| import 'dart:html'; |
| |
| /// Copyright 2013 the V8 project authors. All rights reserved. |
| /// Copyright 2009 Oliver Hunt <http://nerget.com> |
| /// |
| /// Permission is hereby granted, free of charge, to any person |
| /// obtaining a copy of this software and associated documentation |
| /// files (the 'Software'), to deal in the Software without |
| /// restriction, including without limitation the rights to use, |
| /// copy, modify, merge, publish, distribute, sublicense, and/or sell |
| /// copies of the Software, and to permit persons to whom the |
| /// Software is furnished to do so, subject to the following |
| /// conditions: |
| /// |
| /// The above copyright notice and this permission notice shall be |
| /// included in all copies or substantial portions of the Software. |
| /// |
| /// THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, |
| /// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES |
| /// OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| /// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
| /// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| /// WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
| /// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| /// OTHER DEALINGS IN THE SOFTWARE. |
| |
| // Ported from the v8 benchmark suite by Google 2013. |
| // Uses Float64List for data. |
| |
| import 'dart:math'; |
| import 'dart:typed_data'; |
| |
| import 'package:benchmark_harness/benchmark_harness.dart'; |
| |
| /// Copyright 2009 Oliver Hunt <http://nerget.com> |
| /// |
| /// Permission is hereby granted, free of charge, to any person |
| /// obtaining a copy of this software and associated documentation |
| /// files (the 'Software'), to deal in the Software without |
| /// restriction, including without limitation the rights to use, |
| /// copy, modify, merge, publish, distribute, sublicense, and/or sell |
| /// copies of the Software, and to permit persons to whom the |
| /// Software is furnished to do so, subject to the following |
| /// conditions: |
| /// |
| /// The above copyright notice and this permission notice shall be |
| /// included in all copies or substantial portions of the Software. |
| /// |
| /// THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, |
| /// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES |
| /// OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| /// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
| /// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| /// WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
| /// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| /// OTHER DEALINGS IN THE SOFTWARE. |
| // Ported from the v8 benchmark suite by Google 2013. |
| // Uses Float64List for data. |
| |
| /// Copyright 2009 Oliver Hunt <http://nerget.com> |
| /// |
| /// Permission is hereby granted, free of charge, to any person |
| /// obtaining a copy of this software and associated documentation |
| /// files (the 'Software'), to deal in the Software without |
| /// restriction, including without limitation the rights to use, |
| /// copy, modify, merge, publish, distribute, sublicense, and/or sell |
| /// copies of the Software, and to permit persons to whom the |
| /// Software is furnished to do so, subject to the following |
| /// conditions: |
| /// |
| /// The above copyright notice and this permission notice shall be |
| /// included in all copies or substantial portions of the Software. |
| /// |
| /// THE SOFTWARE IS PROVIDED 'AS IS', WITHOUT WARRANTY OF ANY KIND, |
| /// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES |
| /// OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| /// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
| /// HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| /// WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
