[Python] How to Efficiently Solve Millions of Independent 2x2 Linear Systems with Augmented...

I need to solve millions to tens of millions of independent 2-variable linear systems represented in augmented matrix form. The input consists of coefficient vectors A, B, C, A', B', C' (each containing millions to hundreds of millions of elements). At each index i, the system is defined as:

Ax + By = -C
A'x + B'y = -C'


Current Python Approach (Inefficient): Using a for loop with Sympy to solve each system sequentially:

from sympy import symbols, Eq, solve

x = []
y = []
for i in range(len(A)):
sol = solve([
Eq(A*x_sym + B*y_sym, -C),
Eq(A_prime*x_sym + B_prime*y_sym, -C_prime)
], [x_sym, y_sym])
x.append(sol[x_sym])
y.append(sol[y_sym])


This approach becomes impractical for large datasets (e.g., 10M+ entries) due to excessive computation time.

How can I write a DolphinDB script to:

• Accept input vectors A, B, C, A_prime, B_prime, C_prime.
• Compute x and y using vectorized/parallelized operations? Handle edge cases (e.g., singular matrices)?
• Handle edge cases (e.g., singular matrices)?

Continue reading...