Create assembler driver script for recording word addresses

This will allow the profiler to sample the current return stack and
instruction pointer.  It can't just use the ip and rsp globals
directly, as I had initially planned, since these cannot be sent
in between threads.

README.md

@@ -1,31 +1,24 @@
 # Wipforth
 
-Wipforth is a simple Forth implementation that runs in the WebAssembly
-virtual machine.  It does I/O via memory-mapped peripherals, which are
-emulated in JavaScript.
+Wipforth is a Forth implementation that runs in the WebAssembly
+virtual machine.  The system is bootstrapped from source on page load:
+the only non-text file is the favicon :)
 
-- For the Forth kernel, see [wipforth.wat](./wipforth.wat)
-- For the JavaScript emulator, see [emu.js](./emu.js)
-- For the Forth prelude, which is loaded at start-up, see
-  [prelude.f](./prelude.f)
+I/O is done via memory-mapped peripherals, which are emulated in
+JavaScript.
+
+- For the Forth kernel, see [wipforth.ws](./wipforth.ws)
+- For the emulator, see [emu.js](./emu.js)
+- For the assembler, see [asm.js](./asm.js)
+- For the prelude (Forth code loaded right after the kernel boots),
+  see [prelude.f](./prelude.f)
 - For a description of the peripherals, see the
   [Peripherals](#peripherals) section below.
 
 ## Building and Running Locally
 
-You'll need:
-
-- [WABT](https://github.com/WebAssembly/wabt) (not for long mwahaha)
-- [Guile](https://www.gnu.org/software/guile/) (or bring your own HTTP
-  server -- see note below)
-
-To run, first compile the WebAssembly module:
-
-```
-wat2wasm --enable-threads wipforth.wat
-```
-
-Then run the development server:
+There's a [Guile](https://www.gnu.org/software/guile/) script in the
+repo you can use for this:
 
 ```
 guile server.scm
@@ -34,14 +27,20 @@ guile server.scm
 You should then be able to open <http://localhost:8080> in a browser
 and use the system from there.
 
-**NOTE**: The server is very simple and just serves the files with the
-cross-origin isolation headers required for `SharedMemoryBuffer` use.
-You could use any HTTP server that sets these headers.
+However, since everything is bootstrapped on the client, basically any
+HTTP server will do as long as it sets the appropriate response
+headers for `SharedMemoryBuffer` use:
 
-You should **definitely not** use the development server to serve the
-application on the open internet; I just hacked it together for
-testing on localhost during development and it's probably hilariously
-insecure.
+- `Cross-Origin-Opener-Policy: same-origin`
+- `Cross-Origin-Embedder-Policy: require-corp`
+
+So, if you don't have Guile on your system you can use something else
+like Python's `http.server`.
+
+**NOTE**: You should **definitely not** use `server.scm` to serve the
+application on the open internet or anything like that; I just hacked
+it together for testing on localhost during development and it's
+probably hilariously insecure.
 
 ## End-to-End Tests
 
@@ -62,16 +61,13 @@ Given that's all sorted, you should be able to run:
 guile tests.scm
 ```
 
-It will print a JUnit XML report to standard out, you can pretty-print
-it with:
+It will print a JUnit XML report to standard out.  You can
+pretty-print it with `xmllint` if you have it installed:
 
 ```
 guile tests.scm | xmllint --format -
 ```
 
-Though, of course, this will require that you have `xmllint` on your
-system.
-
 ## Peripherals
 
 ### Terminal

asm.js

@@ -59,12 +59,14 @@ class Tokenizer {
         let token;
         while (token = this.next()) {
             if (token.string == undefined
-				&& token.startsWith(this.comment_start))
+				&& token.startsWith(this.comment_start)) {
                 this.comment = true;
-            else if (this.comment && token == LINE_END)
+            } else if (this.comment && token == LINE_END) {
                 this.comment = false;
-            else if (!this.comment)
+				yield token;
+            } else if (!this.comment) {
                 yield token;
+			}
         }
 	}
 }
@@ -106,6 +108,7 @@ const State = Object.freeze({
     ELEM_TABLE:   33,
     ELEM_ELEM:    34,
     ELEM_LABEL:   35,
+    ZERO:         36,
 });
 
 const Action = Object.freeze({
@@ -141,54 +144,61 @@ const types = {
 };
 
