Vitis 2025.2

Documentation changes

• Add introduction section
• Examples moved to separated files that can be compiled individually
• Expanded section on block floating point vectors
• Misc aesthetic improvements

Global AIE API changes

• Improved support for pipelined_loop(s) including constexpr detection of peeled sections (in_front, in_back, in_loop member functions)
• Leverage vector-scalar muls across the API on AIE-MLv2

Changes to data types

• Tensor buffer streams: Enable user-configuration of dimension splits
• Tensor buffer streams: Add block type support on XDNA2 (bfp16ebs8) and AIE-MLv2 (mx9)
• Tensor buffer streams: Add acquire/release semantics when created with async io_buffers
• block_vector: Add support for arbitrary sized block_vectors
• vector: Improve simple element pushed to large vectors

Changes to operations

• add_reduce: Accept accumulator inputs
• add_reduce: Optimize bfloat16 reductions
• add_reduce: Implemented tree reduction for inputs composed of multiple vector registers
• broadcast: Add broadcast_vector that initializes a large vector from a repeating sequence of values
• broadcast: Add overload accepting a list of scalar values, allowing vector initialization from a small repeating sequence of values
• cast: Reduce overhead of accumulator-vector conversions with same element representation (e.g. acc32-int32)
• compare: Improve efficiency of scalar floating point comparisons
• fft: Add cbfloat16 radix 3 and radix 5 support on AIE-ML and AIE-MLv2
• fft: Add cfloat radix 4 support on AIE-ML and AIE-MLv2
• filter: Expand filter_odd/filter_even to accept precomputed shuffle mode, allowing runtime selection of filter mode and step
• mmul: Add 2*8*2 modes for cbfloat16 and cfloat on AIE-ML
• mmul: Add 2*8*4 modes for cbfloat16 and cfloat on AIE-MLv2
• mac/mscmac_square: Add dynamic zeroization support with aie::op_zero
• sliding_mul: Add partial_sliding_mul, an object oriented interface that stores intermediate results in an optimal internal representation. This is specially beneficial when complex multiplications are emulated
• sliding_mul: Improved performance for cint16 x int16
• sliding_mul: Improved performance of emulated int32 and cint32 sliding_mul implementations on AIE-MLv2
• sliding_mul: Add dynamic zeroization support with aie::op_zero

ADF integration

• Add APIs to create tensor buffer streams directly from ADF objects

Vitis 2025.1

Documentation changes

• Document minimum point size for FFTs
• Add notes on emulated mmul modes
• Clarify lane requirements for complex * real sliding_muls
• Clarify requirement of using only native accum types with ADF APIs
• Update CSS rules for tables
• Clarify sparse_vector::extract_data behavior
• Invert AMD logo when viewing docs in dark mode
• Correct rounding modes in to_fixed documentation
• Move examples into separate files
• Document rounding/saturation mode APIs

Global AIE API changes

• Expose the following helper functions
  • unroll_for
  • unroll_times
  • locate_in_register
  • pipelined_loop
  • pipelined_loops
• Add store_floor_v and store_floor_bytes_v APIs
• Add compile flag to disable default scalar implementations: AIE_API_PROVIDE_DEFAULT_SCALAR_IMPLEMENTATION
• Change default behavior of to_fixed for bfloat16 to be in line with other types
• Add to_fixed_floor API for AIE-ML onwards
• Update inv/invsqrt implementation of XDNA2 and AIE-MLv2 to leverage scalar hardware implementation for improved accuracy
• Add zeroization support using aie::op_zero to mac, msc, mac_square, msc_square, sliding_mac, and sliding_msc
• Optimize unaligned loads for AIE-ML onwards

Changes to data types

• Unaligned input stream: Fix fifo-based implementation on XDNA2 and AIE-MLv2
• Unaligned input stream: Add arbitrary vector size support
• Add float8 and float16 support on AIE-MLv2
• Add cbfloat16 support on AIE-MLv2
• Add scoped_mode RAII types for temporarily setting saturation/rounding modes

