Arithmetic Intensity and Classify Bottleneck

Implement a function that computes the arithmetic intensity of a computational kernel and classifies whether it is memory bound or compute bound using the Roofline Model. The Roofline Model is a widely used framework for understanding the performance bottleneck of computational workloads on hardware. It relates the operational performance a machine can achieve to how data intensive (or compute intensive) a given operation is. Your function should take the following inputs: flops: Total number of floating point operations performed by the kernel bytes accessed: Total bytes read from and written to memory peak performance: The hardware's peak computational throughput (in FLOP/s) peak bandwidth: The hardware's peak memory bandwidth (in bytes/s) Your function should return a dictionary with: 'arithmetic intensity': The ratio of computation to memory traffic (in FLOP/byte), rounded to 4 decimal places 'ridge point': The hardware's critical threshold that separates memory bound from compute bound regimes (in FLOP/byte), rounded to 4 decimal places 'bottleneck': A string, either 'memory bound' or 'compute bound' 'achieved performance': The theoretical attainable performance under the roofline model (in FLOP/s), rounded to 4 decimal places 'utilization percent': What percentage of the hardware's peak compute is theoretically achievable, rounded to 4 decimal places When the arithmetic intensity is exactly equal to the ridge point, classify it as 'compute bound'.