 const opcodes = {
-    "block":         0x02,
-    "loop":          0x03,
-    "if":            0x04,
-    "else":          0x05,
-    "end":           0x0b,
-    "br":            0x0c,
-    "br_if":         0x0d,
-    "call":          0x10,
-    "call_indirect": 0x11,
-    "drop":          0x0a,
-    "local.get":     0x20,
-    "local.set":     0x21,
-    "local.tee":     0x22,
-    "global.get":    0x23,
-    "global.set":    0x24,
-    "i32.load":      0x28,
-    "i32.load8_u":   0x2d,
-    "i32.store":     0x36,
-    "i32.store8":    0x3a,
-    "i32.const":     0x41,
-    "i32.eqz":       0x45,
-    "i32.eq":        0x46,
-    "i32.ne":        0x47,
-    "i32.lt_s":      0x48,
-    "i32.lt_u":      0x49,
-    "i32.gt_s":      0x4a,
-    "i32.gt_u":      0x4b,
-    "i32.le_s":      0x4c,
-    "i32.le_u":      0x4d,
-    "i32.ge_s":      0x4e,
-    "i32.ge_u":      0x4f,
-    "i32.add":       0x6a,
-    "i32.sub":       0x6b,
-    "i32.mul":       0x6c,
-    "i32.and":       0x71,
-    "i32.or":        0x72,
-    "i32.xor":       0x73,
-    "i32.shl":       0x74,
-    "i32.shr_s":     0x75,
-    "i32.shr_u":     0x76,
+    "block":            0x02,
+    "loop":             0x03,
+    "if":               0x04,
+    "else":             0x05,
+    "end":              0x0b,
+    "br":               0x0c,
+    "br_if":            0x0d,
+    "call":             0x10,
+    "call_indirect":    0x11,
+    "drop":             0x0a,
+    "local.get":        0x20,
+    "local.set":        0x21,
+    "local.tee":        0x22,
+    "global.get":       0x23,
+    "global.set":       0x24,
+    "i32.load":         0x28,
+    "i32.load8_u":      0x2d,
+    "i32.store":        0x36,
+    "i32.store8":       0x3a,
+    "i32.const":        0x41,
+    "i64.const":        0x42,
+    "i32.eqz":          0x45,
+    "i32.eq":           0x46,
+    "i32.ne":           0x47,
+    "i32.lt_s":         0x48,
+    "i32.lt_u":         0x49,
+    "i32.gt_s":         0x4a,
+    "i32.gt_u":         0x4b,
+    "i32.le_s":         0x4c,
+    "i32.le_u":         0x4d,
+    "i32.ge_s":         0x4e,
+    "i32.ge_u":         0x4f,
+    "i32.add":          0x6a,
+    "i32.sub":          0x6b,
+    "i32.mul":          0x6c,
+	"i32.div_s":        0x6d,
+	"i32.rem_s":        0x6f,
+    "i32.and":          0x71,
+    "i32.or":           0x72,
+    "i32.xor":          0x73,
+    "i32.shl":          0x74,
+    "i32.shr_s":        0x75,
+    "i32.shr_u":        0x76,
+    "i64.or":           0x84,
+    "i64.shl":          0x86,
+    "i64.extend_i32_u": 0xad,
 
     // Threads instructions
-    "memory.atomic.notify":  [ 0xfe, 0x00 ],
-    "memory.atomic.wait32":  [ 0xfe, 0x01 ],
-    "memory.atomic.load":    [ 0xfe, 0x10 ],
-    "memory.atomic.load8_u": [ 0xfe, 0x12 ],
-    "memory.atomic.store":   [ 0xfe, 0x17 ],
-    "memory.atomic.store8":  [ 0xfe, 0x19 ],
+    "memory.atomic.notify": [ 0xfe, 0x00 ],
+    "memory.atomic.wait32": [ 0xfe, 0x01 ],
+    "i32.atomic.load":      [ 0xfe, 0x10 ],
+    "i32.atomic.load8_u":   [ 0xfe, 0x12 ],
+    "i32.atomic.store":     [ 0xfe, 0x17 ],
+    "i64.atomic.store":     [ 0xfe, 0x18 ],
+    "i32.atomic.store8":    [ 0xfe, 0x19 ],
 };
 
 const mem_flags = {
@@ -225,6 +235,7 @@ class Parser {
             ".type":   State.TYPE_NAME,
             ".table":  State.TABLE_NAME,
             ".elem":   State.ELEM_TABLE,
+            ".zero":   State.ZERO,
         };
         this.blocks = new Set(["block", "loop", "if"]);
         this.handlers = {
@@ -264,6 +275,7 @@ class Parser {
             [State.ELEM_TABLE]:   (token) => this.token_elem_table(token),
             [State.ELEM_ELEM]:    (token) => this.token_elem_elem(token),
             [State.ELEM_LABEL]:   (token) => this.token_elem_label(token),
+            [State.ZERO]:         (token) => this.token_zero(token),
         };
 