| /// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
| /// OTHER DEALINGS IN THE SOFTWARE. |
| // Ported from the v8 benchmark suite by Google 2013. |
| // Uses Float64List for data. |
| |
| void main() { |
| const FluidMotion().report(); |
| } |
| |
| class FluidMotion extends BenchmarkBase { |
| static FluidField solver; |
| static int framesTillAddingPoints = 0; |
| static int framesBetweenAddingPoints = 5; |
| |
| const FluidMotion() : super('FluidMotion'); |
| |
| static void setupFluidMotion() { |
| framesTillAddingPoints = 0; |
| framesBetweenAddingPoints = 5; |
| solver = FluidField.create(null, 128, 128, 20); |
| solver.setDisplayFunction((a) {}); |
| solver.setUICallback(prepareFrame); |
| } |
| |
| static void runFluidMotion() { |
| setupFluidMotion(); |
| for (var i = 0; i < 10; i++) { |
| solver.update(); |
| } |
| solver.validate(758.9012130174812, -352.56376676179076, -357.3690235879736); |
| } |
| |
| static void main() { |
| runFluidMotion(); |
| } |
| |
| static void addPoints(Field field) { |
| var n = 64; |
| for (var i = 1; i <= n; i++) { |
| field.setVelocity(i, i, n.toDouble(), n.toDouble()); |
| field.setDensity(i, i, 5.0); |
| field.setVelocity(i, n - i, -n.toDouble(), -n.toDouble()); |
| field.setDensity(i, n - i, 20.0); |
| field.setVelocity(128 - i, n + i, -n.toDouble(), -n.toDouble()); |
| field.setDensity(128 - i, n + i, 30.0); |
| } |
| } |
| |
| static void prepareFrame(Field field) { |
| if (framesTillAddingPoints == 0) { |
| addPoints(field); |
| framesTillAddingPoints = framesBetweenAddingPoints; |
| framesBetweenAddingPoints++; |
| } else { |
| framesTillAddingPoints--; |
| } |
| } |
| |
| // Overrides of BenchmarkBase. |
| |
| @override |
| void warmup() { |
| runFluidMotion(); |
| } |
| |
| @override |
| void exercise() { |
| runFluidMotion(); |
| } |
| } |
| |
| // Code from Oliver Hunt (http://nerget.com/fluidSim/pressure.js) starts here. |
| |
| class FluidField { |
| final CanvasElement canvas; |
| final int iterations; |
| final double dt = 0.1; |
| final int size; |
| Float64List dens, dens_prev; |
| Float64List u, u_prev; |
| Float64List v, v_prev; |
| final int width, height; |
| final int rowSize; |
| void Function(Field) displayFunc; |
| |
| static FluidField _lastCreated; |
| |
| static bool approxEquals(double a, double b) => (a - b).abs() < 0.000001; |
| |
| void validate(double expectedDens, double expectedU, double expectedV) { |
| var sumDens = 0.0; |
| var sumU = 0.0; |
| var sumV = 0.0; |
| for (var i = 0; i < dens.length; i++) { |
| sumDens += dens[i]; |
| sumU += u[i]; |
| sumV += v[i]; |
| } |
| |
| if (!approxEquals(sumDens, expectedDens) || |
| !approxEquals(sumU, expectedU) || |
| !approxEquals(sumV, expectedV)) { |
| throw 'Incorrect result'; |
| } |
| } |
| |
| // Allocates a new FluidField or return previously allocated field if the |
| // size is too large. |
| factory FluidField.create( |
| CanvasElement canvas, int hRes, int wRes, int iterations) { |
| final res = wRes * hRes; |
| if ((res > 0) && (res < 1000000)) { |
| _lastCreated = FluidField(canvas, hRes, wRes, iterations); |
| } |
| _lastCreated ??= FluidField(canvas, 64, 64, iterations); |
| assert((canvas == _lastCreated.canvas) && |
| (iterations == _lastCreated.iterations)); |
| return _lastCreated; |
| } |
| |
| FluidField(this.canvas, this.height, this.width, this.iterations) |
| : rowSize = width + 2, |
| size = (width + 2) * (height + 2) { |
| reset(); |
| } |
| |
| void reset() { |
| // All Float64List elements are initialized to 0.0. |
| dens = Float64List(size); |
| dens_prev = Float64List(size); |
| u = Float64List(size); |
| u_prev = Float64List(size); |
| v = Float64List(size); |
| v_prev = Float64List(size); |
| } |
| |
| void addFields(Float64List x, Float64List s, double dt) { |
