noodle_config.h File Reference

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Macros
#define	NOODLE_USE_SDFAT
#define	NOODLE_FILE_FORMAT_TEXT   0
#define	NOODLE_FILE_FORMAT_BIN   1
#define	NOODLE_FILE_FORMAT   NOODLE_FILE_FORMAT_BIN
#define	NOODLE_POOL_NONE   0
#define	NOODLE_POOL_MAX   1
#define	NOODLE_POOL_MEAN   2
#define	NOODLE_POOL_MODE   NOODLE_POOL_MEAN
#define	NOODLE_FCN_BLOCK   128
	Number of float weights buffered per file-backed FCN read block.
#define	NOODLE_MAX_K   5
	Largest convolution kernel width copied into fixed stack scratch.

Macro Definition Documentation

NOODLE_FCN_BLOCK

#define NOODLE_FCN_BLOCK   128

Number of float weights buffered per file-backed FCN read block.

Used by float-input fully connected layers with FCNFile parameters. Binary file mode reads each block as raw float32 data; text file mode fills the same block by parsing scalar values one at a time. Higher values can reduce file read calls, but increase stack usage by NOODLE_FCN_BLOCK * sizeof(float) bytes.

NOODLE_FILE_FORMAT

#define NOODLE_FILE_FORMAT   NOODLE_FILE_FORMAT_BIN

NOODLE_FILE_FORMAT_BIN

#define NOODLE_FILE_FORMAT_BIN   1

NOODLE_FILE_FORMAT_TEXT

#define NOODLE_FILE_FORMAT_TEXT   0

NOODLE_MAX_K

#define NOODLE_MAX_K   5

Largest convolution kernel width copied into fixed stack scratch.

Convolution and depthwise-convolution paths use stack arrays sized from this macro when copying one kernel at a time. Set it to the largest K used by the firmware.

NOODLE_POOL_MAX

#define NOODLE_POOL_MAX   1

NOODLE_POOL_MEAN

#define NOODLE_POOL_MEAN   2

NOODLE_POOL_MODE

#define NOODLE_POOL_MODE   NOODLE_POOL_MEAN

NOODLE_POOL_NONE

#define NOODLE_POOL_NONE   0

NOODLE_USE_SDFAT

#define NOODLE_USE_SDFAT