Merge pull request #1181 from benaadams/simd-fixes

SIMD fixes

Tools/Gulp/package.json

@@ -8,7 +8,7 @@
   "license": "(Apache-2.0)",
   "devDependencies": {
     "gulp": "^3.8.11",
-    "gulp-uglify": "~1.2.0",
+    "gulp-uglify": "~1.5.3",
     "gulp-sourcemaps": "~1.5.2",
     "typescript": "^1.7.5",
     "gulp-typescript": "~2.13.0",

src/Math/babylon.math.SIMD.ts

@@ -1,255 +1,236 @@
 module BABYLON {
 
-    declare var SIMD;
-
-    export class SIMDVector3 {
+    declare var SIMD: SIMD;
+    var f4 = SIMD.Float32x4;
+    var load = f4.load;
+    var add = f4.add;
+    var sub = f4.sub;
+    var mul = f4.mul;
+    var div = f4.div;
+    var swizzle = f4.swizzle;
+    var splat = f4.splat;
+    var store = f4.store;
+    var shuffle = f4.shuffle;
+    var extract = f4.extractLane;
+    var replaceLane = f4.replaceLane;
+    var neg = f4.neg;
+    var reciprocalApproximation = f4.reciprocalApproximation;
+    var reciprocalSqrtApproximation = f4.reciprocalSqrtApproximation;
+
+    class SIMDVector3 {
         public static TransformCoordinatesToRefSIMD(vector: Vector3, transformation: Matrix, result: Vector3): void {
-            var v = SIMD.float32x4.loadXYZ((<any>vector)._data, 0);
-            var m0 = SIMD.float32x4.load(transformation.m, 0);
-            var m1 = SIMD.float32x4.load(transformation.m, 4);
-            var m2 = SIMD.float32x4.load(transformation.m, 8);
-            var m3 = SIMD.float32x4.load(transformation.m, 12);
-            var r = SIMD.float32x4.add(SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(v, 0, 0, 0, 0), m0), SIMD.float32x4.mul(SIMD.float32x4.swizzle(v, 1, 1, 1, 1), m1)), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(v, 2, 2, 2, 2), m2), m3));
-            r = SIMD.float32x4.div(r, SIMD.float32x4.swizzle(r, 3, 3, 3, 3));
-            SIMD.float32x4.storeXYZ((<any>result)._data, 0, r);
+            SIMDVector3.TransformCoordinatesFromFloatsToRefSIMD(vector.x, vector.y, vector.z, transformation, result)
         }
 
         public static TransformCoordinatesFromFloatsToRefSIMD(x: number, y: number, z: number, transformation: Matrix, result: Vector3): void {
-            var v0 = SIMD.float32x4.splat(x);
-            var v1 = SIMD.float32x4.splat(y);
-            var v2 = SIMD.float32x4.splat(z);
-            var m0 = SIMD.float32x4.load(transformation.m, 0);
-            var m1 = SIMD.float32x4.load(transformation.m, 4);
-            var m2 = SIMD.float32x4.load(transformation.m, 8);
-            var m3 = SIMD.float32x4.load(transformation.m, 12);
-            var r = SIMD.float32x4.add(SIMD.float32x4.add(SIMD.float32x4.mul(v0, m0), SIMD.float32x4.mul(v1, m1)), SIMD.float32x4.add(SIMD.float32x4.mul(v2, m2), m3));
-            r = SIMD.float32x4.div(r, SIMD.float32x4.swizzle(r, 3, 3, 3, 3));
-            SIMD.float32x4.storeXYZ((<any>result)._data, 0, r);
+            var m = transformation.m;
+
+            var m0 = load(m, 0);
+            var m1 = load(m, 4);
+            var m2 = load(m, 8);
+            var m3 = load(m, 12);
+            var r = add(
+                add(mul(splat(x), m0), mul(splat(y), m1)),
+                add(mul(splat(z), m2), m3)
+            );
+            r = div(r, swizzle(r, 3, 3, 3, 3));
+            result.x = extract(r, 0);
+            result.y = extract(r, 1);
+            result.z = extract(r, 2);
         }
     }
 
-    export class SIMDMatrix {
-        public multiplyToArraySIMD(other: Matrix, result: Matrix, offset = 0): void {
-            var tm = (<any>this).m;
+    class SIMDMatrix {
+        public multiplyToArraySIMD(other: Matrix, result: Float32Array, offset: number): Matrix {
+            var tm: Float32Array = (<any>this).m;
             var om = other.m;
-            var om0 = SIMD.float32x4.load(om, 0);
-            var om1 = SIMD.float32x4.load(om, 4);
-            var om2 = SIMD.float32x4.load(om, 8);
-            var om3 = SIMD.float32x4.load(om, 12);
-
-            var tm0 = SIMD.float32x4.load(tm, 0);
-            SIMD.float32x4.store(result, offset + 0, SIMD.float32x4.add(
-                SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm0, 0, 0, 0, 0), om0),
-                SIMD.float32x4.add(
-                    SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm0, 1, 1, 1, 1), om1),
-                    SIMD.float32x4.add(
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm0, 2, 2, 2, 2), om2),
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm0, 3, 3, 3, 3), om3)))));
-
-            var tm1 = SIMD.float32x4.load(tm, 4);
-            SIMD.float32x4.store(result, offset + 4, SIMD.float32x4.add(
-                SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm1, 0, 0, 0, 0), om0),
-                SIMD.float32x4.add(
-                    SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm1, 1, 1, 1, 1), om1),
-                    SIMD.float32x4.add(
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm1, 2, 2, 2, 2), om2),
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm1, 3, 3, 3, 3), om3)))));
-
-            var tm2 = SIMD.float32x4.load(tm, 8);
-            SIMD.float32x4.store(result, offset + 8, SIMD.float32x4.add(
-                SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm2, 0, 0, 0, 0), om0),
-                SIMD.float32x4.add(
-                    SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm2, 1, 1, 1, 1), om1),
-                    SIMD.float32x4.add(
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm2, 2, 2, 2, 2), om2),
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm2, 3, 3, 3, 3), om3)))));
-
-            var tm3 = SIMD.float32x4.load(tm, 12);
-            SIMD.float32x4.store(result, offset + 12, SIMD.float32x4.add(
-                SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm3, 0, 0, 0, 0), om0),
-                SIMD.float32x4.add(
-                    SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm3, 1, 1, 1, 1), om1),
-                    SIMD.float32x4.add(
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm3, 2, 2, 2, 2), om2),
-                        SIMD.float32x4.mul(SIMD.float32x4.swizzle(tm3, 3, 3, 3, 3), om3)))));
+
+            var m0 = load(om, 0);
+            var m1 = load(om, 4);
+            var m2 = load(om, 8);
+            var m3 = load(om, 12);
+
+            for (var i = 0; i < 16; i += 4) {
+                store(result, i + offset, add(
+                    mul(splat(tm[i]), m0),
+                    add(
+                        mul(splat(tm[i+1]), m1),
+                        add(
+                            mul(splat(tm[i+2]), m2),
+                            mul(splat(tm[i+3]), m3)))));
+            }
+
+            return (<any>this);
         }
 
         public invertToRefSIMD(other: Matrix): Matrix {
-            var src = (<any>this).m;
+            var src: Float32Array = (<any>this).m;
             var dest = other.m;
-            var row0, row1, row2, row3;
-            var tmp1;
-            var minor0, minor1, minor2, minor3;
-            var det;
 
