fix build?

It needs some more thought before committing to it.

.github/workflows/ci.yml

@@ -2,7 +2,7 @@ name: Rust
 
 on:
   push:
-    branches: [ master ]
+    branches: [ master, testing ]
 
 env:
   CARGO_TERM_COLOR: always
@@ -27,7 +27,7 @@ jobs:
 
     steps:
     - name: Install dependencies
-      run: sudo apt-get install -y libxkbcommon-dev
+      run: sudo apt-get install -y libxkbcommon-dev libasound2-dev
       if: matrix.os == 'ubuntu-latest'
     - name: Checkout
       uses: actions/checkout@v2

Cargo.toml

@@ -1,22 +1,24 @@
 [package]
 name = "uw8"
-version = "0.1.1"
+version = "0.2.1"
 edition = "2021"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [features]
 default = ["native", "browser"]
-native = ["wasmtime"]
+native = ["wasmtime", "uw8-window", "cpal", "rubato" ]
 browser = ["warp", "tokio", "tokio-stream", "webbrowser"]
 
 [dependencies]
-wasmtime = { version = "0.30", optional = true }
+wasmtime = { version = "0.37.0", optional = true }
 anyhow = "1"
-minifb = { version = "0.20", default-features = false, features = ["x11"] }
+env_logger = "0.9"
+log = "0.4"
+uw8-window = { path = "uw8-window", optional = true }
 notify = "4"
 pico-args = "0.4"
-curlywas = { git = "https://github.com/exoticorn/curlywas.git", rev = "196719b" }
+curlywas = { git = "https://github.com/exoticorn/curlywas.git", rev = "0e7ea50" }
 wat = "1"
 uw8-tool = { path = "uw8-tool" }
 same-file = "1"
@@ -24,3 +26,6 @@ warp = { version = "0.3.2", optional = true }
 tokio = { version = "1.17.0", features = ["sync", "rt"], optional = true }
 tokio-stream = { version = "0.1.8", features = ["sync"], optional = true }
 webbrowser = { version = "0.6.0", optional = true }
+ansi_term = "0.12.1"
+cpal = { version = "0.13.5", optional = true }
+rubato = { version = "0.11.0", optional = true }

README.md

@@ -15,9 +15,9 @@ See [here](https://exoticorn.github.io/microw8/) for more information and docs.
 
 ## Downloads
 
-* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-linux.tgz)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-linux.tgz)
-* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-macos.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-macos.tgz)
-* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-windows.zip)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-windows.zip)
 
 The download includes
 
@@ -35,6 +35,7 @@ Runs <file> which can be a binary WebAssembly module, an `.uw8` cart, a wat (Web
 
 Options:
 
+-b, --browser           : Run in browser instead of using native runtime
 -t, --timeout FRAMES    : Sets the timeout in frames (1/60s)
 -w, --watch             : Reloads the given file every time it changes on disk.
 -p, --pack              : Pack the file into an .uw8 cart before running it and print the resulting size.
@@ -42,6 +43,27 @@ Options:
 -l LEVEL, --level LEVEL : Compression level (0-9). Higher compression levels are really slow.
 -o FILE, --output FILE  : Write the loaded and optionally packed cart back to disk.
 
+when using the native runtime:
+
+-m, --no-audio          : Disable audio, also reduces cpu load a bit
+--no-gpu                : Force old cpu-only window code
+--filter FILTER         : Select an upscale filter at startup
+--fullscreen            : Start in fullscreen mode
+
+Note that the cpu-only window does not support fullscreen nor upscale filters.
+
+Unless --no-gpu is given, uw8 will first try to open a gpu accelerated window, falling back to the old cpu-only window if that fails.
+Therefore you should rarely need to manually pass --no-gpu. If you prefer the old pixel doubling look to the now default crt filter,
+you can just pass "--filter nearest" or "--filter 1".
+
+The upscale filter options are:
+1, nearest              : Anti-aliased nearest filter
+2, fast_crt             : Very simple, cheap crt filter, not very good below a window size of 960x720
+3, ss_crt               : Super sampled crt filter, a little more demanding on the GPU but scales well to smaller window sizes
+4, chromatic_crt        : Variant of fast_crt with a slight offset of the three color dots of a pixel, still pretty cheap
+5, auto_crt (default)   : ss_crt below 960x720, chromatic_crt otherwise
+
+You can switch the upscale filter at any time using the keys 1-5. You can toggle fullscreen with F.
 
 uw8 pack [<options>] <infile> <outfile>
 

examples/curlywas/control.cwa

@@ -1,5 +1,4 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-import "env.printString" fn printString(i32);
 
 export fn upd() {
   printString(0x20000);

examples/curlywas/cracklebass.cwa

@@ -0,0 +1,55 @@
+// port of cracklebass by pestis (originally on TIC-80)
+
+include "../include/microw8-api.cwa"
+
+const MUSIC_DATA = 0x20000;
+
+export fn upd() {
+    let inline t = 32!32 * 6 / 100;
+    let inline p = t / 1024;
+
+    let channel:i32;
+
+    loop channels {
+        let inline e = t * channel?MUSIC_DATA / 8;
+        let lazy pattern = (8 * channel + p)?(MUSIC_DATA + 56);
+        let lazy n = !!pattern * (8 * pattern + e / 16 % 8)?MUSIC_DATA;
+        let inline prev_ctrl = (channel * 6)?80;
+        (channel * 6)?80 = if n {
+            let inline base_note = 12 + 12 * channel?(MUSIC_DATA + 4) + n;
+            let inline pitch_drop = e % 16 * channel?(MUSIC_DATA + 94);
+            let inline key_pattern = p?(MUSIC_DATA + 8*4 + 56);
+            let inline key = select(key_pattern, (8 * key_pattern + t / 128 % 8)?MUSIC_DATA, 1);
+            (channel * 6)?83 = base_note - pitch_drop / 4 + key;
+            prev_ctrl & 0xfc | (e / 8 & 2) | 1
+        } else {
+            prev_ctrl & 0xfe
+        };
+
+        branch_if (channel := channel + 1) < 4: channels;
+    }
+}
+
+data 80 {
+    i8(
+        0x44, 0, 0, 0, 0x50, 0x40,
+        0x4, 0x50, 0, 0, 0x80, 0x80,
+        0x40, 0x80, 0, 0, 0x40, 0x40,
+        0, 0, 0, 0, 0x50, 0x50
+    )
+}
+
+data MUSIC_DATA {
+    i8(
+ 16, 2, 8, 8, 1, 2, 2, 3, 1, 0,
+  1,13,16, 0, 1, 8, 1, 0, 1,13,
+ 16, 1, 1, 8, 1, 0, 8,13,13, 0,
+ 16,13, 1, 0, 1, 0, 1, 0, 1, 1,
+  1, 0, 0, 0, 1, 0,13, 1, 1, 1,
+  6, 8, 1, 1, 6, 8, 1, 1, 2, 1,
+  2, 1, 2, 0, 0, 0, 0, 3, 3, 3,
+  5, 0, 0, 2, 1, 2, 1, 2, 1, 2,
+  0, 4, 4, 0, 4, 4, 4, 4, 0, 0,
+  0, 0, 6, 6, 0, 0, 0, 8
+    )
+}

examples/curlywas/fireworks.cwa

@@ -1,11 +1,4 @@
-import "env.time" fn time() -> f32;
+include "../include/microw8-api.cwa"
-import "env.circle" fn circle(f32, f32, f32, i32);
-import "env.cls" fn cls(i32);
-import "env.randomSeed" fn seed(i32);
-import "env.randomf" fn randomf() -> f32;
-import "env.sin" fn sin(f32) -> f32;
-import "env.cos" fn cos(f32) -> f32;
-import "env.fmod" fn fmod(f32, f32) -> f32;
 
 export fn upd() {
     cls(0);
@@ -15,10 +8,10 @@ export fn upd() {
         let inline rocket = i #>> 9;
         let lazy local_time = fmod(time() + rocket as f32 / 5 as f32, 2 as f32);
         let lazy rocket = rocket + nearest(time() - local_time) as i32 * 10;
-        seed(rocket);
+        randomSeed(rocket);
         let inline x = randomf() * 645 as f32;
         let y = randomf() * 133 as f32;
-        let lazy angle = { seed(i); randomf() } * 44 as f32;
+        let lazy angle = { randomSeed(i); randomf() } * 44 as f32;
         let inline dx = sin(angle);
         let inline dy = cos(angle);
         let lazy dist = local_time * (randomf() * 44 as f32);

examples/curlywas/font_palette.cwa

@@ -1,38 +1,30 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-
-import "env.cls" fn cls(i32);
-import "env.printString" fn printString(i32);
-import "env.printChar" fn printChar(i32);
-import "env.setCursorPosition" fn setCursor(i32, i32);
-import "env.setTextColor" fn setTextColor(i32);
-import "env.line" fn line(f32, f32, f32, f32, i32);
-import "env.isButtonTriggered" fn triggered(i32) -> i32;
 
 global mut mode: i32 = 0;
 
 export fn upd() {
     cls(0);
 
-    if triggered(4) {
+    if isButtonTriggered(BUTTON_A) {
         mode = !mode;
     }
 
     setTextColor(15);
-    printString(mode * 0x20000);
+    printString(mode * USER_MEM);
 
     let y: i32;
     loop y {
         line(0 as f32, (y * 9 + 39) as f32, (14+16*9) as f32, (y * 9 + 39) as f32, 1);
         line((y * 9 + 15) as f32, 24 as f32, (y * 9 + 15) as f32, (38+16*9) as f32, 1);
         setTextColor(15);
-        setCursor(y * 9 + 16, 24);
+        setCursorPosition(y * 9 + 16, 24);
         let lazy hexChar = select(y < 10, y + 48, y + 87);
         printChar(hexChar);
-        setCursor(0, y * 9 + 24+16);
+        setCursorPosition(0, y * 9 + 24+16);
         printChar(hexChar);
         let x = 0;
         loop x {
-            setCursor(x * 9 + 16, y * 9 + 24+16);
+            setCursorPosition(x * 9 + 16, y * 9 + 24+16);
             setTextColor(select(mode, x + y * 16, -9));
             if y >= 2 | mode {
                 printChar(select(mode, 0xa4, x + y * 16));
@@ -47,6 +39,6 @@ data 0 {
     "Default font:      (press " i8(0xcc) " for palette)" i8(5, 0)
 }
 
-data 0x20000 {
+data USER_MEM {
     "Default palette:   (press " i8(0xcc) " for font)" i8(5, 0)
 }

examples/curlywas/line.cwa

@@ -1,10 +1,4 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-
-import "env.cls" fn cls(i32);
-import "env.time" fn time() -> f32;
-import "env.line" fn line(f32, f32, f32, f32, i32);
-import "env.sin" fn sin(f32) -> f32;
-import "env.cos" fn cos(f32) -> f32;
 
 export fn upd() {
     cls(0);

examples/curlywas/simple_music.cwa

@@ -0,0 +1,38 @@
+include "../include/microw8-api.cwa"
+
+global mut frame = 0;
+
+export fn upd() {
+    if frame % 16 == 0 {
+        let ch: i32;
+        loop channels {
+            playNote(ch, (ch * 32 + (frame / 16) % 32)?0x20000);
+            branch_if ch := (ch + 1) % 4: channels;
+        }
+    }
+    frame = frame + 1;
+}
+
+data 0x20000 {
+    i8(
+        0x4e, 0x0, 0x0, 0x4c, 0x49, 0x0, 0x45, 0x47,
+        0x49, 0x47, 0x45, 0x44, 0x42, 0x0, 0x3d, 0x41,
+        0x44, 0x0, 0x0, 0x47, 0x49, 0x47, 0x45, 0x41,
+        0x44, 0x0, 0x0, 0x0, 0x42, 0x0, 0x0, 0x0,
+
+        0x25, 0, 0x49, 0x25, 0x25, 0, 0x49, 0x38,
+        0x25, 0, 0x49, 0x25, 0x25, 0, 0x49, 0x38,
+        0x25, 0, 0x49, 0x25, 0x25, 0, 0x49, 0x38,
+        0x25, 0, 0x49, 0x25, 0x25, 0, 0x49, 0x38,
+
+        0x2a, 0x0, 0x0, 0x0, 0x2d, 0x0, 0x0, 0x0,
+        0x2c, 0x0, 0x28, 0x0, 0x2a, 0x0, 0x0, 0x0,
+        0x25, 0x0, 0x0, 0x0, 0x29, 0x0, 0x0, 0x0,
+        0x2c, 0x0, 0x2d, 0x0, 0x2a, 0x0, 0x25, 0x0,
+
+        0x0, 0x0, 0x31, 0x0, 0x34, 0x0, 0x0, 0x36,
+        0x38, 0x39, 0x38, 0x34, 0x36, 0x0, 0x0, 0x0,
+        0x0, 0x3d, 0x3b, 0x39, 0x38, 0x0, 0x0, 0x0,
+        0x0, 0x39, 0x38, 0x39, 0x38, 0x0, 0x36, 0x0
+    )
+}

examples/curlywas/skipahead.cwa

@@ -1,12 +1,4 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-
-import "env.rectangle" fn rect(f32, f32, f32, f32, i32);
-import "env.circle" fn circle(f32, f32, f32, i32);
-import "env.isButtonPressed" fn btn(i32) -> i32;
-import "env.random" fn random() -> i32;
-import "env.randomSeed" fn randomSeed(i32);
-import "env.cls" fn cls(i32);
-import "env.printInt" fn printInt(i32);
 
 global mut pz: i32 = 4;
 global mut px: f32 = 2.0;
@@ -16,10 +8,10 @@ global mut f: f32 = 2.0;
 
 export fn upd() {
     let y: i32;
-    let inline zero = 0.0;
+    let inline zero = 0_f;
 
     let lazy control_speed = 0.03125;
-    s = s + 0.1875 - (f + control_speed) * btn(4 <| cls(4)) as f32;
+    s = s + 0.1875 - (f + control_speed) * isButtonPressed(4 <| cls(4)) as f32;
     f = f * 0.5625;
 
     printInt(pz);
@@ -33,11 +25,13 @@ export fn upd() {
         
         let inline c = (z & 1) * -2;
         let inline yf = y as f32;
-        rect(rx, yf, rw, yf / 6 as f32, c + 1);
+        rectangle(rx, yf, rw, yf / 6 as f32, c + 1);
-        rect(rx, yf, rw, 1 as f32, c - 4);
+        rectangle(rx, yf, rw, 1 as f32, c - 4);
 
         if y == 180 & py > zero {
             if x > w | x < zero {
+                0?80 = 0xc3;
+                3?80 = 32;
                 return;
             }
             py = zero;
@@ -51,7 +45,10 @@ export fn upd() {
     circle(160 as f32, 160 as f32 + py, 22 as f32, -28);
     circle((160 - 6) as f32, (160 - 6) as f32 + py, 6 as f32, -26);
 
-    px = px + (btn(3) - btn(2)) as f32 * control_speed;
+    0?86 = py < zero;
+    3?86 = 32 - py as i32;
+
+    px = px + (isButtonPressed(3) - isButtonPressed(2)) as f32 * control_speed;
     py = py + s;
     pz = pz + 1;
 }

examples/curlywas/steadyon.cwa

@@ -0,0 +1,38 @@
+// Steady On Tim, It's Only A Budget Game
+// by Gasman / Hooy-Program
+// ported to MicroW8 by exoticorn/icebird
+
+include "../include/microw8-api.cwa"
+
+fn melody(t: i32, T: i32) -> i32 {
+    let inline riff_pos = abs(((T&31) - 16) as f32) as i32;
+    let lazy shift = ((1-((T>>5)&3))%2-1) as f32 / 6 as f32;
+
+    let inline note_count = 5 - (T >= 512);
+    let inline octave = (riff_pos/5) as f32;
+    let inline riff_note = 5514 >> (riff_pos % note_count * 4) & 15;
+    let inline melody_freq = pow(2 as f32, shift + octave - (riff_note as f32 / 12 as f32));
+    let inline melody = (t as f32 * melody_freq) as i32 & 128;
+    
+    let inline arp_note = ((0x85>>((t>>12)%3*4)) & 15) - 1;
+    let inline arp_freq = pow(2 as f32, shift + (arp_note as f32 / 12 as f32));
+    let inline arp_vol = (T >= 256) * (12-T%12);
+    let inline arpeggio = ((t as f32 * arp_freq) as i32 & 128) * arp_vol / 12;
+    
+    melody + arpeggio
+}
+
+export fn snd(t: i32) -> f32 {
+    let lazy T = t/10000;
+    
+    let inline mel_arp = melody(t, T)/3 + melody(t, T-3)/5;
+
+    let inline bass_vol = (T >= 128) & (197 >> (T % 8));
+    let inline bass_freq = pow(2 as f32, (((T & 4) * ((T & 7) - 1)) as f32 / 24 as f32 - 5 as f32));
+    let inline bass = ((t as f32 * bass_freq) as i32 & 63) * bass_vol;
+
+    let inline snare_ish = (random() & 31) * (8 - (T + 4) % 8) / 8;
+
+    let inline sample = mel_arp + bass + snare_ish;
+    sample as f32 / 255 as f32
+}

examples/curlywas/technotunnel.cwa

@@ -1,7 +1,4 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-import "env.sin" fn sin(f32) -> f32;
-import "env.time" fn time() -> f32;
-import "env.setPixel" fn setPixel(i32, i32, i32);
 
 export fn upd() {
     let x: i32;

examples/curlywas/tim_ges.cwa

@@ -0,0 +1,64 @@
+// Steady On Tim, It's Only A Budget Game
+// original bytebeat by Gasman / Hooy-Program
+// ported to MicroW8/GES by exoticorn/icebird
+
+import "env.memory" memory(4);
+
+fn melody(ch: i32, t: i32, T: i32) {
+    let lazy riff_pos = abs(((T&31) - 16) as f32) as i32;
+    let lazy shift = ((1-((T>>5)&3))%2-1) * 2;
+
+    let inline note_count = 5 - (T >= 512);
+    let inline octave = (riff_pos/5) * 12;
+    let inline riff_note = 5514 >> (riff_pos % note_count * 4) & 15;
+    let inline melody_note = shift + octave - riff_note;
+
+    ch?1 = 230 - riff_pos * 14;
+    ch?3 = melody_note + 64;
+
+    let inline arp_note = shift + ((0x85>>((t/2)%3*4)) & 15) - 1;
+    80?3 = arp_note + 64;
+}
+
+export fn upd() {
+    let lazy t = 32!32 / (1000/60);
+    let lazy T = t / 7;
+    melody(98, t, T - 3);
+    melody(92, t, T);
+
+    80?0 = ((T >= 256) & (T/12+(T-3)/12)) * 2 | 0x48; // arp trigger
+
+    if T >= 128 {
+        let inline bass_step = T % 8;
+        86?3 = if bass_step / 2 == 2 {
+            86?0 = 0xd6;
+            81
+        } else {
+            86?0 = ((197 >> bass_step) & 1) | 0x48;
+            ((T & 4) * ((T & 7) - 1)) / 2 + 28
+        };
+    }
+}
+
+data 80 {
+    i8(
+        0, 0x90, 0, 0, 0, 0x90,
+        0, 0x4c, 0, 0, 0, 0x4c,
+        0x19, 0, 0, 0, 0, 0x4c,
+        0x19, 0, 0, 0, 0, 0x4c,
+        0xfa, 0x84,
+        0xc1, 0xc1, 0, 107, 0, 0x4c
+    )
+}
+
+/*
+include "../../platform/src/ges.cwa"
+
+import "env.pow" fn pow(f32, f32) -> f32;
+import "env.exp" fn exp(f32) -> f32;
+import "env.sin" fn sin(f32) -> f32;
+
+export fn snd(t: i32) -> f32 {
+    gesSnd(t)
+}
+*/

examples/curlywas/trainride.cwa

@@ -1,7 +1,4 @@
-import "env.memory" memory(2);
+include "../include/microw8-api.cwa"
-
-import "env.fmod" fn fmod(f32, f32) -> f32;
-import "env.time" fn time() -> f32;
 
 export fn upd() {
   let i: i32;

examples/curlywas/tunnel.cwa

@@ -1,7 +1,4 @@
-import "env.memory" memory(4);
+include "../include/microw8-api.cwa"
-
-import "env.atan2" fn atan2(f32, f32) -> f32;
-import "env.time" fn time() -> f32;
 
 export fn upd() {
     let i: i32;
@@ -12,7 +9,7 @@ export fn upd() {
         let inline d = 40000 as f32 / sqrt(x * x + y * y);
         let inline u = atan2(x, y) * (512.0 / 3.141);
         let inline c = ((i32.trunc_sat_f32_s(d + t * 2 as f32) ^ i32.trunc_sat_f32_s(u + t)) & 255) >> 4;
-        i?120 = c;
+        i?FRAMEBUFFER = c;
 
         branch_if (i := i + 1) < 320*240: pixels;
     }

examples/include/microw8-api.cwa

@@ -0,0 +1,51 @@
+// MicroW8 APIs, to be `include`d in CurlyWas sources
+import "env.memory" memory(4);
+
+import "env.sin" fn sin(f32) -> f32;
+import "env.cos" fn cos(f32) -> f32;
+import "env.tan" fn tan(f32) -> f32;
+import "env.asin" fn asin(f32) -> f32;
+import "env.acos" fn acos(f32) -> f32;
+import "env.atan" fn atan(f32) -> f32;
+import "env.atan2" fn atan2(f32, f32) -> f32;
+import "env.pow" fn pow(f32, f32) -> f32;
+import "env.log" fn log(f32) -> f32;
+import "env.fmod" fn fmod(f32, f32) -> f32;
+import "env.random" fn random() -> i32;
+import "env.randomf" fn randomf() -> f32;
+import "env.randomSeed" fn randomSeed(i32);
+import "env.cls" fn cls(i32);
+import "env.setPixel" fn setPixel(i32, i32, i32);
+import "env.getPixel" fn getPixel(i32, i32) -> i32;
+import "env.hline" fn hline(i32, i32, i32, i32);
+import "env.rectangle" fn rectangle(f32, f32, f32, f32, i32);
+import "env.circle" fn circle(f32, f32, f32, i32);
+import "env.line" fn line(f32, f32, f32, f32, i32);
+import "env.time" fn time() -> f32;
+import "env.isButtonPressed" fn isButtonPressed(i32) -> i32;
+import "env.isButtonTriggered" fn isButtonTriggered(i32) -> i32;
+import "env.printChar" fn printChar(i32);
+import "env.printString" fn printString(i32);
+import "env.printInt" fn printInt(i32);
+import "env.setTextColor" fn setTextColor(i32);
+import "env.setBackgroundColor" fn setBackgroundColor(i32);
+import "env.setCursorPosition" fn setCursorPosition(i32, i32);
+import "env.rectangle_outline" fn rectangle_outline(f32, f32, f32, f32, i32);
+import "env.circle_outline" fn circle_outline(f32, f32, f32, i32);
+import "env.exp" fn exp(f32) -> f32;
+import "env.playNote" fn playNote(i32, i32);
+
+const TIME_MS = 0x40;
+const GAMEPAD = 0x44;
+const FRAMEBUFFER = 0x78;
+const PALETTE = 0x13000;
+const FONT = 0x13400;
+const USER_MEM = 0x14000;
+const BUTTON_UP = 0x0;
+const BUTTON_DOWN = 0x1;
+const BUTTON_LEFT = 0x2;
+const BUTTON_RIGHT = 0x3;
+const BUTTON_A = 0x4;
+const BUTTON_B = 0x5;
+const BUTTON_X = 0x6;
+const BUTTON_Y = 0x7;