Changes to operations

• abs_square: Add support for AIE-ML onwards
• add/sub: Support arbitrary vector sizes
• add_reduce: Support arbitrary vector sizes
• compare: Add support for cbfloat16 and cfloat
• conj: Add support for cbfloat16 and cfloat
• fft: Optimize output stage for radix 4 FFTs taking cint16 input on AIE
• fft: Fix radix 2 and radix 4 upscaler stages on AIE
• fft: Leverage sub_mask for cbfloat16 FFTs improving performance
• fft: Add cfloat radix 2 support on AIE-ML and AIE-MLv2
• fft: Add integral mixed radix support on AIE-MLv2
• fft: Add cbfloat16 radix 2, 3, 4, and 5 support on AIE-ML
• min/max: Support arbitrary vector sizes
• min/max: Fix 4b implementations for dynamic sign and reductions
• mul/mac/msc: Add cbfloat16 support where support is available
• mmul: Add 4x2x4, 4x4x4 modes for c32b*c16b types and 4x2x4 for c32b*c32b types on AIE-ML
• mmul: Add 2x8x8 mode for 8b*8b types
• mmul: Add 8x8x8 mode for bfloat16*bfloat16 on XDNA2
• mmul: Add 8x8x8 mode for bfloat16*bfp16ebs8 on XDNA2
• mmul: Add cbfloat16 support where support is available
• mmul: Add 4*16*16 mode for 8b*4b on AIE-MLv2
• mmul: Add 4*8*8, 8x4x8 modes for 16b*8b on AIE-MLv2
• mmul: Support mixed bfloat16 * float16 modes on AIE-MLv2
• neg: Support arbitrary vector sizes
• parallel_lookup: Fix optimization application for large unsigned keys
• print: Enable for block_vectors on AIE-MLv2
• sliding_mul: Optimize complex * real operations on AIE-ML
• sliding_mul: Add float16 support on AIE-MLv2
• to_fixed: Change to default behaviour for bfloat16: default rounding mode was floor and is now conv_even
• to_fixed_floor: New function to expose floor rounding mode for bfloat16 values

ADF integration

Vitis 2024.2

Documentation changes

Global AIE API changes

• Implement arbitrary vector and accum size support
• Add cbfloat16 support on AIE-ML

Changes to data types

• accum: Implement construction from block_vector
• accum: Implement grow_replicate
• mask: Enable get_submask to work with ElemsOut less than word size
• mask: Implement insert and extract

Changes to operations

• fft: Implement dynamic vectorization support for AIE
• fft: Implement radix 2 and radix 4 cbfloat16 support on AIE-ML
• mmul: Implement 8x8x8 bfloat16 mmul mode on AIE-ML
• shuffle: Fix optimized code paths for when the shuffle mode is known at compile time

ADF integration

• Enable cbfloat16 streams for AIE-ML
• Implement vector reads and writes on cascades

Vitis 2024.1

Documentation changes

• Further refactoring to remove detail namespace from documentation
• Document default accumulator type for given mul inputs
• Document float conversion implementations and behaviour

Global AIE API changes

• Add vector and accum template deduction guides
• Leverage aie::zeros in place of aie:broadcast(0) internally to prevent non-trivial conversions
• Several refactors to isolate external interfaces from detail namespace
• Add clamp function
• Add aie::saturating_add aie::saturating_sub functions
• Add aie::real and aie::imag helpers to get real and imaginary parts of vectors

Changes to data types

• accum: Fix sub-accum insertion
• vector: Optimize vector::get for 1K vectors
• vector: Extend pack/unpack to work for arbitrary conversions
• tensor_buffer_stream: Fix native implementations for non-native vector sizes

Changes to operations

• accumulate: Add array-based interface as alternative to variadic interface
• compare: Add workaround for IEEE incompliance to equality comparisons for bfloat16 on AIE-ML
• compare: Optimize comparisons with zero
• compare: Fix scalar/vector comparisons on AIE
• concat: Add support for concatenating tuples of vectors/accumulators
• elementary: Optimize vector unrolling for scalar functions
• fft: Fix vectorization limits for odd-radix implementations
• fft: Add radix2 combiner for cint16 input/output with cint32 twiddles
• fft: Add radix3 and radix5 stage0 implementations on AIE-ML
• fft: Add cfloat radix3 and radix5 stage0 implementations on AIE
• fft: Add 32b twiddle up/down dit FFTs
• mmul: Add dynamic sign support for sparse_vector
• mmul: Implemented additional modes
  • AIE-ML: 16b x 8b - 8x4x8 , 32b x 16b - 4x4x8, bf16 x bf16 - 4x8x8
• mmul: Add some outer product modes as aie::mmul<M,1,N>
• mmul: Enable zeroization for fp32 mmuls
• mul: Implement cint32 * int16 and int16 * cint32 muls
• neg: Add bfloat16 support for AIE-ML
• sliding_mul: Optimize complex x real implementations
• sliding_mul: Add bfloat16 support on AIE-ML
• to_fixed/to_float: Add support for unsigned types on AIE-ML