| for (var i = 0; i < size; i++) { |
| x[i] += dt * s[i]; |
| } |
| } |
| |
| void set_bnd(int b, Float64List x) { |
| if (b == 1) { |
| var i = 1; |
| for (; i <= width; i++) { |
| x[i] = x[i + rowSize]; |
| x[i + (height + 1) * rowSize] = x[i + height * rowSize]; |
| } |
| |
| for (var j = 1; j <= height; j++) { |
| x[j * rowSize] = x[1 + j * rowSize]; |
| x[(width + 1) + j * rowSize] = -1 * x[width + j * rowSize]; |
| } |
| } else if (b == 2) { |
| for (var i = 1; i <= width; i++) { |
| x[i] = -1 * x[i + rowSize]; |
| x[i + (height + 1) * rowSize] = -1 * x[i + height * rowSize]; |
| } |
| |
| for (var j = 1; j <= height; j++) { |
| x[j * rowSize] = x[1 + j * rowSize]; |
| x[(width + 1) + j * rowSize] = x[width + j * rowSize]; |
| } |
| } else { |
| for (var i = 1; i <= width; i++) { |
| x[i] = x[i + rowSize]; |
| x[i + (height + 1) * rowSize] = x[i + height * rowSize]; |
| } |
| |
| for (var j = 1; j <= height; j++) { |
| x[j * rowSize] = x[1 + j * rowSize]; |
| x[(width + 1) + j * rowSize] = x[width + j * rowSize]; |
| } |
| } |
| var maxEdge = (height + 1) * rowSize; |
| x[0] = 0.5 * (x[1] + x[rowSize]); |
| x[maxEdge] = 0.5 * (x[1 + maxEdge] + x[height * rowSize]); |
| x[(width + 1)] = 0.5 * (x[width] + x[(width + 1) + rowSize]); |
| x[(width + 1) + maxEdge] = |
| 0.5 * (x[width + maxEdge] + x[(width + 1) + height * rowSize]); |
| } |
| |
| void lin_solve(int b, Float64List x, Float64List x0, int a, int c) { |
| if (a == 0 && c == 1) { |
| for (var j = 1; j <= height; j++) { |
| var currentRow = j * rowSize; |
| ++currentRow; |
| for (var i = 0; i < width; i++) { |
| x[currentRow] = x0[currentRow]; |
| ++currentRow; |
| } |
| } |
| set_bnd(b, x); |
| } else { |
| var invC = 1 / c; |
| for (var k = 0; k < iterations; k++) { |
| for (var j = 1; j <= height; j++) { |
| var lastRow = (j - 1) * rowSize; |
| var currentRow = j * rowSize; |
| var nextRow = (j + 1) * rowSize; |
| var lastX = x[currentRow]; |
| ++currentRow; |
| for (var i = 1; i <= width; i++) { |
| lastX = x[currentRow] = (x0[currentRow] + |
| a * |
| (lastX + |
| x[++currentRow] + |
| x[++lastRow] + |
| x[++nextRow])) * |
| invC; |
| } |
| } |
| set_bnd(b, x); |
| } |
| } |
| } |
| |
| void diffuse(int b, Float64List x, Float64List x0, double dt) { |
| var a = 0; |
| lin_solve(b, x, x0, a, 1 + 4 * a); |
| } |
| |
| void lin_solve2(Float64List x, Float64List x0, Float64List y, Float64List y0, |
| int a, int c) { |
| if (a == 0 && c == 1) { |
| for (var j = 1; j <= height; j++) { |
| var currentRow = j * rowSize; |
| ++currentRow; |
| for (var i = 0; i < width; i++) { |
| x[currentRow] = x0[currentRow]; |
| y[currentRow] = y0[currentRow]; |
| ++currentRow; |
| } |
| } |
| set_bnd(1, x); |
| set_bnd(2, y); |
| } else { |
| var invC = 1 / c; |
| for (var k = 0; k < iterations; k++) { |
| for (var j = 1; j <= height; j++) { |
| var lastRow = (j - 1) * rowSize; |
| var currentRow = j * rowSize; |
| var nextRow = (j + 1) * rowSize; |
| var lastX = x[currentRow]; |
| var lastY = y[currentRow]; |
| ++currentRow; |
| for (var i = 1; i <= width; i++) { |
| lastX = x[currentRow] = (x0[currentRow] + |
| a * (lastX + x[currentRow] + x[lastRow] + x[nextRow])) * |
| invC; |
| lastY = y[currentRow] = (y0[currentRow] + |
| a * |
| (lastY + |
| y[++currentRow] + |
| y[++lastRow] + |
| y[++nextRow])) * |
| invC; |
| } |
| } |
| set_bnd(1, x); |
| set_bnd(2, y); |
| } |
| } |
| } |
| |
| void diffuse2( |
| Float64List x, Float64List x0, Float64List y, Float64List y0, double dt) { |
| var a = 0; |
| lin_solve2(x, x0, y, y0, a, 1 + 4 * a); |
| } |
| |
| void advect(int b, Float64List d, Float64List d0, Float64List u, |
| Float64List v, double dt) { |
| var Wdt0 = dt * width; |
| var Hdt0 = dt * height; |
| var Wp5 = width + 0.5; |
| var Hp5 = height + 0.5; |
| for (var j = 1; j <= height; j++) { |
| var pos = j * rowSize; |