examples/include/microw8-api.wat

@@ -0,0 +1,53 @@
+;; MicroW8 APIs, in WAT (Wasm Text) format
+(import "env" "memory" (memory 4))
+
+(import "env" "sin" (func $sin (param f32) (result f32)))
+(import "env" "cos" (func $cos (param f32) (result f32)))
+(import "env" "tan" (func $tan (param f32) (result f32)))
+(import "env" "asin" (func $asin (param f32) (result f32)))
+(import "env" "acos" (func $acos (param f32) (result f32)))
+(import "env" "atan" (func $atan (param f32) (result f32)))
+(import "env" "atan2" (func $atan2 (param f32) (param f32) (result f32)))
+(import "env" "pow" (func $pow (param f32) (param f32) (result f32)))
+(import "env" "log" (func $log (param f32) (result f32)))
+(import "env" "fmod" (func $fmod (param f32) (param f32) (result f32)))
+(import "env" "random" (func $random (result i32)))
+(import "env" "randomf" (func $randomf (result f32)))
+(import "env" "randomSeed" (func $randomSeed (param i32)))
+(import "env" "cls" (func $cls (param i32)))
+(import "env" "setPixel" (func $setPixel (param i32) (param i32) (param i32)))
+(import "env" "getPixel" (func $getPixel (param i32) (param i32) (result i32)))
+(import "env" "hline" (func $hline (param i32) (param i32) (param i32) (param i32)))
+(import "env" "rectangle" (func $rectangle (param f32) (param f32) (param f32) (param f32) (param i32)))
+(import "env" "circle" (func $circle (param f32) (param f32) (param f32) (param i32)))
+(import "env" "line" (func $line (param f32) (param f32) (param f32) (param f32) (param i32)))
+(import "env" "time" (func $time (result f32)))
+(import "env" "isButtonPressed" (func $isButtonPressed (param i32) (result i32)))
+(import "env" "isButtonTriggered" (func $isButtonTriggered (param i32) (result i32)))
+(import "env" "printChar" (func $printChar (param i32)))
+(import "env" "printString" (func $printString (param i32)))
+(import "env" "printInt" (func $printInt (param i32)))
+(import "env" "setTextColor" (func $setTextColor (param i32)))
+(import "env" "setBackgroundColor" (func $setBackgroundColor (param i32)))
+(import "env" "setCursorPosition" (func $setCursorPosition (param i32) (param i32)))
+(import "env" "rectangle_outline" (func $rectangle_outline (param f32) (param f32) (param f32) (param f32) (param i32)))
+(import "env" "circle_outline" (func $circle_outline (param f32) (param f32) (param f32) (param i32)))
+(import "env" "exp" (func $exp (param f32) (result f32)))
+(import "env" "playNote" (func $playNote (param i32) (param i32)))
+
+;; to use defines, include this file with a preprocessor
+;; like gpp (https://logological.org/gpp).
+#define TIME_MS 0x40;
+#define GAMEPAD 0x44;
+#define FRAMEBUFFER 0x78;
+#define PALETTE 0x13000;
+#define FONT 0x13400;
+#define USER_MEM 0x14000;
+#define BUTTON_UP 0x0;
+#define BUTTON_DOWN 0x1;
+#define BUTTON_LEFT 0x2;
+#define BUTTON_RIGHT 0x3;
+#define BUTTON_A 0x4;
+#define BUTTON_B 0x5;
+#define BUTTON_X 0x6;
+#define BUTTON_Y 0x7;

examples/wat/tunnel_opt.wat

@@ -0,0 +1,74 @@
+(module
+  (import "env" "atan2" (func $atan2 (param f32 f32) (result f32)))
+  (import "env" "time" (func $time (result f32)))
+  (import "env" "memory" (memory 4))
+  (func (export "upd")
+    (local $y i32)
+    (local $i i32) 
+    (local $x i32)
+
+    (loop $pixels
+      i32.const 1
+      local.get $i
+
+      local.get $i
+
+      i32.const 36928
+      f32.convert_i32_s
+      local.get $i
+      i32.const 320
+      i32.rem_s
+      i32.const 160
+      i32.sub
+      local.tee $x
+      local.get $x
+      i32.mul
+      local.get $i
+      i32.const 320
+      i32.div_s
+      i32.const 120
+      i32.sub
+      local.tee $y
+      local.get $y
+      i32.mul
+      i32.add
+      f32.convert_i32_s
+      f32.sqrt
+      f32.div
+      call $time
+      i32.const 163
+      f32.convert_i32_s
+      f32.mul
+      f32.add
+      i32.trunc_sat_f32_s
+
+      local.get $x
+      f32.convert_i32_s
+      local.get $y
+      f32.convert_i32_s
+      call $atan2
+      i32.const 163
+      f32.convert_i32_s
+      f32.mul
+      call $time
+      i32.const 64
+      f32.convert_i32_s
+      f32.mul
+      f32.add
+      i32.trunc_f32_s
+
+      i32.xor
+      i32.const 4
+      i32.shr_s
+      i32.const 15
+      i32.and
+      i32.store8 offset=120
+
+      i32.add
+      local.tee $i
+      i32.const 76800
+      i32.rem_s
+      br_if $pixels
+  )
+ )
+)

platform/Cargo.lock

@@ -79,9 +79,9 @@ checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
 
 [[package]]
 name = "chumsky"
-version = "0.5.0"
+version = "0.8.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c2d3efff85e8572b1c3fa0127706af58c4fff8458f8d9436d54b1e97573c7a3f"
+checksum = "8d02796e4586c6c41aeb68eae9bfb4558a522c35f1430c14b40136c3706e09e4"
 dependencies = [
  "ahash 0.3.8",
 ]
@@ -146,14 +146,14 @@ checksum = "7a81dae078cea95a014a339291cec439d2f232ebe854a9d672b796c6afafa9b7"
 [[package]]
 name = "curlywas"
 version = "0.1.0"
-source = "git+https://github.com/exoticorn/curlywas.git?rev=196719b#196719b35ef377cb7e001554b27ac5de013dcf2b"
+source = "git+https://github.com/exoticorn/curlywas.git?rev=0e7ea50#0e7ea508cd0e76836283ae68a44c9097df83c8ac"
 dependencies = [
  "anyhow",
  "ariadne",
  "chumsky",
  "pico-args",
- "wasm-encoder",
+ "wasm-encoder 0.10.0",
- "wasmparser",
+ "wasmparser 0.83.0",
 ]
 
 [[package]]
@@ -197,6 +197,21 @@ dependencies = [
  "ahash 0.7.6",
 ]
 
+[[package]]
+name = "heck"
+version = "0.3.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6d621efb26863f0e9924c6ac577e8275e5e6b77455db64ffa6c65c904e9e132c"
+dependencies = [
+ "unicode-segmentation",
+]
+
+[[package]]
+name = "id-arena"
+version = "2.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "25a2bc672d1148e28034f176e01fffebb08b35768468cc954630da77a1449005"
+
 [[package]]
 name = "lazy_static"
 version = "1.4.0"
@@ -227,6 +242,15 @@ dependencies = [
  "rgb",
 ]
 
+[[package]]
+name = "log"
+version = "0.4.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "51b9bbe6c47d51fc3e1a9b945965946b4c44142ab8792c50835a980d362c2710"
+dependencies = [
+ "cfg-if",
+]
+
 [[package]]
 name = "miniz_oxide"
 version = "0.4.4"
@@ -286,6 +310,24 @@ version = "0.5.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "dbf0c48bc1d91375ae5c3cd81e3722dff1abcf81a30960240640d223f59fe0e5"
 
+[[package]]
+name = "proc-macro2"
+version = "1.0.36"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c7342d5883fbccae1cc37a2353b09c87c9b0f3afd73f5fb9bba687a1f733b029"
+dependencies = [
+ "unicode-xid",
+]
+
+[[package]]
+name = "quote"
+version = "1.0.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "864d3e96a899863136fc6e99f3d7cae289dafe43bf2c5ac19b70df7210c0a145"
+dependencies = [
+ "proc-macro2",
+]
+
 [[package]]
 name = "rgb"
 version = "0.8.31"
@@ -304,6 +346,17 @@ dependencies = [
  "num-traits",
 ]
 
+[[package]]
+name = "syn"
+version = "1.0.86"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8a65b3f4ffa0092e9887669db0eae07941f023991ab58ea44da8fe8e2d511c6b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-xid",
+]
+
 [[package]]
 name = "time"
 version = "0.1.43"
@@ -323,6 +376,18 @@ dependencies = [
  "crunchy",
 ]
 
+[[package]]
+name = "unicode-segmentation"
+version = "1.9.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7e8820f5d777f6224dc4be3632222971ac30164d4a258d595640799554ebfd99"
+
+[[package]]
+name = "unicode-xid"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8ccb82d61f80a663efe1f787a51b16b5a51e3314d6ac365b08639f52387b33f3"
+
 [[package]]
 name = "upkr"
 version = "0.1.0"
@@ -342,8 +407,9 @@ dependencies = [
  "pbr",
  "pico-args",
  "upkr",
- "wasm-encoder",
+ "walrus",
- "wasmparser",
+ "wasm-encoder 0.8.0",
+ "wasmparser 0.81.0",
 ]
 
 [[package]]
@@ -352,6 +418,32 @@ version = "0.9.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5fecdca9a5291cc2b8dcf7dc02453fee791a280f3743cb0905f8822ae463b3fe"
 
+[[package]]
+name = "walrus"
+version = "0.19.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4eb08e48cde54c05f363d984bb54ce374f49e242def9468d2e1b6c2372d291f8"
+dependencies = [
+ "anyhow",
+ "id-arena",
+ "leb128",
+ "log",
+ "walrus-macro",
+ "wasmparser 0.77.0",
+]
+
+[[package]]
+name = "walrus-macro"
+version = "0.19.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0a6e5bd22c71e77d60140b0bd5be56155a37e5bd14e24f5f87298040d0cc40d7"
+dependencies = [
+ "heck",
+ "proc-macro2",
+ "quote",
+ "syn",
+]
+
 [[package]]
 name = "wasi"
 version = "0.10.2+wasi-snapshot-preview1"
@@ -367,12 +459,33 @@ dependencies = [
  "leb128",
 ]
 
+[[package]]
+name = "wasm-encoder"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aa9d9bf45fc46f71c407837c9b30b1e874197f2dc357588430b21e5017d290ab"
+dependencies = [
+ "leb128",
+]
+
+[[package]]
+name = "wasmparser"
+version = "0.77.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b35c86d22e720a07d954ebbed772d01180501afe7d03d464f413bb5f8914a8d6"
+
 [[package]]
 name = "wasmparser"
 version = "0.81.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "98930446519f63d00a836efdc22f67766ceae8dbcc1571379f2bcabc6b2b9abc"
 
+[[package]]
+name = "wasmparser"
+version = "0.83.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "718ed7c55c2add6548cca3ddd6383d738cd73b892df400e96b9aa876f0141d7a"
+
 [[package]]
 name = "winapi"
 version = "0.3.9"

platform/Cargo.toml

@@ -6,7 +6,7 @@ edition = "2021"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-curlywas = { git="https://github.com/exoticorn/curlywas.git", rev="196719b" }
+curlywas = { git="https://github.com/exoticorn/curlywas.git", rev="0e7ea50" }
 uw8-tool = { path="../uw8-tool" }
 anyhow = "1"
 lodepng = "3.4"

platform/src/ges.cwa

@@ -0,0 +1,222 @@
+const GesChannelState.Trigger = 0;
+const GesChannelState.EnvState = 1;
+const GesChannelState.EnvVol = 2;
+const GesChannelState.Phase = 4;
+const GesChannelState.Size = 8;
+
+const GesState.Filter = GesChannelState.Size * 4;
+const GesState.Size = GesState.Filter + 8*4;
+
+const GesStateOffset = 32;
+const GesBufferOffset = 32 + GesState.Size;
+
+export fn gesSnd(t: i32) -> f32 {
+    let baseAddr = 0!0x12c78;
+    if !(t & 127) {
+        let i: i32;
+        loop clearLoop {
+            (baseAddr + i)!GesBufferOffset = 0;
+            branch_if (i := i + 4) < 128*4: clearLoop;
+        }
+
+        let ch: i32;
+        loop channelLoop {
+            let lazy channelState = baseAddr + GesStateOffset + ch * GesChannelState.Size;
+            let lazy channelReg = baseAddr + ch * 6;
+            let envState = channelState?GesChannelState.EnvState;
+            let envVol = i32.load16_u(channelState, GesChannelState.EnvVol);
+
+            let lazy oldTrigger = channelState?GesChannelState.Trigger;
+            let lazy ctrl = channelReg?0;
+            if (oldTrigger ^ ctrl) & (ctrl | 2) & 3 {
+                envState = 1;
+                envVol = 0;
+            }
+            channelState?GesChannelState.Trigger = ctrl;
+
+            if envState {
+                let lazy attack = channelReg?4 & 15;
+                envVol = envVol + 12 * pow(1.675, (15 - attack) as f32) as i32;
+                if envVol >= 65535 {
+                    envVol = 65535;
+                    envState = 0;
+                }
+            } else {
+                let inline decay = (channelReg - (ctrl & 1))?5 >> 4;
+                let inline dec = 8 * pow(1.5625, (15 - decay) as f32) as i32;
+                envVol = envVol - ((dec * (envVol + 8192)) >> 16);
+                let inline sustain = (channelReg?5 & 15) << 12;
+                let lazy targetVol = (ctrl & 1) * sustain;
+                if envVol < targetVol {
+                    envVol = targetVol;
+                }
+            }
+            channelState?GesChannelState.EnvState = envState;
+
+            i32.store16(envVol, channelState, GesChannelState.EnvVol);
+
+            let inline note = i32.load16_u(channelReg, 2);
+            let lazy freq = 440 as f32 * pow(2.0, (note - 69*256) as f32 / (12*256) as f32);
+            let phaseInc = (freq * (65536.0 / 44100.0)) as i32;
+
+            let phase = channelState!GesChannelState.Phase;
+
+            let inline pulseWidth = channelReg?1;
+            let phaseShift = (pulseWidth - 128) * 255;
+            let invPhaseInc = 1 as f32 / phaseInc as f32;
+
+            i = 0;
+            let wave = ctrl >> 6;
+            if wave < 2 {
+                if wave {
+                    let pulsePhase1 = pulseWidth << 23;
+                    let pulsePhase2 = (511 - pulseWidth) << 23;
+                    loop sawLoop {
+                        let p = (phase ^ 32768) << 16;
+                        let saw = (p >> 16) - polyBlep(phase, invPhaseInc, -32767);
+                        let saw2 = select(p #>= pulsePhase1 & p #< pulsePhase2, -saw, saw);
+                        let saw2 = saw2 -
+                            polyBlep((p - pulsePhase1) >> 16, invPhaseInc, -saw) -
+                            polyBlep((p - pulsePhase2) >> 16, invPhaseInc, saw);
+                        (baseAddr + i)!(GesBufferOffset + 128*4) = saw2;
+                        phase = phase + phaseInc;
+                        branch_if (i := i + 4) < 64*4: sawLoop;
+                    }
+                }
+                else
+                {
+                    let pulsePhase = 32768 + pulseWidth * 128;
+                    loop rectLoop {
+                        (baseAddr + i)!(GesBufferOffset + 128*4) = select((phase & 65535) < pulsePhase, -32768, 32767) -
+                            polyBlep(phase, invPhaseInc, -32767) -
+                            polyBlep(phase - pulsePhase, invPhaseInc, 32767);
+                        phase = phase + phaseInc;
+                        branch_if (i := i + 4) < 64*4: rectLoop;
+                    }
+                }
+            } else {
+                if wave == 2 {
+                    let scale = pulseWidth + 256;
+                    loop triLoop {
+                        let s = phase << 16;
+                        s = (s ^ (s >> 31));
+                        s = (s >> 8) * scale;
+                        s = (s ^ (s >> 31));
+                        (baseAddr + i)!(GesBufferOffset + 128*4) = (s >> 15) - 32768;
+                        phase = phase + phaseInc;
+                        branch_if (i := i + 4) < 64*4: triLoop;
+                    }
+                } else {
+                    loop noiseLoop {
+                        let s = phase >> 12;
+                        let inline pulse = ((phase >> 8) & 255) >= pulseWidth;
+                        s = s * 0x6746ba73;
+                        s = s ^ (s >> 15) * pulse;
+                        (baseAddr + i)!(GesBufferOffset + 128*4) = (s * 0x835776c7) >> 16;
+                        phase = phase + phaseInc;
+                        branch_if (i := i + 4) < 64*4: noiseLoop;
+                    }
+                }
+            }
+
+            channelState!GesChannelState.Phase = phase;
+
+            if ctrl & 32 {
+                let lazy modSrc = (ch - 1) & 3;
+                let inline channelState = baseAddr + GesStateOffset + modSrc * GesChannelState.Size;
+                let inline channelReg = baseAddr + modSrc * 6;
+
+                let inline note = i32.load16_u(channelReg, 2);
+                let inline freq = 440 as f32 * pow(2.0, (note - 69*256) as f32 / (12*256) as f32);
+                let phaseInc = (freq * (65536.0 / 44100.0)) as i32;
+
+                let phase = channelState!GesChannelState.Phase;
+                if modSrc > ch {
+                    phase = phase - (phaseInc << 6);
+                }
+
+                i = 0;
+                loop ringLoop {
+                    let s = phase << 16;
+                    s = (s ^ (s >> 31));
+                    (baseAddr + i)!(GesBufferOffset + 128*4) = ((baseAddr + i)!(GesBufferOffset + 128*4) * ((s >> 15) - 32768)) >> 15;
+                    phase = phase + phaseInc;
+                    branch_if (i := i + 4) < 64*4: ringLoop;
+                }
+            }
+
+            let channelVol = ((baseAddr + (ch >> 1))?24 >> ((ch & 1) * 4)) & 15;
+            envVol = envVol * channelVol / 15;
+
+            let leftVol = (select(ctrl & 16, 0x3d5b, 0x6a79) >> (ch * 4)) & 15;
+            let rightVol = 16 - leftVol;
+
+            let lazy filter = (ctrl >> 2) & 3;
+
+            i = 0;
+            if filter < 2 {
+                if filter {
+                    let f = (4096 as f32 - min(4096 as f32, 4096 as f32 * exp(freq * (-8.0 * 3.141 / 44100.0)))) as i32;
+                    let low = (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter);
+                    loop filterLoop {
+                        let in = ((baseAddr + i)!(GesBufferOffset + 128*4) * envVol) >> 18;
+                        low = low + (((in - low) * f) >> 12);
+                        (baseAddr + i * 2)!GesBufferOffset = (baseAddr + i * 2)!GesBufferOffset + ((low * leftVol) >> 4);
+                        (baseAddr + i * 2)!(GesBufferOffset + 4) = (baseAddr + i * 2)!(GesBufferOffset + 4) + ((low * rightVol) >> 4);
+                        branch_if (i := i + 4) < 64*4: filterLoop;
+                        (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter) = low;
+                        (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter + 4) = 0;
+                    }
+                } else {
+                    loop mixLoop {
+                        let sample = ((baseAddr + i)!(GesBufferOffset + 128*4) * envVol) >> 18;
+                        (baseAddr + i * 2)!GesBufferOffset = (baseAddr + i * 2)!GesBufferOffset + ((sample * leftVol) >> 4);
+                        (baseAddr + i * 2)!(GesBufferOffset + 4) = (baseAddr + i * 2)!(GesBufferOffset + 4) + ((sample * rightVol) >> 4);
+                        branch_if (i := i + 4) < 64*4: mixLoop;
+                        (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter) = sample;
+                        (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter + 4) = 0;
+                    }
+                }
+            } else {
+                filter = filter - 2;
+                let ctrl = (baseAddr + filter)?26;
+                let note = i32.load16_u(baseAddr + filter * 2, 28);
+                let inline freq = 440 as f32 * pow(2.0, (note - 69*256) as f32 / (12*256) as f32);
+                let F = (8192 as f32 * sin(min(0.25, freq / 44100 as f32) * 3.1415)) as i32;
+                let Q = 8192 - (ctrl >> 4) * (7000/15);
+                let Qlimit = (8192*4096/F - F/2) * 3 / 4;
+                if Q > Qlimit {
+                    Q = Qlimit;
+                }
+                let low_out = ctrl & 1;
+                let high_out = (ctrl >> 1) & 1;
+                let band_out = (ctrl >> 2) & 1;
+                let low = (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter);
+                let band = (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter + 4);
+                loop filterLoop {
+                    let in = ((baseAddr + i)!(GesBufferOffset + 128*4) * envVol) >> 18;
+
+                    let high = in - low - ((band * Q) >> 12);
+                    band = band + ((F * high) >> 12);
+                    low = low + ((F * band) >> 12);
+
+                    let sample = low * low_out + high * high_out + band * band_out;
+                    (baseAddr + i * 2)!GesBufferOffset = (baseAddr + i * 2)!GesBufferOffset + ((sample * leftVol) >> 4);
+                    (baseAddr + i * 2)!(GesBufferOffset + 4) = (baseAddr + i * 2)!(GesBufferOffset + 4) + ((sample * rightVol) >> 4);
+                    branch_if (i := i + 4) < 64*4: filterLoop;
+                    (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter) = low;
+                    (baseAddr + ch * 8)!(GesStateOffset + GesState.Filter + 4) = band;
+                }
+            }
+
+            branch_if (ch := ch + 1) < 4: channelLoop;
+        }
+    }
+    ((baseAddr + (t & 127) * 4)!GesBufferOffset) as f32 / 32768 as f32
+}
+
+fn polyBlep(transientPhase: i32, invPhaseInc: f32, magnitude: i32) -> i32 {
+    let lazy t = ((transientPhase << 16) >> 16) as f32 * invPhaseInc;
+    let lazy x = max(0 as f32, 1 as f32 - abs(t));
+    (f32.copysign(x * x, t) * magnitude as f32) as i32
+}

platform/src/ges_only.cwa

@@ -0,0 +1,6 @@
+import "env.memory" memory(1);
+import "env.sin" fn sin(f32) -> f32;
+import "env.pow" fn pow(f32, f32) -> f32;
+import "env.exp" fn exp(f32) -> f32;
+
+include "ges.cwa"

platform/src/main.rs

@@ -11,13 +11,13 @@ fn main() -> Result<()> {
     convert_font()?;
 
     println!("Compiling loader module");
-    let loader = curlywas::compile_file("src/loader.cwa", curlywas::Options::default())?;
+    let loader = curlywas::compile_file("src/loader.cwa", curlywas::Options::default()).0?;
     File::create("bin/loader.wasm")?.write_all(&loader)?;
 
     println!("Loader (including base module): {} bytes", loader.len());
 
     println!("Compiling platform module");
-    let platform = curlywas::compile_file("src/platform.cwa", curlywas::Options::default())?;
+    let platform = curlywas::compile_file("src/platform.cwa", curlywas::Options::default()).0?;
     println!("Compressing platform module");
     let platform = uw8_tool::pack(
         &platform,

platform/src/platform.cwa

@@ -1,6 +1,10 @@
 import "env.memory" memory(4);
 
+import "env.sin" fn sin(f32) -> f32;
 import "env.cos" fn cos(f32) -> f32;
+import "env.pow" fn pow(f32, f32) -> f32;
+import "env.exp" fn exp(f32) -> f32;
+import "env.logChar" fn logChar(i32);
 
 export fn time() -> f32 {
     (0!64) as f32 / 1000 as f32
@@ -29,11 +33,9 @@ export fn random() -> i32 {
 }
 
 export fn random64() -> i64 {
-    let state: i64;
+    let lazy state = randomState ^ (randomState #>> 12i64);
-    randomState = (state := (
+    let lazy state = state ^ (state << 25i64);
-                          state := randomState ^ (randomState #>> 12i64)
+    randomState = state ^ (state #>> 27i64);
-                      ) ^ (state << 25i64)
-                  ) ^ (state #>> 27i64);
     randomState * 0x2545f4914f6cdd1di64
 }
 
@@ -59,12 +61,8 @@ export fn cls(col: i32) {
     let i: i32;
     textCursorX = 0;
     textCursorY = 0;
-    graphicsText = 0;
+    outputChannel = 0;
-    col = (col & 255) * 0x1010101;
+    memory.fill(120, col, 320*240);
-    loop pixels {
-        i!120 = col;
-        branch_if (i := i + 4) < 320*240: pixels;
-    }
 }
 
 export fn setPixel(x: i32, y: i32, col: i32) {
@@ -88,12 +86,70 @@ fn clamp(v: i32, min: i32, max: i32) -> i32 {
 export fn hline(x1: i32, x2: i32, y: i32, col: i32) {
     x1 = clamp(x1, 0, 320);
     x2 = clamp(x2, 0, 320);
-    if x1 < x2 & y #< 240 {
+    if  y #>= 240 {
+        return;
+    }
+    let word_start = (x1 + 3) & -4;
+    let word_end = x2 & -4;
+    if word_end > word_start {
+        col = (col & 255) * 0x1010101;
+        let ptr = y * 320 + x1;
+        let end = ptr + word_start - x1;
+        if ptr + 2 <= end {
+            ptr?120 = col;
+            ptr?121 = col;
+            ptr += 2;
+        }
+        if ptr < end {
+            ptr?120 = col;
+            ptr += 1;
+        }
+        end += word_end - word_start;
+        loop words {
+            if ptr + 16 <= end {
+                ptr!120 = col;
+                ptr!124 = col;
+                ptr!128 = col;
+                ptr!132 = col;
+                ptr += 16;
+                branch words;
+            }
+            if ptr + 8 <= end {
+                ptr!120 = col;
+                ptr!124 = col;
+                ptr += 8;
+            }
+            if ptr < end {
+                ptr!120 = col;
+                ptr += 4;
+            }
+        }
+        end += x2 - word_end;
+        if ptr + 2 <= end {
+            ptr?120 = col;
+            ptr?121 = col;
+            ptr += 2;
+        }
+        if ptr < end {
+            ptr?120 = col;
+        }
+    } else {
         let ptr = y * 320 + x1;
         let end = ptr + x2 - x1;
-        loop pixels {
+        if ptr + 4 <= end {
+            ptr?120 = col;
+            ptr?121 = col;
+            ptr?122 = col;
+            ptr?123 = col;
+            ptr += 4;
+        }
+        if ptr + 2 <= end {
+            ptr?120 = col;
+            ptr?121 = col;
+            ptr += 2;
+        }
+        if ptr < end {
             ptr?120 = col;
-            branch_if (ptr := ptr + 1) < end: pixels;
         }
     }
 }
@@ -256,6 +312,11 @@ export fn line(x1: f32, y1: f32, x2: f32, y2: f32, col: i32) {
         p = y1;
     }
 