             // Load the 4 rows
-            var src0 = SIMD.float32x4.load(src, 0);
-            var src1 = SIMD.float32x4.load(src, 4);
-            var src2 = SIMD.float32x4.load(src, 8);
-            var src3 = SIMD.float32x4.load(src, 12);
+            var src0 = load(src, 0);
+            var src1 = load(src, 4);
+            var src2 = load(src, 8);
+            var src3 = load(src, 12);
 
             // Transpose the source matrix.  Sort of.  Not a true transpose operation
 
-            tmp1 = SIMD.float32x4.shuffle(src0, src1, 0, 1, 4, 5);
-            row1 = SIMD.float32x4.shuffle(src2, src3, 0, 1, 4, 5);
-            row0 = SIMD.float32x4.shuffle(tmp1, row1, 0, 2, 4, 6);
-            row1 = SIMD.float32x4.shuffle(row1, tmp1, 1, 3, 5, 7);
+            var tmp1 = shuffle(src0, src1, 0, 1, 4, 5);
+            var row1 = shuffle(src2, src3, 0, 1, 4, 5);
+            var row0 = shuffle(tmp1, row1, 0, 2, 4, 6);
+            row1 = shuffle(row1, tmp1, 1, 3, 5, 7);
 
-            tmp1 = SIMD.float32x4.shuffle(src0, src1, 2, 3, 6, 7);
-            row3 = SIMD.float32x4.shuffle(src2, src3, 2, 3, 6, 7);
-            row2 = SIMD.float32x4.shuffle(tmp1, row3, 0, 2, 4, 6);
-            row3 = SIMD.float32x4.shuffle(row3, tmp1, 1, 3, 5, 7);
+            tmp1 = shuffle(src0, src1, 2, 3, 6, 7);
+            var row3 = shuffle(src2, src3, 2, 3, 6, 7);
+            var row2 = shuffle(tmp1, row3, 0, 2, 4, 6);
+            row3 = shuffle(row3, tmp1, 1, 3, 5, 7);
 
             // This is a true transposition, but it will lead to an incorrect result
 
-            //tmp1 = SIMD.float32x4.shuffle(src0, src1, 0, 1, 4, 5);
-            //tmp2 = SIMD.float32x4.shuffle(src2, src3, 0, 1, 4, 5);
-            //row0  = SIMD.float32x4.shuffle(tmp1, tmp2, 0, 2, 4, 6);
-            //row1  = SIMD.float32x4.shuffle(tmp1, tmp2, 1, 3, 5, 7);
+            //tmp1 = shuffle(src0, src1, 0, 1, 4, 5);
+            //tmp2 = shuffle(src2, src3, 0, 1, 4, 5);
+            //row0  = shuffle(tmp1, tmp2, 0, 2, 4, 6);
+            //row1  = shuffle(tmp1, tmp2, 1, 3, 5, 7);
 
-            //tmp1 = SIMD.float32x4.shuffle(src0, src1, 2, 3, 6, 7);
-            //tmp2 = SIMD.float32x4.shuffle(src2, src3, 2, 3, 6, 7);
-            //row2  = SIMD.float32x4.shuffle(tmp1, tmp2, 0, 2, 4, 6);
-            //row3  = SIMD.float32x4.shuffle(tmp1, tmp2, 1, 3, 5, 7);
+            //tmp1 = shuffle(src0, src1, 2, 3, 6, 7);
+            //tmp2 = shuffle(src2, src3, 2, 3, 6, 7);
+            //row2  = shuffle(tmp1, tmp2, 0, 2, 4, 6);
+            //row3  = shuffle(tmp1, tmp2, 1, 3, 5, 7);
 
             // ----
-            tmp1 = SIMD.float32x4.mul(row2, row3);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            minor0 = SIMD.float32x4.mul(row1, tmp1);
-            minor1 = SIMD.float32x4.mul(row0, tmp1);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor0 = SIMD.float32x4.sub(SIMD.float32x4.mul(row1, tmp1), minor0);
-            minor1 = SIMD.float32x4.sub(SIMD.float32x4.mul(row0, tmp1), minor1);
-            minor1 = SIMD.float32x4.swizzle(minor1, 2, 3, 0, 1); // 0x4E = 01001110
+            tmp1 = mul(row2, row3);
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            var minor0 = mul(row1, tmp1);
+            var minor1 = mul(row0, tmp1);
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor0 = sub(mul(row1, tmp1), minor0);
+            minor1 = sub(mul(row0, tmp1), minor1);
+            minor1 = swizzle(minor1, 2, 3, 0, 1); // 0x4E = 01001110
 
             // ----
-            tmp1 = SIMD.float32x4.mul(row1, row2);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            minor0 = SIMD.float32x4.add(SIMD.float32x4.mul(row3, tmp1), minor0);
-            minor3 = SIMD.float32x4.mul(row0, tmp1);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor0 = SIMD.float32x4.sub(minor0, SIMD.float32x4.mul(row3, tmp1));
-            minor3 = SIMD.float32x4.sub(SIMD.float32x4.mul(row0, tmp1), minor3);
-            minor3 = SIMD.float32x4.swizzle(minor3, 2, 3, 0, 1); // 0x4E = 01001110
+            tmp1 = mul(row1, row2);
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            minor0 = add(mul(row3, tmp1), minor0);
+            var minor3 = mul(row0, tmp1);
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor0 = sub(minor0, mul(row3, tmp1));
+            minor3 = sub(mul(row0, tmp1), minor3);
+            minor3 = swizzle(minor3, 2, 3, 0, 1); // 0x4E = 01001110
 
