 
 // TODO(important): mach.core has a lot of standard Zig APIs, and some global variables, which are
@@ -242,8 +243,10 @@ pub inline fn deinit(entities: *mach.Entities.Mod, core: *Mod) !void {
     if (builtin.os.tag == .linux and
         state.linux_gamemode != null and
         state.linux_gamemode.?)
+    {
         deinitLinuxGamemode();
-
+    }
+    
     state.platform.deinit();
     state.swap_chain.release();
     state.queue.release();

src/core/win32.zig

@@ -0,0 +1,618 @@
+const std = @import("std");
+const w = @import("win32/win32.zig");
+const mach = @import("../main.zig");
+const Core = @import("../Core.zig");
+const InputState = @import("InputState.zig");
+const Frequency = @import("Frequency.zig");
+const unicode = @import("unicode.zig");
+
+const gpu = mach.gpu;
+const InitOptions = Core.InitOptions;
+const Event = Core.Event;
+const KeyEvent = Core.KeyEvent;
+const MouseButtonEvent = Core.MouseButtonEvent;
+const MouseButton = Core.MouseButton;
+const Size = Core.Size;
+const DisplayMode = Core.DisplayMode;
+const CursorShape = Core.CursorShape;
+const VSyncMode = Core.VSyncMode;
+const CursorMode = Core.CursorMode;
+const Position = Core.Position;
+const Key = Core.Key;
+const KeyMods = Core.KeyMods;
+const Joystick = Core.Joystick;
+
+const log = std.log.scoped(.mach);
+
+const EventQueue = std.fifo.LinearFifo(Event, .Dynamic);
+const Win32 = @This();
+
+/////////////////////////
+// Module state
+////////////////////////
+allocator: std.mem.Allocator,
+core: *Core,
+
+// Core platform interface
+title: [:0]u8,
+display_mode: DisplayMode,
+vsync_mode: VSyncMode,
+cursor_mode: CursorMode,
+cursor_shape: CursorShape,
+border: bool,
+headless: bool,
+refresh_rate: u32,
+size: Size,
+surface_descriptor: gpu.Surface.Descriptor,
+surface_descriptor_from_hwnd: gpu.Surface.DescriptorFromWindowsHWND,
+
+// Internals
+window: w.HWND,
+surrogate: u16 = 0,
+dinput: *w.IDirectInput8W,
+
+events: EventQueue,
+oom: std.Thread.ResetEvent = .{},
+
+////////////////////////////
+/// Internals
+////////////////////////////
+
+const RequestAdapterResponse = struct {
+    status: gpu.RequestAdapterStatus,
+    adapter: ?*gpu.Adapter,
+    message: ?[*:0]const u8,
+};
+
+inline fn requestAdapterCallback(
+    context: *RequestAdapterResponse,
+    status: gpu.RequestAdapterStatus,
+    adapter: ?*gpu.Adapter,
+    message: ?[*:0]const u8,
+) void {
+    context.* = RequestAdapterResponse{
+        .status = status,
+        .adapter = adapter,
+        .message = message,
+    };
+}
+
+inline fn printUnhandledErrorCallback(_: void, ty: gpu.ErrorType, message: [*:0]const u8) void {
+    switch (ty) {
+        .validation => std.log.err("gpu: validation error: {s}\n", .{message}),
+        .out_of_memory => std.log.err("gpu: out of memory: {s}\n", .{message}),
+        .unknown => std.log.err("gpu: unknown error: {s}\n", .{message}),
+        else => unreachable,
+    }
+    std.process.exit(1);
+}
+
+////////////////////////////
+/// Platform interface 
+////////////////////////////
+pub fn init(
+    self: *Win32,
+    options: InitOptions,
+) !void {
+    self.allocator = options.allocator;
+    self.core = @fieldParentPtr("platform", self);
+    self.events = EventQueue.init(self.allocator);
+    self.size = options.size;
+
+    const hInstance = w.GetModuleHandleW(null);
+    const class_name = w.L("mach");
+    const class = std.mem.zeroInit(w.WNDCLASSW, .{
+        .style = w.CS_OWNDC,
+        .lpfnWndProc = wndProc,
+        .hInstance = hInstance,
+        .hIcon = w.LoadIconW(null, @as([*:0]align(1) const u16, @ptrFromInt(@as(u32, w.IDI_APPLICATION)))),
+        .hCursor = w.LoadCursorW(null, @as([*:0]align(1) const u16, @ptrFromInt(@as(u32, w.IDC_ARROW)))),
+        .lpszClassName = class_name,
+    });
+    if (w.RegisterClassW(&class) == 0) return error.Unexpected;
+
+    const title = try std.unicode.utf8ToUtf16LeAllocZ(self.allocator, options.title);
+    defer self.allocator.free(title);
+    // TODO set title , copy to self.title
+
+    const window = w.CreateWindowExW(