+    if max_axis == 0 as f32 {
+        setPixel(x1 as i32, y1 as i32, col);
+        return;
+    }
+
     let steps = floor(p + max_axis) as i32 - floor(p) as i32;
     p = floor(p) + 0.5 - p;
     if max_axis < 0 as f32 {
@@ -267,7 +328,7 @@ export fn line(x1: f32, y1: f32, x2: f32, y2: f32, col: i32) {
     dy = dy / max_axis;
 
     let f = min(max_axis, max(0 as f32, p));
-    setPixel((x1 + f * dx) as i32, (y1 + f * dy) as i32, col);
+    setPixel(i32.trunc_sat_f32_s(x1 + f * dx), i32.trunc_sat_f32_s(y1 + f * dy), col);
 
     if !steps {
         return;
@@ -280,7 +341,7 @@ export fn line(x1: f32, y1: f32, x2: f32, y2: f32, col: i32) {
 
     loop pixels {
         if steps := steps - 1 {
-            setPixel(x1 as i32, y1 as i32, col);
+            setPixel(i32.trunc_sat_f32_s(x1), i32.trunc_sat_f32_s(y1), col);
             x1 = x1 + dx;
             y1 = y1 + dy;
             branch pixels;
@@ -288,7 +349,7 @@ export fn line(x1: f32, y1: f32, x2: f32, y2: f32, col: i32) {
     }
 
     f = min(max_axis, p) - p;
-    setPixel((x1 + f * dx) as i32, (y1 + f * dy) as i32, col);
+    setPixel(i32.trunc_sat_f32_s(x1 + f * dx), i32.trunc_sat_f32_s(y1 + f * dy), col);
 }
 
 //////////
@@ -299,7 +360,7 @@ global mut textCursorX = 0;
 global mut textCursorY = 0;
 global mut textColor = 15;
 global mut bgColor = 0;
-global mut graphicsText = 0;
+global mut outputChannel = 0;
 
 export fn printChar(char: i32) {
     loop chars {
@@ -311,6 +372,20 @@ export fn printChar(char: i32) {
 global mut controlCodeLength = 0;
 
 fn printSingleChar(char: i32) {
+    if char >= 4 & char <= 6 {
+        outputChannel = char - 4;
+        if !outputChannel {
+            textCursorX = 0;
+            textCursorY = 0;
+        }
+        return;
+    }
+
+    if outputChannel >= 2 {
+        logChar(char);
+        return;
+    }
+
     controlCodeLength?0x12d20 = char;
     controlCodeLength = controlCodeLength + 1;
     char = 0x12d20?0;
@@ -324,16 +399,14 @@ fn printSingleChar(char: i32) {
         return;
     }
 
-    if char == 4 | char == 5 {
+    if char == 7 {
-        graphicsText = char == 5;
+        80?0 = 80?0 ^ 2;
-	    textCursorX = 0;
-        textCursorY = 0;
         return;
     }
 
     if char == 8 {
         textCursorX = textCursorX - 8;
-        if !graphicsText & textCursorX < 0 {
+        if !outputChannel & textCursorX < 0 {
             textCursorX = 320-8;
             printSingleChar(11);
         }
@@ -341,7 +414,7 @@ fn printSingleChar(char: i32) {
     }
 
     if char == 9 {
-        if !graphicsText & textCursorX >= 320 {
+        if !outputChannel & textCursorX >= 320 {
             printChar(0xd0a);
         }
         textCursorX = textCursorX + 8;
@@ -350,7 +423,7 @@ fn printSingleChar(char: i32) {
 
     if char == 10 {
         textCursorY = textCursorY + 8;
-        if !graphicsText & textCursorY >= 240 {
+        if !outputChannel & textCursorY >= 240 {
             textCursorY = 240 - 8;
             let i: i32;
             loop scroll_copy {
@@ -364,7 +437,7 @@ fn printSingleChar(char: i32) {
 
     if char == 11 {
         textCursorY = textCursorY - 8;
-        if !graphicsText & textCursorY < 0 {
+        if !outputChannel & textCursorY < 0 {
             textCursorY = 0;
             let i = 320 * (240 - 8);
             loop scroll_copy {
@@ -404,8 +477,8 @@ fn printSingleChar(char: i32) {
     }
 
     if char == 31 {
-        textCursorX = 0x12d20?1 * (8 - graphicsText * 6);
+        textCursorX = 0x12d20?1 * (8 - outputChannel * 6);
-        textCursorY = 0x12d20?2 * (8 - graphicsText * 7);
+        textCursorY = 0x12d20?2 * (8 - outputChannel * 7);
         return;
     }
 
@@ -430,7 +503,7 @@ data(0x12d00) {
 }
 
 fn drawChar(char: i32) {
-    if !graphicsText & textCursorX >= 320 {
+    if !outputChannel & textCursorX >= 320 {
         printChar(0xd0a);
     }
 
@@ -438,7 +511,7 @@ fn drawChar(char: i32) {
     loop rows {
         let bits = (char * 8 + y)?0x13400;
         let x = 0;
-        if graphicsText {
+        if outputChannel {
             loop pixels {
                 if (bits := bits << 1) & 256 {
                     setPixel(textCursorX + x, textCursorY + y, textColor);
@@ -490,11 +563,37 @@ export fn setBackgroundColor(col: i32) {
 }
 
 export fn setCursorPosition(x: i32, y: i32) {
-    let lazy scale = select(graphicsText, 1, 8);
+    let lazy scale = select(outputChannel, 1, 8);
     textCursorX = x * scale;
     textCursorY = y * scale;
 }
 
+///////////
+// SOUND //
+///////////
+
+include "ges.cwa"
+
+export fn playNote(channel: i32, note: i32) {
+    (channel * 6)?80 = (channel * 6)?80 & 0xfe ^ if note {
+        (channel * 6)?83 = note & 127;
+        2 | !(note >> 7)
+    } else {
+        0
+    };
+}
+
+data 80 {
+    i8(
+        0x80, 0xc0, 0, 81, 0xa0, 0x50,
+        0xc4, 0, 0, 69, 0x60, 0x40,
+        0x44, 0xb0, 0, 69, 0x90, 0x43,
+        0x4, 0xf0, 0, 69, 0xa4, 0x44,
+        0xff, 0xff,
+        1, 1, 0, 100, 0, 100
+    )
+}
+
 ///////////
 // SETUP //
 ///////////
@@ -535,10 +634,19 @@ start fn setup() {
         branch_if (i := i - 1) >= 0: expand_sweetie;
     }
 
+    memory.fill(0, 0, 64);
+    memory.fill(112, 0, 8);
+    memory.fill(0x14000, 0, 0x2c000);
+
+
     cls(0);
     randomSeed(random());
 }
 
+data 0x12c78 {
+    i32(80)
+}
+
 data 0x13000+192*4 {
     i32(
         0x2c1c1a,

release/make-release

@@ -36,6 +36,8 @@ for dir in build/*; do
         cp $example $dir/examples
     done
 
+    cp -r ../examples/include $dir/include
+
     mkdir $dir/carts
     for example in $dir/examples/*; do
         build/microw8-linux/uw8 pack -l 9 $example $dir/carts/$(basename ${example%.*}).uw8

site/content/_index.md

@@ -15,11 +15,11 @@ The initial motivation behind MicroW8 was to explore whether there was a way to
 * Gamepad input (D-Pad + 4 Buttons)
 
 ## Examples
-* [Fireworks](v0.1.1#AgwvgP+M59snqjl4CMKw5sqm1Zw9yJCbSviMjeLUdHus2a3yl/a99+uiBeqZgP/2jqSjrLjRk73COMM6OSLpsxK8ugT1kuk/q4hQUqqPpGozHoa0laulzGGcahzdfdJsYaK1sIdeIYS9M5PnJx/Wk9H+PvWEPy2Zvv7I6IW7Fg==) (127 bytes): Some fireworks to welcome 2022.
+* [Skip Ahead](v0.2.0#AgVfq24KI2Ok2o8qVtPYj27fSuGnfeSKgbOkIOsaEQMov8TDYQ6UjdjwkZrYcM1i9alo4/+Bhm1PRFEa0YHJlJAk/PGoc2K41rejv9ZSqJqIHNjr7cappqhOR2jT+jk+0b0+U6hO+geRCTP2aufWs7L+f/Z27NFY8LKlqPSv+C6Rd6+ohoKi6sYl5Kcrlf1cyTinV7jTTnmbcXWVDBA5rRKxAGMUTDS8rHxqSztRITOaQVP1pSdYgi/BDdOJOxSOIkeaId84S+Ycls5na7EgwSfVIpgqF+tcfkUecb8t2mQrXA7pyKrh/wzHn5N6Oe5aOgmzY2YpTIct) (249 bytes): A port of my [TIC-80 256byte game](http://tic80.com/play?cart=1735) from LoveByte'21, now with sound
-* [Skip Ahead](v0.1.1#AgyfpZ80wkW28kiUZ9VIK4v+RPnVxqjK1dz2BcDoNyQPsS2g4OgEzkTe6jyoAfFOmqKrS8SM2aRljBal9mjNn8i4fP9eBK+RehQKxxGtJa9FqftvqEnh3ez1YaYxqj7jgTdzJ/WAYVmKMovBT1myrX3FamqKSOgMsNedLhVTLAhQup3sNcYEjGNo8b0HZ5+AgMgCwYRGCe//XQOMAaAAzqDILgmpEZ/43RKHcQpHEQwbURfNQJpadJe2sz3q5FlQnTGXQ9oSMokidhlC+aR/IpNHieuBGLhFZ2GfnwVQ0geBbQpTPA==) (229 bytes): A port of my [TIC-80 256byte game](http://tic80.com/play?cart=1735) from LoveByte'21
+* [Fireworks](v0.2.0#AgwvgP+M59snqjl4CMKw5sqm1Zw9yJCbSviMjeLUdHus2a3yl/a99+uiBeqZgP/2jqSjrLjRk73COMM6OSLpsxK8ugT1kuk/q4hQUqqPpGozHoa0laulzGGcahzdfdJsYaK1sIdeIYS9M5PnJx/Wk9H+PvWEPy2Zvv7I6IW7Fg==) (127 bytes): Some fireworks to welcome 2022.
-* [OhNoAnotherTunnel](v0.1.1#AgPP1oEFvPzY/rBZwTumtYn37zeMFgpir1Bkn91jsNcp26VzoUpkAOOJTtnzVBfW+/dGnnIdbq/irBUJztY5wuua80DORTYZndgdwZHcSk15ajc4nyO0g1A6kGWyW56oZk0iPYJA9WtUmoj0Plvy1CGwIZrMe57X7QZcdqc3u6zjTA41Tpiqi9vnO3xbhi8o594Vx0XPXwVzpYq1ZCTYenfAGaXKkDmAFJqiVIsiCg==) (175 bytes): A port of my [entry](http://tic80.com/play?cart=1871) in the Outline'21 bytebattle final
+* [OhNoAnotherTunnel](v0.2.0#AgPP1oEFvPzY/rBZwTumtYn37zeMFgpir1Bkn91jsNcp26VzoUpkAOOJTtnzVBfW+/dGnnIdbq/irBUJztY5wuua80DORTYZndgdwZHcSk15ajc4nyO0g1A6kGWyW56oZk0iPYJA9WtUmoj0Plvy1CGwIZrMe57X7QZcdqc3u6zjTA41Tpiqi9vnO3xbhi8o594Vx0XPXwVzpYq1ZCTYenfAGaXKkDmAFJqiVIsiCg==) (175 bytes): A port of my [entry](http://tic80.com/play?cart=1871) in the Outline'21 bytebattle final
-* [Technotunnel](v0.1.1#AhPXpq894LaUhp5+HQf39f39/Jc8g5zUrBSc0uyKh36ivskczhY84h55zL8gWpkdvKuRQI+KIt80isKzh8jkM8nILcx0RUvyk8yjE8TgNsgkcORVI0RY5k3qE4ySjaycxa2DVZH61UWZuLsCouuwT7I80TbmmetQSbMywJ/avrrCZIAH0UzQfvOiCJNG48NI0FFY1vjB7a7dcp8Uqg==) (157 bytes): A port of my [entry](https://tic80.com/play?cart=1873) in the Outline'21 bytebattle quater final
+* [Technotunnel](v0.2.0#AhPXpq894LaUhp5+HQf39f39/Jc8g5zUrBSc0uyKh36ivskczhY84h55zL8gWpkdvKuRQI+KIt80isKzh8jkM8nILcx0RUvyk8yjE8TgNsgkcORVI0RY5k3qE4ySjaycxa2DVZH61UWZuLsCouuwT7I80TbmmetQSbMywJ/avrrCZIAH0UzQfvOiCJNG48NI0FFY1vjB7a7dcp8Uqg==) (157 bytes): A port of my [entry](https://tic80.com/play?cart=1873) in the Outline'21 bytebattle quater final
-* [Font & Palette](v0.1.1#At/p39+IBnj6ry1TRe7jzVy2A4tXgBvmoW2itzoyF2aM28pGy5QDiKxqrk8l9sbWZLtnAb+jgOfU+9QhpuyCAkhN6gPOU481IUL/df96vNe3h288Dqwhd3sfFpothIVFsMwRK72kW2hiR7zWsaXyy5pNmjR6BJk4piWx9ApT1ZwoUajhk6/zij6itq/FD1U3jj/J3MOwqZ2ef8Bv6ZPQlJIYVf62icGa69wS6SI1qBpIFiF14F8PcztRVbKIxLpT4ArCS6nz6FPnyUkqATGSBNPJ): Just a simple viewer for the default font and palette.
+* [Font & Palette](v0.2.0#At/p39+IBnj6ry1TRe7jzVy2A4tXgBvmoW2itzoyF2aM28pGy5QDiKxqrk8l9sbWZLtnAb+jgOfU+9QhpuyCAkhN6gPOU481IUL/df96vNe3h288Dqwhd3sfFpothIVFsMwRK72kW2hiR7zWsaXyy5pNmjR6BJk4piWx9ApT1ZwoUajhk6/zij6itq/FD1U3jj/J3MOwqZ2ef8Bv6ZPQlJIYVf62icGa69wS6SI1qBpIFiF14F8PcztRVbKIxLpT4ArCS6nz6FPnyUkqATGSBNPJ): Just a simple viewer for the default font and palette.
 
 Examplers for older versions:
 
@@ -29,27 +29,24 @@ Examplers for older versions:
 
 ## Versions
 
-### v0.1.1
+### v0.2.0
 
-* [Web runtime](v0.1.1)
+* [Web runtime](v0.2.0)
-* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-linux.tgz)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-linux.tgz)
-* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-macos.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-macos.tgz)
-* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-windows.zip)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-windows.zip)
 
 Changes:
 
-* implement more robust file watcher
+* [add sound support!](docs#sound)
-* add basic video recording on F10 in web runtime
+* add support to redirect text output to the console for debugging using control code 6
-* add screenshot on F9
+* update curlywas:
-* add watchdog to interrupt hanging update in native runtime
+  * add support for `else if`
-* add devkit mode to web runtime
+  * add support for escape sequences in strings
-* add unpack and compile commands to uw8
+  * add support for char literals
-* add support for table/element section in pack command
+  * add support for binop-assignment, eg. `+=`, `^=`, `<<=` etc. (also support for the tee operator: `+:=`)
-* disable wayland support (caused missing window decorations in gnome)
+  * "integer constant cast to float" literal syntax in CurlyWas (ex. `1_f` is equivalent to `1 as f32`)
 
-### v0.1.0
+### Older versions
 
-* [Web runtime](v0.1.0)
+[Find older versions here.](versions)
-* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-linux.tgz)
-* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-macos.tgz)
-* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-windows.zip)

site/content/docs.md

@@ -18,9 +18,12 @@ The memory has to be imported as `env` `memory` and has a maximum size of 256kb
 00000-00040: user memory
 00040-00044: time since module start in ms
 00044-0004c: gamepad state
-0004c-00078: reserved
+0004c-00050: reserved
+00050-00070: sound data (synced to sound thread)
+00070-00078: reserved
 00078-12c78: frame buffer
-12c78-13000: reserved
+12c78-12c7c: sound registers/work area base address (for sndGes function)
+12c7c-13000: reserved
 13000-13400: palette
 13400-13c00: font
 13c00-14000: reserved
@@ -47,7 +50,7 @@ Returns the arccosine of `x`.
 
 Returns the arctangent of `x`.
 
-### fn atan2(y: f32, y: f32) -> f32
+### fn atan2(y: f32, x: f32) -> f32
 
 Returns the angle between the point `(x, y)` and the positive x-axis.
 
@@ -222,29 +225,45 @@ Moving/printing past any border does not cause any special operation, the cursor
 Characters 0-31 are control characters and don't print by default. They take the next 0-2 following characters as parameters.
 Avoid the reserved control chars, they are currently NOPs but their behavior can change in later MicroW8 versions.
 
-| Code  | Parameters | Operation                            |
+| Code  | Parameters | Operation                                  |
-| ----- | ---------- | ------------------------------------ |
+| ----- | ---------- | ------------------------------------------ |
-| 0     | -          | Nop                                  |
+| 0     | -          | Nop                                        |
-| 1     | char       | Print char (including control chars) |
+| 1     | char       | Print char (including control chars)       |
-| 2-3   | -          | Reserved                             |
+| 2-3   | -          | Reserved                                   |
-| 4     | -          | Switch to normal mode                |
+| 4     | -          | Switch to normal mode, reset cursor to 0,0 |
-| 5     | -          | Switch to graphics mode              |
+| 5     | -          | Switch to graphics mode                    |
-| 6-7   | -          | Reserved                             |
+| 6     | -          | Switch output to (debug) console           |
-| 8     | -          | Move cursor left                     |
+| 7     | -          | Bell / trigger sound channel 0             |
-| 9     | -          | Move cursor right                    |
+| 8     | -          | Move cursor left                           |
-| 10    | -          | Move cursor down                     |
+| 9     | -          | Move cursor right                          |
-| 11    | -          | Move cursor up                       |
+| 10    | -          | Move cursor down                           |
-| 12    | -          | do `cls(background_color)`           |
+| 11    | -          | Move cursor up                             |
-| 13    | -          | Move cursor to the left border       |
+| 12    | -          | do `cls(background_color)`                 |
-| 14    | color      | Set the background color             |
+| 13    | -          | Move cursor to the left border             |
-| 15    | color      | Set the text color                   |
+| 14    | color      | Set the background color                   |
-| 16-23 | -          | Reserved                             |
+| 15    | color      | Set the text color                         |
-| 24    | -          | Swap text/background colors          |
+| 16-23 | -          | Reserved                                   |
-| 25-30 | -          | Reserved                             |
+| 24    | -          | Swap text/background colors                |
-| 31    | x, y       | Set cursor position (*)              |
+| 25-30 | -          | Reserved                                   |
+| 31    | x, y       | Set cursor position (*)                    |
 
 (*) In graphics mode, the x coordinate is doubled when using control char 31 to be able to cover the whole screen with one byte.
 
+#### Debug output
+
+Control code 6 switches all text output (except codes 4 and 5 to switch output back to the screen) to the console. Where exactly this ends
+up (if at all) is an implementation detail of the runtimes. The native dev-runtime writes the debug output to `stdout`, the web runtime to
+the debug console using `console.log`. Both implementation buffer the output until they encounter a newline character (10) in the output stream.
+
+There may be future runtimes that ignore the debug output completely.
+
+In CurlyWas, a simple way to log some value might look like this:
+```
+printChar('\06V: '); // switch to console out, print some prefix
+printInt(some_value);
+printChar('\n\4'); // newline and switch back to screen
+```
+
 ### fn printChar(char: i32)
 
 Prints the character in the lower 8 bits of `char`. If the upper 24 bits are non-zero, right-shifts `char` by 8 bits and loops back to the beginning.
@@ -269,6 +288,130 @@ Sets the background color.
 
 Sets the cursor position. In normal mode `x` and `y` are multiplied by 8 to get the pixel position, in graphics mode they are used as is.
 
+## Sound
+
+### Low level operation
+
+MicroW8 actually runs two instances of your module. On the first instance, it calls `upd` and displays the framebuffer found in its memory. On the
+second instance, it calls `snd` instead with an incrementing sample index and expects that function to return sound samples for the left and right
+channel at 44100 Hz. If your module does not export a `snd` function, it calls the api function `sndGes` instead.
+
+As the only means of communication, 32 bytes starting at address 0x00050 are copied from main to sound memory after `upd` returns.
+
+By default, the `sndGes` function generates sound based on the 32 bytes at 0x00050. This means that in the default configuration those 32 bytes act
+as sound registers. See the `sndGes` function for the meaning of those registers.
+
+### export fn snd(sampleIndex: i32) -> f32
+
+If the module exports a `snd` function, it is called 88200 times per second to provide PCM sample data for playback (44.1kHz stereo).
+The `sampleIndex` will start at 0 and increments by 1 for each call. On even indices the function is expected to return a sample value for
+the left channel, on odd indices for the right channel.
+
+### fn playNote(channel: i32, note: i32)
+
+Triggers a note (1-127) on the given channel (0-3). Notes are semitones with 69 being A4 (same as MIDI). A note value of 0 stops the
+sound playing on that channel. A note value 128-255 will trigger note-128 and immediately stop it (playing attack+release parts of envelope).
+
+This function assumes the default setup, with the `sndGes` registers located at 0x00050.
+
+### fn sndGes(sampleIndex: i32) -> f32
+
+This implements a sound chip, generating sound based on 32 bytes of sound registers.
+
+The spec of this sound chip are:
+
+- 4 channels with individual volume control (0-15)
+- rect, saw, tri, noise wave forms selectable per channel
+- each wave form supports some kind of pulse width modulation
+- each channel has an optional automatic low pass filter, or can be sent to one of two manually controllable filters
+- each channel can select between a narrow and a wide stereo positioning. The two stereo positions of each channel are fixed.
+- optional ring modulation
+
+This function requires 1024 bytes of working memory, the first 32 bytes of which are interpreted as the sound registers.
+The base address of its working memory can be configured by writing the address to 0x12c78. It defaults to 0x00050.
+
+Here is a short description of the 32 sound registers.
+
+```
+00 - CTRL0
+06 - CTRL1
+0c - CTRL2
+12 - CTRL3
+  | 7  6 |   5  |   4  |  3  2  |    1    |    0    |
+  | wave | ring | wide | filter | trigger | note on |
+
+  note on: stay in decay/sustain part of envelope
+  trigger: the attack part of the envlope is triggered when either this changes
+           or note on is changed from 0 to 1.
+  filter : 0 - no filter
+           1 - fixed 6db 1-pole filter with cutoff two octaves above note
+           2 - programmable filter 0
+           3 - programmable filter 1
+  wide   : use wide stereo panning
+  ring   : ring modulate with triangle wave at frequency of previous channel
+  wave   : 0 - rectangle
+           1 - saw
+           2 - triangle
+           3 - noise
+
+01 - PULS0
+07 - PULS1
+0d - PULS2
+13 - PULS3
+  Pulse width 0-255, with 0 being the plain version of each wave form.
+  rectangle - 50%-100% pulse width
+  saw       - inverts 0%-100% of the saw wave form around the center
+  triangle  - morphs into an octave up triangle wave
+  noise     - blends into a decimated saw wave (just try it out)
+
+02 - FINE0
+08 - FINE1
+0e - FINE2
+14 - FINE3
+  Fractional note
+
+03 - NOTE0
+09 - NOTE1
+0f - NOTE2
+15 - NOTE3
+  Note, 69 = A4
+
+04 - ENVA0
+0a - ENVA1
+10 - ENVA2
+16 - ENVA3
+  | 7 6 5 4 | 3 2 1 0 |
+  | decay   | attack  |
+
+05 - ENVB0
+0b - ENVB1
+11 - ENVB2
+17 - ENVB3
+  | 7 6 5 4 | 3 2 1 0 |
+  | release | sustain |
+
+18 - VO01
+  | 7 6 5 4  | 3 2 1 0  |
+  | volume 1 | volume 0 |
+
+19 - VO23
+  | 7 6 5 4  | 3 2 1 0  |
+  | volume 3 | volume 2 |
+
+1a - FCTR0
+1b - FCTR1
+  | 7 6 5 4   | 3 | 2    | 1    | 0   |
+  | resonance | 0 | band | high | low |
+
+1c - FFIN0
+1e - FFIN1
+  cutoff frequency - fractional note
+
+1d - FNOT0
+1f - FNOT1
+  cutoff frequency - note
+```
+
 # The `uw8` tool
 
 The `uw8` tool included in the MicroW8 download includes a number of useful tools for developing MicroW8 carts. For small productions written in
@@ -284,7 +427,8 @@ Runs `<file>` which can be a binary WebAssembly module, an `.uw8` cart, a wat (W
 
 Options:
 
-* `-t FRAMES`, `--timeout FRAMES` : Sets the timeout in frames (1/60s). If the start or update function runs longer than this it is forcibly interupted
+* `-b`, `--browser`: Run in browser instead of using native runtime
+* `-t FRAMES`, `--timeout FRAMES`: Sets the timeout in frames (1/60s). If the start or update function runs longer than this it is forcibly interupted
 and execution of the cart is stopped. Defaults to 30 (0.5s)
 * `-w`, `--watch`: Reloads the given file every time it changes on disk.
 * `-p`, `--pack`: Pack the file into an `.uw8` cart before running it and print the resulting size.
@@ -292,6 +436,30 @@ and execution of the cart is stopped. Defaults to 30 (0.5s)
 * `-l LEVEL`, `--level LEVEL`: Compression level (0-9). Higher compression levels are really slow.
 * `-o FILE`, `--output FILE`: Write the loaded and optionally packed cart back to disk.
 