             // ----
-            tmp1 = SIMD.float32x4.mul(SIMD.float32x4.swizzle(row1, 2, 3, 0, 1), row3); // 0x4E = 01001110
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            row2 = SIMD.float32x4.swizzle(row2, 2, 3, 0, 1);  // 0x4E = 01001110
-            minor0 = SIMD.float32x4.add(SIMD.float32x4.mul(row2, tmp1), minor0);
-            minor2 = SIMD.float32x4.mul(row0, tmp1);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor0 = SIMD.float32x4.sub(minor0, SIMD.float32x4.mul(row2, tmp1));
-            minor2 = SIMD.float32x4.sub(SIMD.float32x4.mul(row0, tmp1), minor2);
-            minor2 = SIMD.float32x4.swizzle(minor2, 2, 3, 0, 1); // 0x4E = 01001110
+            tmp1 = mul(swizzle(row1, 2, 3, 0, 1), row3); // 0x4E = 01001110
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            row2 = swizzle(row2, 2, 3, 0, 1);  // 0x4E = 01001110
+            minor0 = add(mul(row2, tmp1), minor0);
+            var minor2 = mul(row0, tmp1);
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor0 = sub(minor0, mul(row2, tmp1));
+            minor2 = sub(mul(row0, tmp1), minor2);
+            minor2 = swizzle(minor2, 2, 3, 0, 1); // 0x4E = 01001110
 
             // ----
-            tmp1 = SIMD.float32x4.mul(row0, row1);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            minor2 = SIMD.float32x4.add(SIMD.float32x4.mul(row3, tmp1), minor2);
-            minor3 = SIMD.float32x4.sub(SIMD.float32x4.mul(row2, tmp1), minor3);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor2 = SIMD.float32x4.sub(SIMD.float32x4.mul(row3, tmp1), minor2);
-            minor3 = SIMD.float32x4.sub(minor3, SIMD.float32x4.mul(row2, tmp1));
+            tmp1 = mul(row0, row1);
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            minor2 = add(mul(row3, tmp1), minor2);
+            minor3 = sub(mul(row2, tmp1), minor3);
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor2 = sub(mul(row3, tmp1), minor2);
+            minor3 = sub(minor3, mul(row2, tmp1));
 
             // ----
-            tmp1 = SIMD.float32x4.mul(row0, row3);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            minor1 = SIMD.float32x4.sub(minor1, SIMD.float32x4.mul(row2, tmp1));
-            minor2 = SIMD.float32x4.add(SIMD.float32x4.mul(row1, tmp1), minor2);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor1 = SIMD.float32x4.add(SIMD.float32x4.mul(row2, tmp1), minor1);
-            minor2 = SIMD.float32x4.sub(minor2, SIMD.float32x4.mul(row1, tmp1));
+            tmp1 = mul(row0, row3);
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            minor1 = sub(minor1, mul(row2, tmp1));
+            minor2 = add(mul(row1, tmp1), minor2);
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor1 = add(mul(row2, tmp1), minor1);
+            minor2 = sub(minor2, mul(row1, tmp1));
 
             // ----
-            tmp1 = SIMD.float32x4.mul(row0, row2);
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
-            minor1 = SIMD.float32x4.add(SIMD.float32x4.mul(row3, tmp1), minor1);
-            minor3 = SIMD.float32x4.sub(minor3, SIMD.float32x4.mul(row1, tmp1));
-            tmp1 = SIMD.float32x4.swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
-            minor1 = SIMD.float32x4.sub(minor1, SIMD.float32x4.mul(row3, tmp1));
-            minor3 = SIMD.float32x4.add(SIMD.float32x4.mul(row1, tmp1), minor3);
+            tmp1 = mul(row0, row2);
+            tmp1 = swizzle(tmp1, 1, 0, 3, 2); // 0xB1 = 10110001
+            minor1 = add(mul(row3, tmp1), minor1);
+            minor3 = sub(minor3, mul(row1, tmp1));
+            tmp1 = swizzle(tmp1, 2, 3, 0, 1); // 0x4E = 01001110
+            minor1 = sub(minor1, mul(row3, tmp1));
+            minor3 = add(mul(row1, tmp1), minor3);
 
             // Compute determinant
-            det = SIMD.float32x4.mul(row0, minor0);
-            det = SIMD.float32x4.add(SIMD.float32x4.swizzle(det, 2, 3, 0, 1), det); // 0x4E = 01001110
-            det = SIMD.float32x4.add(SIMD.float32x4.swizzle(det, 1, 0, 3, 2), det); // 0xB1 = 10110001
-            tmp1 = SIMD.float32x4.reciprocalApproximation(det);
-            det = SIMD.float32x4.sub(SIMD.float32x4.add(tmp1, tmp1), SIMD.float32x4.mul(det, SIMD.float32x4.mul(tmp1, tmp1)));
-            det = SIMD.float32x4.swizzle(det, 0, 0, 0, 0);
+            var det = mul(row0, minor0);
+            det = add(swizzle(det, 2, 3, 0, 1), det); // 0x4E = 01001110
+            det = add(swizzle(det, 1, 0, 3, 2), det); // 0xB1 = 10110001
+            tmp1 = reciprocalApproximation(det);
+            det = sub(add(tmp1, tmp1), mul(det, mul(tmp1, tmp1)));
+            det = swizzle(det, 0, 0, 0, 0);
 
             // These shuffles aren't necessary if the faulty transposition is done
             // up at the top of this function.
-            //minor0 = SIMD.float32x4.swizzle(minor0, 2, 1, 0, 3);
-            //minor1 = SIMD.float32x4.swizzle(minor1, 2, 1, 0, 3);
-            //minor2 = SIMD.float32x4.swizzle(minor2, 2, 1, 0, 3);
-            //minor3 = SIMD.float32x4.swizzle(minor3, 2, 1, 0, 3);
+            //minor0 = swizzle(minor0, 2, 1, 0, 3);
+            //minor1 = swizzle(minor1, 2, 1, 0, 3);
+            //minor2 = swizzle(minor2, 2, 1, 0, 3);
+            //minor3 = swizzle(minor3, 2, 1, 0, 3);
 
             // Compute final values by multiplying with 1/det
-            minor0 = SIMD.float32x4.mul(det, minor0);
-            minor1 = SIMD.float32x4.mul(det, minor1);
-            minor2 = SIMD.float32x4.mul(det, minor2);
-            minor3 = SIMD.float32x4.mul(det, minor3);
-
-            SIMD.float32x4.store(dest, 0, minor0);
-            SIMD.float32x4.store(dest, 4, minor1);
-            SIMD.float32x4.store(dest, 8, minor2);
-            SIMD.float32x4.store(dest, 12, minor3);
+            store(dest, 0, mul(det, minor0));
+            store(dest, 4, mul(det, minor1));
+            store(dest, 8, minor2 = mul(det, minor2));
+            store(dest, 12, mul(det, minor3));
 
             return (<any>this);
         }
 
         public static LookAtLHToRefSIMD(eyeRef: Vector3, targetRef: Vector3, upRef: Vector3, result: Matrix): void {
             var out = result.m;
-            var center = SIMD.float32x4(targetRef.x, targetRef.y, targetRef.z, 0);
-            var eye = SIMD.float32x4(eyeRef.x, eyeRef.y, eyeRef.z, 0);
-            var up = SIMD.float32x4(upRef.x, upRef.y, upRef.z, 0);
+            var center = f4(targetRef.x, targetRef.y, targetRef.z, 0.0);
+            var eye = f4(eyeRef.x, eyeRef.y, eyeRef.z, 0.0);
+            var up = f4(upRef.x, upRef.y, upRef.z, 0.0);
 
