Perform the product between two matrix.
[CPU GPU dA dB]=gpuMult(A,B,[option1 option2])
Can be :
Pointer on the matrix stored in GPU memory.
Pointer on the matrix stored in CPU memory.
Can be :
Pointer on the matrix stored in GPU memory.
Pointer on the matrix stored in CPU memory.
option1 :
0 : the computation result is store in CPU memory.
1 : the computation result is store in GPU memory.
2 : the computation result is store in GPU and GPU memory.
option2 :
0 : If input data are stored in CPU memory, the data load in GPU memory will be erased.
1 : Only the first argument will be preserved.
2 : Only the second argument will be preserved.
12 : Both will be preserved.
CPU is the result of computation in CPU memory.
GPU is the result of computation in GPU memory.
dA is data of A locate in GPU memory.
dB is data of B locate in GPU memory.
gpuMult perform an multiplication between two matrix or between a number and a matrix.
Call this function with option [0 0] is equal to call it without option.
The number of output parameters depends of options.
This function use cuBLAS to perform operations.
A=rand(300,300)+%i*rand(300,300); B=rand(300,300)+%i*rand(300,300); [GPU dA dB]=gpuMult(A,B,[1 12]); GPU1=gpuMult(GPU,dB,[1 0]); GPU=gpuFree(GPU); GPU2=gpuMult(GPU1,dA,[1 0]); CPU=gpuMult(GPU2,GPU1); GPU1=gpuFree(GPU1); GPU2=gpuFree(GPU2); clear GPU1; clear GPU2; clear CPU; clear dA; clear dB; clear A; clear B; | ![]() | ![]() |