| for (var i = 1; i <= width; i++) { |
| var x = i - Wdt0 * u[++pos]; |
| var y = j - Hdt0 * v[pos]; |
| if (x < 0.5) { |
| x = 0.5; |
| } else if (x > Wp5) x = Wp5; |
| var i0 = x.toInt(); |
| var i1 = i0 + 1; |
| if (y < 0.5) { |
| y = 0.5; |
| } else if (y > Hp5) y = Hp5; |
| var j0 = y.toInt(); |
| var j1 = j0 + 1; |
| var s1 = x - i0; |
| var s0 = 1 - s1; |
| var t1 = y - j0; |
| var t0 = 1 - t1; |
| var row1 = j0 * rowSize; |
| var row2 = j1 * rowSize; |
| d[pos] = s0 * (t0 * d0[i0 + row1] + t1 * d0[i0 + row2]) + |
| s1 * (t0 * d0[i1 + row1] + t1 * d0[i1 + row2]); |
| } |
| } |
| set_bnd(b, d); |
| } |
| |
| void project(Float64List u, Float64List v, Float64List p, Float64List div) { |
| var h = -0.5 / sqrt(width * height); |
| for (var j = 1; j <= height; j++) { |
| var row = j * rowSize; |
| var previousRow = (j - 1) * rowSize; |
| var prevValue = row - 1; |
| var currentRow = row; |
| var nextValue = row + 1; |
| var nextRow = (j + 1) * rowSize; |
| for (var i = 1; i <= width; i++) { |
| div[++currentRow] = h * |
| (u[++nextValue] - u[++prevValue] + v[++nextRow] - v[++previousRow]); |
| p[currentRow] = 0.0; |
| } |
| } |
| set_bnd(0, div); |
| set_bnd(0, p); |
| |
| lin_solve(0, p, div, 1, 4); |
| var wScale = 0.5 * width; |
| var hScale = 0.5 * height; |
| for (var j = 1; j <= height; j++) { |
| var prevPos = j * rowSize - 1; |
| var currentPos = j * rowSize; |
| var nextPos = j * rowSize + 1; |
| var prevRow = (j - 1) * rowSize; |
| var nextRow = (j + 1) * rowSize; |
| |
| for (var i = 1; i <= width; i++) { |
| u[++currentPos] -= wScale * (p[++nextPos] - p[++prevPos]); |
| v[currentPos] -= hScale * (p[++nextRow] - p[++prevRow]); |
| } |
| } |
| set_bnd(1, u); |
| set_bnd(2, v); |
| } |
| |
| void dens_step( |
| Float64List x, Float64List x0, Float64List u, Float64List v, double dt) { |
| addFields(x, x0, dt); |
| diffuse(0, x0, x, dt); |
| advect(0, x, x0, u, v, dt); |
| } |
| |
| void vel_step( |
| Float64List u, Float64List v, Float64List u0, Float64List v0, double dt) { |
| addFields(u, u0, dt); |
| addFields(v, v0, dt); |
| var temp = u0; |
| u0 = u; |
| u = temp; |
| temp = v0; |
| v0 = v; |
| v = temp; |
| diffuse2(u, u0, v, v0, dt); |
| project(u, v, u0, v0); |
| temp = u0; |
| u0 = u; |
| u = temp; |
| temp = v0; |
| v0 = v; |
| v = temp; |
| advect(1, u, u0, u0, v0, dt); |
| advect(2, v, v0, u0, v0, dt); |
| project(u, v, u0, v0); |
| } |
| |
| void Function(Field) uiCallback; |
| |
| void setDisplayFunction(void Function(Field) func) { |
| displayFunc = func; |
| } |
| |
| void setUICallback(void Function(Field) callback) { |
| uiCallback = callback; |
| } |
| |
| void queryUI(Float64List d, Float64List u, Float64List v) { |
| for (var i = 0; i < size; i++) { |
| u[i] = v[i] = d[i] = 0.0; |
| } |
| uiCallback(Field(d, u, v, rowSize)); |
| } |
| |
| void update() { |
| queryUI(dens_prev, u_prev, v_prev); |
| vel_step(u, v, u_prev, v_prev, dt); |
| dens_step(dens, dens_prev, u, v, dt); |
| displayFunc(Field(dens, u, v, rowSize)); |
| } |
| } |
| |
| // Difference from JS version: Field takes an argument rowSize, but this |
| // used for display purpose only. |
| class Field { |
| final Float64List dens, u, v; |
| final int rowSize; |
| |
| Field(this.dens, this.u, this.v, this.rowSize); |
| |
| void setDensity(int x, int y, double d) { |
| dens[(x + 1) + (y + 1) * rowSize] = d; |
| } |
| |
| double getDensity(int x, int y) { |
| return dens[(x + 1) + (y + 1) * rowSize]; // rowSize from FluidField? |
| } |
| |
| void setVelocity(int x, int y, double xv, double yv) { |
| u[(x + 1) + (y + 1) * rowSize] = xv; |
| v[(x + 1) + (y + 1) * rowSize] = yv; |
| } |
| |
| double getXVelocity(int x, int y) { |
| return u[(x + 1) + (y + 1) * rowSize]; |
| } |
| |
| double getYVelocity(int x, int y) { |
| return v[(x + 1) + (y + 1) * rowSize]; |
| } |
| } |