Accelerate provides two functions for performing matrix multiplication. The first is the vDSP_mmulD
function which is demonstrated below.
// Arrays represent 2D matrices, rows and columns are defined for each array
// Note that if matrix A is m x p then matrix B must be p x n thus matrix C is m x n
let a: [Double] = [5, 2, 3,
4, 5, 6,
7, 8, 9]
let b: [Double] = [1, 2, 3, 4,
4, 5, 6, 7,
1, 3, 2, 1]
let m: vDSP_Length = 3 // rows in matrices A and C
let n: vDSP_Length = 4 // columns in matrices B and C
let p: vDSP_Length = 3 // columns in matrix A, rows in matrix B
let stride: vDSP_Stride = 1
var c = [Double](repeating: 0, count: Int(m * n))
vDSP_mmulD(a, stride, b, stride, &c, stride, m, n, p)
// Print result
for i in 0..<m {
for j in 0..<n {
let idx = Int(i * n + j)
print(c[idx], terminator: " ")
}
print("")
}
Compile and run the example using the commands shown here. Notice you must compile with the new LAPACK headers to avoid deprecation warnings.
swiftc -Xcc -DACCELERATE_NEW_LAPACK -Ounchecked matmul-accel.swift
./matmul-accel
The result is:
16.0 29.0 33.0 37.0
30.0 51.0 54.0 57.0
48.0 81.0 87.0 93.0
Accelerate also provides the cblas_dgemm
function for matrix multiplication which is demonstrated here. This example is compiled the same way as the previous example and provides the same result.
// Arrays represent 2D matrices, rows and columns are defined for each array
// Note that if matrix A is m x k then matrix B must be k x n thus matrix C is m x n
var a: [Double] = [5, 2, 3,
4, 5, 6,
7, 8, 9]
var b: [Double] = [1, 2, 3, 4,
4, 5, 6, 7,
1, 3, 2, 1]
let m: Int32 = 3 // rows in matrices A and C
let n: Int32 = 4 // columns in matrices B and C
let k: Int32 = 3 // columns in matrix A, rows in matrix B
let alpha = 1.0
let beta = 0.0
var c = [Double](repeating: 0, count: Int(m * n))
cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, m, n, k, alpha, &a, k, &b, n, beta, &c, n)
// Print result
for i in 0..<m {
for j in 0..<n {
let idx = Int(i * n + j)
print(c[idx], terminator: " ")
}
print("")
}
Gavin Wiggins © 2024.
Made on a Mac with Genja. Hosted on GitHub Pages.