             // cc.kmVec3Subtract(f, pCenter, pEye);
-            var f = SIMD.float32x4.sub(center, eye);
+            var f = sub(center, eye);
             // cc.kmVec3Normalize(f, f);    
-            var tmp = SIMD.float32x4.mul(f, f);
-            tmp = SIMD.float32x4.add(tmp, SIMD.float32x4.add(SIMD.float32x4.swizzle(tmp, 1, 2, 0, 3), SIMD.float32x4.swizzle(tmp, 2, 0, 1, 3)));
-            f = SIMD.float32x4.mul(f, SIMD.float32x4.reciprocalSqrtApproximation(tmp));
+            var tmp = mul(f, f);
+            tmp = add(tmp, add(swizzle(tmp, 1, 2, 0, 3), swizzle(tmp, 2, 0, 1, 3)));
+            f = mul(f, reciprocalSqrtApproximation(tmp));
 
             // cc.kmVec3Assign(up, pUp);
             // cc.kmVec3Normalize(up, up);
-            tmp = SIMD.float32x4.mul(up, up);
-            tmp = SIMD.float32x4.add(tmp, SIMD.float32x4.add(SIMD.float32x4.swizzle(tmp, 1, 2, 0, 3), SIMD.float32x4.swizzle(tmp, 2, 0, 1, 3)));
-            up = SIMD.float32x4.mul(up, SIMD.float32x4.reciprocalSqrtApproximation(tmp));
+            tmp = mul(up, up);
+            tmp = add(tmp, add(swizzle(tmp, 1, 2, 0, 3), swizzle(tmp, 2, 0, 1, 3)));
+            up = mul(up, reciprocalSqrtApproximation(tmp));
             // cc.kmVec3Cross(s, f, up);
-            var s = SIMD.float32x4.sub(SIMD.float32x4.mul(SIMD.float32x4.swizzle(f, 1, 2, 0, 3), SIMD.float32x4.swizzle(up, 2, 0, 1, 3)), SIMD.float32x4.mul(SIMD.float32x4.swizzle(f, 2, 0, 1, 3), SIMD.float32x4.swizzle(up, 1, 2, 0, 3)));
+            var s = sub(mul(swizzle(f, 1, 2, 0, 3), swizzle(up, 2, 0, 1, 3)), mul(swizzle(f, 2, 0, 1, 3), swizzle(up, 1, 2, 0, 3)));
             // cc.kmVec3Normalize(s, s);
-            tmp = SIMD.float32x4.mul(s, s);
-            tmp = SIMD.float32x4.add(tmp, SIMD.float32x4.add(SIMD.float32x4.swizzle(tmp, 1, 2, 0, 3), SIMD.float32x4.swizzle(tmp, 2, 0, 1, 3)));
-            s = SIMD.float32x4.mul(s, SIMD.float32x4.reciprocalSqrtApproximation(tmp));
+            tmp = mul(s, s);
+            tmp = add(tmp, add(swizzle(tmp, 1, 2, 0, 3), swizzle(tmp, 2, 0, 1, 3)));
+            s = mul(s, reciprocalSqrtApproximation(tmp));
             // cc.kmVec3Cross(u, s, f);
-            var u = SIMD.float32x4.sub(SIMD.float32x4.mul(SIMD.float32x4.swizzle(s, 1, 2, 0, 3), SIMD.float32x4.swizzle(f, 2, 0, 1, 3)), SIMD.float32x4.mul(SIMD.float32x4.swizzle(s, 2, 0, 1, 3), SIMD.float32x4.swizzle(f, 1, 2, 0, 3)));
+            var u = sub(mul(swizzle(s, 1, 2, 0, 3), swizzle(f, 2, 0, 1, 3)), mul(swizzle(s, 2, 0, 1, 3), swizzle(f, 1, 2, 0, 3)));
             // cc.kmVec3Normalize(s, s);
-            tmp = SIMD.float32x4.mul(s, s);
-            tmp = SIMD.float32x4.add(tmp, SIMD.float32x4.add(SIMD.float32x4.swizzle(tmp, 1, 2, 0, 3), SIMD.float32x4.swizzle(tmp, 2, 0, 1, 3)));
-            s = SIMD.float32x4.mul(s, SIMD.float32x4.reciprocalSqrtApproximation(tmp));
-
-            var zero = SIMD.float32x4.splat(0.0);
-            s = SIMD.float32x4.neg(s);
-            var tmp01 = SIMD.float32x4.shuffle(s, u, 0, 1, 4, 5);
-            var tmp23 = SIMD.float32x4.shuffle(f, zero, 0, 1, 4, 5);
-            var a0 = SIMD.float32x4.shuffle(tmp01, tmp23, 0, 2, 4, 6);
-            var a1 = SIMD.float32x4.shuffle(tmp01, tmp23, 1, 3, 5, 7);
-            tmp01 = SIMD.float32x4.shuffle(s, u, 2, 3, 6, 7);
-            tmp23 = SIMD.float32x4.shuffle(f, zero, 2, 3, 6, 7);
-            var a2 = SIMD.float32x4.shuffle(tmp01, tmp23, 0, 2, 4, 6);
-            var a3 = SIMD.float32x4(0.0, 0.0, 0.0, 1.0);
-            var b0 = SIMD.float32x4(1.0, 0.0, 0.0, 0.0);
-            var b1 = SIMD.float32x4(0.0, 1.0, 0.0, 0.0);
-            var b2 = SIMD.float32x4(0.0, 0.0, 1.0, 0.0);
-            var b3 = SIMD.float32x4.neg(eye);
-            b3 = SIMD.float32x4.withW(b3, 1.0);
-
-            SIMD.float32x4.store(out, 0, SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b0, 0, 0, 0, 0), a0), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b0, 1, 1, 1, 1), a1), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b0, 2, 2, 2, 2), a2), SIMD.float32x4.mul(SIMD.float32x4.swizzle(b0, 3, 3, 3, 3), a3)))));
-            SIMD.float32x4.store(out, 4, SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b1, 0, 0, 0, 0), a0), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b1, 1, 1, 1, 1), a1), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b1, 2, 2, 2, 2), a2), SIMD.float32x4.mul(SIMD.float32x4.swizzle(b1, 3, 3, 3, 3), a3)))));
-            SIMD.float32x4.store(out, 8, SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b2, 0, 0, 0, 0), a0), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b2, 1, 1, 1, 1), a1), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b2, 2, 2, 2, 2), a2), SIMD.float32x4.mul(SIMD.float32x4.swizzle(b2, 3, 3, 3, 3), a3)))));
-            SIMD.float32x4.store(out, 12, SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b3, 0, 0, 0, 0), a0), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b3, 1, 1, 1, 1), a1), SIMD.float32x4.add(SIMD.float32x4.mul(SIMD.float32x4.swizzle(b3, 2, 2, 2, 2), a2), SIMD.float32x4.mul(SIMD.float32x4.swizzle(b3, 3, 3, 3, 3), a3)))));
+            tmp = mul(s, s);
+            tmp = add(tmp, add(swizzle(tmp, 1, 2, 0, 3), swizzle(tmp, 2, 0, 1, 3)));
+            s = mul(s, reciprocalSqrtApproximation(tmp));
+
+            var zero = splat(0.0);
+            s = neg(s);
+            var tmp01 = shuffle(s, u, 0, 1, 4, 5);
+            var tmp23 = shuffle(f, zero, 0, 1, 4, 5);
+            var a0 = shuffle(tmp01, tmp23, 0, 2, 4, 6);
+            var a1 = shuffle(tmp01, tmp23, 1, 3, 5, 7);
+            var a2 = shuffle(shuffle(s, u, 2, 3, 6, 7), shuffle(f, zero, 2, 3, 6, 7), 0, 2, 4, 6);
+            var a3 = f4(0.0, 0.0, 0.0, 1.0);
+
+            var b = f4(1.0, 0.0, 0.0, 0.0);
+            store(out, 0, add(mul(swizzle(b, 0, 0, 0, 0), a0), add(mul(swizzle(b, 1, 1, 1, 1), a1), add(mul(swizzle(b, 2, 2, 2, 2), a2), mul(swizzle(b, 3, 3, 3, 3), a3)))));
+
+            b = f4(0.0, 1.0, 0.0, 0.0);
+            store(out, 4, add(mul(swizzle(b, 0, 0, 0, 0), a0), add(mul(swizzle(b, 1, 1, 1, 1), a1), add(mul(swizzle(b, 2, 2, 2, 2), a2), mul(swizzle(b, 3, 3, 3, 3), a3)))));
+
+            b = f4(0.0, 0.0, 1.0, 0.0);
+            store(out, 8, add(mul(swizzle(b, 0, 0, 0, 0), a0), add(mul(swizzle(b, 1, 1, 1, 1), a1), add(mul(swizzle(b, 2, 2, 2, 2), a2), mul(swizzle(b, 3, 3, 3, 3), a3)))));
+
+            b = replaceLane(neg(eye), 3, 1.0);
+            store(out, 12, add(mul(swizzle(b, 0, 0, 0, 0), a0), add(mul(swizzle(b, 1, 1, 1, 1), a1), add(mul(swizzle(b, 2, 2, 2, 2), a2), mul(swizzle(b, 3, 3, 3, 3), a3)))));
+
         }
     }
 