+when using the native runtime:
+
+* `-m`, `--no-audio`: Disable audio, also reduces cpu load a bit
+* `--no-gpu`:  Force old cpu-only window code
+* `--filter FILTER`:  Select an upscale filter at startup
+* `--fullscreen`:  Start in fullscreen mode
+
+Note that the cpu-only window does not support fullscreen nor upscale filters.
+
+Unless --no-gpu is given, uw8 will first try to open a gpu accelerated window, falling back to the old cpu-only window if that fails.
+Therefore you should rarely need to manually pass --no-gpu. If you prefer the old pixel doubling look to the now default crt filter,
+you can just pass `--filter nearest` or `--filter 1`.
+
+The upscale filter options are:
+```
+1, nearest              : Anti-aliased nearest filter
+2, fast_crt             : Very simple, cheap crt filter, not very good below a window size of 960x720
+3, ss_crt               : Super sampled crt filter, a little more demanding on the GPU but scales well to smaller window sizes
+4, chromatic_crt        : Variant of fast_crt with a slight offset of the three color dots of a pixel, still pretty cheap
+5, auto_crt (default)   : ss_crt below 960x720, chromatic_crt otherwise
+```
+
+You can switch the upscale filter at any time using the keys 1-5. You can toggle fullscreen with F.
+
 ## `uw8 pack`
 
 Usage:

site/content/versions.md

@@ -0,0 +1,105 @@
++++
+description = "Versions"
++++
+
+### v0.2.0
+
+* [Web runtime](v0.2.0)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.2.0/microw8-0.2.0-windows.zip)
+
+Changes:
+
+* [add sound support!](docs#sound)
+* add support to redirect text output to the console for debugging using control code 6
+* update curlywas:
+  * add support for `else if`
+  * add support for escape sequences in strings
+  * add support for char literals
+  * add support for binop-assignment, eg. `+=`, `^=`, `<<=` etc. (also support for the tee operator: `+:=`)
+  * "integer constant cast to float" literal syntax in CurlyWas (ex. `1_f` is equivalent to `1 as f32`)
+
+### v0.2.0-rc3
+
+* [Web runtime](v0.2.0-rc3)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc3/microw8-0.2.0-rc3-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc3/microw8-0.2.0-rc3-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc3/microw8-0.2.0-rc3-windows.zip)
+
+Changes:
+
+* improve timing stability some more. essentially now guaranteeing that "frame = time_ms * 6 / 100" returns
+  consecutive frame numbers, provided the module can be run at 60 fps
+* add support to redirect text output to the console for debugging using control code 6
+* update curlywas:
+  * add support for `else if`
+  * add support for escape sequences in strings
+  * add support for char literals
+  * add support for binop-assignment, eg. `+=`, `^=`, `<<=` etc. (also support for the tee operator: `+:=`)
+
+### v0.2.0-rc2
+
+* [Web runtime](v0.2.0-rc2)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc2/microw8-0.2.0-rc2-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc2/microw8-0.2.0-rc2-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc2/microw8-0.2.0-rc2-windows.zip)
+
+Changes:
+
+* fix timing issues of sound playback, especially on systems with large sound buffers
+
+### v0.2.0-rc1
+
+* [Web runtime](v0.2.0-rc1)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc1/microw8-0.2.0-rc1-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc1/microw8-0.2.0-rc1-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.2.0-rc1/microw8-0.2.0-rc1-windows.zip)
+
+Changes:
+
+* [add sound support](docs#sound)
+* "integer constant cast to float" literal syntax in CurlyWas (ex. `1_f` is equivalent to `1 as f32`)
+
+Known issues:
+
+* timing accuracy/update frequency of sound support currently depends on sound buffer size
+
+### v0.1.2
+
+* [Web runtime](v0.1.2)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.2/microw8-0.1.2-windows.zip)
+
+Changes:
+
+* add option to `uw8 run` to run the cart in the browser using the web runtime
+* CurlyWas: implement `include` support
+* CurlyWas: implement support for constants
+* fix crash when trying to draw zero sized line
+
+### v0.1.1
+
+* [Web runtime](v0.1.1)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.1/microw8-0.1.1-windows.zip)
+
+Changes:
+
+* implement more robust file watcher
+* add basic video recording on F10 in web runtime
+* add screenshot on F9
+* add watchdog to interrupt hanging update in native runtime
+* add devkit mode to web runtime
+* add unpack and compile commands to uw8
+* add support for table/element section in pack command
+* disable wayland support (caused missing window decorations in gnome)
+
+### v0.1.0
+
+* [Web runtime](v0.1.0)
+* [Linux](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-linux.tgz)
+* [MacOS](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-macos.tgz)
+* [Windows](https://github.com/exoticorn/microw8/releases/download/v0.1.0/microw8-0.1.0-windows.zip)

site/templates/index.html

@@ -4,7 +4,7 @@
         <section>
             <h1 class="text-center heading-text">A WebAssembly based fantasy console</h1>
         </section>
-        <a href="v0.1.1">
+        <a href="v0.2.0">
             <img class="demonstration-gif" style="width:640px;height:480px;image-rendering:pixelated" src="img/technotunnel.png"></img>
         </a>
     </div>

src/filewatcher.rs

@@ -1,23 +1,38 @@
-use anyhow::{bail, Result};
+use anyhow::{anyhow, bail, Result};
-use notify::{DebouncedEvent, Watcher, RecommendedWatcher};
+use notify::{DebouncedEvent, RecommendedWatcher, Watcher};
-use std::{
+use std::{collections::BTreeSet, path::PathBuf, sync::mpsc, time::Duration};
-    collections::BTreeSet,
-    path::{Path, PathBuf},
-    sync::mpsc,
-    time::Duration,
-};
 
 pub struct FileWatcher {
-    _watcher: RecommendedWatcher,
+    watcher: RecommendedWatcher,
     watched_files: BTreeSet<PathBuf>,
+    directories: BTreeSet<PathBuf>,
     rx: mpsc::Receiver<DebouncedEvent>,
 }
 
-pub struct FileWatcherBuilder(BTreeSet<PathBuf>);
-
 impl FileWatcher {
-    pub fn builder() -> FileWatcherBuilder {
+    pub fn new() -> Result<FileWatcher> {
-        FileWatcherBuilder(BTreeSet::new())
+        let (tx, rx) = mpsc::channel();
+        let watcher = notify::watcher(tx, Duration::from_millis(100))?;
+        Ok(FileWatcher {
+            watcher,
+            watched_files: BTreeSet::new(),
+            directories: BTreeSet::new(),
+            rx,
+        })
+    }
+
+    pub fn add_file<P: Into<PathBuf>>(&mut self, path: P) -> Result<()> {
+        let path = path.into();
+        let parent = path.parent().ok_or_else(|| anyhow!("File has no parent"))?;
+
+        if !self.directories.contains(parent) {
+            self.watcher
+                .watch(parent, notify::RecursiveMode::NonRecursive)?;
+            self.directories.insert(parent.to_path_buf());
+        }
+
+        self.watched_files.insert(path);
+        Ok(())
     }
 
     pub fn poll_changed_file(&self) -> Result<Option<PathBuf>> {
@@ -38,33 +53,3 @@ impl FileWatcher {
         Ok(None)
     }
 }
-
-impl FileWatcherBuilder {
-    pub fn add_file<P: Into<PathBuf>>(&mut self, path: P) -> &mut Self {
-        self.0.insert(path.into());
-        self
-    }
-
-    pub fn build(self) -> Result<FileWatcher> {
-        let mut directories: BTreeSet<&Path> = BTreeSet::new();
-
-        for file in &self.0 {
-            if let Some(directory) = file.parent() {
-                directories.insert(directory);
-            }
-        }
-
-        let (tx, rx) = mpsc::channel();
-        let mut watcher = notify::watcher(tx, Duration::from_millis(100))?;
-
-        for directory in directories {
-            watcher.watch(directory, notify::RecursiveMode::NonRecursive)?;
-        }
-
-        Ok(FileWatcher {
-            _watcher: watcher,
-            watched_files: self.0,
-            rx,
-        })
-    }
-}

src/lib.rs

@@ -14,9 +14,6 @@ use anyhow::Result;
 
 pub trait Runtime {
     fn is_open(&self) -> bool;
-    fn set_timeout(&mut self, _timeout: u32) {
-        eprintln!("Warning: runtime doesn't support timeout");
-    }
     fn load(&mut self, module_data: &[u8]) -> Result<()>;
     fn run_frame(&mut self) -> Result<()>;
 }

src/main.rs

@@ -13,8 +13,13 @@ use uw8::RunWebServer;
 use uw8::Runtime;
 
 fn main() -> Result<()> {
+    env_logger::Builder::from_env(env_logger::Env::default()).init();
     let mut args = Arguments::from_env();
 
+    // try to enable ansi support in win10 cmd shell
+    #[cfg(target_os = "windows")]
+    let _ = ansi_term::enable_ansi_support();
+
     match args.subcommand()?.as_deref() {
         Some("version") => {
             println!("{}", env!("CARGO_PKG_VERSION"));
@@ -48,6 +53,7 @@ fn main() -> Result<()> {
 #[cfg(any(feature = "native", feature = "browser"))]
 fn run(mut args: Arguments) -> Result<()> {
     let watch_mode = args.contains(["-w", "--watch"]);
+    #[allow(unused)]
     let timeout: Option<u32> = args.opt_value_from_str(["-t", "--timeout"])?;
 
     let mut config = Config::default();
@@ -75,23 +81,36 @@ fn run(mut args: Arguments) -> Result<()> {
     #[cfg(not(feature = "native"))]
     let run_browser = args.contains(["-b", "--browser"]) || true;
 
+    let disable_audio = args.contains(["-m", "--no-audio"]);
+
+    #[cfg(feature = "native")]
+    let window_config = {
+        let mut config = WindowConfig::default();
+        if !run_browser {
+            config.parse_arguments(&mut args);
+        }
+        config
+    };
+
     let filename = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
 
-    let mut watcher = uw8::FileWatcher::builder();
+    let mut watcher = uw8::FileWatcher::new()?;
-
-    if watch_mode {
-        watcher.add_file(&filename);
-    }
-
-    let watcher = watcher.build()?;
 
     use std::process::exit;
 
+    use uw8_window::WindowConfig;
+
     let mut runtime: Box<dyn Runtime> = if !run_browser {
         #[cfg(not(feature = "native"))]
         unimplemented!();
         #[cfg(feature = "native")]
-        Box::new(MicroW8::new()?)
+        {
+            let mut microw8 = MicroW8::new(timeout, window_config)?;
+            if disable_audio {
+                microw8.disable_audio();
+            }
+            Box::new(microw8)
+        }
     } else {
         #[cfg(not(feature = "browser"))]
         unimplemented!();
@@ -99,22 +118,23 @@ fn run(mut args: Arguments) -> Result<()> {
         Box::new(RunWebServer::new())
     };
 
-    if let Some(timeout) = timeout {
+    let mut first_run = true;
-        runtime.set_timeout(timeout);
-    }
-
-    if let Err(err) = start_cart(&filename, &mut *runtime, &config) {
-        eprintln!("Load error: {}", err);
-        if !watch_mode {
-            exit(1);
-        }
-    }
 
     while runtime.is_open() {
-        if watcher.poll_changed_file()?.is_some() {
+        if first_run || watcher.poll_changed_file()?.is_some() {
-            if let Err(err) = start_cart(&filename, &mut *runtime, &config) {
+            let (result, dependencies) = start_cart(&filename, &mut *runtime, &config);
-                eprintln!("Load error: {}", err);
+            if watch_mode {
+                for dep in dependencies {
+                    watcher.add_file(dep)?;
+                }
             }
+            if let Err(err) = result {
+                eprintln!("Load error: {}", err);
+                if !watch_mode {
+                    exit(1);
+                }
+            }
+            first_run = false;
         }
 
         if let Err(err) = runtime.run_frame() {
@@ -134,40 +154,103 @@ struct Config {
     output_path: Option<PathBuf>,
 }
 
-fn load_cart(filename: &Path, config: &Config) -> Result<Vec<u8>> {
+fn load_cart(filename: &Path, config: &Config) -> (Result<Vec<u8>>, Vec<PathBuf>) {
-    let mut cart = vec![];
+    let mut dependencies = Vec::new();
-    File::open(filename)?.read_to_end(&mut cart)?;
+    fn inner(filename: &Path, config: &Config, dependencies: &mut Vec<PathBuf>) -> Result<Vec<u8>> {
-
+        let mut cart = match SourceType::of_file(filename)? {
-    if cart[0] >= 10 {
+            SourceType::Binary => {
-        let src = String::from_utf8(cart)?;
+                let mut cart = vec![];
-        cart = if src.chars().find(|c| !c.is_whitespace()) == Some('(') {
+                File::open(filename)?.read_to_end(&mut cart)?;
-            wat::parse_str(src)?
+                cart
-        } else {
+            }
-            curlywas::compile_str(&src, filename, curlywas::Options::default())?
+            SourceType::Wat => {
+                let cart = wat::parse_file(filename)?;
+                cart
+            }
+            SourceType::CurlyWas => {
+                let (module, deps) = curlywas::compile_file(filename, curlywas::Options::default());
+                *dependencies = deps;
+                module?
+            }
         };
+
+        if let Some(ref pack_config) = config.pack {
+            cart = uw8_tool::pack(&cart, pack_config)?;
+            println!(
+                "\npacked size: {} bytes ({:.2})",
+                cart.len(),
+                uw8_tool::compressed_size(&cart)
+            );
+        }
+
+        if let Some(ref path) = config.output_path {
+            File::create(path)?.write_all(&cart)?;
+        }
+
+        Ok(cart)
     }
 
-    if let Some(ref pack_config) = config.pack {
+    let result = inner(filename, config, &mut dependencies);
-        cart = uw8_tool::pack(&cart, pack_config)?;
+
-        println!("packed size: {} bytes", cart.len());
+    if dependencies.is_empty() {
+        dependencies.push(filename.to_path_buf());
     }
 
-    if let Some(ref path) = config.output_path {
+    (result, dependencies)
-        File::create(path)?.write_all(&cart)?;
+}
-    }
 
-    Ok(cart)
+enum SourceType {
+    Binary,
+    Wat,
+    CurlyWas,
+}
+
+impl SourceType {
+    fn of_file(filename: &Path) -> Result<SourceType> {
+        if let Some(extension) = filename.extension() {
+            if extension == "uw8" || extension == "wasm" {
+                return Ok(SourceType::Binary);
+            } else if extension == "wat" || extension == "wast" {
+                return Ok(SourceType::Wat);
+            } else if extension == "cwa" {
+                return Ok(SourceType::CurlyWas);
+            }
+        }
+
+        let mut cart = vec![];
+        File::open(filename)?.read_to_end(&mut cart)?;
+
+        let ty = if cart[0] < 10 {
+            SourceType::Binary
+        } else {
+            let src = String::from_utf8(cart)?;
+            if src.chars().find(|&c| !c.is_whitespace() && c != ';') == Some('(') {
+                SourceType::Wat
+            } else {
+                SourceType::CurlyWas
+            }
+        };
+        Ok(ty)
+    }
 }
 
 #[cfg(any(feature = "native", feature = "browser"))]
-fn start_cart(filename: &Path, runtime: &mut dyn Runtime, config: &Config) -> Result<()> {
+fn start_cart(
-    let cart = load_cart(filename, config)?;
+    filename: &Path,
+    runtime: &mut dyn Runtime,
+    config: &Config,
+) -> (Result<()>, Vec<PathBuf>) {
+    let (cart, dependencies) = load_cart(filename, config);
+    let cart = match cart {
+        Ok(cart) => cart,
+        Err(err) => return (Err(err), dependencies),
+    };
 
     if let Err(err) = runtime.load(&cart) {
         eprintln!("Load error: {}", err);
-        Err(err)
+        (Err(err), dependencies)
     } else {
-        Ok(())
+        (Ok(()), dependencies)
     }
 }
 
@@ -192,7 +275,8 @@ fn pack(mut args: Arguments) -> Result<()> {
             pack: Some(pack_config),
             output_path: None,
         },
-    )?;
+    )
+    .0?;
 
     File::create(out_file)?.write_all(&cart)?;
 
@@ -215,7 +299,7 @@ fn compile(mut args: Arguments) -> Result<()> {
     let in_file = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
     let out_file = args.free_from_os_str::<PathBuf, bool>(|s| Ok(s.into()))?;
 
-    let module = curlywas::compile_file(in_file, options)?;
+    let module = curlywas::compile_file(in_file, options).0?;
     File::create(out_file)?.write_all(&module)?;
 
     Ok(())

src/run_native.rs

@@ -1,41 +1,34 @@
-use std::sync::{Arc, Mutex};
+use std::sync::{mpsc, Arc, Mutex};
 use std::time::Duration;
 use std::{thread, time::Instant};
 
-use anyhow::Result;
+use anyhow::{anyhow, Result};
-use minifb::{Key, Window, WindowOptions};
+use cpal::traits::*;
+use rubato::Resampler;
+use uw8_window::{Window, WindowConfig};
 use wasmtime::{
     Engine, GlobalType, Memory, MemoryType, Module, Mutability, Store, TypedFunc, ValType,
 };
 
-static GAMEPAD_KEYS: &[Key] = &[
-    Key::Up,
-    Key::Down,
-    Key::Left,
-    Key::Right,
-    Key::Z,
-    Key::X,
-    Key::A,
-    Key::S,
-];
-
 pub struct MicroW8 {
+    window: Window,
+    stream: Option<cpal::Stream>,
     engine: Engine,
     loader_module: Module,
-    window: Window,
+    disable_audio: bool,
-    window_buffer: Vec<u32>,
+    module_data: Option<Vec<u8>>,
-    instance: Option<UW8Instance>,
     timeout: u32,
+    instance: Option<UW8Instance>,
 }
 
 struct UW8Instance {
     store: Store<()>,
     memory: Memory,
     end_frame: TypedFunc<(), ()>,
-    update: TypedFunc<(), ()>,
+    update: Option<TypedFunc<(), ()>>,
     start_time: Instant,
-    module: Vec<u8>,
     watchdog: Arc<Mutex<UW8WatchDog>>,
+    sound_tx: Option<mpsc::SyncSender<RegisterUpdate>>,
 }
 
 impl Drop for UW8Instance {
@@ -47,58 +40,52 @@ impl Drop for UW8Instance {
 }
 
 struct UW8WatchDog {
-    interupt: wasmtime::InterruptHandle,
+    engine: Engine,
-    timeout: u32,
     stop: bool,
 }
 
 impl MicroW8 {
-    pub fn new() -> Result<MicroW8> {
+    pub fn new(timeout: Option<u32>, window_config: WindowConfig) -> Result<MicroW8> {
-        let engine = wasmtime::Engine::new(wasmtime::Config::new().interruptable(true))?;
+        let mut config = wasmtime::Config::new();
+        config.cranelift_opt_level(wasmtime::OptLevel::Speed);
+        if timeout.is_some() {
+            config.epoch_interruption(true);
+        }
+        let engine = wasmtime::Engine::new(&config)?;
 
         let loader_module =
             wasmtime::Module::new(&engine, include_bytes!("../platform/bin/loader.wasm"))?;
 
-        let options = WindowOptions {
+        let window = Window::new(window_config)?;
-            scale: minifb::Scale::X2,
-            scale_mode: minifb::ScaleMode::AspectRatioStretch,
-            resize: true,
-            ..Default::default()
-        };
-        let mut window = Window::new("MicroW8", 320, 240, options)?;
-        window.limit_update_rate(Some(std::time::Duration::from_micros(16666)));
 
         Ok(MicroW8 {
+            window,
+            stream: None,
             engine,
             loader_module,
-            window,
+            disable_audio: false,
-            window_buffer: vec![0u32; 320 * 240],
+            module_data: None,
+            timeout: timeout.unwrap_or(0),
             instance: None,
-            timeout: 30,
         })
     }
 
-    fn reset(&mut self) {
+    pub fn disable_audio(&mut self) {
-        self.instance = None;
+        self.disable_audio = true;
-        for v in &mut self.window_buffer {
-            *v = 0;
-        }
     }
 }
 
 impl super::Runtime for MicroW8 {
     fn is_open(&self) -> bool {
-        self.window.is_open() && !self.window.is_key_down(Key::Escape)
+        self.window.is_open()
-    }
-
-    fn set_timeout(&mut self, timeout: u32) {
-        self.timeout = timeout;
     }
 
     fn load(&mut self, module_data: &[u8]) -> Result<()> {
-        self.reset();
+        self.stream = None;
+        self.instance = None;
 
         let mut store = wasmtime::Store::new(&self.engine, ());
+        store.set_epoch_deadline(60);
 
         let memory = wasmtime::Memory::new(&mut store, MemoryType::new(4, Some(4)))?;
 
@@ -119,46 +106,12 @@ impl super::Runtime for MicroW8 {
         let module_length = load_uw8.call(&mut store, module_data.len() as i32)? as u32 as usize;
         let module = wasmtime::Module::new(&self.engine, &memory.data(&store)[..module_length])?;
 
-        linker.func_wrap("env", "acos", |v: f32| v.acos())?;
+        add_native_functions(&mut linker, &mut store)?;
-        linker.func_wrap("env", "asin", |v: f32| v.asin())?;
-        linker.func_wrap("env", "atan", |v: f32| v.atan())?;
-        linker.func_wrap("env", "atan2", |x: f32, y: f32| x.atan2(y))?;
-        linker.func_wrap("env", "cos", |v: f32| v.cos())?;
-        linker.func_wrap("env", "exp", |v: f32| v.exp())?;
-        linker.func_wrap("env", "log", |v: f32| v.ln())?;
-        linker.func_wrap("env", "sin", |v: f32| v.sin())?;
-        linker.func_wrap("env", "tan", |v: f32| v.tan())?;
-        linker.func_wrap("env", "pow", |a: f32, b: f32| a.powf(b))?;
-        for i in 9..64 {
-            linker.func_wrap("env", &format!("reserved{}", i), || {})?;
-        }
-        for i in 0..16 {
-            linker.define(
-                "env",
-                &format!("g_reserved{}", i),
-                wasmtime::Global::new(
-                    &mut store,
-                    GlobalType::new(ValType::I32, Mutability::Const),
-                    0.into(),
-                )?,
-            )?;
-        }
 
