Question

A data science team is deciding which of their pipeline stages will benefit most from GPU acceleration before purchasing hardware. They list four candidate stages. Which stage is the strongest candidate for GPU acceleration because it is inherently parallelizable across thousands of identical operations?

Accepted Answer

Applying the same matrix multiplication across every image in a large batch during neural network training

Answer

Parsing a configuration file line-by-line where each line depends on the state set by the previous line

Answer

Reading a sequence of log entries and writing each one to disk in strict chronological order

Answer

Executing a recursive business-rules script where each decision branch must complete before the next begins