Any fastest way to compute 3x3 Matrix inverse and Matrix multiplication?

[Logan.B]

I have an DELL computer with Intel Duo-Core 64-bit processors and server support is Dell Poweredge 2950 as the workstation along with with 2 dualcore, hyperthreading 3.73 GHz Xeon processors. I usually do 3x3 Matrix multiplication, but not able for multithreading for using in Matrix. Because of that currently I am using the single thread. And also the program is running quite slow than the usual process. IS there any way to compute the Matrix multiplication in a fast way?

[Broad-Man]

Try for this script; it is the cluster of the Matrix multiplication. It done by the using of the NSD, which can stored in the common block of the method due to the legacy of the Matrix dimension script.
The script using NSD:

  do i=1,NSD
    do j = 1,NSD
      C(i,j) = 0
      do k = 1,NSD
        C(i,j) = C(i,j) + A(i,k)*B(k,j)
      enddo !k
    enddo !j
  enddo !i
 
and
 
 
  do i=1,NSD
    do j = 1,NSD
      C(i,j) = sum( A(i,:)*B(:,j) )
    enddo !j
  enddo !i
}
}

[TEcH GuRuJI]

In the both 32 and 64bit the username is not required for the terminating. If you have 32 bit then you need for ifort, and if you have 64 bit you need 'gfortran -m32'. Observe the variables like GOMP_CPU_AFFINITY and KMP_AFFINITY. SSE is parallel with 4x4. With the low level source code you can able to find the reference.

[newtech]

The MKL does not help you till a large dimension. I would suggest you to use to disable hyper threading in the BIOS. This particularly runs in the operating system. It should make a difference between the compiler and the “ifort”. The LOOP COUNT and UNROLL of the code may be sometimes gives the similar result.

[Shen]

The most efficient sub routine for creating the arbitrary 3x3 matrix inversion multiplication.Consider the code for passing some 3x3 prebuilt pointer array.
type_3x3

  real(8), target :: x(3,3)
end type type_3x3
type type_3x3_pointer
  type(type_3x3), pointer :: p
end type_3x3_pointer

subroutine matmult_3x3_ListOf_3x3(inA, ListOf_B, ListOf_C, n)
real(8) :: inA(3,3)
integer :: n
type(type_3x3_pointer) :: ListOf_B(n), ListOf_C(n)
real(8) :: A(3,3)
real(8) :: temp
A=inA
do iVector=1,n
   do i=1,3
     do j=1,3
       temp = 0.0D
       do k=1,3
         temp = temp + ListOf_B(iVector)%p%x(i,k) * A(k,j)
       end do
       ListOf_C(iVector)%p%x(i,j) = temp
    end do
  end do
end do
}
}