fz::GInterpConfig Struct Reference

#include <ginterp_stage.h>

Public Attributes
double	error_bound
uint32_t	quant_radius
	Quantization radius (codes lie in [0, 2*radius)).
uint32_t	num_elements
	Total element count (= dim_x*dim_y*dim_z).
uint32_t	outlier_count
	Actual outlier count (post-execute).
DataType	input_type
	Float input type (1 B).
DataType	code_type
	Quant code type (1 B).
uint8_t	ndim
	Spatial dimensionality (3 in MVP).
uint8_t	eb_mode
	ErrorBoundMode cast to uint8_t.
uint32_t	dim_x
	X (fast) dimension.
uint32_t	dim_y
	Y dimension.
uint32_t	dim_z
	Z dimension.
uint32_t	anchor_dim_x
	Anchor grid X extent.
uint32_t	anchor_dim_y
	Anchor grid Y extent.
uint32_t	anchor_dim_z
	Anchor grid Z extent.
float	user_eb
	Original user-specified bound (before mode conversion).
float	value_base
	value_range (NOA) / max(|data|) (REL) used in conversion.
double	intp_alpha
	Resolved alpha (auto-tuned from rel_eb or fixed).
double	intp_beta
	Resolved beta (default 4.0).
uint8_t	intp_use_md [6]
	Resolved use_md[level], booleans as u8.
uint8_t	auto_tuning_mode
	0=off, 1=cheap, 3=full, 4=full+alpha sweep, 5+=manual α/β.
uint8_t	pad [8]
	Reserved for future fields (alignment also).

Detailed Description

Serialized G-Interp configuration stored in FZMBufferEntry.stage_config. Fits within the 128-byte FZM_STAGE_CONFIG_SIZE limit.

Member Data Documentation

error_bound

double fz::GInterpConfig::error_bound

Absolute bound used by the decompressor. Double so double inputs keep full precision.