Struct CudaEncodeStageAccelerator 

pub struct CudaEncodeStageAccelerator { /* private fields */ }

CUDA implementation of selected JPEG 2000 encode stages.

Implementations

impl CudaEncodeStageAccelerator

pub fn for_auto_host_output() -> Self

Create the measured Auto route for host-output HTJ2K encode.

CUDA keeps the DWT and HT code-block stages, while forward RCT and Tier-2 packetization stay on the CPU for the current host-pixel path.

pub fn prefer_cpu_forward_rct(self, prefer_cpu_forward_rct: bool) -> Self

Prefer scalar CPU forward RCT while keeping later CUDA stages enabled.

pub fn prefer_cpu_packetization(self, prefer_cpu_packetization: bool) -> Self

Prefer scalar CPU Tier-2 packetization while keeping CUDA Tier-1/HT block coding enabled.

This is useful for batches of many small tiles where launching a CUDA packetization kernel and copying several tiny descriptor buffers per tile costs more than forming the packet body on the host.

pub fn prefer_cpu_ht_subband(self, prefer_cpu_ht_subband: bool) -> Self

Prefer host sub-band quantization while keeping batched CUDA HT code-block encode enabled.

This avoids launching one CUDA quantize/subband path for every prepared subband in multi-resolution precomputed transcode outputs, where the many tiny launches cost more than CPU quantization.

pub fn prefer_cpu_quantize_subband( self, prefer_cpu_quantize_subband: bool, ) -> Self

Prefer host sub-band quantization while keeping CUDA HT code-block encode enabled.

Multi-resolution transcode workloads can contain thousands of small subbands; for those, CPU quantization plus one batched HT code-block encode per tile is currently faster than launching CUDA quantization for every subband.

pub const fn collected_stage_timings(&self) -> CudaEncodeStageTimings

Return cumulative CUDA encode stage timings collected by this accelerator.

pub fn reset_collected_stage_timings(&mut self)

Clear cumulative CUDA encode stage timings without changing dispatch counters.

Trait Implementations

impl Clone for CudaEncodeStageAccelerator

fn clone(&self) -> CudaEncodeStageAccelerator

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for CudaEncodeStageAccelerator

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

Formats the value using the given formatter.

impl Default for CudaEncodeStageAccelerator

fn default() -> CudaEncodeStageAccelerator

Returns the “default value” for a type.