         this.results = [];
@@ -272,15 +284,14 @@ class Parser {
 	}
 
     integer(token) {
-        let base;
+        let base, regex;
         switch (token.slice(-1)) {
-        case "b": base = 2; break;
-        case "o": base = 8; break;
-        case "h": base = 16; break;
-        default:  base = 10; break;
+        case "b": base = 2;  regex = /^-?[01]+b$/;     break;
+        case "o": base = 8;  regex = /^-?[0-7]+o$/;    break;
+        case "h": base = 16; regex = /^-?[0-9A-F]+h$/; break;
+        default:  base = 10; regex = /^-?[0-9]+d?$/;   break;
         }
-        const x = parseInt(token, base);
-        return Number.isNaN(x) ? null : x;
+		return regex.test(token) ? parseInt(token, base) : null;
     }
 
     translate_code(token) {
@@ -547,15 +558,10 @@ class Parser {
 
     token_at_addr(token) {
         const value = this.integer(token);
-        if (value == null) {
-            console.error(
-                `ERROR: Unexpected token ${token} in .mem: `
-                    + "expected address");
-            this.at = undefined;
-            return;
-        }
-
-        this.at.addr = value;
+        if (value != null)
+            this.at.addr = value;
+        else
+            this.at.addr_symbol = token;
         const action = { type: Action.AT, at: this.at };
         this.at = undefined;
         this.state = State.TOP;
@@ -823,6 +829,26 @@ class Parser {
         return action;
     }
 
+	token_zero(token) {
+		if (token == LINE_END) {
+			console.error(
+				"ERROR: Unexpected newline in .zero, expected count")
+			this.state = State.TOP;
+			return;
+		}
+
+		const count = this.integer(token);
+		if (count == null) {
+			console.error(
+				`ERROR: Unexpected token ${token} in .zero, expected count`);
+			this.state = State.TOP;
+			return;
+		}
+
+		this.state = State.TOP;
+		return { type: Action.DATA, size: count, value: 0 }
+	}
+
     mem_action() {
         const action = {
             type: Action.MEM,
@@ -864,8 +890,9 @@ const Section = Object.freeze({
 });
 
 const Kind = Object.freeze({
-    FUNC: 0x00,
-    MEM:  0x02,
+    FUNC:   0x00,
+    MEM:    0x02,
+    GLOBAL: 0x03,
 });
 
 export class Assembler {
@@ -909,6 +936,7 @@ export class Assembler {
         this.types = [];
         this.type_bindings = {};
         this.tables = {};
+		this.unresolved = [];
 	}
 
     action_append(action) {
@@ -919,8 +947,28 @@ export class Assembler {
     }
 
     action_export(action) {
-        const index = Object.keys(this.funcs).indexOf(action.name);
-        this.exports[action.name] = { kind: Kind.FUNC, index };
+        const func_index = Object.keys(this.funcs).indexOf(action.name);
+        if (func_index != -1) {
+            this.exports[action.name] = {
+                kind: Kind.FUNC,
+                index: func_index,
+            };
+            return;
+        }
+
+        const global_index = Object.keys(this.globals).indexOf(action.name);
+        if (global_index != -1) {
+            this.exports[action.name] = {
+                kind: Kind.GLOBAL,
+                index: global_index,
+            };
+            return;
+        }
+
+        console.error(
+            `ERROR: Unable to resolve export ${action.name} `
+                + "(only functions and globals currently supported)"
+        );
     }
 
     action_func(action) {
@@ -963,7 +1011,7 @@ export class Assembler {
             }
             func.body.push(...this.leb128(def_value));
         } else {
-            func.body.push(...this.leb128(value));
+            func.body.push(...this.uleb128(value));
         }
     }
 
@@ -995,7 +1043,8 @@ export class Assembler {
             return;
         }
         this.pos.mem = mem;
-        this.pos.addr = action.at.addr;
+        this.pos.addr = action.at.addr
+            ?? this.lookup_def(action.at.addr_symbol);
         this.data.push({ loc: { ...this.pos }, data: [] })
     }
 