@@ -268,54 +249,49 @@
 
         public static DisableSIMD(): void {
             // Replace functions
-            Matrix.prototype.multiplyToArray = <any>previousMultiplyToArray;
-            Matrix.prototype.invertToRef = <any>previousInvertToRef;
-            Matrix.LookAtLHToRef = <any>previousLookAtLHToRef;
-            Vector3.TransformCoordinatesToRef = <any>previousTransformCoordinatesToRef;
-            Vector3.TransformCoordinatesFromFloatsToRef = <any>previousTransformCoordinatesFromFloatsToRef;
+            Matrix.prototype.multiplyToArray = previousMultiplyToArray;
+            Matrix.prototype.invertToRef = previousInvertToRef;
+            Matrix.LookAtLHToRef = previousLookAtLHToRef;
+            Vector3.TransformCoordinatesToRef = previousTransformCoordinatesToRef;
+            Vector3.TransformCoordinatesFromFloatsToRef = previousTransformCoordinatesFromFloatsToRef;
 
             SIMDHelper._isEnabled = false;
         }
 
         public static EnableSIMD(): void {
-            if (window.SIMD === undefined) {
+            if (self.SIMD === undefined) {
                 return;
             }
 
+            // check if polyfills needed
+            if (!self.Math.fround) {
+                self.Math.fround = (function (array) {
+                    return function (x) {
+                        return array[0] = x, array[0];
+                    };
+                })(new Float32Array(1));
+            }
+
+            if (!self.Math.imul) {
+                self.Math.imul = function (a, b) {
+                    var ah = (a >>> 16) & 0xffff;
+                    var al = a & 0xffff;
+                    var bh = (b >>> 16) & 0xffff;
+                    var bl = b & 0xffff;
+                    // the shift by 0 fixes the sign on the high part
+                    // the final |0 converts the unsigned value into a signed value
+                    return ((al * bl) + (((ah * bl + al * bh) << 16) >>> 0) | 0);
+                };
+            }
+
             // Replace functions
-            Matrix.prototype.multiplyToArray = <any>SIMDMatrix.prototype.multiplyToArraySIMD;
-            Matrix.prototype.invertToRef = <any>SIMDMatrix.prototype.invertToRefSIMD;
-            Matrix.LookAtLHToRef = <any>SIMDMatrix.LookAtLHToRefSIMD;
-            Vector3.TransformCoordinatesToRef = <any>SIMDVector3.TransformCoordinatesToRefSIMD;
-            Vector3.TransformCoordinatesFromFloatsToRef = <any>SIMDVector3.TransformCoordinatesFromFloatsToRefSIMD;
-
-            Object.defineProperty(Vector3.prototype, "x", {
-                get() { return this._data[0]; },
-                set(value: number) {
-                    if (!this._data) {
-                        this._data = new Float32Array(3);
-                    }
-                    this._data[0] = value;
-                }
-            });
-
-            Object.defineProperty(Vector3.prototype, "y", {
-                get() { return this._data[1]; },
-                set(value: number) {
-                    this._data[1] = value;
-                }
-            });
-
-            Object.defineProperty(Vector3.prototype, "z", {
-                get() { return this._data[2]; },
-                set(value: number) {
-                    this._data[2] = value;
-                }
-            });
+            Matrix.prototype.multiplyToArray = SIMDMatrix.prototype.multiplyToArraySIMD;
+            Matrix.prototype.invertToRef = SIMDMatrix.prototype.invertToRefSIMD;
+            Matrix.LookAtLHToRef = SIMDMatrix.LookAtLHToRefSIMD;
+            Vector3.TransformCoordinatesToRef = SIMDVector3.TransformCoordinatesToRefSIMD;
+            Vector3.TransformCoordinatesFromFloatsToRef = SIMDVector3.TransformCoordinatesFromFloatsToRefSIMD;
 
