The following features have been added to the library in this release.
- TDM FIR - New library element.
Function | Namespace and class name |
---|---|
TDM FIR | xf::dsp::aie::fir::tdm::fir_tdm_graph |
This element adds an implementation of the Time-Division Multiplexing (TDM) variant of finite impulse response (FIR) filter.
Supports AIE and AIE-ML devices.
- Convolution / Correlation - New library element.
Function | Namespace and class name |
---|---|
Convolution / Correlation | xf::dsp::aie::conv_corr::conv_corr_graph |
This element adds an implementation of the Convolution and Correlation, depending on the specified FUNCT_TYPE template parameter.
Supports AIE and AIE-ML devices.
- Hadamard Product - New library element.
Function | Namespace and Class Name |
---|---|
Hadamard Product | xf::dsp::aie::hadamard::hadamard_graph |
This element adds an implementation of the Hadamard Product.
Supports AIE and AIE-ML devices.
- Outer Tensor Product - New library element.
Function | Namespace and Class Name |
---|---|
Outer Tensor Product | xf::dsp::aie::outer_tensor::outer_tensor_graph |
This element adds an implementation of the Outer Tensor Product.
Supports AIE devices.
- Kronecker Matrix Product - New library element.
Function | Namespace and Class Name |
---|---|
Kronecker Matrix Product | xf::dsp::aie::kronecker::kronecker_graph |
This element adds an implementation of the Kronecker Matrix Product.
Supports AIE devices.
- Example design of FFT on AIE and PL
This element illustrates the implementation of the Fast Fourier transform that spans across the AIE and Programmable Logic.
Supports AIE and AIE-ML devices.
- Matrix-Vector Multiplication
Added support for AIE-ML devices.
Added Super Sample Rate (SSR) feature. Allows the matrix-vector multiplication to be computed using parallel paths for increased throughput.
- Matrix Multiplication
Added Super Sample Rate (SSR) feature. Allows the matrix multiplication to be computed using parallel paths for increased throughput.
- DFT
Added Super Sample Rate (SSR) feature. Allows the Discrete Fourier Transform (DFT) to be computed using parallel paths for increased throughput.
- FFT IFFT
This element now supports cint32 type twiddles for integer data types. This element also now supports a new parameter TP_TWIDDLE_MODE.
- Mixed-Radix FFT
This element now supports AIE-ML devices.