solving Ax=b in parallel using GPU cores

I want to solve Ax=b using gpu cores. The matrix A=(f_i(x_j)) and b=(b(x_j)) is constructed by evaluating function over some points. Please suggest how can I use the gpu cores to solve it? Any book article related is also helpful. thanks...

gpuDevice()

ans =

CUDADevice with properties:

                      Name: 'GeForce GT 740M'
                     Index: 1
         ComputeCapability: '3.5'
            SupportsDouble: 1
             DriverVersion: 6.5000
            ToolkitVersion: 5.5000
        MaxThreadsPerBlock: 1024
          MaxShmemPerBlock: 49152
        MaxThreadBlockSize: [1024 1024 64]
               MaxGridSize: [2.1475e+09 65535 65535]
                 SIMDWidth: 32
               TotalMemory: 2.1475e+09
                FreeMemory: 2.0033e+09
       MultiprocessorCount: 2
              ClockRateKHz: 1032500
               ComputeMode: 'Default'
      GPUOverlapsTransfers: 1
    KernelExecutionTimeout: 1
          CanMapHostMemory: 1
           DeviceSupported: 1
            DeviceSelected: 1

Answer 1

[Joss Knight]

If you can form a dense A then do so. Then call

x = gpuArray(A) \ b;

If you need A to remain a function (perhaps it is too big to fit in GPU memory) then currently you can only do this on the CPU using one of the iterative solvers: gmres, cgs, bicg, lsqr etc