fz::StageTimingResult Struct Reference

#include <perf.h>

Public Member Functions
float	throughput_gbs () const noexcept
	Input throughput in GB/s (input_bytes / elapsed_ms, not host time).

Public Attributes
std::string	name
	Stage name (e.g. "lorenzo", "rle")
int	level
	DAG execution level (0 = source stages)
float	elapsed_ms
size_t	input_bytes
	Total bytes across all input buffers.
size_t	output_bytes
	Total bytes across all output buffers.

Detailed Description

Timing and throughput for a single stage in the DAG. Populated by CompressionDAG::collectTimings() and stored in PipelinePerfResult.

Member Data Documentation

elapsed_ms

float fz::StageTimingResult::elapsed_ms

GPU execution time measured by CUDA events (milliseconds). Covers the interval [cudaEventRecord(start) → cudaEventRecord(completion)], which is the time the stage occupied its CUDA stream.