AI Engine-ML v2 Intrinsics User Guide  v2025.1
Loading...
Searching...
No Matches
Scalar Data Types

All the standard C scalar data-types are supported. More...

Data Structures

class  bfloat16
 Brain floating point 16-bit: 1 bit sign, 8 bits exponent, 7 bits mantissa. More...
 
class  bfloat8
 Brain floating point 8-bit: 1 bit sign, 5 bits exponent, 2 bits mantissa. More...
 
struct  cbfloat16
 
struct  cfloat
 
class  float
 Emulated data type following IEEE-754 standard: 1 bit sign, 8 bits exponent, 23 bits mantissa. More...
 
class  float16
 Following IEEE-754 standard half precision: 5 bits exponent, 10 bits mantissa. More...
 
class  float8
 Floating point 8-bit: 1 bit sign, 4 bits exponent, 3 bits mantissa. More...
 
struct  mx4
 Shared exponent (formerly block floating-point or bfp) 3-bit mantissa, but represented as mx6 internally. More...
 
struct  mx6
 Shared exponent (formerly block floating-point or bfp) 5-bit mantissa. More...
 
struct  mx9
 Shared exponent (formerly block floating-point or bfp) 8-bit mantissa. More...
 

Detailed Description

All the standard C scalar data-types are supported.

Emulated complex floating point 32-bit: 1 bit sign, 8 bits exponent, 23 bits mantissa.

Emulated complex brain floating point 16-bit: 1 bit sign, 8 bits exponent, 7 bits mantissa.

Note
All types except float/double support signed and unsigned prefixes

Data Structure Documentation

◆ bfloat16

class bfloat16

Brain floating point 16-bit: 1 bit sign, 8 bits exponent, 7 bits mantissa.

◆ bfloat8

class bfloat8

Brain floating point 8-bit: 1 bit sign, 5 bits exponent, 2 bits mantissa.

◆ cbfloat16

struct cbfloat16

Public Member Functions

 cbfloat16 (bfloat16 a, bfloat16 b)
 

Data Fields

bfloat16 imag
 
bfloat16 real
 

Constructor & Destructor Documentation

◆ cbfloat16()

cbfloat16::cbfloat16 ( bfloat16 a,
bfloat16 b )

Field Documentation

◆ imag

bfloat16 cbfloat16::imag

◆ real

bfloat16 cbfloat16::real

◆ cfloat

struct cfloat

Public Member Functions

 cfloat (float a, float b)
 

Data Fields

float imag
 
float real
 

Constructor & Destructor Documentation

◆ cfloat()

cfloat::cfloat ( float a,
float b )

Field Documentation

◆ imag

float cfloat::imag

◆ real

float cfloat::real

◆ float

class float

Emulated data type following IEEE-754 standard: 1 bit sign, 8 bits exponent, 23 bits mantissa.

◆ float16

class float16

Following IEEE-754 standard half precision: 5 bits exponent, 10 bits mantissa.

◆ float8

class float8

Floating point 8-bit: 1 bit sign, 4 bits exponent, 3 bits mantissa.

◆ mx4

struct mx4

Shared exponent (formerly block floating-point or bfp) 3-bit mantissa, but represented as mx6 internally.

◆ mx6

struct mx6

Shared exponent (formerly block floating-point or bfp) 5-bit mantissa.

◆ mx9

struct mx9

Shared exponent (formerly block floating-point or bfp) 8-bit mantissa.