             SIMDHelper._isEnabled = true;
         }
     }
 }
-
-

src/Math/babylon.math.ts

@@ -737,7 +737,6 @@
     export class Vector3 {
 
         constructor(public x: number, public y: number, public z: number) {
-
         }
 
         public toString(): string {
@@ -757,7 +756,7 @@
 
         // Operators
         public asArray(): number[] {
-            var result = [];
+            var result: number[] = [];
 
             this.toArray(result, 0);
 

src/Math/simd.ts

@@ -0,0 +1,435 @@
+interface Window {
+    Math: Math;
+    Uint8Array: Uint8ArrayConstructor;
+    Float32Array: Float32ArrayConstructor;
+}
+
+interface Math {
+    fround(x: number): number
+    imul(a: number, b: number): number
+}
+
+interface SIMDglobal {
+    SIMD: SIMD;
+    Math: Math;
+    Uint8Array: Uint8ArrayConstructor;
+    Float32Array: Float32ArrayConstructor;
+}
+
+interface SIMD {
+    Float32x4: SIMD.Float32x4Constructor;
+    Int32x4: SIMD.Int32x4Constructor;
+    Int16x8: SIMD.Int16x8Constructor;
+    Int8x16: SIMD.Int8x16Constructor;
+    Uint32x4: SIMD.Uint32x4Constructor;
+    Uint16x8: SIMD.Uint16x8Constructor;
+    Uint8x16: SIMD.Uint8x16Constructor;
+    Bool32x4: SIMD.Bool32x4Constructor;
+    Bool16x8: SIMD.Bool16x8Constructor;
+    Bool8x16: SIMD.Bool8x16Constructor;
+}
+
+declare namespace SIMD {
+    interface Float32x4 {
+        constructor: Float32x4Constructor;
+        valueOf(): Float32x4;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Float32x4Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number): Float32x4;
+        prototype: Float32x4;
+        extractLane(simd: SIMD.Float32x4, lane: number): number;
+        swizzle(a: SIMD.Float32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Float32x4;
+        shuffle(a: SIMD.Float32x4, b: SIMD.Float32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Float32x4;
+        check(a: SIMD.Float32x4): SIMD.Float32x4;
+        splat(n: number): SIMD.Float32x4;
+        replaceLane(simd: SIMD.Float32x4, lane: number, value: number): SIMD.Float32x4;
+        select(selector: SIMD.Bool32x4, a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        equal(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        notEqual(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        lessThan(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        lessThanOrEqual(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        greaterThan(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        greaterThanOrEqual(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Bool32x4;
+        add(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        sub(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        mul(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        div(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        neg(a: SIMD.Float32x4): SIMD.Float32x4;
+        abs(a: SIMD.Float32x4): SIMD.Float32x4;
+        min(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        max(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        minNum(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        maxNum(a: SIMD.Float32x4, b: SIMD.Float32x4): SIMD.Float32x4;
+        reciprocalApproximation(a: SIMD.Float32x4): SIMD.Float32x4;
+        reciprocalSqrtApproximation(a: SIMD.Float32x4): SIMD.Float32x4;
+        sqrt(a: SIMD.Float32x4): SIMD.Float32x4;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Float32x4;
+        load1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Float32x4;
+        load2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Float32x4;
+        load3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Float32x4;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Float32x4): SIMD.Float32x4;
+        store1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Float32x4): SIMD.Float32x4;
+        store2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Float32x4): SIMD.Float32x4;
+        store3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Float32x4): SIMD.Float32x4;
+        fromInt32x4(value: SIMD.Int32x4): SIMD.Float32x4;
+        fromUint32x4(value: SIMD.Uint32x4): SIMD.Float32x4;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Float32x4;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Float32x4;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Float32x4;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Float32x4;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Float32x4;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Float32x4;
+    }
+
+    interface Int32x4 {
+        constructor: Int32x4Constructor;
+        valueOf(): Int32x4;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Int32x4Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number): Int32x4;
+        prototype: Int32x4;
+        extractLane(simd: SIMD.Int32x4, lane: number): number;
+        swizzle(a: SIMD.Int32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Int32x4;
+        shuffle(a: SIMD.Int32x4, b: SIMD.Int32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Int32x4;
+        check(a: SIMD.Int32x4): SIMD.Int32x4;
+        splat(n: number): SIMD.Int32x4;
+        replaceLane(simd: SIMD.Int32x4, lane: number, value: number): SIMD.Int32x4;
+        select(selector: SIMD.Bool32x4, a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        equal(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        notEqual(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        lessThan(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        lessThanOrEqual(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        greaterThan(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        greaterThanOrEqual(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Bool32x4;
+        and(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        or(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        xor(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        not(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        add(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        sub(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        mul(a: SIMD.Int32x4, b: SIMD.Int32x4): SIMD.Int32x4;
+        neg(a: SIMD.Int32x4): SIMD.Int32x4;
+        shiftLeftByScalar(a: SIMD.Int32x4, bits: number): SIMD.Int32x4;
+        shiftRightByScalar(a: SIMD.Int32x4, bits: number): SIMD.Int32x4;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int32x4;
+        load1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int32x4;
+        load2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int32x4;
+        load3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int32x4;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int32x4): SIMD.Int32x4;
+        store1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int32x4): SIMD.Int32x4;
+        store2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int32x4): SIMD.Int32x4;
+        store3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int32x4): SIMD.Int32x4;
+        fromFloat32x4(value: SIMD.Float32x4): SIMD.Int32x4;
+        fromUint32x4(value: SIMD.Uint32x4): SIMD.Int32x4;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Int32x4;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Int32x4;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Int32x4;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Int32x4;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Int32x4;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Int32x4;
+    }
+
+    interface Int16x8 {
+        constructor: Int16x8Constructor;
+        valueOf(): Int16x8;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Int16x8Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number, s4?: number, s5?: number, s6?: number, s7?: number): Int16x8;
+        prototype: Int16x8;
+        extractLane(simd: SIMD.Int16x8, lane: number): number;
+        swizzle(a: SIMD.Int16x8, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number): SIMD.Int16x8;
+        shuffle(a: SIMD.Int16x8, b: SIMD.Int16x8, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number): SIMD.Int16x8;
+        check(a: SIMD.Int16x8): SIMD.Int16x8;
+        splat(n: number): SIMD.Int16x8;