+        .{.APPWINDOW = 1, .WINDOWEDGE = 1, .CLIENTEDGE = 1},
+        class_name,
+        title,
+        if (options.border) w.WS_OVERLAPPEDWINDOW else w.WS_POPUPWINDOW,
+        w.CW_USEDEFAULT,
+        w.CW_USEDEFAULT,
+        @bitCast(self.size.width),
+        @bitCast(self.size.height),
+        null,
+        null,
+        hInstance,
+        null,
+    ) orelse return error.Unexpected;
+    self.window = window;
+
+    var dinput: ?*w.IDirectInput8W = undefined;
+    //@ptrCast(*anyopaque
+
+    const ptr: ?*?*anyopaque = @ptrCast(&dinput);
+
+    //    if (w.DirectInput8Create(instance, w.DIRECTINPUT_VERSION, &w.IID_IDirectInput8W, &dinput, null) != w.DI_OK) return error.Unexpected;
+    if (w.DirectInput8Create(hInstance, w.DIRECTINPUT_VERSION, w.IID_IDirectInput8W, ptr, null) != w.DI_OK) {
+        return error.Unexpected;
+    }
+    self.dinput = dinput.?;
+
+    self.surface_descriptor_from_hwnd = .{
+                .hinstance = std.os.windows.kernel32.GetModuleHandleW(null).?,
+                .hwnd = window,
+            };
+
+    self.surface_descriptor = .{
+        .next_in_chain = .{
+            .from_windows_hwnd = &self.surface_descriptor_from_hwnd,
+        }
+    };
+
+    _ = w.SetWindowLongPtrW(window, w.GWLP_USERDATA, @bitCast(@intFromPtr(self)));
+    if (!options.headless) {
+        setDisplayMode(self, options.display_mode);
+    }
+}
+
+pub fn deinit(self: *Win32) void {
+    self.events.deinit();
+    _ = self.dinput.IUnknown_Release();
+}
+
+pub fn update(self: *Win32) !void {
+    _ = self;
+    var msg: w.MSG = undefined;
+    while (w.PeekMessageW(&msg, null, 0, 0, w.PM_REMOVE) != 0) {
+        _ = w.TranslateMessage(&msg);
+        _ = w.DispatchMessageW(&msg);
+    }   
+}
+
+pub const EventIterator = struct {
+    queue: *EventQueue,
+
+    pub fn next(self: *EventIterator) ?Event {
+        return self.queue.readItem();
+    }
+};
+
+pub fn pollEvents(self: *Win32) EventIterator {
+    return .{ .queue = &self.events };
+}
+
+pub fn setTitle(self: *Win32, title: [:0]const u8) void {
+    const wtitle = std.unicode.utf8ToUtf16LeAllocZ(self.allocator, title) catch {
+        self.oom.set();
+        return;
+    };
+    // TODO update self.title
+    defer self.allocator.free(wtitle);
+    _ = w.SetWindowTextW(self.window, wtitle);
+}
+
+pub fn setDisplayMode(self: *Win32, mode: DisplayMode) void {
+    // TODO update self.displayMode
+    switch (mode) {
+        .windowed => _ = w.ShowWindow(self.window, w.SW_RESTORE),
+        //        .maximized => _ = w.ShowWindow(self.window, w.SW_MAXIMIZE),
+        .fullscreen => {},
+        .borderless => {},
+    }
+}
+
+pub fn setBorder(self: *Win32, value: bool) void {
+    _ = w.SetWindowLongW(self.window, w.GWL_STYLE, if (value) w.WS_OVERLAPPEDWINDOW else w.WS_POPUPWINDOW);
+    self.border = value;
+}
+
+pub fn setHeadless(self: *Win32, value: bool) void {
+    _ = w.ShowWindow(self.window, if (value) w.SW_HIDE else w.SW_SHOW);
+    self.headless = value;
+}
+
+pub fn setVSync(self: *Win32, mode: VSyncMode) void {
+    self.vsync_mode = mode;
+}
+
+pub fn setSize(self: *Win32, value: Size) void {
+    _ = w.SetWindowPos(self.window, 0, 0, 0, value.width, value.height, w.SWP_NOMOVE | w.SWP_NOZORDER | w.SWP_NOACTIVATE);
+    self.size = value;
+}
+// pub fn size(self: *Core) core.Size {
+//     var rect: w.RECT = undefined;
+//     _ = w.GetClientRect(self.window, &rect);
+
+//     const width:u32 = @intCast(rect.right - rect.left);
+//     const height:u32 = @intCast(rect.bottom - rect.top);
+
+//     return .{ .width = width, .height = height };
+// }
+
+// pub fn setSizeLimit(self: *Core, value: core.SizeLimit) void {
+//     self.mutex.lock();
+//     defer self.mutex.unlock();
+//     self.limits = value;
+//     // trigger WM_GETMINMAXINFO
+//     _ = w.SetWindowPos(self.window, 0, 0, 0, 0, 0, w.SWP_NOMOVE | w.SWP_NOSIZE | w.SWP_NOZORDER | w.SWP_NOACTIVATE);
+// }
+
+// pub fn sizeLimit(self: *Core) core.SizeLimit {
+//     self.mutex.lock();
+//     defer self.mutex.unlock();
+//     return self.limits;
+// }
+
