I compare here the C-language implementation and Julia.
I noticed that the number of sweeps executed by these two implementations was different.
I tracked it down to the return line in timeStep4_Jacobi: this C function fails to return the square root of the error accumulator, which the Julia functions do. Hence, when the error is compared with the tolerance, the C-language implementation exits the while loop much sooner. The number of sweeps is consequently not equal between the implementations, which invalidates the comparison.
I think it would be best for the benchmark to fix the number of sweeps and not test the error in the loop condition.
I compare here the C-language implementation and Julia.
I noticed that the number of sweeps executed by these two implementations was different. I tracked it down to the return line in
timeStep4_Jacobi:this C function fails to return the square root of the error accumulator, which the Julia functions do. Hence, when the error is compared with the tolerance, the C-language implementation exits the while loop much sooner. The number of sweeps is consequently not equal between the implementations, which invalidates the comparison.I think it would be best for the benchmark to fix the number of sweeps and not test the error in the loop condition.