const { assert, skip, test, module: describe, only } = require('qunit'); const { GPU, input } = require('../../src'); describe('features: dev mode'); test('are added to GPU instance .kernels property', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value; }, { output: [1] }); assert.equal(gpu.kernels.length, 1); assert.deepEqual(kernel(1), new Float32Array([1])); gpu.destroy(); }); test('works with integer', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value; }, { output: [1] }); assert.deepEqual(kernel(1), new Float32Array([1])); gpu.destroy(); }); test('works with float', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value; }, { output: [1] }); assert.deepEqual(kernel(1.5), new Float32Array([1.5])); gpu.destroy(); }); test('works with array', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.x]; }, { output: [4] }); assert.deepEqual(kernel([1,2,3,4]), new Float32Array([1,2,3,4])); gpu.destroy(); }); test('works with matrix', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.y][this.thread.x]; }, { output: [4, 2] }); assert.deepEqual(kernel( [ [1,2,3,4], [5,6,7,8] ] ), [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ]); gpu.destroy(); }); test('works with cube', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.z][this.thread.y][this.thread.x]; }, { output: [4, 2, 2] }); assert.deepEqual(kernel( [ [ [1,2,3,4], [5,6,7,8] ], [ [9,10,11,12], [13,14,15,16] ] ] ), [ [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ],[ new Float32Array([9,10,11,12]), new Float32Array([13,14,15,16]), ] ]); gpu.destroy(); }); test('works with input array', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.x]; }, { output: [4] }); assert.deepEqual(kernel(input([1,2,3,4], [4])), new Float32Array([1,2,3,4])); gpu.destroy(); }); test('works with input matrix', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.y][this.thread.x]; }, { output: [4, 2] }); assert.deepEqual(kernel(input([1,2,3,4,5,6,7,8], [4, 2])), [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ]); gpu.destroy(); }); test('works with input cube', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.z][this.thread.y][this.thread.x]; }, { output: [4, 2, 2] }); assert.deepEqual(kernel( input([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16], [4,2,2]) ), [ [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ],[ new Float32Array([9,10,11,12]), new Float32Array([13,14,15,16]), ] ]); gpu.destroy(); }); test('works with texture', () => { const texture = ((new GPU()).createKernel(function (cube) { return cube[this.thread.z][this.thread.y][this.thread.x]; }, { output: [4,2,2], pipeline: true }))([ [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ],[ new Float32Array([9,10,11,12]), new Float32Array([13,14,15,16]), ] ]); assert.ok(texture.constructor.name.match('Texture')); const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function(value) { return value[this.thread.z][this.thread.y][this.thread.x]; }, { output: [4, 2, 2] }); assert.deepEqual(kernel( texture ), [ [ new Float32Array([1,2,3,4]), new Float32Array([5,6,7,8]), ],[ new Float32Array([9,10,11,12]), new Float32Array([13,14,15,16]), ] ]); gpu.destroy(); }); test('works with adding functions', () => { const gpu = new GPU({ mode: 'dev' }); function addOne(value) { return value + 1; } gpu.addFunction(addOne); const kernel = gpu.createKernel(function(value) { return addOne(value); }, { output: [1] }); assert.deepEqual(kernel(1), new Float32Array([2])); gpu.destroy(); }); // issue #719: kernels returning an array produced NaN in dev mode, because // gpu-mock.js assigned the returned array into a Float32Array slot test('works with array return', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function() { return [1, 2]; }, { output: [2] }); assert.deepEqual(kernel(), [ new Float32Array([1, 2]), new Float32Array([1, 2]), ]); gpu.destroy(); }); test('works with array return in matrix', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function() { return [this.thread.x, this.thread.y]; }, { output: [2, 2] }); assert.deepEqual(kernel(), [ [new Float32Array([0, 0]), new Float32Array([1, 0])], [new Float32Array([0, 1]), new Float32Array([1, 1])], ]); gpu.destroy(); }); test('works with array return in cube', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function() { return [this.thread.z, 9]; }, { output: [1, 1, 2] }); assert.deepEqual(kernel(), [ [[new Float32Array([0, 9])]], [[new Float32Array([1, 9])]], ]); gpu.destroy(); }); test('array returns match cpu mode', () => { function build(mode) { const gpu = new GPU({ mode }); const kernel = gpu.createKernel(function() { return [1, 2, 3]; }, { output: [2, 2] }); const result = kernel(); gpu.destroy(); return result; } assert.deepEqual(build('dev'), build('cpu')); }); test('number returns stay a flat Float32Array', () => { const gpu = new GPU({ mode: 'dev' }); const kernel = gpu.createKernel(function() { return 7; }, { output: [3] }); const result = kernel(); assert.ok(result instanceof Float32Array, 'expected a Float32Array, got ' + result.constructor.name); assert.deepEqual(result, new Float32Array([7, 7, 7])); gpu.destroy(); });