+pub fn setCursorMode(self: *Win32, mode: CursorMode) void {
+    switch (mode) {
+        .normal => while (w.ShowCursor(true) < 0) {},
+        .hidden => while (w.ShowCursor(false) >= 0) {},
+        .disabled => {},
+    }
+    self.cursor_mode = mode;
+}
+
+pub fn setCursorShape(self: *Win32, shape: CursorShape) void {
+    const name = switch (shape) {
+        .arrow => w.IDC_ARROW,
+        .ibeam => w.IDC_IBEAM,
+        .crosshair => w.IDC_CROSS,
+        .pointing_hand => w.IDC_HAND,
+        .resize_ew => w.IDC_SIZEWE,
+        .resize_ns => w.IDC_SIZENS,
+        .resize_nwse => w.IDC_SIZENWSE,
+        .resize_nesw => w.IDC_SIZENESW,
+        .resize_all => w.IDC_SIZEALL,
+        .not_allowed => w.IDC_NO,
+    };
+    _ = w.SetCursor(w.LoadCursorW(null, @ptrFromInt(name)));
+    self.cursor_shape = shape;
+}
+
+pub fn keyPressed(self: *Win32, key: Key) bool {
+    _ = self;
+    _ = key;
+    // self.input_mutex.lockShared();
+    // defer self.input_mutex.unlockShared();
+    // return self.input_state.isKeyPressed(key);
+}
+
+pub fn keyReleased(self: *Win32, key: Key) bool {
+    _ = self;
+    _ = key;
+    // self.input_mutex.lockShared();
+    // defer self.input_mutex.unlockShared();
+    // return self.input_state.isKeyReleased(key);
+}
+
+pub fn mousePressed(self: *Win32, button: MouseButton) bool {
+    _ = self;
+    _ = button;
+    // self.input_mutex.lockShared();
+    // defer self.input_mutex.unlockShared();
+    // return self.input_state.isMouseButtonPressed(button);
+}
+
+pub fn mouseReleased(self: *Win32, button: MouseButton) bool {
+    _ = self;
+    _ = button;
+    // self.input_mutex.lockShared();
+    // defer self.input_mutex.unlockShared();
+    // return self.input_state.isMouseButtonReleased(button);
+}
+
+pub fn mousePosition(self: *Win32) Position {
+    _ = self;
+    // self.input_mutex.lockShared();
+    // defer self.input_mutex.unlockShared();
+    // return self.input_state.mouse_position;
+}
+
+pub fn outOfMemory(self: *Win32) bool {
+    if (self.oom.isSet()) {
+        self.oom.reset();
+        return true;
+    }
+    return false;
+}
+
+fn pushEvent(self: *Win32, event: Event) void {
+    //self.event_mutex.lock();
+    //defer self.event_mutex.unlock();
+    self.events.writeItem(event) catch self.oom.set();
+}
+
+fn wndProc(wnd: w.HWND, msg: u32, wParam: w.WPARAM, lParam: w.LPARAM) callconv(w.WINAPI) w.LRESULT {
+    const self = blk: {
+        const userdata: usize = @bitCast(w.GetWindowLongPtrW(wnd, w.GWLP_USERDATA));
+        const ptr: ?*Win32 = @ptrFromInt(userdata);
+        break :blk ptr orelse return w.DefWindowProcW(wnd, msg, wParam, lParam);
+    };
+
+    switch (msg) {
+        w.WM_CLOSE => {
+            self.pushEvent(.close);
+            return 0;
+        },
+        w.WM_GETMINMAXINFO => {
+            //self.mutex.lock();
+            //defer self.mutex.unlock();
+
+            // TODO: SizeLimit is no longer in mach core or has changed
+            // const info: *w.MINMAXINFO = blk: {
+            //     const int: usize = @bitCast(lParam);
+            //     break :blk @ptrFromInt(int);
+            // };
+            // if (self.limits.min.width) |width| info.ptMinTrackSize.x = @bitCast(width);
+            // if (self.limits.min.height) |height| info.ptMinTrackSize.y = @bitCast(height);
+            // if (self.limits.max.width) |width| info.ptMaxTrackSize.x = @bitCast(width);
+            // if (self.limits.max.height) |height| info.ptMaxTrackSize.y = @bitCast(height);
+            return 0;
+        },
+        w.WM_KEYDOWN, w.WM_KEYUP => {
+            const vkey: w.VIRTUAL_KEY = @enumFromInt(wParam);
+            if (vkey == w.VK_PROCESSKEY) return 0;
+
+            if (msg == w.WM_SYSKEYDOWN and vkey == w.VK_F4) {
+                self.pushEvent(.close);
+                return 0;
+            }
+
+            const flags = lParam >> 16;
+            const scancode: u9 = @intCast(flags & 0x1FF);
+
+            if (scancode == 0x1D) {
+                // right alt sends left control first
+                var next: w.MSG = undefined;
