AI Engine-ML v2 Intrinsics User Guide (UG1639) - 2025.2 English - UG1639

Document ID: UG1639
Release Date: 2025-11-20
Version: 2025.2 English

Datatype and instrinsic documentation
Deprecated List
API Reference
Accumulator Data Types
FP Complex Accumulator Types
Floating-Point Accumulator Types
Integer Accumulator Types
1024-bit accumulator types
2048-bit accumulator types
256-bit accumulator types
512-bit accumulator types
Integer Complex Accumulator Types
Basic Types: Summary
Compiler optimizations
Code annotations
Loop annotations
Loop transformations
Optimization annotations
Resource allocation annotations
Scheduling annotations
Code optimizations
Broadcast-extract
Chain simplification
Chain simplification (different accumulators)
Extract and concat
Extract and concat all elements
Insert and extract generating new vector
Merge of Multiply Accumulate operations
Remove extract and set for same accumulator component
Set-extract
Set-update-extract
Shuffles
Simplification between complex and real accumulators
Skip intermediate extractions
Skip intermediate extractions for 2k accumulators
Undefined simplifications
Conversion Functions
Load/Store Operations
Addressing intrinsics
Load 4x Operations
Streams
Cascade read
Cascade write
Stream read
Stream write
Load/store to FIFO registers
Nonlinear functions
Scalar nonlinear functions
Vector nonlinear functions
Scalar Data Types
Scalar Operations
Configuration
Mode Settings
Control registers
Status registers
Core ID
Cycle Counter
Events
Initialization
Integer Operations
Locks
Scalar updates and extracts
Stream access
Vector Conversions
Brain float (BF16) to integer conversions
Broadcast
Broadcast from scalar
Broadcast to vector
Updating all elements with element extracted from vector
Updating all elements with one
Updating all elements with zero
Casting
Concatenate vectors
Concatenate four vectors
Concatenate two vectors
Extract vector
Extract MX vector components
Extract sparse vector components
Extract/insert element
Insert vector
Insert MX vector components
Insert a vector into a vector
Insert sparse vector components
Pack/Unpack
Set vector
Set MX vector components
Set sparse vector components
Shift-Round-Saturate
AIE interface
Floating-point interface
Size interface
Upshift
AIE interface
Floating-point
Size interface
Vector Data Types
Brain Floating-Point Vector Types
16-bit Brain Floating-Point Vector Types
8-bit Brain Floating-Point Vector Types
Complex 16-bit Brain Floating-Point Vector Types
Complex Floating-Point Vector Types
Complex Vector Types
Floating-Point Vector Types
16-bit Floating-Point Vector Types
32-bit Floating-Point Vector Types
8-bit Floating-Point Vector Types
Integer Vector Types
1024-bit vector types
128-bit vector types
16-bit vector types
256-bit vector types
32-bit vector types
512-bit vector types
64-bit vector types
8-bit vector types
MX Vector Types
Sparse Vector Types
Sparse bfloat16/float16 vector types
Sparse float8 vector types
Sparse integer vector types
Vector Operations
Add/Subtract
Bitwise logical
Compare/Select
Initialization
Multiply Accumulate
Emulated Multiply-accumulate of 16-bit Complex Brain Floating-Point
Emulated Multiply-accumulate of 16b x 32b datatypes
Emulated Multiply-accumulate of 32-bit Floating-Point
Emulated Multiply-accumulate of 32b x 16b datatypes
Emulated Multiply-accumulate of 32b x 32b datatypes
Emulated Multiply-accumulate of Complex 32b x Complex 32b datatypes
Emulated Multiply-accumulate of Complex Float and Float datatypes
Multiply-accumulate of 16b x 16b complex integer datatypes
Multiply-accumulate of 16b x 16b integer datatypes
Multiply-accumulate of 32b x 16b complex integer datatypes
Multiply-accumulate of 32b x 16b integer datatypes
Multiply-accumulate of 8b x 8b integer datatypes
Multiply-accumulate of MX4 datatypes
Multiply-accumulate of MX6 datatypes
Multiply-accumulate of MX9 datatypes
Multiply-accumulate of bfloat16 * float16 datatypes
Multiply-accumulate of bfloat16 datatypes
Multiply-accumulate of float16 * bfloat16 datatypes
Multiply-accumulate of float16 datatypes
Multiply-accumulate of float8 and bfloat8 datatypes
Multiply-accumulate with a sparse matrix
Negation control in complex multiplication modes
Vector x scalar multiply-accumulate of 16b x 16b complex integer datatypes
Vector x scalar multiply-accumulate of 16b x 16b datatypes
Vector x scalar multiply-accumulate of 32b x 16b complex integer datatypes
Vector x scalar multiply-accumulate of 8b x 8b datatypes
Vector x scalar multiply-accumulate of float16 and bfloat16 datatypes
Shift
Shift element
Shuffle
Illustration of Shuffle Modes

Back to home page