-        let platform_instance = linker.instantiate(&mut store, &platform_module)?;
+        let platform_instance = instantiate_platform(&mut linker, &mut store, &platform_module)?;
-
-        for export in platform_instance.exports(&mut store) {
-            linker.define(
-                "env",
-                export.name(),
-                export
-                    .into_func()
-                    .expect("platform surely only exports functions"),
-            )?;
-        }
 
         let watchdog = Arc::new(Mutex::new(UW8WatchDog {
-            interupt: store.interrupt_handle()?,
+            engine: self.engine.clone(),
-            timeout: self.timeout,
             stop: false,
         }));
 
@@ -166,16 +119,11 @@ impl super::Runtime for MicroW8 {
             let watchdog = watchdog.clone();
             thread::spawn(move || loop {
                 thread::sleep(Duration::from_millis(17));
-                if let Ok(mut watchdog) = watchdog.lock() {
+                if let Ok(watchdog) = watchdog.lock() {
                     if watchdog.stop {
                         break;
                     }
-                    if watchdog.timeout > 0 {
+                    watchdog.engine.increment_epoch();
-                        watchdog.timeout -= 1;
-                        if watchdog.timeout == 0 {
-                            watchdog.interupt.interrupt();
-                        }
-                    }
                 } else {
                     break;
                 }
@@ -183,11 +131,23 @@ impl super::Runtime for MicroW8 {
         }
 
         let instance = linker.instantiate(&mut store, &module)?;
-        if let Ok(mut watchdog) = watchdog.lock() {
-            watchdog.timeout = 0;
-        }
         let end_frame = platform_instance.get_typed_func::<(), (), _>(&mut store, "endFrame")?;
-        let update = instance.get_typed_func::<(), (), _>(&mut store, "upd")?;
+        let update = instance.get_typed_func::<(), (), _>(&mut store, "upd").ok();
+
+        let (sound_tx, stream) = if self.disable_audio {
+            (None, None)
+        } else {
+            match init_sound(&self.engine, &platform_module, &module) {
+                Ok(sound) => {
+                    sound.stream.play()?;
+                    (Some(sound.tx), Some(sound.stream))
+                }
+                Err(err) => {
+                    eprintln!("Failed to init sound: {}", err);
+                    (None, None)
+                }
+            }
+        };
 
         self.instance = Some(UW8Instance {
             store,
@@ -195,66 +155,323 @@ impl super::Runtime for MicroW8 {
             end_frame,
             update,
             start_time: Instant::now(),
-            module: module_data.into(),
             watchdog,
+            sound_tx,
         });
-
+        self.stream = stream;
+        self.module_data = Some(module_data.into());
         Ok(())
     }
 
     fn run_frame(&mut self) -> Result<()> {
+        let input = self.window.begin_frame();
+
+        if input.reset {
+            if let Some(module_data) = self.module_data.take() {
+                self.load(&module_data)?;
+            }
+        }
+
+        let now = Instant::now();
         let mut result = Ok(());
         if let Some(mut instance) = self.instance.take() {
-            {
+            let time = (now - instance.start_time).as_millis() as i32;
-                let time = instance.start_time.elapsed().as_millis() as i32;
+            let next_frame = {
-                let mut gamepad: u32 = 0;
+                let offset = ((time as u32 as i64 * 6) % 100 - 50) / 6;
-                for key in self.window.get_keys() {
+                let max = now + Duration::from_millis(17);
-                    if let Some(index) = GAMEPAD_KEYS
+                let next_center = now + Duration::from_millis((16 - offset) as u64);
-                        .iter()
+                next_center.min(max)
-                        .enumerate()
+            };
-                        .find(|(_, &k)| k == key)
-                        .map(|(i, _)| i)
-                    {
-                        gamepad |= 1 << index;
-                    }
-                }
 
+            {
                 let mem = instance.memory.data_mut(&mut instance.store);
                 mem[64..68].copy_from_slice(&time.to_le_bytes());
-                mem[68..72].copy_from_slice(&gamepad.to_le_bytes());
+                mem[68..72].copy_from_slice(&input.gamepads);
             }
 
-            if let Ok(mut watchdog) = instance.watchdog.lock() {
+            instance.store.set_epoch_deadline(self.timeout as u64);
-                watchdog.timeout = self.timeout;
+            if let Some(ref update) = instance.update {
-            }
+                if let Err(err) = update.call(&mut instance.store, ()) {
-            result = instance.update.call(&mut instance.store, ());
+                    result = Err(err);
-            if let Ok(mut watchdog) = instance.watchdog.lock() {
+                }
-                watchdog.timeout = 0;
             }
             instance.end_frame.call(&mut instance.store, ())?;
 
             let memory = instance.memory.data(&instance.store);
-            let framebuffer = &memory[120..(120 + 320 * 240)];
+
-            let palette = &memory[0x13000..];
+            let mut sound_regs = [0u8; 32];
-            for (i, &color_index) in framebuffer.iter().enumerate() {
+            sound_regs.copy_from_slice(&memory[80..112]);
-                let offset = color_index as usize * 4;
+            if let Some(ref sound_tx) = instance.sound_tx {
-                self.window_buffer[i] = 0xff000000
+                let _ = sound_tx.send(RegisterUpdate {
-                    | ((palette[offset] as u32) << 16)
+                    time,
-                    | ((palette[offset + 1] as u32) << 8)
+                    data: sound_regs,
-                    | palette[offset + 2] as u32;
+                });
             }
 
-            if self.window.is_key_pressed(Key::R, minifb::KeyRepeat::No) {
+            let framebuffer_mem = &memory[120..(120 + 320 * 240)];
-                self.load(&instance.module)?;
+            let palette_mem = &memory[0x13000..];
-            } else if result.is_ok() {
+            self.window
+                .end_frame(framebuffer_mem, palette_mem, next_frame);
+
+            if result.is_ok() {
                 self.instance = Some(instance);
             }
         }
 
-        self.window
-            .update_with_buffer(&self.window_buffer, 320, 240)?;
-
         result?;
         Ok(())
     }
 }
+
+fn add_native_functions(
+    linker: &mut wasmtime::Linker<()>,
+    store: &mut wasmtime::Store<()>,
+) -> Result<()> {
+    linker.func_wrap("env", "acos", |v: f32| v.acos())?;
+    linker.func_wrap("env", "asin", |v: f32| v.asin())?;
+    linker.func_wrap("env", "atan", |v: f32| v.atan())?;
+    linker.func_wrap("env", "atan2", |x: f32, y: f32| x.atan2(y))?;
+    linker.func_wrap("env", "cos", |v: f32| v.cos())?;
+    linker.func_wrap("env", "exp", |v: f32| v.exp())?;
+    linker.func_wrap("env", "log", |v: f32| v.ln())?;
+    linker.func_wrap("env", "sin", |v: f32| v.sin())?;
+    linker.func_wrap("env", "tan", |v: f32| v.tan())?;
+    linker.func_wrap("env", "pow", |a: f32, b: f32| a.powf(b))?;
+    for i in 10..64 {
+        linker.func_wrap("env", &format!("reserved{}", i), || {})?;
+    }
+    let log_line = std::sync::Mutex::new(String::new());
+    linker.func_wrap("env", "logChar", move |c: i32| {
+        let mut log_line = log_line.lock().unwrap();
+        if c == 10 {
+            println!("{}", log_line);
+            log_line.clear();
+        } else {
+            log_line.push(c as u8 as char);
+        }
+    })?;
+    for i in 0..16 {
+        linker.define(
+            "env",
+            &format!("g_reserved{}", i),
+            wasmtime::Global::new(
+                &mut *store,
+                GlobalType::new(ValType::I32, Mutability::Const),
+                0.into(),
+            )?,
+        )?;
+    }
+
+    Ok(())
+}
+
+fn instantiate_platform(
+    linker: &mut wasmtime::Linker<()>,
+    store: &mut wasmtime::Store<()>,
+    platform_module: &wasmtime::Module,
+) -> Result<wasmtime::Instance> {
+    let platform_instance = linker.instantiate(&mut *store, &platform_module)?;
+
+    for export in platform_instance.exports(&mut *store) {
+        linker.define(
+            "env",
+            export.name(),
+            export
+                .into_func()
+                .expect("platform surely only exports functions"),
+        )?;
+    }
+
+    Ok(platform_instance)
+}
+
+struct RegisterUpdate {
+    time: i32,
+    data: [u8; 32],
+}
+
+struct Uw8Sound {
+    stream: cpal::Stream,
+    tx: mpsc::SyncSender<RegisterUpdate>,
+}
+
+fn init_sound(
+    engine: &wasmtime::Engine,
+    platform_module: &wasmtime::Module,
+    module: &wasmtime::Module,
+) -> Result<Uw8Sound> {
+    let mut store = wasmtime::Store::new(engine, ());
+    store.set_epoch_deadline(60);
+
+    let memory = wasmtime::Memory::new(&mut store, MemoryType::new(4, Some(4)))?;
+
+    let mut linker = wasmtime::Linker::new(engine);
+    linker.define("env", "memory", memory)?;
+    add_native_functions(&mut linker, &mut store)?;
+
+    let platform_instance = instantiate_platform(&mut linker, &mut store, platform_module)?;
+    let instance = linker.instantiate(&mut store, module)?;
+
+    let snd = instance
+        .get_typed_func::<(i32,), f32, _>(&mut store, "snd")
+        .or_else(|_| platform_instance.get_typed_func::<(i32,), f32, _>(&mut store, "gesSnd"))?;
+
+    let host = cpal::default_host();
+    let device = host
+        .default_output_device()
+        .ok_or_else(|| anyhow!("No audio output device available"))?;
+    let mut configs: Vec<_> = device
+        .supported_output_configs()?
+        .filter(|config| {
+            config.channels() == 2 && config.sample_format() == cpal::SampleFormat::F32
+        })
+        .collect();
+    configs.sort_by_key(|config| {
+        let rate = 44100
+            .max(config.min_sample_rate().0)
+            .min(config.max_sample_rate().0);
+        if rate >= 44100 {
+            rate - 44100
+        } else {
+            (44100 - rate) * 1000
+        }
+    });
+    let config = configs
+        .into_iter()
+        .next()
+        .ok_or_else(|| anyhow!("Could not find float output config"))?;
+    let sample_rate = cpal::SampleRate(44100)
+        .max(config.min_sample_rate())
+        .max(config.max_sample_rate());
+    let config = config.with_sample_rate(sample_rate);
+    let buffer_size = match *config.buffer_size() {
+        cpal::SupportedBufferSize::Unknown => cpal::BufferSize::Default,
+        cpal::SupportedBufferSize::Range { min, max } => {
+            cpal::BufferSize::Fixed(256.max(min).min(max))
+        }
+    };
+    let config = cpal::StreamConfig {
+        buffer_size,
+        ..config.config()
+    };
+
+    let sample_rate = config.sample_rate.0 as usize;
+
+    let (tx, rx) = mpsc::sync_channel::<RegisterUpdate>(30);
+
+    struct Resampler {
+        resampler: rubato::FftFixedIn<f32>,
+        input_buffers: Vec<Vec<f32>>,
+        output_buffers: Vec<Vec<f32>>,
+        output_index: usize,
+    }
+    let mut resampler: Option<Resampler> = if sample_rate == 44100 {
+        None
+    } else {
+        let rs = rubato::FftFixedIn::new(44100, sample_rate, 128, 1, 2)?;
+        let input_buffers = rs.input_buffer_allocate();
+        let output_buffers = rs.output_buffer_allocate();
+        Some(Resampler {
+            resampler: rs,
+            input_buffers,
+            output_buffers,
+            output_index: usize::MAX,
+        })
+    };
+
+    let mut sample_index = 0;
+    let mut pending_updates: Vec<RegisterUpdate> = Vec::with_capacity(30);
+    let mut current_time = 0;
+    let stream = device.build_output_stream(
+        &config,
+        move |mut outer_buffer: &mut [f32], _| {
+            let mut first_update = true;
+            while let Ok(update) = rx.try_recv() {
+                if first_update {
+                    current_time += update.time.wrapping_sub(current_time) / 8;
+                    first_update = false;
+                }
+                pending_updates.push(update);
+            }
+
+            while !outer_buffer.is_empty() {
+                store.set_epoch_deadline(30);
+                while pending_updates
+                    .first()
+                    .into_iter()
+                    .any(|u| u.time.wrapping_sub(current_time) <= 0)
+                {
+                    let update = pending_updates.remove(0);
+                    memory.write(&mut store, 80, &update.data).unwrap();
+                }
+
+                let duration = if let Some(update) = pending_updates.first() {
+                    ((update.time.wrapping_sub(current_time) as usize) * sample_rate + 999) / 1000
+                } else {
+                    outer_buffer.len()
+                };
+                let step_size = (duration.max(64) * 2).min(outer_buffer.len());
+
+                let mut buffer = &mut outer_buffer[..step_size];
+
+                {
+                    let mem = memory.data_mut(&mut store);
+                    mem[64..68].copy_from_slice(&current_time.to_le_bytes());
+                }
+
+                if let Some(ref mut resampler) = resampler {
+                    while !buffer.is_empty() {
+                        let copy_size = resampler.output_buffers[0]
+                            .len()
+                            .saturating_sub(resampler.output_index)
+                            .min(buffer.len() / 2);
+                        if copy_size == 0 {
+                            resampler.input_buffers[0].clear();
+                            resampler.input_buffers[1].clear();
+                            for _ in 0..resampler.resampler.input_frames_next() {
+                                resampler.input_buffers[0]
+                                    .push(snd.call(&mut store, (sample_index,)).unwrap_or(0.0));
+                                resampler.input_buffers[1]
+                                    .push(snd.call(&mut store, (sample_index + 1,)).unwrap_or(0.0));
+                                sample_index = sample_index.wrapping_add(2);
+                            }
+
+                            resampler
+                                .resampler
+                                .process_into_buffer(
+                                    &resampler.input_buffers,
+                                    &mut resampler.output_buffers,
+                                    None,
+                                )
+                                .unwrap();
+                            resampler.output_index = 0;
+                        } else {
+                            for i in 0..copy_size {
+                                buffer[i * 2] =
+                                    resampler.output_buffers[0][resampler.output_index + i];
+                                buffer[i * 2 + 1] =
+                                    resampler.output_buffers[1][resampler.output_index + i];
+                            }
+                            resampler.output_index += copy_size;
+                            buffer = &mut buffer[copy_size * 2..];
+                        }
+                    }
+                } else {
+                    for v in buffer {
+                        *v = snd.call(&mut store, (sample_index,)).unwrap_or(0.0);
+                        sample_index = sample_index.wrapping_add(1);
+                    }
+                }
+
+                outer_buffer = &mut outer_buffer[step_size..];
+                current_time =
+                    current_time.wrapping_add((step_size * 500 / sample_rate).max(1) as i32);
+            }
+        },
+        move |err| {
+            dbg!(err);
+        },
+    )?;
+
+    Ok(Uw8Sound { stream, tx })
+}

test/drawing_test.cwa

@@ -0,0 +1,62 @@
+include "../examples/include/microw8-api.cwa"
+
+global mut counter = 0;
+
+export fn upd() {
+    cls(0);
+    
+    let col: i32 = 1;
+    
+    loop colors {
+        if !testCircle(counter, col) {
+            printInt(counter);
+            return;
+        }
+        counter += 1;
+        branch_if (col +:= 1) < 256: colors;
+    }
+}
+
+fn testCircle(seed: i32, col: i32) -> i32 {
+    randomSeed(seed);
+    let cx = randomf() * 640_f - 160_f;
+    let cy = randomf() * 480_f - 120_f;
+    let radius = randomf() * 4_f;
+    radius *= radius;
+    radius *= radius;
+    
+    circle(cx, cy, radius, col);
+    
+    let min_x = max(0_f, floor(cx - radius - 1_f)) as i32;
+    let min_y = max(0_f, floor(cy - radius - 1_f)) as i32;
+    let max_x = min(320_f, ceil(cx + radius + 1_f)) as i32;
+    let max_y = min(240_f, ceil(cy + radius + 1_f)) as i32;
+    
+    let x = min_x;
+    loop xloop {
+        if x < max_x {
+            let y = min_y;
+            loop yloop {
+                if y < max_y {
+                    let rx = x as f32 + 0.5 - cx;
+                    let ry = y as f32 + 0.5 - cy;
+                    let d = sqrt(rx*rx + ry*ry) - radius;
+                    if abs(d) > 0.001 {
+                        let is_inside = d < 0_f;
+                        let is_plotted = getPixel(x, y) == col;
+                        if is_inside != is_plotted {
+                            return 0;
+                        }
+                    }                    
+                
+                    y += 1;
+                    branch yloop;
+                }
+            }
+            x += 1;
+            branch xloop;
+        }
+    }
+    
+    1
+}

test/frame_time.cwa

@@ -0,0 +1,20 @@
+include "../examples/include/microw8-api.cwa"
+
+global mut pos = 0;
+global mut next = 0;
+
+export fn upd() {
+    let lazy t = 32!32;
+    let lazy tick = t * 6 / 100;
+    let lazy rel = t - tick * 100 / 6;
+
+    setBackgroundColor(select(tick == next, 0, select(tick < next, 0x35, 0x55)));
+    setCursorPosition(pos % 13 * 3, pos / 13 % 30);
+    if rel < 10 {
+        printChar(32);
+    }
+    printInt(rel);
+
+    pos = pos + 1;
+    next = tick + 1;
+}

test/ges_test.cwa

@@ -0,0 +1,30 @@
+import "env.memory" memory(4);
+import "env.pow" fn pow(f32, f32) -> f32;
+import "env.sin" fn sin(f32) -> f32;
+import "env.cls" fn cls(i32);
+import "env.rectangle" fn rectangle(f32, f32, f32, f32, i32);
+
+include "../platform/src/ges.cwa"
+
+export fn snd(t: i32) -> f32 {
+    gesSnd(t)
+}
+
+export fn upd() {
+    80?0 = 32!32 / 200 & 2 | 0x41;
+    80?3 = (32!32 / 400)%7*12/7 + 40;
+    let pulse = (32!32 * 256 / 2000) & 511;
+    if pulse >= 256 {
+        pulse = 511 - pulse;
+    }
+    80?1 = pulse;
+
+    cls(0);
+    rectangle(0.0, 100.0, (pulse * 320 / 256) as f32, 16.0, 15);
+}
+
+data 80 {
+    i8(
+        0x41, 0, 0, 80, 0x70, 0
+    )
+}

test/log.cwa

@@ -0,0 +1,7 @@
+include "../examples/include/microw8-api.cwa"
+
+export fn upd() {
+    printChar('\06f: ');
+    printInt(32!32 * 6 / 100);
+    printChar('\n\4');
+}

test/plot_ges.cwa

@@ -0,0 +1,59 @@
+include "../examples/include/microw8-api.cwa"
+
+export fn upd() {
+    80?0 = (32!32 >> 11 << 6) | 5;
+    80?1 = (sin(time() * 6 as f32) * 95 as f32) as i32 + 128;
+    plotGes();
+}
+
+data 80 { i8 (
+    1, 128, 0, 69, 0, 15,
+    0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0,
+    0, 0, 0, 0, 0, 0,
+    0xff, 0xff,
+    0xc1, 0, 0, 110, 0, 0
+) }
+
+//import "env.gesSnd" fn gesSnd(i32) -> f32;
+
+include "../platform/src/ges.cwa"
+
+export fn snd(t: i32) -> f32 {
+    gesSnd(t)
+}
+
+global mut samplePos: i32 = 0;
+
+const SoundBuffer = 0x30000;
+
+fn plotGes() {
+    rectangle(0 as f32, 10 as f32, 320 as f32, 320 as f32, 0);
+    let count = (time() * 44100 as f32) as i32 * 2 - samplePos;
+    let i: i32;
+    loop genLoop {
+        (i*4)$SoundBuffer = gesSnd(samplePos + i);
+        branch_if (i := i + 1) < count: genLoop;
+    }
+    samplePos = samplePos + count;
+
+    let ch: i32;
+    loop channelLoop {
+        let offset = 159;
+        i = 0;
+
+        loop searchLoop {
+            offset = offset + 1;
+            branch_if (offset * 8 + ch - 8)$SoundBuffer < 0 as f32 | (offset * 8 + ch)$SoundBuffer >= 0 as f32 & offset + 160 < count: searchLoop;
+        }
+
+        offset = ch + (offset - 160) * 8;
+        i = 0;
+        loop plotLoop {
+            setPixel(i, floor((i * 8 + offset)$SoundBuffer * 127 as f32) as i32 + 60 + ch * (120/8), 15);
+            branch_if (i := i + 1) < 320: plotLoop;
+        }
+
+        branch_if (ch := ch + 8) < 16: channelLoop;
+    }
+}

uw8-tool/Cargo.lock

@@ -189,7 +189,7 @@ checksum = "8ccb82d61f80a663efe1f787a51b16b5a51e3314d6ac365b08639f52387b33f3"
 [[package]]
 name = "upkr"
 version = "0.1.0"
-source = "git+https://github.com/exoticorn/upkr.git?rev=2e7983fc#2e7983fc650788d98da2eecef2d16f63e849e4a0"
+source = "git+https://github.com/exoticorn/upkr.git?rev=d93aec186c9fb91d962c488682a2db125c61306c#d93aec186c9fb91d962c488682a2db125c61306c"
 dependencies = [
  "anyhow",
  "cdivsufsort",

uw8-tool/Cargo.toml

@@ -11,5 +11,5 @@ wasm-encoder = "0.8"
 walrus = "0.19"
 anyhow = "1"
 pico-args = "0.4"
-upkr = { git = "https://github.com/exoticorn/upkr.git", rev = "2e7983fc" }
+upkr = { git = "https://github.com/exoticorn/upkr.git", rev = "d93aec186c9fb91d962c488682a2db125c61306c" }
 pbr = "1"

uw8-tool/src/base_module.rs

@@ -71,26 +71,102 @@ impl BaseModule {
         add_function(&mut functions, &type_map, "randomSeed", &[I32], None);
 
         add_function(&mut functions, &type_map, "cls", &[I32], None);
-        add_function(&mut functions, &type_map, "setPixel", &[I32, I32, I32], None);
+        add_function(
-        add_function(&mut functions, &type_map, "getPixel", &[I32, I32], Some(I32));
+            &mut functions,
-        add_function(&mut functions, &type_map, "hline", &[I32, I32, I32, I32], None);
+            &type_map,
-        add_function(&mut functions, &type_map, "rectangle", &[F32, F32, F32, F32, I32], None);
+            "setPixel",
-        add_function(&mut functions, &type_map, "circle", &[F32, F32, F32, I32], None);
+            &[I32, I32, I32],
-        add_function(&mut functions, &type_map, "line", &[F32, F32, F32, F32, I32], None);
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "getPixel",
+            &[I32, I32],
+            Some(I32),
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "hline",
+            &[I32, I32, I32, I32],
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "rectangle",
+            &[F32, F32, F32, F32, I32],
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "circle",
+            &[F32, F32, F32, I32],
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "line",
+            &[F32, F32, F32, F32, I32],
+            None,
+        );
 
         add_function(&mut functions, &type_map, "time", &[], Some(F32));
-        add_function(&mut functions, &type_map, "isButtonPressed", &[I32], Some(I32));
+        add_function(
-        add_function(&mut functions, &type_map, "isButtonTriggered", &[I32], Some(I32));
+            &mut functions,
+            &type_map,
+            "isButtonPressed",
+            &[I32],
+            Some(I32),
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "isButtonTriggered",
+            &[I32],
+            Some(I32),
+        );
 
         add_function(&mut functions, &type_map, "printChar", &[I32], None);
         add_function(&mut functions, &type_map, "printString", &[I32], None);
         add_function(&mut functions, &type_map, "printInt", &[I32], None);
         add_function(&mut functions, &type_map, "setTextColor", &[I32], None);
-        add_function(&mut functions, &type_map, "setBackgroundColor", &[I32], None);
+        add_function(
-        add_function(&mut functions, &type_map, "setCursorPosition", &[I32, I32], None);
+            &mut functions,
+            &type_map,
+            "setBackgroundColor",
+            &[I32],
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "setCursorPosition",
+            &[I32, I32],
+            None,
+        );
 