+                const time = w.GetMessageTime();
+                if (w.PeekMessageW(&next, self.window, 0, 0, w.PM_NOREMOVE) != 0 and
+                    next.time == time and
+                    (next.message == msg or (msg == w.WM_SYSKEYDOWN and next.message == w.WM_KEYUP)) and
+                    ((next.lParam >> 16) & 0x1FF) == 0x138)
+                {
+                    return 0;
+                }
+            }
+
+            const key = keyFromScancode(scancode);
+            if (msg == w.WM_KEYDOWN) {
+                if (flags & w.KF_REPEAT == 0) {
+                    self.pushEvent(.{ .key_press = .{ .key = key, .mods = undefined } });
+                } else {
+                    self.pushEvent(.{ .key_repeat = .{ .key = key, .mods = undefined } });
+                }
+            } else {
+                self.pushEvent(.{ .key_release = .{ .key = key, .mods = undefined } });
+            }
+
+            return 0;
+        },
+        w.WM_CHAR => {
+            const char: u16 = @truncate(wParam);
+            var chars: []const u16 = undefined;
+            if (self.surrogate != 0) {
+                chars = &.{ self.surrogate, char };
+                self.surrogate = 0;
+            } else if (std.unicode.utf16IsHighSurrogate(char)) {
+                self.surrogate = char;
+                return 0;
+            } else {
+                chars = &.{char};
+            }
+            var iter = std.unicode.Utf16LeIterator.init(chars);
+            if (iter.nextCodepoint()) |codepoint| {
+                self.pushEvent(.{ .char_input = .{ .codepoint = codepoint.? } });
+            } else |err| {
+                err catch {};
+            }
+            return 0;
+        },
+        w.WM_LBUTTONDOWN,
+        w.WM_LBUTTONUP,
+        w.WM_RBUTTONDOWN,
+        w.WM_RBUTTONUP,
+        w.WM_MBUTTONDOWN,
+        w.WM_MBUTTONUP,
+        w.WM_XBUTTONDOWN,
+        w.WM_XBUTTONUP,
+        => {
+            const x:f64 = @floatFromInt(@as(i16, @truncate(lParam & 0xFFFF))); 
+            const y:f64 = @floatFromInt(@as(i16, @truncate((lParam >> 16) & 0xFFFF))); 
+            const xbutton: u32 = @truncate(wParam >> 16);
+            const button: MouseButton = switch (msg) {
+                w.WM_LBUTTONDOWN, w.WM_LBUTTONUP => .left,
+                w.WM_RBUTTONDOWN, w.WM_RBUTTONUP => .right,
+                w.WM_MBUTTONDOWN, w.WM_MBUTTONUP => .middle,
+                else => if (xbutton == @as(u32, @bitCast(w.XBUTTON1))) .four else .five,
+            };
+
+            switch (msg) {
+                w.WM_LBUTTONDOWN,
+                w.WM_MBUTTONDOWN,
+                w.WM_RBUTTONDOWN,
+                w.WM_XBUTTONDOWN,
+                => self.pushEvent(.{ .mouse_press = .{ .button = button, .mods = undefined, .pos = .{.x = x, .y = y }}}),
+                else => self.pushEvent(.{ .mouse_release = .{ .button = button, .mods = undefined, .pos = .{.x = x, .y = y}}}),
+            }
+
+            return if (msg == w.WM_XBUTTONDOWN or msg == w.WM_XBUTTONUP) w.TRUE else 0;
+        },
+        w.WM_MOUSEMOVE => {
+            self.pushEvent(.{
+                .mouse_motion = .{
+                    .pos = .{
+                        .x = 0,
+                        .y = 0,
+                    },
+                },
+            });
+            return 0;
+        },
+        w.WM_MOUSEWHEEL => {
+            self.pushEvent(.{
+                .mouse_scroll = .{
+                    .xoffset = 0,
+                    .yoffset = 0,
+                },
+            });
+            return 0;
+        },
+        else => return w.DefWindowProcW(wnd, msg, wParam, lParam),
+    }
+}
+
+fn keyFromScancode(scancode: u9) Key {
+    comptime var table: [0x15D]Key = undefined;
+    comptime for (&table, 1..) |*ptr, i| {
+        ptr.* = switch (i) {
+            0x1 => .escape,
+            0x2 => .one,
+            0x3 => .two,
+            0x4 => .three,
+            0x5 => .four,
+            0x6 => .five,
+            0x7 => .six,
+            0x8 => .seven,
