2023-08-07 gpu js 研究

1. gpu相比于cpu,有强大的内存访问带宽和并行单元运算的能力
2. WebGL GPU.js

const { GPU } = require("gpu.js")
function generateMatrices() {
  this.matrices = [[], []]
  this.matrixSize = 4
  for (let y = 0; y < this.matrixSize; y++) {
    this.matrices[0].push([])
    this.matrices[1].push([])
    for (let x = 0; x < this.matrixSize; x++) {
      const value1 = parseInt((Math.random() * 10).toString())
      const value2 = parseInt((Math.random() * 10).toString())
      this.matrices[0][y].push(value1)
      this.matrices[1][y].push(value2)
    }
  }
}

var m = new generateMatrices()

function gpuMultiplyMatrix(m) {
  const gpu = new GPU()
  this.matrixSize = m.matrixSize
  const multiplyMatrix = gpu
    .createKernel(function (a, b, matrixSize) {
      let sum = 0

      for (let i = 0; i < matrixSize; i++) {
        sum += a[this.thread.y][i] * b[i][this.thread.x]
      }
      return sum
    })
    .setOutput([this.matrixSize, this.matrixSize])
  const startTime = performance.now()
  const resultMatrix = multiplyMatrix(
    m.matrices[0],
    m.matrices[1],
    this.matrixSize
  )
  //   console.log(resultMatrix, "resultMatrix")
  const endTime = performance.now()
  this.gpuTime = endTime - startTime + " ms"
  console.log("GPU TIME : " + this.gpuTime)
}
gpuMultiplyMatrix(m)
cpuMutiplyMatrix(m)
function cpuMutiplyMatrix(m) {
  const startTime = performance.now()
  const a = m.matrices[0]
  const b = m.matrices[1]
  let productRow = Array.apply(null, new Array(m.matrixSize)).map(
    Number.prototype.valueOf,
    0
  )
  let product = new Array(m.matrixSize)

  for (let p = 0; p < m.matrixSize; p++) {
    product[p] = productRow.slice()
  }

  for (let i = 0; i < m.matrixSize; i++) {
    for (let j = 0; j < m.matrixSize; j++) {
      for (let k = 0; k < m.matrixSize; k++) {
        product[i][j] += a[i][k] * b[k][j]
      }
    }
  }
  console.log(product)
  const endTime = performance.now()
  this.cpuTime = endTime - startTime + "ms"
  console.log("CPU TIME : " + this.cpuTime)
}

const { GPU } = require("gpu.js")

const gpu = new GPU()

let kernel = gpu
  .createKernel(function (a, b, c) {   //1. 函数体内不能用a.length,不被识别,应该作为参数传进来
    let sum = 0

    for (var i = 0; i < c; i++) {  //c 如果大于1000,则i只能循环到999,搞不懂
      sum += a[0] + b[0] 
    }
    return sum
  })
  .setOutput([1]) //2.  x,y,z 对应for(z){for(y){for(x){     函数体内的this.thread.x|y|z      }}}

let a = new Array(100)
let b = new Array(100)
for (let i = 0; i < 100; i++) {
  a[i] = Math.random()
  b[i] = Math.random()
}

console.log(a.length, "length")
console.time()
let result = kernel(a, b, a.length)
console.log(result)
console.timeEnd()

let kernel = gpu
  .createKernel(function (a, b, cc) {
    var sum = 0
 
    do {
      ++sum
    } while (sum != this.constants.size)
     
    do {
      ++sum
    } while (sum != 10000)
    do {
      ++sum
    } while (sum != 10000)
    return sum
  },{
    constants:{size:2000},  // 传参
    pipeline :true,  //搞不懂
    loopMaxIterations :2000,  // 设置循环限制,默认1000
    output:[1]
  })
  
 console.log( kernel(1,1,1).toArray())