ADF integration

• Fix cfloat stream reads

Vitis 2023.2

Documentation changes

• Integrate AIE-ML documentation
• Document rounding modes
• Expand accumulate documentation
• Clarify limitations on 8b parallel lookup
• Fix mmul member documentation
• Clarify requirement for linear_approx step bits
• Improve documentation of vector, accum, and mask
• Highlight architecture requirements of functions using C++ requires clauses
• Document FFT twiddle factor generation
• Clarify internal rounding mode for bfloat16 to integer conversion
• Clarify native and emulated modes for mmul
• Clarify native and emulated modes for sliding_mul
• Document sparse_vector_input_buffer_stream with memory layout and GEMM example
• Document tensor_buffer_stream with a GEMM example

Global AIE API changes

• Add cfloat support for AIE-ML

Changes to data types

• vector: Optimize grow_replicate on AIE-ML
• mmul: Support reinitialization from an accum
• DM resources: Add compound aie_dm_resource variants
• streams: Add sparse_vector_input_buffer_stream for loading sparse data on AIE-ML
• streams: Add tensor_buffer_stream to handle multi-dimensional addressing for AIE-ML
• bfloat16: Add specialization for std::numeric_limits on AIE-ML

Changes to operations

• abs: Fix for float input
• add_reduce: Optimize for 8b and 16b types on AIE-ML
• div: Implement vector-vector and vector-scalar division
• downshift: Implement logical_downshift for AIE
• fft: Add support for 32 bit twiddles on AIE
• fft: Fix for radix-3 and radix-5 FFTs on AIE
• fft: Fix radix-5 performance for low vectorizations on AIE
• fft: Add stage-based FFT functions and deprecate iterator interface
• mul: Fix for vector * vector_elem_ref on AIE
• print_fixed: Support printing Q format data
• print_matrix: Added accumulator support
• sliding_mul: Add support float
• sliding_mul: Add support for remaining 32b modes for AIE-ML
• sliding_mul: Add support for Points < Native Points
• sliding_mul_ch: Fix DataStepX == DataStepY requirement
• sincos: Optimize AIE implementation
• to_fixed: Fix for AIE-ML
• to_fixed/to_float: Add vectorized float conversions for AIE
• to_fixed/to_float: Add generic conversions ((int8, int16, int32) <-> (bfloat16, float)) for AIE-ML

ADF integration

• Add TLAST support for stream reads on AIE-ML
• Add support for input_cascade and output_cascade types
• Deprecate accum reads from input_stream and output_stream

Vitis 2023.1

Documentation changes

• Add explanation of FFT inputs
• Use block_size in FFT docs
• Clarify matrix data layout expectations
• Clarify downshift being arithmetic
• Correct description of bfloat16 linear_approx lookup table

Global AIE API changes

• Do not explicitly initialize inferred template arguments
• More aggressive inlining of internal functions
• Avoid using 128b vectors in stream helper functions for AIE-ML

Changes to data types

• iterator: Do not declare iterator data members as const
• mask: Optimized implementation for 64b masks on AIE-ML
• mask: New constructors added to initialize the mask from uint32 or uint64 values
• vector: Fix 1024b inserts
• vector: Use 128b concats in upd_all
• vector: Fix 8b unsigned to_vector for AIE-ML