+            0x9 => .eight,
+            0xA => .nine,
+            0xB => .zero,
+            0xC => .minus,
+            0xD => .equal,
+            0xE => .backspace,
+            0xF => .tab,
+            0x10 => .q,
+            0x11 => .w,
+            0x12 => .e,
+            0x13 => .r,
+            0x14 => .t,
+            0x15 => .y,
+            0x16 => .u,
+            0x17 => .i,
+            0x18 => .o,
+            0x19 => .p,
+            0x1A => .left_bracket,
+            0x1B => .right_bracket,
+            0x1C => .enter,
+            0x1D => .left_control,
+            0x1E => .a,
+            0x1F => .s,
+            0x20 => .d,
+            0x21 => .f,
+            0x22 => .g,
+            0x23 => .h,
+            0x24 => .j,
+            0x25 => .k,
+            0x26 => .l,
+            0x27 => .semicolon,
+            0x28 => .apostrophe,
+            0x29 => .grave,
+            0x2A => .left_shift,
+            0x2B => .backslash,
+            0x2C => .z,
+            0x2D => .x,
+            0x2E => .c,
+            0x2F => .v,
+            0x30 => .b,
+            0x31 => .n,
+            0x32 => .m,
+            0x33 => .comma,
+            0x34 => .period,
+            0x35 => .slash,
+            0x36 => .right_shift,
+            0x37 => .kp_multiply,
+            0x38 => .left_alt,
+            0x39 => .space,
+            0x3A => .caps_lock,
+            0x3B => .f1,
+            0x3C => .f2,
+            0x3D => .f3,
+            0x3E => .f4,
+            0x3F => .f5,
+            0x40 => .f6,
+            0x41 => .f7,
+            0x42 => .f8,
+            0x43 => .f9,
+            0x44 => .f10,
+            0x45 => .pause,
+            0x46 => .scroll_lock,
+            0x47 => .kp_7,
+            0x48 => .kp_8,
+            0x49 => .kp_9,
+            0x4A => .kp_subtract,
+            0x4B => .kp_4,
+            0x4C => .kp_5,
+            0x4D => .kp_6,
+            0x4E => .kp_add,
+            0x4F => .kp_1,
+            0x50 => .kp_2,
+            0x51 => .kp_3,
+            0x52 => .kp_0,
+            0x53 => .kp_decimal,
+            //0x56 => .europe2,
+            0x57 => .f11,
+            0x58 => .f12,
+            //0x5C => .international6,
+            0x64 => .f13,
+            0x65 => .f14,
+            0x66 => .f15,
+            0x67 => .f16,
+            0x68 => .f17,
+            0x69 => .f18,
+            0x6A => .f19,
+            0x6B => .f20,
+            0x6C => .f21,
+            0x6D => .f22,
+            0x6E => .f23,
+            //0x70 => .international2,
+            //0x73 => .international1,
+            //0x76 => .lang5,
+            //0x77 => .lang4,
+            //0x78 => .lang3,
+            //0x79 => .international4,
+            //0x7B => .international5,
+            //0x7D => .international3,
+            //0x7E => .kp_comma,
+            //0xF1 => .lang2,
+            //0xF2 => .lang1,
+            //0x11C => .kp_enter,
+            0x11D => .right_control,
+            //0x135 => .kp_divide,
+            0x138 => .right_alt,
+            0x145 => .num_lock,
+            0x146 => .pause,
+            0x147 => .home,
+            0x148 => .up,
+            0x149 => .page_up,
+            0x14B => .left,
+            0x14D => .right,
+            0x14F => .end,
+            0x150 => .down,
+            0x151 => .page_down,
+            0x152 => .insert,
+            0x153 => .delete,
+            0x15B => .left_super,
+            0x15C => .right_super,
+            0x15D => .menu,
+            else => .unknown,
+        };
+    };
+    return if (scancode > 0 and scancode <= table.len) table[scancode - 1] else .unknown;
+}
+
+pub fn joystickPresent(_: *Win32, _: Joystick) bool {
+    @panic("NOT IMPLEMENTED");
+}
+pub fn joystickName(_: *Win32, _: Joystick) ?[:0]const u8 {
+    @panic("NOT IMPLEMENTED");
+}
+pub fn joystickButtons(_: *Win32, _: Joystick) ?[]const bool {
+    @panic("NOT IMPLEMENTED");
+}
+// May be called from any thread.
+pub fn joystickAxes(_: *Win32, _: Joystick) ?[]const f32 {
+    @panic("NOT IMPLEMENTED");
+}

