Struct AdaptiveOptimizer 

pub struct AdaptiveOptimizer { /* private fields */ }

Adaptive optimizer for runtime optimization

Implementations

impl AdaptiveOptimizer

pub fn new() -> Self

Create a new adaptive optimizer

pub fn update_performance_data(&mut self, data: &KernelPerformance)

Update performance data by recording an OptimizationResult synthesised from the observed KernelPerformance. The result is stored under the special key "__perf_trends__" so that optimize_kernel can later inspect aggregate patterns when no per-kernel history is available.

Strategy selection heuristic:

• High throughput (> 1 GOp/s) → prefer Vectorization
• Low execution count (< 10 executions) → prefer ConstantFolding
• Otherwise → cycle through the available strategies in insertion order

pub fn optimize_kernel( &self, kernel_id: &str, config: &JitConfig, ) -> Result<String, JitError>

Suggest an optimization directive for the kernel identified by kernel_id.

If the kernel has dedicated history (populated by callers that use kernel_id as the history key), the most successful OptimizationResult is selected. Otherwise the method falls back to the aggregate performance trends recorded by update_performance_data. The returned String is a human-readable directive (e.g. "vectorize", "unroll") that the JIT back-end may interpret as a compilation hint.

Trait Implementations

impl Debug for AdaptiveOptimizer

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.