Changes to operations

• add/sub: Support for dynamic accumulator zeroization
• begin_restrict_vector: Add implementation for io_buffer
• eq: Add support for complex numbers
• fft: Correctly set radix configuration in fft::begin_stage calls
• inv/invsqrt: Add implementation for AIE-ML
• linear_approx: Performance optimization for AIE-ML
• logical_downshift: New function that implements a logical downshift (as opposed to aie::downshift, which is arithmetic)
• max/min/maxdiff: Add support for dynamic sign
• mmul: Implement 16b 8x2x8 mode for AIE-ML
• mmul: Implement 8b 8x8x8 mode for AIE-ML
• mmul: Implemet missing 16b x 8b and 8b x 4b sparse multiplication modes for AIE-ML
• neq: Add support for complex numbers
• parallel_lookup: Optimize implementation for signed truncation
• print_matrix: New function that prints vectors with the specified matrix shape
• shuffle_up/down: Minor optimization for 16b
• shuffle_up/down: Optimized implementation for AIE-ML
• sliding_mul: Support data_start/coeff_start values larger than vector size
• sliding_mul: Add support for 32b modes for AIE-ML
• sliding_mul: Add 2 point 16b 16 channel for AIE-ML
• sliding_mul_ch: New function for multi-channel multiplication modes for AIE-ML
• sliding_mul_sym_uct: Fix for 16b two-buffer implementation
• store_unaligned_v: Optimized implementation for AIE-ML
• transpose: Add support for 64b and 32b types
• transpose: Enable transposition of 256 element 4b vectors (scalar implementation for now)
• to_fixed: Add bfloat16 to int32 conversion on AIE-ML

Vitis 2022.2

Documentation changes

• Add code samples for load_v/store_v and load_unaligned_v/store_unaligned_v
• Enhanced documentation for parallel_lookup and linear_approx
• Clarify coeff vector size limit on AIE-ML

Global AIE API changes

• Remove usage of srs in compare functions, to avoid compilation warnings as it is deprecated
• Add support for stream ADF vector types on AIE-ML

Changes to data types

• mask: add shift operators
• saturation_mode: add saturate value. It was previously named truncate, which is not correct. The old name is also kept until it is deprecated

Changes to operations

• add: support accumulator addition on AIE-ML
• add_reduce: add optimized implementation for cfloat on AIE
• add_reduce: add optimized implementation for bfloat16 on AIE-ML
• eq/neq: enhanced implementation on AIE-ML
• le: enhanced implementation on AIE-ML
• load_unaligned_v: leverage pointer truncation to 128b done by HW on AIE
• fft: add support for radix 3/5 on AIE
• mmul: add matrix x vector multiplicatio modes on AIE
• mmul: add support for dynamic accumulator zeroization
• to_fixed: added implementation for AIE-ML
• to_fixed: provide a default return type
• to_float: added implementation for AIE-ML
• reverse: optimized implementation for 32b and 64b on AIE-ML
• zeros: include fixes on AIE

Vitis 2022.1

Documentation changes

• Small documentation fixes for operators
• Issues of documentation on msc_square and mmul
• Enhance documentation for sliding_mul operations
• Change logo in documentation
• Add documentation for ADF stream operators

Global AIE API changes

• Add support for emulated FP32 data types and operations on AIE-ML

Changes to data types

• unaligned_vector_iterator: add new type and helper functions
• random_circular_vector_iterator: add new type and helper functions
• iterator: add linear iterator type and helper functions for scalar values
• accum: add support for dynamic sign in to/from_vector on AIE-ML
• accum: add implicit conversion to float on AIE-ML
• vector: add support for dynamic sign in pack/unpack
• vector: optimization of initialization by value on AIE-ML
• vector: add constructor from 1024b native types on AIE-ML
• vector: fixes and optimizations for unaligned_load/store

Changes to operations

• adf::buffer_port: add many wrapper iterators
• adf::stream: annotate read/write functions with stream resource so they can be scheduled in parallel
• adf::stream: add stream operator overloading
• fft: performance fixes on AIE-ML
• max/min/maxdiff: add support for bfloat16 and float on AIE-ML
• mul/mmul: add support for bfloat16 and float on AIE-ML
• mul/mmul: add support for dynamic sign AIE-ML
• parallel_lookup: expanded to int16->bfloat, performance optimisations, and softmax kernel
• print: add support to print accumulators
• add/max/min_reduce: add support for float on AIE-ML
• reverse: add optimized implementation on AIE-ML using matrix multiplications
• shuffle_down_replicate: add new function
• sliding_mul: add 32b for 8b * 8b and 16b * 16b on AIE-ML
• transpose: add new function and implementation for AIE-ML
• upshift/downshift: add implementation for AIE-ML

