Benchmarking $(30$ points$)$

$[5$ Points$]$ Write a C code for addition of two $1$D matrices with the Number of elements in an array $($N$)$ equal to $10000000.$ You will be rewriting this C code.

$$ Hint: You may need to use dynamic memory allocation function malloc$()$

$//$ Allocate host memory

$$ a $=($float$*)$malloc$($sizeof$($float$)*$ N$)$;

$[5$ Points$]$ Write an equivalent python code.

$[5$ Points$]$ Compile the C code and Compare the performance between C and python codes. Which code performs better and why? Also, without dynamic allocation, report the error if any while running the C code. What$$s the eason behind that error?

$[5$ Points$]$ Apply Numba JIT $($Just in Time$)$ decorator in Python Code script$)$and compare the performance between two python codes. Benchmarking Cont$$d

$[5$ Points$]$ Employ OpenACC pragma to optimize the C code $($matrix$/$vector addition$)$ you wrote earlier and compare its performance with serial implementation.

$[5$ Points$]$ Implement the C code $($matrix addition$)$ in CUDA using single and multiple cuda$/$GPU threads and compare the performance.