-        add_function(&mut functions, &type_map, "rectangle_outline", &[F32, F32, F32, F32, I32], None);
+        add_function(
-        add_function(&mut functions, &type_map, "circle_outline", &[F32, F32, F32, I32], None);
+            &mut functions,
+            &type_map,
+            "rectangle_outline",
+            &[F32, F32, F32, F32, I32],
+            None,
+        );
+        add_function(
+            &mut functions,
+            &type_map,
+            "circle_outline",
+            &[F32, F32, F32, I32],
+            None,
+        );
+
+        add_function(&mut functions, &type_map, "exp", &[F32], Some(F32));
+
+        add_function(&mut functions, &type_map, "playNote", &[I32, I32], None);
 
         for i in functions.len()..64 {
             add_function(
@@ -210,10 +286,75 @@ impl BaseModule {
 
     pub fn create_binary(path: &Path) -> Result<()> {
         let base1 = BaseModule::for_format_version(1)?.to_wasm();
-        let data = upkr::pack(&base1, 4, None);
+        let data = upkr::pack(&base1, 4, false, None);
         File::create(path)?.write_all(&data)?;
         Ok(())
     }
+
+    pub fn write_as_cwa<P: AsRef<Path>>(&self, path: P) -> Result<()> {
+        fn inner(mut file: File, base: &BaseModule) -> Result<()> {
+            writeln!(
+                file,
+                "// MicroW8 APIs, to be `include`d in CurlyWas sources"
+            )?;
+            writeln!(file, "import \"env.memory\" memory({});", base.memory)?;
+            writeln!(file)?;
+            for &(module, ref name, type_id) in &base.function_imports {
+                if !name.contains("reserved") {
+                    let ty = &base.types[type_id as usize];
+                    let params: Vec<&str> = ty.params.iter().copied().map(type_to_str).collect();
+                    write!(
+                        file,
+                        "import \"{}.{}\" fn {}({})",
+                        module,
+                        name,
+                        name,
+                        params.join(", ")
+                    )?;
+                    if let Some(result) = ty.result {
+                        write!(file, " -> {}", type_to_str(result))?;
+                    }
+                    writeln!(file, ";")?;
+                }
+            }
+
+            writeln!(file)?;
+            for &(name, value) in CONSTANTS {
+                writeln!(file, "const {} = 0x{:x};", name, value)?;
+            }
+            Ok(())
+        }
+        inner(File::create(path)?, self)
+    }
+
+    pub fn write_as_wat<P: AsRef<Path>>(&self, path: P) -> Result<()> {
+        fn inner(mut file: File, base: &BaseModule) -> Result<()> {
+            writeln!(file, ";; MicroW8 APIs, in WAT (Wasm Text) format")?;
+            writeln!(file, "(import \"env\" \"memory\" (memory {}))", base.memory)?;
+            writeln!(file)?;
+            for &(module, ref name, type_id) in &base.function_imports {
+                if !name.contains("reserved") {
+                    let ty = &base.types[type_id as usize];
+                    write!(file, "(import \"{}\" \"{}\" (func ${}", module, name, name)?;
+                    for &param in &ty.params {
+                        write!(file, " (param {})", type_to_str(param))?;
+                    }
+                    if let Some(result) = ty.result {
+                        write!(file, " (result {})", type_to_str(result))?;
+                    }
+                    writeln!(file, "))")?;
+                }
+            }
+
+            writeln!(file)?;
+            writeln!(file, ";; to use defines, include this file with a preprocessor\n;; like gpp (https://logological.org/gpp).")?;
+            for &(name, value) in CONSTANTS {
+                writeln!(file, "#define {} 0x{:x};", name, value)?;
+            }
+            Ok(())
+        }
+        inner(File::create(path)?, self)
+    }
 }
 
 fn add_function(
@@ -241,3 +382,30 @@ fn lookup_type(
     };
     *type_map.get(&key).unwrap()
 }
+
+fn type_to_str(ty: ValType) -> &'static str {
+    match ty {
+        ValType::I32 => "i32",
+        ValType::I64 => "i64",
+        ValType::F32 => "f32",
+        ValType::F64 => "f64",
+        _ => unimplemented!(),
+    }
+}
+
+const CONSTANTS: &[(&str, u32)] = &[
+    ("TIME_MS", 0x40),
+    ("GAMEPAD", 0x44),
+    ("FRAMEBUFFER", 0x78),
+    ("PALETTE", 0x13000),
+    ("FONT", 0x13400),
+    ("USER_MEM", 0x14000),
+    ("BUTTON_UP", 0),
+    ("BUTTON_DOWN", 1),
+    ("BUTTON_LEFT", 2),
+    ("BUTTON_RIGHT", 3),
+    ("BUTTON_A", 4),
+    ("BUTTON_B", 5),
+    ("BUTTON_X", 6),
+    ("BUTTON_Y", 7),
+];

uw8-tool/src/lib.rs

@@ -1,7 +1,15 @@
 mod base_module;
-mod pack;
 mod filter_exports;
+mod pack;
 
 pub use base_module::BaseModule;
-pub use pack::{pack, pack_file, unpack, unpack_file, PackConfig};
 pub use filter_exports::filter_exports;
+pub use pack::{pack, pack_file, unpack, unpack_file, PackConfig};
+
+pub fn compressed_size(cart: &[u8]) -> f32 {
+    if cart[0] != 2 {
+        cart.len() as f32
+    } else {
+        upkr::compressed_size(&cart[1..]) + 1.
+    }
+}

uw8-tool/src/main.rs

@@ -1,8 +1,8 @@
 use std::path::PathBuf;
 
 use anyhow::Result;
-use uw8_tool::BaseModule;
 use pico_args::Arguments;
+use uw8_tool::BaseModule;
 
 fn main() -> Result<()> {
     let mut args = Arguments::from_env();
@@ -32,6 +32,14 @@ fn main() -> Result<()> {
                 let dest: PathBuf = args.free_from_str()?;
                 uw8_tool::filter_exports(&source, &dest)?;
             }
+            "base-cwa" => {
+                let path: PathBuf = args.free_from_str()?;
+                BaseModule::for_format_version(1)?.write_as_cwa(path)?;
+            }
+            "base-wat" => {
+                let path: PathBuf = args.free_from_str()?;
+                BaseModule::for_format_version(1)?.write_as_wat(path)?;
+            }
             _ => {
                 eprintln!("Unknown subcommand '{}'", cmd);
                 print_help();

uw8-tool/src/pack.rs

@@ -63,6 +63,7 @@ pub fn pack(data: &[u8], config: &PackConfig) -> Result<Vec<u8>> {
         uw8.extend_from_slice(&upkr::pack(
             &result[8..],
             level,
+            false,
             Some(&mut |pos| {
                 pb.set(pos as u64);
             }),
@@ -89,7 +90,7 @@ pub fn unpack(data: Vec<u8>) -> Result<Vec<u8>> {
     let (version, data) = match data[0] {
         0 => return Ok(data),
         1 => (1, data[1..].to_vec()),
-        2 => (1, upkr::unpack(&data[1..])),
+        2 => (1, upkr::unpack(&data[1..], false)),
         other => bail!("Uknown format version {}", other),
     };
 
@@ -962,6 +963,8 @@ fn remap_function(
             De::I64TruncSatF32U => En::I64TruncSatF32U,
             De::I64TruncSatF64S => En::I64TruncSatF64S,
             De::I64TruncSatF64U => En::I64TruncSatF64U,
+            De::MemoryCopy { src, dst } => En::MemoryCopy { src, dst },
+            De::MemoryFill { mem } => En::MemoryFill(mem),
             other => bail!("Unsupported instruction {:?}", other),
         });
     }

uw8-window/.gitignore

@@ -0,0 +1 @@
+/target

uw8-window/Cargo.toml

@@ -0,0 +1,18 @@
+[package]
+name = "uw8-window"
+version = "0.1.0"
+edition = "2021"
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+winit = "0.26.1"
+env_logger = "0.9"
+log = "0.4"
+pico-args = "0.4"
+wgpu = "0.13.1"
+pollster = "0.2"
+bytemuck = { version = "1.4", features = [ "derive" ] }
+anyhow = "1"
+minifb = { version = "0.23.0", default-features = false, features = ["x11"] }
+winapi = "0.3.9"

uw8-window/src/cpu.rs

@@ -0,0 +1,83 @@
+use std::time::Instant;
+
+use crate::{Input, WindowImpl};
+use anyhow::Result;
+use minifb::{Key, WindowOptions};
+
+static GAMEPAD_KEYS: &[Key] = &[
+    Key::Up,
+    Key::Down,
+    Key::Left,
+    Key::Right,
+    Key::Z,
+    Key::X,
+    Key::A,
+    Key::S,
+];
+
+pub struct Window {
+    window: minifb::Window,
+    buffer: Vec<u32>,
+}
+
+impl Window {
+    pub fn new() -> Result<Window> {
+        #[cfg(target_os = "windows")]
+        unsafe {
+            winapi::um::timeapi::timeBeginPeriod(1);
+        }
+
+        let buffer: Vec<u32> = vec![0; 320 * 240];
+
+        let options = WindowOptions {
+            scale: minifb::Scale::X2,
+            scale_mode: minifb::ScaleMode::AspectRatioStretch,
+            resize: true,
+            ..Default::default()
+        };
+        let window = minifb::Window::new("MicroW8", 320, 240, options).unwrap();
+
+        Ok(Window { window, buffer })
+    }
+}
+
+impl WindowImpl for Window {
+    fn begin_frame(&mut self) -> Input {
+        let mut gamepads = [0u8; 4];
+        for key in self.window.get_keys() {
+            if let Some(index) = GAMEPAD_KEYS
+                .iter()
+                .enumerate()
+                .find(|(_, &k)| k == key)
+                .map(|(i, _)| i)
+            {
+                gamepads[0] |= 1 << index;
+            }
+        }
+
+        Input {
+            gamepads,
+            reset: self.window.is_key_pressed(Key::R, minifb::KeyRepeat::No),
+        }
+    }
+
+    fn end_frame(&mut self, framebuffer: &[u8], palette: &[u8], next_frame: Instant) {
+        for (i, &color_index) in framebuffer.iter().enumerate() {
+            let offset = color_index as usize * 4;
+            self.buffer[i] = 0xff000000
+                | ((palette[offset] as u32) << 16)
+                | ((palette[offset + 1] as u32) << 8)
+                | palette[offset + 2] as u32;
+        }
+        self.window
+            .update_with_buffer(&self.buffer, 320, 240)
+            .unwrap();
+        if let Some(sleep) = next_frame.checked_duration_since(Instant::now()) {
+            std::thread::sleep(sleep);
+        }
+    }
+
+    fn is_open(&self) -> bool {
+        self.window.is_open() && !self.window.is_key_down(Key::Escape)
+    }
+}

uw8-window/src/gpu/crt.rs

@@ -0,0 +1,143 @@
+use wgpu::util::DeviceExt;
+use winit::dpi::PhysicalSize;
+
+use super::Filter;
+
+pub struct CrtFilter {
+    uniform_buffer: wgpu::Buffer,
+    bind_group: wgpu::BindGroup,
+    pipeline: wgpu::RenderPipeline,
+}
+
+impl CrtFilter {
+    pub fn new(
+        device: &wgpu::Device,
+        screen: &wgpu::TextureView,
+        resolution: PhysicalSize<u32>,
+        surface_format: wgpu::TextureFormat,
+    ) -> CrtFilter {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(resolution),
+        };
+
+        let uniform_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+            label: None,
+            contents: bytemuck::cast_slice(&[uniforms]),
+            usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
+        });
+
+        let crt_bind_group_layout =
+            device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+                entries: &[
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 0,
+                        visibility: wgpu::ShaderStages::FRAGMENT,
+                        ty: wgpu::BindingType::Texture {
+                            multisampled: false,
+                            view_dimension: wgpu::TextureViewDimension::D2,
+                            sample_type: wgpu::TextureSampleType::Float { filterable: false },
+                        },
+                        count: None,
+                    },
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 1,
+                        visibility: wgpu::ShaderStages::VERTEX_FRAGMENT,
+                        ty: wgpu::BindingType::Buffer {
+                            ty: wgpu::BufferBindingType::Uniform,
+                            has_dynamic_offset: false,
+                            min_binding_size: None,
+                        },
+                        count: None,
+                    },
+                ],
+                label: None,
+            });
+
+        let crt_bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+            layout: &crt_bind_group_layout,
+            entries: &[
+                wgpu::BindGroupEntry {
+                    binding: 0,
+                    resource: wgpu::BindingResource::TextureView(&screen),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 1,
+                    resource: uniform_buffer.as_entire_binding(),
+                },
+            ],
+            label: None,
+        });
+
+        let crt_shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+            label: None,
+            source: wgpu::ShaderSource::Wgsl(include_str!("crt.wgsl").into()),
+        });
+
+        let render_pipeline_layout =
+            device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
+                label: None,
+                bind_group_layouts: &[&crt_bind_group_layout],
+                push_constant_ranges: &[],
+            });
+
+        let render_pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+            label: None,
+            layout: Some(&render_pipeline_layout),
+            vertex: wgpu::VertexState {
+                module: &crt_shader,
+                entry_point: "vs_main",
+                buffers: &[],
+            },
+            fragment: Some(wgpu::FragmentState {
+                module: &crt_shader,
+                entry_point: "fs_main",
+                targets: &[Some(wgpu::ColorTargetState {
+                    format: surface_format,
+                    blend: None,
+                    write_mask: wgpu::ColorWrites::ALL,
+                })],
+            }),
+            primitive: Default::default(),
+            depth_stencil: None,
+            multisample: Default::default(),
+            multiview: None,
+        });
+
+        CrtFilter {
+            uniform_buffer,
+            bind_group: crt_bind_group,
+            pipeline: render_pipeline,
+        }
+    }
+}
+
+impl Filter for CrtFilter {
+    fn resize(&mut self, queue: &wgpu::Queue, new_size: PhysicalSize<u32>) {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(new_size),
+        };
+        queue.write_buffer(&self.uniform_buffer, 0, bytemuck::cast_slice(&[uniforms]));
+    }
+
+    fn render<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>) {
+        render_pass.set_pipeline(&self.pipeline);
+        render_pass.set_bind_group(0, &self.bind_group, &[]);
+        render_pass.draw(0..6, 0..1);
+    }
+}
+
+fn texture_scale_from_resolution(res: PhysicalSize<u32>) -> [f32; 4] {
+    let scale = ((res.width as f32) / 160.0).min((res.height as f32) / 120.0);
+    [
+        res.width as f32 / scale,
+        res.height as f32 / scale,
+        2.0 / scale,
+        0.0,
+    ]
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
+struct Uniforms {
+    texture_scale: [f32; 4],
+}

uw8-window/src/gpu/crt.wgsl

@@ -0,0 +1,75 @@
+struct VertexOutput {
+    @builtin(position) clip_position: vec4<f32>,
+    @location(0) tex_coords: vec2<f32>,
+}
+
+struct Uniforms {
+    texture_scale: vec4<f32>,
+}
+
+@group(0) @binding(1) var<uniform> uniforms: Uniforms;
+
+@vertex
+fn vs_main(
+    @builtin(vertex_index) in_vertex_index: u32,
+) -> VertexOutput {
+    var out: VertexOutput;
+    let i = in_vertex_index / 3u + in_vertex_index % 3u;
+    let x = -1.0 + f32(i % 2u) * 322.0;
+    let y = -1.0 + f32(i / 2u) * 242.0;
+    out.clip_position = vec4<f32>((vec2<f32>(x, y) - vec2<f32>(160.0, 120.0)) / uniforms.texture_scale.xy, 0.0, 1.0);
+    out.tex_coords = vec2<f32>(x, y);
+    return out;
+}
+
+@group(0) @binding(0) var screen_texture: texture_2d<f32>;
+
+fn sample_pixel(coords: vec2<i32>, offset: vec4<f32>) -> vec3<f32> {
+    let is_outside = any(vec2<u32>(coords) >= vec2<u32>(320u, 240u));
+    if(is_outside) {
+        return vec3<f32>(0.0);
+    } else {
+        let f = max(vec4<f32>(0.008) / offset - vec4<f32>(0.0024), vec4<f32>(0.0));
+        return textureLoad(screen_texture, coords, 0).rgb * (f.x + f.y + f.z + f.w);
+    }
+}
+
+@fragment
+fn fs_main(in: VertexOutput) -> @location(0) vec4<f32> {
+    let pixel = floor(in.tex_coords);
+    let o = vec2<f32>(0.5) - (in.tex_coords - pixel);
+    let pixel = vec2<i32>(pixel);
+    
+    let offset_x = o.xxxx + vec4<f32>(-0.125, 0.375, 0.125, -0.375) * uniforms.texture_scale.z;
+    let offset_y = o.yyyy + vec4<f32>(-0.375, -0.125, 0.375, 0.125) * uniforms.texture_scale.z;
+    
+    let offset_x0 = max(abs(offset_x + vec4<f32>(-1.0)) - vec4<f32>(0.5), vec4<f32>(0.0));
+    let offset_x1 = max(abs(offset_x) - vec4<f32>(0.5), vec4<f32>(0.0));
+    let offset_x2 = max(abs(offset_x + vec4<f32>(1.0)) - vec4<f32>(0.5), vec4<f32>(0.0));
+    
+    let offset_x0 = offset_x0 * offset_x0;
+    let offset_x1 = offset_x1 * offset_x1;
+    let offset_x2 = offset_x2 * offset_x2;
+    
+    let offset_yr = offset_y + vec4<f32>(-1.0);
+    let offset_yr = vec4<f32>(0.02) + offset_yr * offset_yr;
+    
+    var acc = sample_pixel(pixel + vec2<i32>(-1, -1), offset_x0 + offset_yr);
+    acc = acc + sample_pixel(pixel + vec2<i32>(0, -1), offset_x1 + offset_yr);
+    acc = acc + sample_pixel(pixel + vec2<i32>(1, -1), offset_x2 + offset_yr);
+
+    let offset_yr = vec4<f32>(0.02) + offset_y * offset_y;
+    
+    acc = acc + sample_pixel(pixel + vec2<i32>(-1, 0), offset_x0 + offset_yr);
+    acc = acc + sample_pixel(pixel, offset_x1 + offset_yr);
+    acc = acc + sample_pixel(pixel + vec2<i32>(1, 0), offset_x2 + offset_yr);
+
+    let offset_yr = offset_y + vec4<f32>(1.0);
+    let offset_yr = vec4<f32>(0.02) + offset_yr * offset_yr;
+    
+    acc = acc + sample_pixel(pixel + vec2<i32>(-1, 1), offset_x0 + offset_yr);
+    acc = acc + sample_pixel(pixel + vec2<i32>(0, 1), offset_x1 + offset_yr);
+    acc = acc + sample_pixel(pixel + vec2<i32>(1, 1), offset_x2 + offset_yr);
+
+    return vec4<f32>(acc, 1.0);
+}

uw8-window/src/gpu/fast_crt.rs

@@ -0,0 +1,162 @@
+use wgpu::util::DeviceExt;
+use winit::dpi::PhysicalSize;
+
+use super::Filter;
+
+pub struct FastCrtFilter {
+    uniform_buffer: wgpu::Buffer,
+    bind_group: wgpu::BindGroup,
+    pipeline: wgpu::RenderPipeline,
+}
+
+impl FastCrtFilter {
+    pub fn new(
+        device: &wgpu::Device,
+        screen: &wgpu::TextureView,
+        resolution: PhysicalSize<u32>,
+        surface_format: wgpu::TextureFormat,
+        chromatic: bool,
+    ) -> FastCrtFilter {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(resolution),
+        };
+
+        let uniform_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+            label: None,
+            contents: bytemuck::cast_slice(&[uniforms]),
+            usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
+        });
+
+        let bind_group_layout = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+            entries: &[
+                wgpu::BindGroupLayoutEntry {
+                    binding: 0,
+                    visibility: wgpu::ShaderStages::FRAGMENT,
+                    ty: wgpu::BindingType::Texture {
+                        multisampled: false,
+                        view_dimension: wgpu::TextureViewDimension::D2,
+                        sample_type: wgpu::TextureSampleType::Float { filterable: true },
+                    },
+                    count: None,
+                },
+                wgpu::BindGroupLayoutEntry {
+                    binding: 1,
+                    visibility: wgpu::ShaderStages::FRAGMENT,
+                    ty: wgpu::BindingType::Sampler(wgpu::SamplerBindingType::Filtering),
+                    count: None,
+                },
+                wgpu::BindGroupLayoutEntry {
+                    binding: 2,
+                    visibility: wgpu::ShaderStages::VERTEX_FRAGMENT,
+                    ty: wgpu::BindingType::Buffer {
+                        ty: wgpu::BufferBindingType::Uniform,
+                        has_dynamic_offset: false,
+                        min_binding_size: None,
+                    },
+                    count: None,
+                },
+            ],
+            label: None,
+        });
+
+        let sampler = device.create_sampler(&wgpu::SamplerDescriptor {
+            mag_filter: wgpu::FilterMode::Linear,
+            ..Default::default()
+        });
+
+        let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+            layout: &bind_group_layout,
+            entries: &[
+                wgpu::BindGroupEntry {
+                    binding: 0,
+                    resource: wgpu::BindingResource::TextureView(&screen),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 1,
+                    resource: wgpu::BindingResource::Sampler(&sampler),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 2,
+                    resource: uniform_buffer.as_entire_binding(),
+                },
+            ],
+            label: None,
+        });
+
+        let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+            label: None,
+            source: wgpu::ShaderSource::Wgsl(include_str!("fast_crt.wgsl").into()),
+        });
+
+        let render_pipeline_layout =
+            device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
+                label: None,
+                bind_group_layouts: &[&bind_group_layout],
+                push_constant_ranges: &[],
+            });
+
+        let render_pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+            label: None,
+            layout: Some(&render_pipeline_layout),
+            vertex: wgpu::VertexState {
+                module: &shader,
+                entry_point: "vs_main",
+                buffers: &[],
+            },
+            fragment: Some(wgpu::FragmentState {
+                module: &shader,
+                entry_point: if chromatic {
+                    "fs_main_chromatic"
+                } else {
+                    "fs_main"
+                },
+                targets: &[Some(wgpu::ColorTargetState {
+                    format: surface_format,
+                    blend: None,
+                    write_mask: wgpu::ColorWrites::ALL,
+                })],
+            }),
+            primitive: Default::default(),
+            depth_stencil: None,
+            multisample: Default::default(),
+            multiview: None,
+        });
+
+        FastCrtFilter {
+            uniform_buffer,
+            bind_group,
+            pipeline: render_pipeline,
+        }
+    }
+}
+
+impl Filter for FastCrtFilter {
+    fn resize(&mut self, queue: &wgpu::Queue, new_size: PhysicalSize<u32>) {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(new_size),
+        };
+        queue.write_buffer(&self.uniform_buffer, 0, bytemuck::cast_slice(&[uniforms]));
+    }
+
+    fn render<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>) {
+        render_pass.set_pipeline(&self.pipeline);
+        render_pass.set_bind_group(0, &self.bind_group, &[]);
+        render_pass.draw(0..6, 0..1);
+    }
+}
+
+fn texture_scale_from_resolution(res: PhysicalSize<u32>) -> [f32; 4] {
+    let scale = ((res.width as f32) / 160.0).min((res.height as f32) / 120.0);
+    [
+        scale / res.width as f32,
+        scale / res.height as f32,
+        2.0 / scale,
+        0.0,
+    ]
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
+struct Uniforms {
+    texture_scale: [f32; 4],
+}