Vitis 2021.2

Documentation changes

• Fix description of sliding_mul_sym_uct
• Make return types explicit for better documentation
• Fix documentation for sin/cos so that it says that the input must be in radians
• Add support for concepts
• Add documenttion for missing arguments and fix wrong argument names
• Fixes in documentation for int4/uint4 AIE-ML types
• Add documentation for the mmul class
• Update documentation about supported accumulator sizes
• Update the matrix multiplication example to use the new MxKxN scheme and size_A/size_B/size_C

Global AIE API changes

• Make all entry points always_inline
• Add declaration macros to aie_declaration.hpp so that they can be used in headers parsed by aiecompiler

Changes to data types

• Add support for bfloat16 data type on AIE-ML
• Add support for cint16/cint32 data types on AIE-ML
• Add an argument to vector::grow, to specify where the input vector will be located in the output vector
• Remove copy constructor so that the vector type becomes trivial
• Remove copy constructor so that the mask type becomes trivial
• Make all member functions in circular_index constexpr
• Add tiled_mdspan::begin_vector_dim functions that return vector iterators
• Initial support for sparse vectors on AIE-ML, including iterators to read from memory
• Make vector methods always_inline
• Make vector::push be applied to the object it is called on and return a reference

Changes to operations

• add: Implementation optimization on AIE-ML

• add_reduce: Implement on AIE-ML

• bit/or/xor: Implement scalar x vector variants of bit operations

• equal/not_equal: Add fix in which not all lanes were being compared for certain vector sizes.

• fft: Interface change to enhance portability across AIE/AIE-ML
• fft: Add initial support on AIE-ML
• fft: Add alignment checks for x86sim in FFT iterators

• fft: Make FFT output interface uniform for radix 2 cint16 upscale version on AIE

• filter_even/filter_odd: Functional fixes
• filter_even/filter_odd: Performance improvement for 4b/8b/16b implementations
• filter_even/filter_odd: Performance optimization on AIE-ML

• filter_even/filter_odd: Do not require step argument to be a compile-time constant

• interleave_zip/interleave_unzip: Improve performance when configuration is a run-time value

• interleave_*: Do not require step argument to be a compile-time constant

• load_floor_v/load_floor_bytes_v: New functions that floor the pointer to a requested boundary before performing the load.

• load_unaligned_v/store_unaligned_v: Performance optimization on AIE-ML

• lut/parallel_lookup/linear_approx: First implementation of look-up based linear functions on AIE-ML.

• max_reduce/min_reduce: Add 8b implementation

• max_reduce/min_reduce: Implement on AIE-ML

• mmul: Implement new shapes for AIE-ML
• mmul: Initial support for 4b multiplication
• mmul: Add support for 80b accumulation for 16b x 32b / 32b x 16b cases
• mmul: Change dimension names from MxNxK to MxKxN

• mmul: Add size_A/size_B/size_C data members

• mul: Optimized mul+conj operations to merged into a single intrinsic call on AIE-ML

• sin/cos/sincos: Fix to avoid int -> unsigned conversions that reduce the range
• sin/cos/sincos: Use a compile-time division to compute 1/PI
• sin/cos/sincos: Fix floating-point range

• sin/cos/sincos: Optimized implementation for float vector

• shuffle_up/shuffle_down: Elements don't wrap around anymore. Instead, new elements are undefined.
• shuffle_up_rotate/shuffle_down_rotate: New variants added for the cases in which elements need to wrap-around
• shuffle_up_replicate: Variant added which replicates the first element.
• shuffle_up_fill: Variant added which fills new elements with elements from another vector.

• shuffle_*: Optimization in shuffle primitives on AIE, especially for 8b/16b cases

• sliding_mul: Fixes to handle larger Step values for cfloat variants
• sliding_mul: Initial implementation for 16b x 16b and cint16b x cint16b on AIE-ML

• sliding_mul: Optimized mul+conj operations to merged into a single intrinsic call on AIE-ML

• sliding_mul_sym: Fixes in start computation for filters with DataStepX > 1
• sliding_mul_sym: Add missing int32 x int16 / int16 x int32 type combinations
• sliding_mul_sym: Fix two-buffer sliding_mul_sym acc80

• sliding_mul_sym: Add support for separate left/right start arguments

• store_v: Support pointers annotated with storage attributes