intermediate · simd
When Does SIMD Actually Help?
When does SIMD help, and what stops a loop from being vectorized?
Practical scenario
A numerical loop is expected to be four times faster after a rewrite and is not. Someone suggests the CPU "must not support SIMD", but the report shows the loop was never vectorized.
What it tests
- • Understanding of the independence requirement
- • Knowledge of what blocks the compiler in practice
- • Whether the candidate verifies vectorization rather than assuming it