Fortran语言，do-end do循环，相互包含测试，自动性能优化

1）上代码

    !$omp parallel private(n, j, dx, dy, dz, r, a)
    do n = 1, nsteps
        !$omp do
        do i = 0, nparticles - 1
            x_tmp(i) = x(i) + vx(i) * dt
            y_tmp(i) = y(i) + vy(i) * dt
            z_tmp(i) = z(i) + vz(i) * dt

            do j = 0, nparticles - 1
                dx = x(j) - x(i)
                dy = y(j) - y(i)
                dz = z(j) - z(i)

                R = sqrt(dx ** 2 + dy ** 2 + dz ** 2 + softening)
                a = G * m(j) / R ** 3 * dt

                vx(i) = vx(i) + a * dx
                vy(i) = vy(i) + a * dy
                vz(i) = vz(i) + a * dz
            end do
			

            do j1 = 0, nparticles - 1
                dx = x(j1) - x(i)
            end do
			
            do j2 = 0, nparticles - 1
                dx = x(j2) - x(i)
					
				do j3 = 0, nparticles - 1
					dx = x(j3) - x(i)
				end do
            end do
        end do

        !$omp do
        do i = 0, nparticles - 1
            x(i) = x_tmp(i)
            y(i) = y_tmp(i)
            z(i) = z_tmp(i)
            do j4 = 0, nparticles - 1
                dx = x(j4) - x(i)
				
				do j = 0, nparticles - 1
					dx = x(j) - x(i)
				end do
            end do
        end do
		
		do j = 0, nparticles - 1
			dx = x(j) - x(i)
		end do

        !$omp master
        print *, "n =", n
        do i = 0, nparticles - 1
            print *, i, ":", x(i), y(i), z(i)
        end do
        !$omp end master
    end do
    !$omp end parallel

2)

循环对：

 [[8, 19, -100], [22, 24, -100], [29, 31, -100], [26, 32, -100], [3, 33, -100], [43, 45, -100], [40, 46, -100], [36, 47, -100], [49, 51, -100], [55, 57, -100], [1, 59, -100]]

3）下面这个结果有错

 [[8, 19, 4], [22, 24, 0], [29, 31, 3], [26, 32, 1], [3, 33, 10], [43, 45, 6], [40, 46, 7], [36, 47, 10], [49, 51, 10], [55, 57, 10], [1, 59, -1]]

4）这个才是对的。

 [[8, 19, 4], [22, 24, 4], [29, 31, 3], [26, 32, 4], [3, 33, 10], [43, 45, 6], [40, 46, 7], [36, 47, 10], [49, 51, 10], [55, 57, 10], [1, 59, -1]]

目前看起来是对的。