uw8-window/src/gpu/fast_crt.wgsl

@@ -0,0 +1,66 @@
+struct VertexOutput {
+    @builtin(position) clip_position: vec4<f32>,
+    @location(0) tex_coords: vec2<f32>,
+}
+
+struct Uniforms {
+    texture_scale: vec4<f32>,
+}
+
+@group(0) @binding(2) var<uniform> uniforms: Uniforms;
+
+@vertex
+fn vs_main(
+    @builtin(vertex_index) in_vertex_index: u32,
+) -> VertexOutput {
+    var out: VertexOutput;
+    let i = in_vertex_index / 3u + in_vertex_index % 3u;
+    let x = 0.0 + f32(i % 2u) * 320.0;
+    let y = 0.0 + f32(i / 2u) * 240.0;
+    out.clip_position = vec4<f32>((vec2<f32>(x, y) - vec2<f32>(160.0, 120.0)) * uniforms.texture_scale.xy, 0.0, 1.0);
+    out.tex_coords = vec2<f32>(x, y);
+    return out;
+}
+
+@group(0) @binding(0) var screen_texture: texture_2d<f32>;
+@group(0) @binding(1) var linear_sampler: sampler;
+
+fn row_factor(offset: f32) -> f32 {
+    return 1.0 / (1.0 + offset * offset * 16.0);
+}
+
+fn col_factor(offset: f32) -> f32 {
+    let offset = max(0.0, abs(offset) - 0.4);
+    return 1.0 / (1.0 + offset * offset * 16.0);
+}
+
+fn sample_screen(tex_coords: vec2<f32>) -> vec4<f32> {
+    let base = round(tex_coords) - vec2<f32>(0.5);
+    let frac = tex_coords - base;
+    
+    let top_factor = row_factor(frac.y);
+    let bottom_factor = row_factor(frac.y - 1.0);
+    
+    let v = base.y + bottom_factor / (bottom_factor + top_factor);
+    
+    let left_factor = col_factor(frac.x);
+    let right_factor = col_factor(frac.x - 1.0);
+    
+    let u = base.x + right_factor / (right_factor + left_factor);
+    
+    return textureSample(screen_texture, linear_sampler, vec2<f32>(u, v) / vec2<f32>(320.0, 240.0)) * (top_factor + bottom_factor) * (left_factor + right_factor) * 1.1;
+}
+
+@fragment
+fn fs_main(in: VertexOutput) -> @location(0) vec4<f32> {
+    return sample_screen(in.tex_coords);
+}
+
+@fragment
+fn fs_main_chromatic(in: VertexOutput) -> @location(0) vec4<f32> {
+    let r = sample_screen(in.tex_coords + vec2<f32>(0.2, 0.2)).r;
+    let g = sample_screen(in.tex_coords + vec2<f32>(0.07, -0.27)).g;
+    let b = sample_screen(in.tex_coords + vec2<f32>(-0.27, 0.07)).b;
+    return vec4<f32>(r, g, b, 1.0);
+}
+

uw8-window/src/gpu/mod.rs

@@ -0,0 +1,545 @@
+use crate::{Input, WindowConfig, WindowImpl};
+use anyhow::{anyhow, Result};
+use std::{num::NonZeroU32, time::Instant};
+
+use winit::{
+    dpi::PhysicalSize,
+    event::{Event, VirtualKeyCode, WindowEvent},
+    event_loop::{ControlFlow, EventLoop},
+    window::{Fullscreen, WindowBuilder},
+};
+
+use winit::platform::run_return::EventLoopExtRunReturn;
+
+mod crt;
+mod fast_crt;
+mod square;
+
+use crt::CrtFilter;
+use fast_crt::FastCrtFilter;
+use square::SquareFilter;
+
+pub struct Window {
+    _instance: wgpu::Instance,
+    surface: wgpu::Surface,
+    _adapter: wgpu::Adapter,
+    device: wgpu::Device,
+    queue: wgpu::Queue,
+    palette_screen_mode: PaletteScreenMode,
+    surface_config: wgpu::SurfaceConfiguration,
+    filter: Box<dyn Filter>,
+    event_loop: EventLoop<()>,
+    window: winit::window::Window,
+    gamepads: [u8; 4],
+    next_frame: Instant,
+    is_fullscreen: bool,
+    is_open: bool,
+}
+
+impl Window {
+    pub fn new(window_config: WindowConfig) -> Result<Window> {
+        async fn create(window_config: WindowConfig) -> Result<Window> {
+            let event_loop = EventLoop::new();
+            let window = WindowBuilder::new()
+                .with_inner_size(PhysicalSize::new(640u32, 480))
+                .with_min_inner_size(PhysicalSize::new(320u32, 240))
+                .with_title("MicroW8")
+                .with_fullscreen(if window_config.fullscreen {
+                    Some(Fullscreen::Borderless(None))
+                } else {
+                    None
+                })
+                .build(&event_loop)?;
+
+            window.set_cursor_visible(false);
+
+            let instance = wgpu::Instance::new(wgpu::Backends::all());
+            let surface = unsafe { instance.create_surface(&window) };
+            let adapter = instance
+                .request_adapter(&wgpu::RequestAdapterOptions {
+                    power_preference: wgpu::PowerPreference::LowPower,
+                    compatible_surface: Some(&surface),
+                    force_fallback_adapter: false,
+                })
+                .await
+                .ok_or_else(|| anyhow!("Request adapter failed"))?;
+
+            let (device, queue) = adapter
+                .request_device(&wgpu::DeviceDescriptor::default(), None)
+                .await?;
+
+            let palette_screen_mode = PaletteScreenMode::new(&device);
+
+            let surface_config = wgpu::SurfaceConfiguration {
+                usage: wgpu::TextureUsages::RENDER_ATTACHMENT,
+                format: surface.get_supported_formats(&adapter)[0],
+                width: window.inner_size().width,
+                height: window.inner_size().height,
+                present_mode: wgpu::PresentMode::AutoNoVsync,
+            };
+
+            let filter: Box<dyn Filter> = create_filter(
+                &device,
+                &palette_screen_mode.screen_view,
+                window.inner_size(),
+                surface_config.format,
+                window_config.filter,
+            );
+
+            surface.configure(&device, &surface_config);
+
+            Ok(Window {
+                event_loop,
+                window,
+                _instance: instance,
+                surface,
+                _adapter: adapter,
+                device,
+                queue,
+                palette_screen_mode,
+                surface_config,
+                filter,
+                gamepads: [0; 4],
+                next_frame: Instant::now(),
+                is_fullscreen: window_config.fullscreen,
+                is_open: true,
+            })
+        }
+
+        pollster::block_on(create(window_config))
+    }
+}
+
+impl WindowImpl for Window {
+    fn begin_frame(&mut self) -> Input {
+        let mut reset = false;
+        self.event_loop.run_return(|event, _, control_flow| {
+            *control_flow = ControlFlow::WaitUntil(self.next_frame);
+            let mut new_filter = None;
+            match event {
+                Event::WindowEvent { event, .. } => match event {
+                    WindowEvent::Resized(new_size) => {
+                        self.surface_config.width = new_size.width;
+                        self.surface_config.height = new_size.height;
+                        self.surface.configure(&self.device, &self.surface_config);
+                        self.filter.resize(&self.queue, new_size);
+                    }
+                    WindowEvent::CloseRequested => {
+                        self.is_open = false;
+                        *control_flow = ControlFlow::Exit;
+                    }
+                    WindowEvent::KeyboardInput { input, .. } => {
+                        fn gamepad_button(input: &winit::event::KeyboardInput) -> u8 {
+                            match input.scancode {
+                                44 => 16,
+                                45 => 32,
+                                30 => 64,
+                                31 => 128,
+                                _ => match input.virtual_keycode {
+                                    Some(VirtualKeyCode::Up) => 1,
+                                    Some(VirtualKeyCode::Down) => 2,
+                                    Some(VirtualKeyCode::Left) => 4,
+                                    Some(VirtualKeyCode::Right) => 8,
+                                    _ => 0,
+                                },
+                            }
+                        }
+                        if input.state == winit::event::ElementState::Pressed {
+                            match input.virtual_keycode {
+                                Some(VirtualKeyCode::Escape) => {
+                                    self.is_open = false;
+                                    *control_flow = ControlFlow::Exit;
+                                }
+                                Some(VirtualKeyCode::F) => {
+                                    let fullscreen = if self.window.fullscreen().is_some() {
+                                        None
+                                    } else {
+                                        Some(Fullscreen::Borderless(None))
+                                    };
+                                    self.is_fullscreen = fullscreen.is_some();
+                                    self.window.set_fullscreen(fullscreen);
+                                }
+                                Some(VirtualKeyCode::R) => reset = true,
+                                Some(VirtualKeyCode::Key1) => new_filter = Some(1),
+                                Some(VirtualKeyCode::Key2) => new_filter = Some(2),
+                                Some(VirtualKeyCode::Key3) => new_filter = Some(3),
+                                Some(VirtualKeyCode::Key4) => new_filter = Some(4),
+                                Some(VirtualKeyCode::Key5) => new_filter = Some(5),
+                                _ => (),
+                            }
+
+                            self.gamepads[0] |= gamepad_button(&input);
+                        } else {
+                            self.gamepads[0] &= !gamepad_button(&input);
+                        }
+                    }
+                    _ => (),
+                },
+                Event::RedrawEventsCleared => {
+                    if Instant::now() >= self.next_frame
+                        // workaround needed on Wayland until the next winit release
+                        && self.window.fullscreen().is_some() == self.is_fullscreen
+                    {
+                        *control_flow = ControlFlow::Exit
+                    }
+                }
+                _ => (),
+            }
+            if let Some(new_filter) = new_filter {
+                self.filter = create_filter(
+                    &self.device,
+                    &self.palette_screen_mode.screen_view,
+                    self.window.inner_size(),
+                    self.surface_config.format,
+                    new_filter,
+                );
+            }
+        });
+        Input {
+            gamepads: self.gamepads,
+            reset,
+        }
+    }
+
+    fn end_frame(&mut self, framebuffer: &[u8], palette: &[u8], next_frame: Instant) {
+        self.next_frame = next_frame;
+        self.palette_screen_mode
+            .write_framebuffer(&self.queue, framebuffer);
+        self.palette_screen_mode.write_palette(&self.queue, palette);
+
+        let output = self.surface.get_current_texture().unwrap();
+        let view = output
+            .texture
+            .create_view(&wgpu::TextureViewDescriptor::default());
+        let mut encoder = self
+            .device
+            .create_command_encoder(&wgpu::CommandEncoderDescriptor { label: None });
+
+        self.palette_screen_mode.resolve_screen(&mut encoder);
+
+        {
+            let mut render_pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
+                label: None,
+                color_attachments: &[Some(wgpu::RenderPassColorAttachment {
+                    view: &view,
+                    resolve_target: None,
+                    ops: wgpu::Operations {
+                        load: wgpu::LoadOp::Clear(wgpu::Color {
+                            r: 0.0,
+                            g: 0.0,
+                            b: 0.0,
+                            a: 1.0,
+                        }),
+                        store: true,
+                    },
+                })],
+                depth_stencil_attachment: None,
+            });
+
+            self.filter.render(&mut render_pass);
+        }
+
+        self.queue.submit(std::iter::once(encoder.finish()));
+        output.present();
+    }
+
+    fn is_open(&self) -> bool {
+        self.is_open
+    }
+}
+
+fn create_filter(
+    device: &wgpu::Device,
+    screen_texture: &wgpu::TextureView,
+    window_size: PhysicalSize<u32>,
+    surface_format: wgpu::TextureFormat,
+    filter: u32,
+) -> Box<dyn Filter> {
+    match filter {
+        1 => Box::new(SquareFilter::new(
+            device,
+            screen_texture,
+            window_size,
+            surface_format,
+        )),
+        2 => Box::new(FastCrtFilter::new(
+            device,
+            screen_texture,
+            window_size,
+            surface_format,
+            false,
+        )),
+        3 => Box::new(CrtFilter::new(
+            device,
+            screen_texture,
+            window_size,
+            surface_format,
+        )),
+        4 => Box::new(FastCrtFilter::new(
+            device,
+            screen_texture,
+            window_size,
+            surface_format,
+            true,
+        )),
+        _ => Box::new(AutoCrtFilter::new(
+            device,
+            screen_texture,
+            window_size,
+            surface_format,
+        )),
+    }
+}
+
+trait Filter {
+    fn resize(&mut self, queue: &wgpu::Queue, new_size: PhysicalSize<u32>);
+    fn render<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>);
+}
+
+struct AutoCrtFilter {
+    small: CrtFilter,
+    large: FastCrtFilter,
+    resolution: PhysicalSize<u32>,
+}
+
+impl AutoCrtFilter {
+    fn new(
+        device: &wgpu::Device,
+        screen: &wgpu::TextureView,
+        resolution: PhysicalSize<u32>,
+        surface_format: wgpu::TextureFormat,
+    ) -> AutoCrtFilter {
+        let small = CrtFilter::new(device, screen, resolution, surface_format);
+        let large = FastCrtFilter::new(device, screen, resolution, surface_format, true);
+        AutoCrtFilter {
+            small,
+            large,
+            resolution,
+        }
+    }
+}
+
+impl Filter for AutoCrtFilter {
+    fn resize(&mut self, queue: &wgpu::Queue, new_size: PhysicalSize<u32>) {
+        self.small.resize(queue, new_size);
+        self.large.resize(queue, new_size);
+        self.resolution = new_size;
+    }
+
+    fn render<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>) {
+        if self.resolution.width < 960 || self.resolution.height < 720 {
+            self.small.render(render_pass);
+        } else {
+            self.large.render(render_pass);
+        }
+    }
+}
+
+struct PaletteScreenMode {
+    framebuffer: wgpu::Texture,
+    palette: wgpu::Texture,
+    screen_view: wgpu::TextureView,
+    bind_group: wgpu::BindGroup,
+    pipeline: wgpu::RenderPipeline,
+}
+
+impl PaletteScreenMode {
+    fn new(device: &wgpu::Device) -> PaletteScreenMode {
+        let framebuffer_texture = device.create_texture(&wgpu::TextureDescriptor {
+            size: wgpu::Extent3d {
+                width: 320,
+                height: 240,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: wgpu::TextureDimension::D2,
+            format: wgpu::TextureFormat::R8Uint,
+            usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::COPY_DST,
+            label: None,
+        });
+
+        let palette_texture = device.create_texture(&wgpu::TextureDescriptor {
+            size: wgpu::Extent3d {
+                width: 256,
+                height: 1,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: wgpu::TextureDimension::D1,
+            format: wgpu::TextureFormat::Rgba8UnormSrgb,
+            usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::COPY_DST,
+            label: None,
+        });
+
+        let screen_texture = device.create_texture(&wgpu::TextureDescriptor {
+            size: wgpu::Extent3d {
+                width: 320,
+                height: 240,
+                depth_or_array_layers: 1,
+            },
+            mip_level_count: 1,
+            sample_count: 1,
+            dimension: wgpu::TextureDimension::D2,
+            format: wgpu::TextureFormat::Rgba8UnormSrgb,
+            usage: wgpu::TextureUsages::TEXTURE_BINDING | wgpu::TextureUsages::RENDER_ATTACHMENT,
+            label: None,
+        });
+
+        let framebuffer_texture_view =
+            framebuffer_texture.create_view(&wgpu::TextureViewDescriptor::default());
+        let palette_texture_view =
+            palette_texture.create_view(&wgpu::TextureViewDescriptor::default());
+        let screen_texture_view =
+            screen_texture.create_view(&wgpu::TextureViewDescriptor::default());
+
+        let palette_bind_group_layout =
+            device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+                entries: &[
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 0,
+                        visibility: wgpu::ShaderStages::FRAGMENT,
+                        ty: wgpu::BindingType::Texture {
+                            multisampled: false,
+                            view_dimension: wgpu::TextureViewDimension::D2,
+                            sample_type: wgpu::TextureSampleType::Uint,
+                        },
+                        count: None,
+                    },
+                    wgpu::BindGroupLayoutEntry {
+                        binding: 1,
+                        visibility: wgpu::ShaderStages::FRAGMENT,
+                        ty: wgpu::BindingType::Texture {
+                            multisampled: false,
+                            view_dimension: wgpu::TextureViewDimension::D1,
+                            sample_type: wgpu::TextureSampleType::Float { filterable: false },
+                        },
+                        count: None,
+                    },
+                ],
+                label: None,
+            });
+
+        let palette_bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+            layout: &palette_bind_group_layout,
+            entries: &[
+                wgpu::BindGroupEntry {
+                    binding: 0,
+                    resource: wgpu::BindingResource::TextureView(&framebuffer_texture_view),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 1,
+                    resource: wgpu::BindingResource::TextureView(&palette_texture_view),
+                },
+            ],
+            label: None,
+        });
+
+        let palette_shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+            label: None,
+            source: wgpu::ShaderSource::Wgsl(include_str!("palette.wgsl").into()),
+        });
+
+        let palette_pipeline_layout =
+            device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
+                label: None,
+                bind_group_layouts: &[&palette_bind_group_layout],
+                push_constant_ranges: &[],
+            });
+
+        let palette_pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+            label: None,
+            layout: Some(&palette_pipeline_layout),
+            vertex: wgpu::VertexState {
+                module: &palette_shader,
+                entry_point: "vs_main",
+                buffers: &[],
+            },
+            fragment: Some(wgpu::FragmentState {
+                module: &palette_shader,
+                entry_point: "fs_main",
+                targets: &[Some(wgpu::ColorTargetState {
+                    format: wgpu::TextureFormat::Rgba8UnormSrgb,
+                    blend: None,
+                    write_mask: wgpu::ColorWrites::ALL,
+                })],
+            }),
+            primitive: Default::default(),
+            depth_stencil: None,
+            multisample: Default::default(),
+            multiview: None,
+        });
+
+        PaletteScreenMode {
+            framebuffer: framebuffer_texture,
+            palette: palette_texture,
+            screen_view: screen_texture_view,
+            bind_group: palette_bind_group,
+            pipeline: palette_pipeline,
+        }
+    }
+
+    fn write_framebuffer(&self, queue: &wgpu::Queue, pixels: &[u8]) {
+        queue.write_texture(
+            wgpu::ImageCopyTexture {
+                texture: &self.framebuffer,
+                mip_level: 0,
+                origin: wgpu::Origin3d::ZERO,
+                aspect: wgpu::TextureAspect::All,
+            },
+            &bytemuck::cast_slice(pixels),
+            wgpu::ImageDataLayout {
+                offset: 0,
+                bytes_per_row: NonZeroU32::new(320),
+                rows_per_image: None,
+            },
+            wgpu::Extent3d {
+                width: 320,
+                height: 240,
+                depth_or_array_layers: 1,
+            },
+        );
+    }
+
+    fn write_palette(&self, queue: &wgpu::Queue, palette: &[u8]) {
+        queue.write_texture(
+            wgpu::ImageCopyTexture {
+                texture: &self.palette,
+                mip_level: 0,
+                origin: wgpu::Origin3d::ZERO,
+                aspect: wgpu::TextureAspect::All,
+            },
+            &bytemuck::cast_slice(palette),
+            wgpu::ImageDataLayout {
+                offset: 0,
+                bytes_per_row: NonZeroU32::new(256 * 4),
+                rows_per_image: None,
+            },
+            wgpu::Extent3d {
+                width: 256,
+                height: 1,
+                depth_or_array_layers: 1,
+            },
+        );
+    }
+
+    fn resolve_screen(&self, encoder: &mut wgpu::CommandEncoder) {
+        let mut render_pass = encoder.begin_render_pass(&wgpu::RenderPassDescriptor {
+            label: None,
+            color_attachments: &[Some(wgpu::RenderPassColorAttachment {
+                view: &self.screen_view,
+                resolve_target: None,
+                ops: wgpu::Operations {
+                    load: wgpu::LoadOp::Load,
+                    store: true,
+                },
+            })],
+            depth_stencil_attachment: None,
+        });
+
+        render_pass.set_pipeline(&self.pipeline);
+        render_pass.set_bind_group(0, &self.bind_group, &[]);
+        render_pass.draw(0..3, 0..1);
+    }
+}

uw8-window/src/gpu/palette.wgsl

@@ -0,0 +1,24 @@
+struct VertexOutput {
+    @builtin(position) clip_position: vec4<f32>,
+    @location(0) tex_coords: vec2<f32>,
+}
+
+@vertex
+fn vs_main(@builtin(vertex_index) vertex_index: u32) -> VertexOutput {
+    var out: VertexOutput;
+    let x = (1.0 - f32(vertex_index)) * 3.0;
+    let y = f32(vertex_index & 1u) * 3.0 - 1.0;
+    out.clip_position = vec4<f32>(x, y, 0.0, 1.0);
+    out.tex_coords = vec2<f32>((x + 1.0) * 160.0, (y + 1.0) * 120.0);
+    return out;
+}
+
+@group(0) @binding(0) var framebuffer_texture: texture_2d<u32>;
+@group(0) @binding(1) var palette_texture: texture_1d<f32>;
+
+@fragment
+fn fs_main(in: VertexOutput) -> @location(0) vec4<f32> {
+    let texel = vec2<i32>(floor(in.tex_coords));
+    let index = textureLoad(framebuffer_texture, texel, 0).r;
+    return textureLoad(palette_texture, i32(index), 0);
+}

uw8-window/src/gpu/square.rs

@@ -0,0 +1,157 @@
+use wgpu::util::DeviceExt;
+use winit::dpi::PhysicalSize;
+
+use super::Filter;
+
+pub struct SquareFilter {
+    uniform_buffer: wgpu::Buffer,
+    bind_group: wgpu::BindGroup,
+    pipeline: wgpu::RenderPipeline,
+}
+
+impl SquareFilter {
+    pub fn new(
+        device: &wgpu::Device,
+        screen: &wgpu::TextureView,
+        resolution: PhysicalSize<u32>,
+        surface_format: wgpu::TextureFormat,
+    ) -> SquareFilter {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(resolution),
+        };
+
+        let uniform_buffer = device.create_buffer_init(&wgpu::util::BufferInitDescriptor {
+            label: None,
+            contents: bytemuck::cast_slice(&[uniforms]),
+            usage: wgpu::BufferUsages::UNIFORM | wgpu::BufferUsages::COPY_DST,
+        });
+
+        let bind_group_layout = device.create_bind_group_layout(&wgpu::BindGroupLayoutDescriptor {
+            entries: &[
+                wgpu::BindGroupLayoutEntry {
+                    binding: 0,
+                    visibility: wgpu::ShaderStages::FRAGMENT,
+                    ty: wgpu::BindingType::Texture {
+                        multisampled: false,
+                        view_dimension: wgpu::TextureViewDimension::D2,
+                        sample_type: wgpu::TextureSampleType::Float { filterable: true },
+                    },
+                    count: None,
+                },
+                wgpu::BindGroupLayoutEntry {
+                    binding: 1,
+                    visibility: wgpu::ShaderStages::FRAGMENT,
+                    ty: wgpu::BindingType::Sampler(wgpu::SamplerBindingType::Filtering),
+                    count: None,
+                },
+                wgpu::BindGroupLayoutEntry {
+                    binding: 2,
+                    visibility: wgpu::ShaderStages::VERTEX_FRAGMENT,
+                    ty: wgpu::BindingType::Buffer {
+                        ty: wgpu::BufferBindingType::Uniform,
+                        has_dynamic_offset: false,
+                        min_binding_size: None,
+                    },
+                    count: None,
+                },
+            ],
+            label: None,
+        });
+
+        let sampler = device.create_sampler(&wgpu::SamplerDescriptor {
+            mag_filter: wgpu::FilterMode::Linear,
+            ..Default::default()
+        });
+
+        let bind_group = device.create_bind_group(&wgpu::BindGroupDescriptor {
+            layout: &bind_group_layout,
+            entries: &[
+                wgpu::BindGroupEntry {
+                    binding: 0,
+                    resource: wgpu::BindingResource::TextureView(&screen),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 1,
+                    resource: wgpu::BindingResource::Sampler(&sampler),
+                },
+                wgpu::BindGroupEntry {
+                    binding: 2,
+                    resource: uniform_buffer.as_entire_binding(),
+                },
+            ],
+            label: None,
+        });
+
+        let shader = device.create_shader_module(wgpu::ShaderModuleDescriptor {
+            label: None,
+            source: wgpu::ShaderSource::Wgsl(include_str!("square.wgsl").into()),
+        });
+
+        let render_pipeline_layout =
+            device.create_pipeline_layout(&wgpu::PipelineLayoutDescriptor {
+                label: None,
+                bind_group_layouts: &[&bind_group_layout],
+                push_constant_ranges: &[],
+            });
+
+        let render_pipeline = device.create_render_pipeline(&wgpu::RenderPipelineDescriptor {
+            label: None,
+            layout: Some(&render_pipeline_layout),
+            vertex: wgpu::VertexState {
+                module: &shader,
+                entry_point: "vs_main",
+                buffers: &[],
+            },
+            fragment: Some(wgpu::FragmentState {
+                module: &shader,
+                entry_point: "fs_main",
+                targets: &[Some(wgpu::ColorTargetState {
+                    format: surface_format,
+                    blend: None,
+                    write_mask: wgpu::ColorWrites::ALL,
+                })],
+            }),
+            primitive: Default::default(),
+            depth_stencil: None,
+            multisample: Default::default(),
+            multiview: None,
+        });
+
+        SquareFilter {
+            uniform_buffer,
+            bind_group,
+            pipeline: render_pipeline,
+        }
+    }
+}
+
+impl Filter for SquareFilter {
+    fn resize(&mut self, queue: &wgpu::Queue, new_size: PhysicalSize<u32>) {
+        let uniforms = Uniforms {
+            texture_scale: texture_scale_from_resolution(new_size),
+        };
+        queue.write_buffer(&self.uniform_buffer, 0, bytemuck::cast_slice(&[uniforms]));
+    }
+
+    fn render<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>) {
+        render_pass.set_pipeline(&self.pipeline);
+        render_pass.set_bind_group(0, &self.bind_group, &[]);
+        render_pass.draw(0..6, 0..1);
+    }
+}
+
+fn texture_scale_from_resolution(res: PhysicalSize<u32>) -> [f32; 4] {
+    let scale = ((res.width as f32) / 160.0).min((res.height as f32) / 120.0);
+    [
+        scale / res.width as f32,
+        scale / res.height as f32,
+        2.0 / scale,
+        0.0,
+    ]
+}
+
+#[repr(C)]
+#[derive(Debug, Copy, Clone, bytemuck::Pod, bytemuck::Zeroable)]
+struct Uniforms {
+    texture_scale: [f32; 4],
+}