+        replaceLane(simd: SIMD.Int16x8, lane: number, value: number): SIMD.Int16x8;
+        select(selector: SIMD.Bool16x8, a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        equal(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        notEqual(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        lessThan(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        lessThanOrEqual(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        greaterThan(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        greaterThanOrEqual(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Bool16x8;
+        and(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        or(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        xor(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        not(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        add(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        sub(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        mul(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        neg(a: SIMD.Int16x8): SIMD.Int16x8;
+        shiftLeftByScalar(a: SIMD.Int16x8, bits: number): SIMD.Int16x8;
+        shiftRightByScalar(a: SIMD.Int16x8, bits: number): SIMD.Int16x8;
+        addSaturate(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        subSaturate(a: SIMD.Int16x8, b: SIMD.Int16x8): SIMD.Int16x8;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int16x8;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int16x8): SIMD.Int16x8;
+        fromUint16x8(value: SIMD.Uint16x8): SIMD.Int16x8;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Int16x8;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Int16x8;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Int16x8;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Int16x8;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Int16x8;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Int16x8;
+    }
+
+    interface Int8x16 {
+        constructor: Int8x16Constructor;
+        valueOf(): Int8x16;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Int8x16Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number, s4?: number, s5?: number, s6?: number, s7?: number, s8?: number, s9?: number, s10?: number, s11?: number, s12?: number, s13?: number, s14?: number, s15?: number): Int8x16;
+        prototype: Int8x16;
+        extractLane(simd: SIMD.Int8x16, lane: number): number;
+        swizzle(a: SIMD.Int8x16, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number, l9: number, l10: number, l11: number, l12: number, l13: number, l14: number, l15: number, l16: number): SIMD.Int8x16;
+        shuffle(a: SIMD.Int8x16, b: SIMD.Int8x16, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number, l9: number, l10: number, l11: number, l12: number, l13: number, l14: number, l15: number, l16: number): SIMD.Int8x16;
+        check(a: SIMD.Int8x16): SIMD.Int8x16;
+        splat(n: number): SIMD.Int8x16;
+        replaceLane(simd: SIMD.Int8x16, lane: number, value: number): SIMD.Int8x16;
+        select(selector: SIMD.Bool8x16, a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        equal(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        notEqual(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        lessThan(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        lessThanOrEqual(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        greaterThan(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        greaterThanOrEqual(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Bool8x16;
+        and(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        or(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        xor(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        not(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        add(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        sub(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        mul(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        neg(a: SIMD.Int8x16): SIMD.Int8x16;
+        shiftLeftByScalar(a: SIMD.Int8x16, bits: number): SIMD.Int8x16;
+        shiftRightByScalar(a: SIMD.Int8x16, bits: number): SIMD.Int8x16;
+        addSaturate(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        subSaturate(a: SIMD.Int8x16, b: SIMD.Int8x16): SIMD.Int8x16;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Int8x16;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Int8x16): SIMD.Int8x16;
+        fromUint8x16(value: SIMD.Uint8x16): SIMD.Int8x16;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Int8x16;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Int8x16;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Int8x16;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Int8x16;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Int8x16;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Int8x16;
+    }
+
+    interface Uint32x4 {
+        constructor: Uint32x4Constructor;
+        valueOf(): Uint32x4;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Uint32x4Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number): Uint32x4;
+        prototype: Uint32x4;
+        extractLane(simd: SIMD.Uint32x4, lane: number): number;
+        swizzle(a: SIMD.Uint32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Uint32x4;
+        shuffle(a: SIMD.Uint32x4, b: SIMD.Uint32x4, l1: number, l2: number, l3: number, l4: number): SIMD.Uint32x4;
+        check(a: SIMD.Uint32x4): SIMD.Uint32x4;
+        splat(n: number): SIMD.Uint32x4;
+        replaceLane(simd: SIMD.Uint32x4, lane: number, value: number): SIMD.Uint32x4;
+        select(selector: SIMD.Bool32x4, a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        equal(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        notEqual(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        lessThan(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        lessThanOrEqual(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        greaterThan(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        greaterThanOrEqual(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Bool32x4;
+        and(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        or(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        xor(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        not(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        add(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        sub(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        mul(a: SIMD.Uint32x4, b: SIMD.Uint32x4): SIMD.Uint32x4;
+        shiftLeftByScalar(a: SIMD.Uint32x4, bits: number): SIMD.Uint32x4;
+        shiftRightByScalar(a: SIMD.Uint32x4, bits: number): SIMD.Uint32x4;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint32x4;
+        load1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint32x4;
+        load2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint32x4;
+        load3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint32x4;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint32x4): SIMD.Uint32x4;
+        store1(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint32x4): SIMD.Uint32x4;
+        store2(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint32x4): SIMD.Uint32x4;
+        store3(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint32x4): SIMD.Uint32x4;
+        fromFloat32x4(value: SIMD.Float32x4): SIMD.Uint32x4;
+        fromInt32x4(value: SIMD.Int32x4): SIMD.Uint32x4;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Uint32x4;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Uint32x4;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Uint32x4;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Uint32x4;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Uint32x4;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Uint32x4;
+    }
+
+    interface Uint16x8 {
+        constructor: Uint16x8Constructor;
+        valueOf(): Uint16x8;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Uint16x8Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number, s4?: number, s5?: number, s6?: number, s7?: number): Uint16x8;
+        prototype: Uint16x8;
+        extractLane(simd: SIMD.Uint16x8, lane: number): number;