src/gfx/Sprite.zig

@@ -5,6 +5,9 @@ const gfx = mach.gfx;
 
 const math = mach.math;
 const vec2 = math.vec2;
+const vec3 = math.vec3;
+const vec4 = math.vec4;
+const mat4x4 = math.mat4x4;
 const Vec2 = math.Vec2;
 const Vec3 = math.Vec3;
 const Mat3x3 = math.Mat3x3;
@@ -87,8 +90,14 @@ fn updatePipeline(
             // that all entities with the same pipeline value are stored in contiguous memory, and
             // skip this copy.
             if (sprite_pipeline_id == pipeline_id) {
+                const uv = uv_transform;
                 gfx.SpritePipeline.cp_transforms[i] = transform;
-                gfx.SpritePipeline.cp_uv_transforms[i] = uv_transform;
+                // TODO(d3d12): #1217
+                //   changed the uv_transform to 4x4. The 3x3 causes issues with d3d12/hlsl
+                gfx.SpritePipeline.cp_uv_transforms[i].v[0] = vec4(uv.v[0].x(), uv.v[0].y(), uv.v[0].z(), 0.0);
+                gfx.SpritePipeline.cp_uv_transforms[i].v[1] = vec4(uv.v[1].x(), uv.v[1].y(), uv.v[1].z(), 0.0);
+                gfx.SpritePipeline.cp_uv_transforms[i].v[2] = vec4(uv.v[2].x(), uv.v[2].y(), uv.v[2].z(), 0.0);
+                gfx.SpritePipeline.cp_uv_transforms[i].v[3] = vec4(0.0, 0.0, 0.0, 0.0);
                 gfx.SpritePipeline.cp_sizes[i] = size;
                 i += 1;
                 num_sprites += 1;
@@ -98,8 +107,10 @@ fn updatePipeline(
 