@@ -1013,9 +1062,14 @@ export class Assembler {
 				}
 				value = this.lookup_def(action.symbol);
 				if (value == undefined) {
-					console.error(
-						`ERROR: Unable to resolve symbol ${action.symbol}`);
-					return;
+					this.unresolved.push({
+						type: "data",
+						size: action.size,
+						symbol: action.symbol,
+						target: data,
+						offset: data.length,
+					});
+					value = 0;
 				}
 			}
 			bytes = this.le(value, action.size);
@@ -1303,7 +1357,7 @@ export class Assembler {
                 opcodes["i32.const"],
                 ...this.leb128(loc.addr),
                 opcodes["end"],
-                data.length,
+                ...this.uleb128(data.length),
                 ...data,
             ]
         });
@@ -1311,6 +1365,7 @@ export class Assembler {
     }
 
 	wasm() {
+		this.resolve_refs();
         this.resolve_func_types();
 
         const template = [
@@ -1327,7 +1382,9 @@ export class Assembler {
         ];
         const sections = template.map(([ code, generator ]) => {
             const body = generator();
-            return body == null ? [] : [ code, body.length, body ];
+            if (body == null)
+                return [];
+            return [ code, this.uleb128(body.length), body ];
         });
 
 		return new Uint8Array([ HEADER, sections ].flat(Infinity));
@@ -1377,4 +1434,32 @@ export class Assembler {
 			return acc;
 		}, []);
     }
+
+	resolve_refs() {
+		const failed = [];
+		for (const ref of this.unresolved) {
+			if (ref.type != "data") {
+				console.error(
+					`ERROR: Unsupported ref type ${ref.type} for `
+						+ `symbol ${ref.symbol}`
+				);
+				failed.push(ref.symbol);
+				continue;
+			}
+
+			const value = this.defs[ref.symbol];
+			if (value == undefined) {
+				failed.push(ref.symbol);
+				continue;
+			}
+
+			const bytes = this.le(value, ref.size);
+			ref.target.splice(ref.offset, ref.size, ...bytes);
+		}
+
+		if (failed.length != 0) {
+			const failed_str = failed.join(" ");
+			console.error(`ERROR: Unable to resolve refs: ${failed_str}`);
+		}
+	}
 }

boot.js

@@ -1,7 +1,26 @@
+import { Assembler } from './asm.js';
+
+const assemble = (async () => {
+	const asm = new Assembler();
+	const resp = await fetch('wipforth.ws');
+	for await (const chunk of resp.body) {
+		asm.push(chunk);
+	}
+	return asm.wasm();
+})();
+
 self.onmessage = async (e) => {
-	const exports = { emu: { mem: e.data } };
-	const mod = await WebAssembly.instantiateStreaming(
-		fetch('wipforth.wasm'), exports)
-	mod.instance.exports.reset();
-	console.log('System halt');
+	switch (e.data.type) {
+	case "load":
+		const exports = { emu: { mem: e.data.mem } };
+		const wasm = await assemble;
+		self.mod = await WebAssembly.instantiate(wasm, exports);
+		await self.postMessage('ready');
+		break;
+
+	case "boot":
+		self.mod.instance.exports.reset();
+		console.log('System halt');
+		break;
+	}
 };

deploy-manifest.conf

@@ -3,4 +3,4 @@ emu.js
 index.html
 prelude.f
 styles.css
-wipforth.wasm
+wipforth.ws

emu.js

@@ -11,6 +11,7 @@ const RXBUF_SIZE = 32;
 const PERIPHS_SIZE = 81;
 
 const POLL_INTERVAL_MS = 20;
+const DOT_INTERVAL_MS = 120;
 
 const COLS = 80;
 const TAB_WIDTH = 8;
@@ -50,8 +51,29 @@ class Emulator {
 		document.addEventListener('keydown', (e) => this.handle_keydown(e));
 		window.addEventListener('resize', () => this.handle_resize());
 
-		this.worker = new Worker('boot.js');
-		this.worker.postMessage(this.mem);
+		this.prof = new Worker("prof.js");
+		this.prof.onmessage = (e) => {
+			const blob = new Blob(
+				[JSON.stringify(e.data)],
+				{ type: "application/json" });
+			const url = URL.createObjectURL(blob);
+			const a = document.createElement("a");
+			a.href = url;
+			a.download = "wipforth-profile.json";
+			a.click();
+			URL.revokeObjectURL(url);
+		};
+
+		this.print("Assembling kernel ");
+		const dots = setInterval(() => this.print("."), DOT_INTERVAL_MS);
+		this.worker = new Worker('boot.js', { type: 'module' });
+		this.worker.postMessage({ type: "load", mem: this.mem });
+		this.worker.onmessage = (e) => {
+			clearInterval(dots);
+			this.print(" done\n");
+			this.worker.postMessage({ type: "boot" });
+			this.prof.postMessage({ type: "start", mem: this.mem });
+		};
 