+        swizzle(a: SIMD.Uint16x8, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number): SIMD.Uint16x8;
+        shuffle(a: SIMD.Uint16x8, b: SIMD.Uint16x8, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number): SIMD.Uint16x8;
+        check(a: SIMD.Uint16x8): SIMD.Uint16x8;
+        splat(n: number): SIMD.Uint16x8;
+        replaceLane(simd: SIMD.Uint16x8, lane: number, value: number): SIMD.Uint16x8;
+        select(selector: SIMD.Bool16x8, a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        equal(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        notEqual(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        lessThan(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        lessThanOrEqual(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        greaterThan(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        greaterThanOrEqual(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Bool16x8;
+        and(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        or(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        xor(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        not(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        add(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        sub(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        mul(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        shiftLeftByScalar(a: SIMD.Uint16x8, bits: number): SIMD.Uint16x8;
+        shiftRightByScalar(a: SIMD.Uint16x8, bits: number): SIMD.Uint16x8;
+        addSaturate(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        subSaturate(a: SIMD.Uint16x8, b: SIMD.Uint16x8): SIMD.Uint16x8;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint16x8;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint16x8): SIMD.Uint16x8;
+        fromInt16x8(value: SIMD.Int16x8): SIMD.Uint16x8;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Uint16x8;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Uint16x8;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Uint16x8;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Uint16x8;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Uint16x8;
+        fromUint8x16Bits(value: SIMD.Uint8x16): SIMD.Uint16x8;
+    }
+
+    interface Uint8x16 {
+        constructor: Uint8x16Constructor;
+        valueOf(): Uint8x16;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Uint8x16Constructor {
+        (s0?: number, s1?: number, s2?: number, s3?: number, s4?: number, s5?: number, s6?: number, s7?: number, s8?: number, s9?: number, s10?: number, s11?: number, s12?: number, s13?: number, s14?: number, s15?: number): Uint8x16;
+        prototype: Uint8x16;
+        extractLane(simd: SIMD.Uint8x16, lane: number): number;
+        swizzle(a: SIMD.Uint8x16, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number, l9: number, l10: number, l11: number, l12: number, l13: number, l14: number, l15: number, l16: number): SIMD.Uint8x16;
+        shuffle(a: SIMD.Uint8x16, b: SIMD.Uint8x16, l1: number, l2: number, l3: number, l4: number, l5: number, l6: number, l7: number, l8: number, l9: number, l10: number, l11: number, l12: number, l13: number, l14: number, l15: number, l16: number): SIMD.Uint8x16;
+        check(a: SIMD.Uint8x16): SIMD.Uint8x16;
+        splat(n: number): SIMD.Uint8x16;
+        replaceLane(simd: SIMD.Uint8x16, lane: number, value: number): SIMD.Uint8x16;
+        select(selector: SIMD.Bool8x16, a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        equal(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        notEqual(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        lessThan(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        lessThanOrEqual(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        greaterThan(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        greaterThanOrEqual(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Bool8x16;
+        and(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        or(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        xor(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        not(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        add(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        sub(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        mul(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        shiftLeftByScalar(a: SIMD.Uint8x16, bits: number): SIMD.Uint8x16;
+        shiftRightByScalar(a: SIMD.Uint8x16, bits: number): SIMD.Uint8x16;
+        addSaturate(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        subSaturate(a: SIMD.Uint8x16, b: SIMD.Uint8x16): SIMD.Uint8x16;
+        load(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number): SIMD.Uint8x16;
+        store(tarray: Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array, index: number, value: SIMD.Uint8x16): SIMD.Uint8x16;
+        fromInt8x16(value: SIMD.Int8x16): SIMD.Uint8x16;
+        fromFloat32x4Bits(value: SIMD.Float32x4): SIMD.Uint8x16;
+        fromInt32x4Bits(value: SIMD.Int32x4): SIMD.Uint8x16;
+        fromInt16x8Bits(value: SIMD.Int16x8): SIMD.Uint8x16;
+        fromInt8x16Bits(value: SIMD.Int8x16): SIMD.Uint8x16;
+        fromUint32x4Bits(value: SIMD.Uint32x4): SIMD.Uint8x16;
+        fromUint16x8Bits(value: SIMD.Uint16x8): SIMD.Uint8x16;
+    }
+
+    interface Bool32x4 {
+        constructor: Bool32x4Constructor;
+        valueOf(): Bool32x4;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Bool32x4Constructor {
+        (s0?: boolean, s1?: boolean, s2?: boolean, s3?: boolean): Bool32x4;
+        prototype: Bool32x4;
+        extractLane(simd: SIMD.Bool32x4, lane: number): boolean;
+        check(a: SIMD.Bool32x4): SIMD.Bool32x4;
+        splat(n: boolean): SIMD.Bool32x4;
+        replaceLane(simd: SIMD.Bool32x4, lane: number, value: boolean): SIMD.Bool32x4;
+        allTrue(a: SIMD.Bool32x4): boolean;
+        anyTrue(a: SIMD.Bool32x4): boolean;
+        and(a: SIMD.Bool32x4, b: SIMD.Bool32x4): SIMD.Bool32x4;
+        or(a: SIMD.Bool32x4, b: SIMD.Bool32x4): SIMD.Bool32x4;
+        xor(a: SIMD.Bool32x4, b: SIMD.Bool32x4): SIMD.Bool32x4;
+        not(a: SIMD.Bool32x4, b: SIMD.Bool32x4): SIMD.Bool32x4;
+    }
+
+    interface Bool16x8 {
+        constructor: Bool16x8Constructor;
+        valueOf(): Bool16x8;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Bool16x8Constructor {
+        (s0?: boolean, s1?: boolean, s2?: boolean, s3?: boolean, s4?: boolean, s5?: boolean, s6?: boolean, s7?: boolean): Bool16x8;
+        prototype: Bool16x8;
+        extractLane(simd: SIMD.Bool16x8, lane: number): boolean;
+        check(a: SIMD.Bool16x8): SIMD.Bool16x8;
+        splat(n: boolean): SIMD.Bool16x8;
+        replaceLane(simd: SIMD.Bool16x8, lane: number, value: boolean): SIMD.Bool16x8;
+        allTrue(a: SIMD.Bool16x8): boolean;
+        anyTrue(a: SIMD.Bool16x8): boolean;
+        and(a: SIMD.Bool16x8, b: SIMD.Bool16x8): SIMD.Bool16x8;
+        or(a: SIMD.Bool16x8, b: SIMD.Bool16x8): SIMD.Bool16x8;
+        xor(a: SIMD.Bool16x8, b: SIMD.Bool16x8): SIMD.Bool16x8;
+        not(a: SIMD.Bool16x8, b: SIMD.Bool16x8): SIMD.Bool16x8;
+    }
+
+    interface Bool8x16 {
+        constructor: Bool8x16Constructor;
+        valueOf(): Bool8x16;
+        toLocaleString(): string;
+        toString(): string;
+    }
+
+    interface Bool8x16Constructor {
+        (s0?: boolean, s1?: boolean, s2?: boolean, s3?: boolean, s4?: boolean, s5?: boolean, s6?: boolean, s7?: boolean, s8?: boolean, s9?: boolean, s10?: boolean, s11?: boolean, s12?: boolean, s13?: boolean, s14?: boolean, s15?: boolean): Bool8x16;
+        prototype: Bool8x16;
+        extractLane(simd: SIMD.Bool8x16, lane: number): boolean;
+        check(a: SIMD.Bool8x16): SIMD.Bool8x16;
+        splat(n: boolean): SIMD.Bool8x16;
+        replaceLane(simd: SIMD.Bool8x16, lane: number, value: boolean): SIMD.Bool8x16;
+        allTrue(a: SIMD.Bool8x16): boolean;
+        anyTrue(a: SIMD.Bool8x16): boolean;
+        and(a: SIMD.Bool8x16, b: SIMD.Bool8x16): SIMD.Bool8x16;
+        or(a: SIMD.Bool8x16, b: SIMD.Bool8x16): SIMD.Bool8x16;
+        xor(a: SIMD.Bool8x16, b: SIMD.Bool8x16): SIMD.Bool8x16;
+        not(a: SIMD.Bool8x16, b: SIMD.Bool8x16): SIMD.Bool8x16;
+    }
+}