     // Sort sprites back-to-front for draw order, alpha blending
     const Context = struct {
+        // TODO(d3d12): #1217
+        //   changed the uv_transform to 4x4. The 3x3 causes issues with d3d12/hlsl
         transforms: []Mat4x4,
-        uv_transforms: []Mat3x3,
+        uv_transforms: []Mat4x4, 
         sizes: []Vec2,
 
         pub fn lessThan(ctx: @This(), a: usize, b: usize) bool {
@@ -111,7 +122,9 @@ fn updatePipeline(
 
         pub fn swap(ctx: @This(), a: usize, b: usize) void {
             std.mem.swap(Mat4x4, &ctx.transforms[a], &ctx.transforms[b]);
-            std.mem.swap(Mat3x3, &ctx.uv_transforms[a], &ctx.uv_transforms[b]);
+            // TODO(d3d12): #1217
+            //   changed the uv_transform to 4x4. The 3x3 causes issues with d3d12/hlsl
+            std.mem.swap(Mat4x4, &ctx.uv_transforms[a], &ctx.uv_transforms[b]);
             std.mem.swap(Vec2, &ctx.sizes[a], &ctx.sizes[b]);
         }
     };

src/gfx/SpritePipeline.zig

@@ -110,7 +110,9 @@ const sprite_buffer_cap = 1024 * 512; // TODO(sprite): allow user to specify pre
 // TODO(sprite): eliminate these, see Sprite.updatePipeline for details on why these exist
 // currently.
 pub var cp_transforms: [sprite_buffer_cap]math.Mat4x4 = undefined;
-pub var cp_uv_transforms: [sprite_buffer_cap]math.Mat3x3 = undefined;
+// TODO(d3d12): #1217
+//   changed the uv_transform to 4x4. The 3x3 causes issues with d3d12/hlsl
+pub var cp_uv_transforms: [sprite_buffer_cap]math.Mat4x4 = undefined;
 pub var cp_sizes: [sprite_buffer_cap]math.Vec2 = undefined;
 
 /// Which render pass should be used during .render
@@ -209,7 +211,9 @@ fn buildPipeline(
     const uv_transforms = device.createBuffer(&.{
         .label = label ++ " uv_transforms",
         .usage = .{ .storage = true, .copy_dst = true },
-        .size = @sizeOf(math.Mat3x3) * sprite_buffer_cap,
+        // TODO(d3d12): #1217
+        //   changed the uv_transform to 4x4. The 3x3 causes issues with d3d12/hlsl
+        .size = @sizeOf(math.Mat4x4) * sprite_buffer_cap,
         .mapped_at_creation = .false,
     });
     const sizes = device.createBuffer(&.{

src/gfx/Text.zig

@@ -273,6 +273,9 @@ fn updatePipeline(
                         ).divScalar(px_density),
                         .size = size.divScalar(px_density),
                         .text_index = num_texts,
+                        // TODO(d3d12): #1217
+                        //   Added padding for d3d12/hlsl. Having 7 floats before the color vec caused and error.
+                        .text_padding = 0,
                         .uv_pos = vec2(@floatFromInt(r.x), @floatFromInt(r.y)),
                         .color = font_color,
                     });

src/gfx/TextPipeline.zig

@@ -134,6 +134,9 @@ pub const Glyph = extern struct {
 
     /// Which text this glyph belongs to; this is the index for transforms[i], colors[i].
     text_index: u32,
+    // TODO(d3d12): #1217
+    //   Added padding for d3d12/hlsl. Having 7 floats before the color vec caused and error.
+    text_padding: u32,
 
