fixed performance issue on windows. got profiling with visual studio working, found that calling canvas.fill_rect() every pixel was way too slow. now writing pixels to one raw RGB buffer, and between frames, updating a texture with that buffer and copying it to the canvas.

Cargo.toml

@@ -7,3 +7,6 @@ edition = "2018"
 [dependencies]
 sdl2 = "0.32"
 cpuprofiler = "0.0.3"
+
+[profile.release]
+debug = true

src/main.rs

@@ -9,24 +9,33 @@ mod input;
 use cpu::Cpu;
 use ppu::Ppu;
 use cartridge::Cartridge;
-use screen::Screen;
+use screen::{init_window, draw_pixel, draw_to_window};
 use input::poll_buttons;
 
 use sdl2::keyboard::Keycode;
 use sdl2::event::Event;
+use sdl2::pixels::PixelFormatEnum;
 
 // use cpuprofiler::PROFILER;
 
 fn main() -> Result<(), String> {
+    // Set up screen
 	let sdl_context = sdl2::init()?;
 	let mut event_pump = sdl_context.event_pump()?;
-    let mut screen = Screen::new(&sdl_context)?;
+    let (mut canvas, texture_creator) = init_window(&sdl_context).expect("Could not create window");
+    let mut texture = texture_creator.create_texture_streaming(
+        PixelFormatEnum::RGB24, 256*screen::SCALE_FACTOR as u32, 240*screen::SCALE_FACTOR as u32)
+		.map_err(|e| e.to_string())?;
+    let byte_width = 256 * 3 * screen::SCALE_FACTOR; // 256 NES pixels, 3 bytes for each pixel (RGB 24-bit), and NES-to-SDL scale factor
+    let byte_height = 240 * screen::SCALE_FACTOR; // NES image is 240 pixels tall, multiply by scale factor for total number of rows needed
+    let mut screen_buffer = vec![0; byte_width * byte_height]; // contains raw RGB data for the screen
 
+    // Initialize hardware components
     let cart = Cartridge::new();
-
     let ppu = Ppu::new(&cart);
     let mut cpu = Cpu::new(&cart, ppu);
 
+    // For throttling to 60 FPS
     let mut timer = Instant::now();
     let mut fps_timer = Instant::now();
     let mut fps = 0;
@@ -39,12 +48,12 @@ fn main() -> Result<(), String> {
         for _i in 0..num_cycles * 3 {
             let (pixel, end_of_frame) = cpu.ppu.step();
             match pixel {
-                Some((x, y, color)) => screen.draw_pixel(x, y, color),
-                None => Ok(()),
-            }?;
+                Some((x, y, color)) => draw_pixel(&mut screen_buffer, x, y, color),
+                None => (),
+            };
             if end_of_frame {
                 fps += 1;
-                screen.canvas.present();
+                draw_to_window(&mut texture, &mut canvas, &screen_buffer)?;
                 let now = Instant::now();
                 if now < timer + Duration::from_millis(1000/60) {
                     std::thread::sleep(timer + Duration::from_millis(1000/60) - now);

src/screen/mod.rs

@@ -2,38 +2,50 @@ extern crate sdl2;
 
 use sdl2::Sdl;
 use sdl2::pixels::Color;
-use sdl2::rect::Rect;
-use sdl2::render::Canvas;
-use sdl2::video::Window;
+use sdl2::render::{Canvas, Texture, TextureCreator};
+use sdl2::video::{Window, WindowContext};
 
-const SCALE_FACTOR: usize = 4;
+pub const SCALE_FACTOR: usize = 3;
+const BYTES_IN_COL: usize = SCALE_FACTOR * 3;   // 3 bytes per pixel in RGB24. This represents a thick, SCALE_FACTOR-pixel-wide column.
+const BYTES_IN_ROW: usize = BYTES_IN_COL * 256; // 256 = screen width in NES pixels. This represents a thin, one-SDL-pixel-tall row.
 
 type RGBColor = (u8, u8, u8);
 
-pub struct Screen {
-	pub canvas: Canvas<Window>,
+pub fn init_window(context: &Sdl) -> Result<(Canvas<Window>, TextureCreator<WindowContext>), String> {
+	let video_subsystem = context.video()?;
+	let window = video_subsystem.window("NESTUR", (256 * SCALE_FACTOR) as u32, (240 * SCALE_FACTOR) as u32)
+		.position_centered()
+		.opengl()
+		.build()
+		.map_err(|e| e.to_string())?;
+	let mut canvas = window.into_canvas().build().map_err(|e| e.to_string())?;
+	let texture_creator = canvas.texture_creator();
+	canvas.set_draw_color(Color::RGB(0, 0, 0));
+	canvas.clear();
+	canvas.present();
+	Ok((canvas, texture_creator))
 }
 
-impl Screen {
-	pub fn new(context: &Sdl) -> Result<Self, String> {
-		let video_subsystem = context.video()?;
-		let window = video_subsystem.window("NESTUR", (256 * SCALE_FACTOR) as u32, (240 * SCALE_FACTOR) as u32)
-			.position_centered()
-			.opengl()
-			.build()
-			.map_err(|e| e.to_string())?;
-		let mut canvas = window.into_canvas().build().map_err(|e| e.to_string())?;
-		canvas.set_draw_color(Color::RGB(0, 0, 0));
-		canvas.clear();
-		canvas.present();
-		Ok(Screen{canvas})
+pub fn draw_pixel(buffer: &mut Vec<u8>, x: usize, y: usize, color: RGBColor) {
+	let (r, g, b) = color;
+	let nes_y_offset = y * BYTES_IN_ROW * SCALE_FACTOR; // find offset for thick, SCALE_FACTOR-pixel tall row
+	for sdl_row_num in 0..SCALE_FACTOR { // looping over one-pixel tall rows up to SCALE_FACTOR
+		let row_offset = nes_y_offset + (sdl_row_num * BYTES_IN_ROW); // row_offset is the offset within buffer of the thin row we're on
+		let nes_x_offset = x * BYTES_IN_COL; // find horizontal offset within row (in byte terms) of NES x-coordinate
+		for sdl_col_num in 0..SCALE_FACTOR { // for pixels up to SCALE_FACTOR, moving horizontally
+			let col_offset = nes_x_offset + (sdl_col_num * 3); // skip 3 bytes at a time, R/G/B for each pixel
+			let offset = row_offset + col_offset;
+			buffer[offset + 0] = r;
+			buffer[offset + 1] = g;
+			buffer[offset + 2] = b;
+		}
 	}
+}
 
-	pub fn draw_pixel(&mut self, x: usize, y: usize, color: RGBColor) -> Result<(), String> {
-		let (r, g, b) = color;
-		self.canvas.set_draw_color(Color::RGB(r, g, b));
-		self.canvas.fill_rect(Rect::new((x * SCALE_FACTOR) as i32, (y * SCALE_FACTOR) as i32,
-			SCALE_FACTOR as u32, SCALE_FACTOR as u32))?;
-		Ok(())
-	}
-}
+pub fn draw_to_window(texture: &mut Texture, canvas: &mut Canvas<Window>, buffer: &Vec<u8>) -> Result<(), String> {
+	texture.update(None, buffer, 256*3*SCALE_FACTOR)
+		.map_err(|e| e.to_string())?;
+    canvas.copy(&texture, None, None)?;
+	canvas.present();
+	Ok(())
+}