 		fetch('prelude.f')
 			.then(res => res.text())
@@ -77,6 +99,7 @@ class Emulator {
 		if (!this.input_enable) {
 			const sysready = Atomics.load(this.mem_u8, SYSREADY);
 			if (sysready != 0) {
+				this.prof.postMessage({ type: "stop" });
 				this.input_enable = true;
 				this.flush_output();
 				document.getElementById('cursor').classList.add('blinking');

prelude.f

@@ -249,8 +249,8 @@ CHAR . EMIT
 \ Version number
 
 0 CONSTANT VERSION-MAJOR
-1 CONSTANT VERSION-MINOR
-0 CONSTANT VERSION-PATCH
+2 CONSTANT VERSION-MINOR
+1 CONSTANT VERSION-PATCH
 
 : PRINT-VERSION
   CHAR v EMIT VERSION-MAJOR .

prof.js

@@ -0,0 +1,48 @@
+const INTERVAL_MS = 1;
+const RS_TOP_ADDR = 0x10000;
+const PROF_DATA_ADDR = 0x58;
+const PROF_DATA_IDX = PROF_DATA_ADDR / 8;
+
+let mem_8;
+let mem_64;
+let sampler;
+
+const samples = [];
+
+function sample() {
+	const data = Atomics.load(mem_64, PROF_DATA_IDX);
+	const ip = Number(data & 0xffffffffn);
+	const rsp = Number(data >> 32n);
+	samples.push({ ip, rs_bytes: mem_8.slice(rsp, RS_TOP_ADDR) });
+}
+
+function i32(bytes) {
+	return bytes[0]
+		| (bytes[1] << 8)
+		| (bytes[2] << 16)
+		| (bytes[3] << 24);
+}
+function postproc({ ip, rs_bytes }) {
+	const rs = [];
+	for (let i = 0; i < rs_bytes.length; i += 4)
+		rs.push(i32(rs_bytes.slice(i, i + 4)));
+	rs.reverse();
+	return { ip, rs };
+}
+
+self.onmessage = (e) => {
+	switch (e.data.type) {
+	case "start":
+		console.log("Starting profiler");
+		mem_8 = new Uint8Array(e.data.mem.buffer);
+		mem_64 = new BigUint64Array(e.data.mem.buffer);
+		ip = e.data.ip;
+		rsp = e.data.rsp;
+		sampler = setInterval(sample, INTERVAL_MS);
+		break;
+	case "stop":
+		clearInterval(sample);
+		console.log("Stopped profiler");
+		self.postMessage(samples.map(postproc));
+	}
+};

server.scm

@@ -16,9 +16,8 @@
   '(("html" . (text/html))
 	("css"  . (text/css))
 	("js"   . (application/javascript))
-	("wasm" . (application/wasm))
 	("f"    . (text/plain))
-	("wat"  . (text/plain))
+	("ws"   . (text/plain))
 	("png"  . (image/png))))
 
 (define (mime-type path)

tests.scm

@@ -21,11 +21,17 @@
 (navigate client "http://localhost:8080")
 (sleep 5)
 
+(define-test kernel-assembles-successfully
+  (let* ((display (get-display client))
+		 (line (first (lines display))))
+	(assert (string-match "Assembling kernel \\.+ done" line)
+			(format #f "Kernel assemble line: ~s" line))))
+
 (define-test prelude-loads-successfully
   (let* ((display (get-display client))
-		 (first-line (first (lines display))))
-	(assert (string-match "Loading prelude \\.+ done" first-line)
-			(format #f "Prelude load line: ~s" first-line))))
+		 (line (second (lines display))))
+	(assert (string-match "Loading prelude \\.+ done" line)
+			(format #f "Prelude load line: ~s" line))))
 
 (define-test six-seven-times-dot-cr-yields-42
   (input-line client "6 7 * . CR")

words.js

@@ -0,0 +1,16 @@
+import { Assembler } from "./asm.js";
+
+const asm = new Assembler();
+for await (const chunk of Deno.stdin.readable) {
+	asm.push(chunk);
+}
+asm.wasm();
+
+const defs = Object.entries(asm.defs);
+while (defs[0][0] != '_DUP')
+	defs.shift();
+while (defs.at(-1)[0] != 'WNF_HANDLER')
+	defs.pop();
+const words = Object.fromEntries(defs.filter(([k,v]) => !k.startsWith("_")));
+
+console.log(JSON.stringify(words));