     /// Color of the glyph
     color: math.Vec4,

src/gfx/sprite.wgsl

@@ -25,7 +25,9 @@ struct Uniforms {
 
 // Sprite UV coordinate transformation matrices. Sprite UV coordinates are (0, 0) at the top-left
 // corner, and in pixels.
-@group(0) @binding(2) var<storage, read> sprite_uv_transforms: array<mat3x3<f32>>;
+// TODO(d3d12): #1217
+//   changed the uv_transform to 4x4. The 3x3 ca@group(0) @binding(2) var<storage, read> sprite_uv_transforms: array<mat4x4<f32>>;
+@group(0) @binding(2) var<storage, read> sprite_uv_transforms: array<mat4x4<f32>>;
 
 // Sprite sizes, in pixels.
 @group(0) @binding(3) var<storage, read> sprite_sizes: array<vec2<f32>>;
@@ -70,12 +72,12 @@ fn vertMain(
   // Currently, our pos_2d and uv coordinates describe a card that covers 1px by 1px; and the UV
   // coordinates describe using the entire texture. We alter the coordinates to describe the
   // desired sprite location, size, and apply a subset of the texture instead of the entire texture.
-  var pos = vec4<f32>(pos_2d * sprite_size, 0, 1); // normalized -> pixels
+  var pos = vec4<f32>(pos_2d * sprite_size, 0, 1); // normalized -> pixels 
   pos = sprite_transform * pos; // apply sprite transform (pixels)
   pos = uniforms.view_projection * pos; // pixels -> normalized
 
   uv *= sprite_size; // normalized -> pixels
-  uv = (sprite_uv_transform * vec3<f32>(uv, 1)).xy; // apply sprite UV transform (pixels)
+  uv = (sprite_uv_transform * vec4<f32>(uv, 1, 0)).xy; // apply sprite UV transform (pixels)
   uv /= uniforms.texture_size; // pixels -> normalized
 
   var output : VertexOutput;

src/gfx/text.wgsl

@@ -34,6 +34,10 @@ struct Glyph {
   // Which text this glyph belongs to; this is the index for transforms[i], colors[i]
   text_index: u32,
 
+  // TODO(d3d12): #1217
+  //   Added padding for d3d12/hlsl. Having 7 floats before the color vec caused and error.
+  text_index2: u32,   // Padding for struct alignment
+
   // Color of the glyph
   color: vec4<f32>,
 }

src/sysgpu/d3d12.zig

@@ -413,10 +413,6 @@ pub const Device = struct {
     }
 
     pub fn deinit(device: *Device) void {
-        if (device.lost_cb) |lost_cb| {
-            lost_cb(.destroyed, "Device was destroyed.", device.lost_cb_userdata);
-        }
-
         device.queue.waitUntil(device.queue.fence_value);
         device.processQueuedOperations();
 

src/sysgpu/shader/codegen/hlsl.zig

@@ -32,8 +32,6 @@ pub fn gen(allocator: std.mem.Allocator, air: *const Air, debug_info: DebugInfo)
         hlsl.arena.deinit();
     }
 
-    try hlsl.output.appendSlice(allocator, "#pragma pack_matrix( row_major )\n");
-
     for (air.refToList(air.globals_index)) |inst_idx| {
         switch (air.getInst(inst_idx)) {
             .@"var" => try hlsl.emitGlobalVar(inst_idx),
@@ -938,12 +936,14 @@ fn emitBinary(hlsl: *Hlsl, inst: Inst.Binary) !void {
             const rhs_type = hlsl.air.getInst(inst.rhs_type);
 
             if (lhs_type == .matrix or rhs_type == .matrix) {
-                // matrices are transposed
+                // TODO(d3d12): 
+                //   Changed to column major storage because dxc does not apply same ordering to Storage Buffers.
+                // matrices are in column major storage
                 try hlsl.writeAll("mul");
                 try hlsl.writeAll("(");
-                try hlsl.emitExpr(inst.rhs);
-                try hlsl.writeAll(", ");
                 try hlsl.emitExpr(inst.lhs);
+                try hlsl.writeAll(", ");
+                try hlsl.emitExpr(inst.rhs);
                 try hlsl.writeAll(")");
             } else {
                 try hlsl.emitBinaryOp(inst);