uw8-window/src/gpu/square.wgsl

@@ -0,0 +1,44 @@
+struct VertexOutput {
+    @builtin(position) clip_position: vec4<f32>,
+    @location(0) tex_coords: vec2<f32>,
+}
+
+struct Uniforms {
+    texture_scale: vec4<f32>,
+}
+
+@group(0) @binding(2) var<uniform> uniforms: Uniforms;
+
+@vertex
+fn vs_main(
+    @builtin(vertex_index) in_vertex_index: u32,
+) -> VertexOutput {
+    var out: VertexOutput;
+    let i = in_vertex_index / 3u + in_vertex_index % 3u;
+    let x = 0.0 + f32(i % 2u) * 320.0;
+    let y = 0.0 + f32(i / 2u) * 240.0;
+    out.clip_position = vec4<f32>((vec2<f32>(x, y) - vec2<f32>(160.0, 120.0)) * uniforms.texture_scale.xy, 0.0, 1.0);
+    out.tex_coords = vec2<f32>(x, y);
+    return out;
+}
+
+@group(0) @binding(0) var screen_texture: texture_2d<f32>;
+@group(0) @binding(1) var linear_sampler: sampler;
+
+fn aa_tex_coord(c: f32) -> f32 {
+    let low = c - uniforms.texture_scale.z * 0.5;
+    let high = c + uniforms.texture_scale.z * 0.5;
+    let base = floor(low);
+    let center = base + 0.5;
+    let next = base + 1.0;
+    if high > next {
+        return center + (high - next) / (high - low);
+    } else {
+        return center;
+    }
+}
+
+@fragment
+fn fs_main(in: VertexOutput) -> @location(0) vec4<f32> {
+    return textureSample(screen_texture, linear_sampler, vec2<f32>(aa_tex_coord(in.tex_coords.x), aa_tex_coord(in.tex_coords.y)) / vec2<f32>(320.0, 240.0));
+}

uw8-window/src/lib.rs

@@ -0,0 +1,81 @@
+use anyhow::Result;
+use std::time::Instant;
+
+mod cpu;
+mod gpu;
+
+pub struct Window(Box<dyn WindowImpl>);
+
+impl Window {
+    pub fn new(config: WindowConfig) -> Result<Window> {
+        if config.enable_gpu {
+            match gpu::Window::new(config) {
+                Ok(window) => return Ok(Window(Box::new(window))),
+                Err(err) => eprintln!(
+                    "Failed to create gpu window: {}\nFalling back tp cpu window",
+                    err
+                ),
+            }
+        }
+        cpu::Window::new().map(|window| Window(Box::new(window)))
+    }
+
+    pub fn begin_frame(&mut self) -> Input {
+        self.0.begin_frame()
+    }
+    pub fn end_frame(&mut self, framebuffer: &[u8], palette: &[u8], next_frame: Instant) {
+        self.0.end_frame(framebuffer, palette, next_frame)
+    }
+
+    pub fn is_open(&self) -> bool {
+        self.0.is_open()
+    }
+}
+
+#[derive(Debug)]
+pub struct WindowConfig {
+    enable_gpu: bool,
+    filter: u32,
+    fullscreen: bool,
+}
+
+impl Default for WindowConfig {
+    fn default() -> WindowConfig {
+        WindowConfig {
+            enable_gpu: true,
+            filter: 5,
+            fullscreen: false,
+        }
+    }
+}
+
+impl WindowConfig {
+    pub fn parse_arguments(&mut self, args: &mut pico_args::Arguments) {
+        self.enable_gpu = !args.contains("--no-gpu");
+        if let Some(filter) = args.opt_value_from_str::<_, String>("--filter").unwrap() {
+            self.filter = match filter.as_str() {
+                "1" | "nearest" => 1,
+                "2" | "fast_crt" => 2,
+                "3" | "ss_crt" => 3,
+                "4" | "chromatic" => 4,
+                "5" | "auto_crt" => 5,
+                o => {
+                    println!("Unknown --filter '{}'", o);
+                    std::process::exit(1);
+                }
+            }
+        }
+        self.fullscreen = args.contains("--fullscreen");
+    }
+}
+
+pub struct Input {
+    pub gamepads: [u8; 4],
+    pub reset: bool,
+}
+
+trait WindowImpl {
+    fn begin_frame(&mut self) -> Input;
+    fn end_frame(&mut self, framebuffer: &[u8], palette: &[u8], next_frame: Instant);
+    fn is_open(&self) -> bool;
+}

uw8-window/src/main.rs

@@ -0,0 +1,59 @@
+use std::time::Instant;
+use uw8_window::WindowConfig;
+
+fn main() {
+    env_logger::Builder::from_env(env_logger::Env::default().default_filter_or("info")).init();
+
+    let mut args = pico_args::Arguments::from_env();
+
+    let mut framebuffer = vec![0u8; 320 * 240];
+    let mut start_time = Instant::now();
+
+    let mut palette = vec![0u32; 256];
+    for i in 0..256 {
+        let v = i & 15;
+        let r = ((i >> 2) & 12) * v;
+        let g = ((i >> 3) & 12) * v;
+        let b = ((i >> 4) & 12) * v;
+        palette[i as usize] = r + (g << 8) + (b << 16);
+    }
+
+    let mut fps_start = Instant::now();
+    let mut fps_counter = 0;
+
+    let mut window_config = WindowConfig::default();
+    window_config.parse_arguments(&mut args);
+
+    let mut window = uw8_window::Window::new(window_config).unwrap();
+
+    while window.is_open() {
+        let input = window.begin_frame();
+        if input.reset {
+            start_time = Instant::now();
+        }
+        draw_frame(&mut framebuffer, start_time.elapsed().as_secs_f32());
+        window.end_frame(&framebuffer, bytemuck::cast_slice(&palette), Instant::now());
+
+        fps_counter += 1;
+        let elapsed = fps_start.elapsed().as_secs_f32();
+        if elapsed >= 1.0 {
+            println!("{:.1} fps", fps_counter as f32 / elapsed);
+            fps_start = Instant::now();
+            fps_counter = 0;
+        }
+    }
+}
+
+fn draw_frame(framebuffer: &mut [u8], time: f32) {
+    for x in 0..320 {
+        let xr = x as f32 - 160.0;
+        for y in 0..240 {
+            let yr = y as f32 - 120.0;
+            let f = 8192.0 / (xr * xr + yr * yr);
+            let u = xr * f + 512.0 + time * 32.0;
+            let v = yr * f + time * 29.0;
+            let c = (u.floor() as i32 ^ v.floor() as i32) as u32;
+            framebuffer[x + y * 320] = c as u8;
+        }
+    }
+}

web/opus-repro.html

@@ -0,0 +1,57 @@
+<html>
+    <button onclick="go()">Go!</button>
+    <canvas id="screen" width="320" height="240"></canvas>
+    <video id="video"></video>
+    <script>
+        function go() {
+            let audioContext = new AudioContext({sampleRate: 44100});
+
+            let oscillator = new OscillatorNode(audioContext);
+            let gain = new GainNode(audioContext, {gain: 1});
+            oscillator.connect(gain);
+            gain.connect(audioContext.destination);
+            for(let i = 0; i < 8; ++i ) {
+                gain.gain.setValueAtTime(1, i / 2);
+                gain.gain.setValueAtTime(0, i / 2 + 0.3);
+            }
+            oscillator.start();
+            oscillator.stop(4);
+
+            let screen = document.getElementById('screen');
+            let context = screen.getContext('2d');
+            let startTime = Date.now();
+            let drawFrame = () => {
+                let time = Date.now() - startTime;
+                context.fillStyle = 'white';
+                context.fillRect(0, 0, 320, 240);
+                if(time < 4000) {
+                    if(time % 500 < 300) {
+                        context.fillStyle = 'black';
+                        context.fillRect(time / 15, 50, 50, 50);
+                    }
+                    window.requestAnimationFrame(drawFrame);
+                }
+            };
+            drawFrame();
+
+            let stream = screen.captureStream();
+            let audioStreamNode = audioContext.createMediaStreamDestination();
+            gain.connect(audioStreamNode);
+            stream.addTrack(audioStreamNode.stream.getAudioTracks()[0]);
+            let recorder = new MediaRecorder(stream, { mimeType: 'video/webm' });
+
+            let chunks = [];
+            recorder.ondataavailable = e => chunks.push(e.data);
+            recorder.onstop = () => {
+                let blob = new Blob(chunks, {type: 'video/webm'});
+                let url = URL.createObjectURL(blob);
+                let video = document.getElementById('video');
+                video.src = url;
+                video.play();
+            };
+
+            recorder.start();
+            setTimeout(() => recorder.stop(), 4000);
+        }
+    </script>
+</html>

web/run

@@ -0,0 +1,2 @@
+#!/bin/bash
+rm -rf .parcel-cache && yarn parcel src/index.html

web/src/audiolet.js

@@ -0,0 +1,94 @@
+let U8 = (...a) => new Uint8Array(...a);
+class APU extends AudioWorkletProcessor {
+    constructor() {
+        super();
+        this.sampleIndex = 0;
+        this.currentTime = 0;
+        this.isFirstMessage = true;
+        this.pendingUpdates = [];
+        this.port.onmessage = (ev) => {
+            if(this.memory) {
+                if(this.isFirstMessage)
+                {
+                    this.currentTime += (ev.data.t - this.currentTime) / 8;
+                    this.isFirstMessage = false;
+                }
+                this.pendingUpdates.push(ev.data);
+            } else {
+                this.load(ev.data[0], ev.data[1]);
+            }
+        };
+    }
+
+    async load(platform_data, data) {
+        let memory = new WebAssembly.Memory({ initial: 4, maximum: 4 });
+
+        let importObject = {
+            env: {
+                memory
+            },
+        };
+
+        for (let n of ['acos', 'asin', 'atan', 'atan2', 'cos', 'exp', 'log', 'sin', 'tan', 'pow']) {
+            importObject.env[n] = Math[n];
+        }
+
+        for (let i = 9; i < 64; ++i) {
+            importObject.env['reserved' + i] = () => { };
+        }
+
+        let logLine = '';
+        importObject.env['logChar'] = (c) => {
+            if(c == 10) {
+                console.log(logLine);
+                logLine = '';
+            } else {
+                logLine += String.fromCharCode(c);
+            }
+        };
+
+        for (let i = 0; i < 16; ++i) {
+            importObject.env['g_reserved' + i] = 0;
+        }
+
+        let instantiate = async (data) => (await WebAssembly.instantiate(data, importObject)).instance;
+
+        let platform_instance = await instantiate(platform_data);
+
+        for (let name in platform_instance.exports) {
+            importObject.env[name] = platform_instance.exports[name]
+        }
+
+        let instance = await instantiate(data);
+
+        this.memory = memory;
+
+        this.snd = instance.exports.snd || platform_instance.exports.gesSnd;
+
+        this.port.postMessage(2);
+    }
+
+    process(inputs, outputs, parameters) {
+        this.isFirstMessage = true;
+        if(this.snd) {
+            while(this.pendingUpdates.length > 0 && this.pendingUpdates[0].t <= this.currentTime) {
+                U8(this.memory.buffer, 80, 32).set(U8(this.pendingUpdates.shift().r));
+            }
+            let u32Mem = new Uint32Array(this.memory.buffer);
+            u32Mem[16] = this.currentTime;
+            let channels = outputs[0];
+            let index = this.sampleIndex;
+            let numSamples = channels[0].length;
+            for(let i = 0; i < numSamples; ++i) {
+                channels[0][i] = this.snd(index++);
+                channels[1][i] = this.snd(index++);
+            }
+            this.sampleIndex = index & 0xffffffff;
+            this.currentTime += numSamples / 44.1;
+        }
+
+        return true;
+    }
+}
+
+registerProcessor('apu', APU);

web/src/index.html

@@ -10,13 +10,14 @@
     </head>
     <body>
         <div id="uw8">
-            <a href="https://exoticorn.github.io/microw8">MicroW8</a> 0.1.1
+            <a href="https://exoticorn.github.io/microw8">MicroW8</a> 0.2.1
         </div>
         <div id="centered">
-            <canvas id="screen" width="320" height="240">
+            <canvas class="screen" id="screen" width="320" height="240">
             </canvas>
-        <div id="timer" hidden="true"></div>        
+            <button class="screen" id="start" style="display:none">Click to start</button>
-        <div id="message"></div>
+            <div id="timer" hidden="true"></div>        
+            <div id="message"></div>
             <button id="cartButton" style="visibility:hidden">Load cart...</button>
         </div>
         <div id="footer">

web/src/main.js

@@ -12,6 +12,7 @@ let uw8 = MicroW8(document.getElementById('screen'), {
     setMessage,
     keyboardElement: window,
     timerElement: document.getElementById("timer"),
+    startButton: document.getElementById("start")
 });
 
 function runModuleFromHash() {
@@ -79,7 +80,9 @@ if(location.hash.length != 0) {
             url += 'cart.uw8';
         }
         try {
-            await uw8.runModuleFromURL(url, true);
+            if(!await uw8.runModuleFromURL(url, true)) {
+                setupLoad();
+            }
         } catch(e) {
             setupLoad();
         }

web/src/microw8.js

@@ -1,5 +1,15 @@
 import loaderUrl from "data-url:../../platform/bin/loader.wasm";
 import platformUrl from "data-url:../../platform/bin/platform.uw8";
+import audioWorkletUrl from "data-url:./audiolet.js";
+
+class AudioNode extends AudioWorkletNode {
+    constructor(context) {
+        super(context, 'apu', {outputChannelCount: [2]});
+    }
+}
+
+let U8 = (...a) => new Uint8Array(...a);
+let U32 = (...a) => new Uint32Array(...a);
 
 export default function MicroW8(screen, config = {}) {
     if(!config.setMessage) {
@@ -18,9 +28,6 @@ export default function MicroW8(screen, config = {}) {
     
     let currentData;
     
-    let U8 = (d) => new Uint8Array(d);
-    let U32 = (d) => new Uint32Array(d);
-    
     let pad = 0;
     let keyboardElement = config.keyboardElement == undefined ? screen : config.keyboardElement;
     if(keyboardElement) {
@@ -84,12 +91,24 @@ export default function MicroW8(screen, config = {}) {
         keyboardElement.onkeyup = keyHandler;
     }
 
+    let audioContext;
+    let audioNode;
+    
     async function runModule(data, keepUrl) {
         if (cancelFunction) {
             cancelFunction();
             cancelFunction = null;
         }
     
+        audioContext = new AudioContext({sampleRate: 44100});
+        let keepRunning = true;
+        let abortController = new AbortController();
+        cancelFunction = () => {
+            audioContext.close();
+            keepRunning = false;
+            abortController.abort();
+        };
+
         let cartridgeSize = data.byteLength;
     
         config.setMessage(cartridgeSize);
@@ -97,6 +116,39 @@ export default function MicroW8(screen, config = {}) {
             return;
         }
     
+        await audioContext.audioWorklet.addModule(audioWorkletUrl);
+        audioNode = new AudioNode(audioContext);
+
+        let audioReadyFlags = 0;
+        let audioReadyResolve;
+        let audioReadyPromise = new Promise(resolve => audioReadyResolve = resolve);
+        let updateAudioReady = (f) => {
+            audioReadyFlags |= f;
+            if(audioReadyFlags == 3 && audioReadyResolve) {
+                audioReadyResolve(true);
+                audioReadyResolve = null;
+            }
+        };
+        let audioStateChange = () => {
+            if(audioContext.state == 'suspended') {
+                if(config.startButton) {
+                    config.startButton.style = '';
+                    screen.style = 'display:none';
+                }
+                (config.startButton || screen).onclick = () => {
+                    audioContext.resume();
+                };
+            } else {
+                if(config.startButton) {
+                    config.startButton.style = 'display:none';
+                    screen.style = '';
+                }
+                updateAudioReady(1);
+            }
+        };
+        audioContext.onstatechange = audioStateChange;
+        audioStateChange();
+
         currentData = data;
     
         let newURL = window.location.pathname;
@@ -119,7 +171,7 @@ export default function MicroW8(screen, config = {}) {
             if(!devkitMode) {
                 memSize.maximum = 4;
             }
-            let memory = new WebAssembly.Memory({ initial: 4, maximum: devkitMode ? 16 : 4 });
+            let memory = new WebAssembly.Memory(memSize);
             let memU8 = U8(memory.buffer);
     
             let importObject = {
@@ -142,9 +194,9 @@ export default function MicroW8(screen, config = {}) {
     
             let instantiate = async (data) => (await WebAssembly.instantiate(data, importObject)).instance;
     
-            let loadModuleURL = async (url) => instantiate(loadModuleData(await (await fetch(url)).arrayBuffer()));
+            let loadModuleURL = async (url) => loadModuleData(await (await fetch(url)).arrayBuffer());
     
-            loader = await loadModuleURL(loaderUrl);
+            loader = await instantiate(await loadModuleURL(loaderUrl));
     
             for (let n of ['acos', 'asin', 'atan', 'atan2', 'cos', 'exp', 'log', 'sin', 'tan', 'pow']) {
                 importObject.env[n] = Math[n];
@@ -154,13 +206,28 @@ export default function MicroW8(screen, config = {}) {
                 importObject.env['reserved' + i] = () => { };
             }
     
+            let logLine = '';
+            importObject.env['logChar'] = (c) => {
+                if(c == 10) {
+                    console.log(logLine);
+                    logLine = '';
+                } else {
+                    logLine += String.fromCharCode(c);
+                }
+            };
+    
             for (let i = 0; i < 16; ++i) {
                 importObject.env['g_reserved' + i] = 0;
             }
     
             data = loadModuleData(data);
     
-            let platform_instance = await loadModuleURL(platformUrl);
+            let platform_data = await loadModuleURL(platformUrl);
+
+            audioNode.port.onmessage = (e) => updateAudioReady(e.data);
+            audioNode.port.postMessage([platform_data, data]);
+
+            let platform_instance = await instantiate(platform_data);
     
             for (let name in platform_instance.exports) {
                 importObject.env[name] = platform_instance.exports[name]
@@ -170,13 +237,31 @@ export default function MicroW8(screen, config = {}) {
     
             let buffer = U32(imageData.data.buffer);
 
+            await audioReadyPromise;
+    
             let startTime = Date.now();
     
-            let keepRunning = true;
-            cancelFunction = () => keepRunning = false;
-    
             const timePerFrame = 1000 / 60;
-            let nextFrame = startTime;
+
+            audioNode.connect(audioContext.destination);
+
+            let isPaused = false;
+            let pauseTime = startTime;
+            let updateVisibility = isVisible => {
+                let now = Date.now();
+                if(isVisible) {
+                    isPaused = false;
+                    audioContext.resume();
+                    startTime += now - pauseTime;
+                } else {
+                    isPaused = true;
+                    audioContext.suspend();
+                    pauseTime = now;
+                }
+            };
+            window.addEventListener('focus', () => updateVisibility(true), { signal: abortController.signal });
+            window.addEventListener('blur', () => updateVisibility(false), { signal: abortController.signal });
+            updateVisibility(document.hasFocus());
 
             function mainloop() {
                 if (!keepRunning) {
@@ -184,9 +269,9 @@ export default function MicroW8(screen, config = {}) {
                 }
     
                 try {
-                    let now = Date.now();
                     let restart = false;
-                    if (now >= nextFrame) {
+                    let thisFrame;
+                    if (!isPaused) {
                         let gamepads = navigator.getGamepads();
                         let gamepad = 0;
                         for (let i = 0; i < 4; ++i) {
@@ -214,23 +299,36 @@ export default function MicroW8(screen, config = {}) {
                         }
     
                         let u32Mem = U32(memory.buffer);
-                        u32Mem[16] = now - startTime;
+                        let time = Date.now() - startTime;
+                        u32Mem[16] = time;
                         u32Mem[17] = pad | gamepad;
-                        instance.exports.upd();
+                        if(instance.exports.upd) {
+                            instance.exports.upd();
+                        }
                         platform_instance.exports.endFrame();
 
-                        let palette = U32(memory.buffer.slice(0x13000, 0x13000 + 1024));
+                        let soundRegisters = new ArrayBuffer(32);
+                        U8(soundRegisters).set(U8(memory.buffer, 80, 32));
+                        audioNode.port.postMessage({t: time, r: soundRegisters}, [soundRegisters]);
+    
+                        let palette = U32(memory.buffer, 0x13000, 1024);
                         for (let i = 0; i < 320 * 240; ++i) {
                             buffer[i] = palette[memU8[i + 120]] | 0xff000000;
                         }
                         canvasCtx.putImageData(imageData, 0, 0);
-                        nextFrame = Math.max(nextFrame + timePerFrame, now);
+
+                        let timeOffset = ((time * 6) % 100 - 50) / 6;
+                        thisFrame = startTime + time - timeOffset / 8;
+                    } else {
+                        thisFrame = Date.now();
                     }
+                    let now = Date.now();
+                    let nextFrame = Math.max(thisFrame + timePerFrame, now);
     
                     if (restart) {
                         runModule(currentData);
                     } else {
-                        window.requestAnimationFrame(mainloop);
+                        window.setTimeout(mainloop, nextFrame - now)
                     }
                 } catch (err) {
                     config.setMessage(cartridgeSize, err.toString());
@@ -253,14 +351,25 @@ export default function MicroW8(screen, config = {}) {
     
     let videoRecorder;
     let videoStartTime;
+    let videoAudioSourceNode;
+    let videoAudioStreamNode;
     function recordVideo() {
         if(videoRecorder) {
             videoRecorder.stop();
             videoRecorder = null;
+            videoAudioSourceNode.disconnect(videoAudioStreamNode);
+            videoAudioSourceNode = null;
+            videoAudioStreamNode = null;
             return;
         }
 
-        videoRecorder = new MediaRecorder(screen.captureStream(), {
+        let stream = screen.captureStream();
+        videoAudioStreamNode = audioContext.createMediaStreamDestination();
+        videoAudioSourceNode = audioNode;
+        audioNode.connect(videoAudioStreamNode);
+        stream.addTrack(videoAudioStreamNode.stream.getAudioTracks()[0]);
+    
+        videoRecorder = new MediaRecorder(stream, {
             mimeType: 'video/webm',
             videoBitsPerSecond: 25000000
         });
@@ -305,10 +414,11 @@ export default function MicroW8(screen, config = {}) {
     async function runModuleFromURL(url, keepUrl) {
         let response = await fetch(url);
         let type = response.headers.get('Content-Type');
-        if(type && type.includes('html')) {
+        if((type && type.includes('html')) || response.status != 200) {
-            throw false;
+            return false;
         }
         runModule(await response.arrayBuffer(), keepUrl || devkitMode);
+        return true;
     }
 
     return {

web/src/style.css

@@ -37,7 +37,7 @@ a:hover {
     color: #405040;
 }
 
-#screen {
+.screen {
     width: 320px;
     height: 240px;
     image-rendering: pixelated;
@@ -45,9 +45,16 @@ a:hover {
     margin-bottom: 8px;
     border: 4px solid #303040;
     box-shadow: 5px 5px 20px black;
+}
+
+#screen {
     cursor: none;
 }
 
+#start {
+    font-size: 150%;
+}
+
 #timer::before {
     content: '';
     display: inline-block;
@@ -84,21 +91,21 @@ button:active {
 }
 
 @media (min-width: 680px) and (min-height: 620px) {
-    #screen {
+    .screen {
         width: 640px;
         height: 480px;
     }
 }
 
 @media (min-width: 1000px) and (min-height: 800px) {
-    #screen {
+    .screen {
         width: 960px;
         height: 720px;
     }
 }
 
 @media (width:640px) and (height:480px) {
-    #screen {
+    .screen {
         width: 640px;
         height: 480px;
         border: 0;