Struct GpuMetricsComputer 

pub struct GpuMetricsComputer { /* private fields */ }

GPU-accelerated metrics computer with comprehensive hardware detection

Implementations

impl GpuMetricsComputer

pub fn new(config: GpuAccelConfig) -> Result<Self>

Create new GPU metrics computer with hardware detection

pub fn with_parallel_config(self, config: ParallelConfig) -> Self

Configure parallel processing

pub fn should_use_gpu(&self, datasize: usize) -> bool

Check if GPU acceleration should be used for given data size

pub fn is_gpu_available(&self) -> bool

Check if GPU is available

pub fn get_gpu_info(&self) -> Option<&GpuInfo>

Get GPU information if available

pub fn get_capabilities(&self) -> &PlatformCapabilities

Get hardware capabilities information

pub fn gpu_accuracy( &self, y_true: &Array1<i32>, ypred: &Array1<i32>, ) -> Result<f32>

Compute accuracy on GPU with intelligent fallback

pub fn gpu_mse<F>(&self, y_true: &Array1<F>, ypred: &Array1<F>) -> Result<F>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

Compute MSE on GPU with SIMD fallback

pub fn simd_mse<F>(&self, y_true: &Array1<F>, ypred: &Array1<F>) -> Result<F>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

SIMD-accelerated MSE computation

pub fn simd_accuracy( &self, y_true: &Array1<i32>, ypred: &Array1<i32>, ) -> Result<f32>

SIMD-accelerated accuracy computation

pub fn gpu_confusion_matrix( &self, y_true: &Array1<i32>, ypred: &Array1<i32>, num_classes: usize, ) -> Result<Array2<i32>>

Compute confusion matrix on GPU (falls back to CPU)

pub fn gpu_batch_metrics<F>( &self, y_true_batch: ArrayView2<'_, F>, y_pred_batch: ArrayView2<'_, F>, metrics: &[&str], ) -> Result<Vec<HashMap<String, F>>>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

GPU-accelerated batch metric computation with comprehensive fallbacks

pub fn simd_mae<F>(&self, y_true: &Array1<F>, ypred: &Array1<F>) -> Result<F>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

SIMD-accelerated MAE computation

pub fn simd_r2_score<F>( &self, y_true: &Array1<F>, ypred: &Array1<F>, ) -> Result<F>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

SIMD-accelerated R² score computation

pub fn benchmark_implementations<F>( &self, y_true: &Array1<F>, ypred: &Array1<F>, iterations: usize, ) -> Result<BenchmarkResults>

where F: Float + SimdUnifiedOps + Send + Sync + Sum,

Benchmark different implementations to choose the best one