## Ultrasound Library - Level 1 (L1)
The level 1 of Vitis Ultrasound Library contains the aie kernels. For more details information, please reference to L1 User Guide in the document for usage and design information.
## Overview of APIs
The L1 APIs is a set of APIs which are created to create a closest mapping possible between *NumPy* and the *C++* SIMD APIs of the AI Engine. The APIs resulted are a more comprehensive and easy interface for the user to create its own application using numpy-like interfaces.
The APIs can be divided in two main groups:
1. Element-wise operations between vectors and matrices. Those operations varies between basic operations (i.e. sum, multiplication etc...) to more complex ones (i.e. reciprocal, sign etc...);
2. Vector managing and creation. Those operations are intended to be used by the user to create or modify the dimension of the vectorial operands. Two example of those operations are *Tile* and *Ones*.
Follow the details of L1 kernel available in the library:
### Kernel name: absV
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM>
void absV(adf::input_buffer<T>& __restrict in, adf::output_buffer<T>& __restrict out);
```
Element-Wise absolute values of a vector.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: elements of the vector to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: cosV
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM>
void cosV(adf::input_buffer<T>& __restrict in1, adf::output_buffer<T>& __restrict out);
```
Element-Wise cosine values of a vector. Elements must be expressed in radians with the range [0...2k$\pi$]
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: elements of the vector to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: diffMV
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM>
void diffMV(adf::input_buffer<T>& __restrict in1, adf::input_buffer<T>& __restrict in2, adf::output_buffer<T>& __restrict out);
```
Element-Wise difference between the of the row of a matrix and the values of a vector. The number of the column of the matrix and the entry of the vector must have the same size.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: elements of the first matrix to be passed to the kernel.
- `in2`: elements of the array to be passed to the kernel.
- `out`: elements of the result of the operation (matrix) to be passed from the kernel.
### Kernel name: diffSV
```c++
template<typename T, const unsigned int LEN, const unsigned int INCREMENT, const unsigned VECDIM>
void diffSV(adf::input_buffer<T>& __restrict in1, adf::input_buffer<T>& __restrict in2, adf::output_buffer<T>& __restrict out);
```
Element-Wise difference between a scalar and the values of a vector. For every iteration (expressed by `LEN`) we need to pass 4 times the scalar value to the stream of the scalar value.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: scalar element to be passed to the kernel.
- `in2`: elements of the array to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: diffVS
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM>
void diffVS(adf::input_buffer<T>& __restrict in1, adf::input_buffer<T>& __restrict in2, adf::output_buffer<T>& __restrict out);
```
Element-Wise difference between the values of a vector and a scalar. For every iteration (expressed by `LEN`) we need to pass 4 times the scalar value to the stream of the scalar value.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: elements of the array to be passed to the kernel.
- `in2`: scalar element to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: divVS
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM>
void divVSSpeedOfSound(adf::input_buffer<T>& in1, adf::output_buffer<T>& out);
```
Element-Wise division between the values of a vector and a scalar (SpeedOfSound). For every iteration (expressed by `LEN`) we need to pass 4 times the scalar value to the stream of the scalar value.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- **Function params**:
- `in1`: elements of the array to be passed to the kernel.
- `in2`: scalar element to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: equalS
```c++
template<typename T, const unsigned LEN, const unsigned INCREMENT, const unsigned VECDIM, const unsigned SCALAR>
void equalS(adf::input_buffer<T>& __restrict in1, adf::output_buffer<T>& __restrict out);
```
Check whether the element of an array are equal to a specific number. An array of 0s or 1s is returned. 1 means that the element at that specific position is equal to the scalar, otherwise 0 is returned.
- **Template params**:
- `T`: type of the operation;
- `LEN`: number of elements to be processed in the kernel per iteration;
- `INCREMENT`: parameter which indicates how much iterations have been performed by the SIMD with respect to the intended total length;
- `VECDIM`: dimension of the SIMD to be performed. Addressed in the Xilinx UG1076, it depends on the type chosen;
- `SCALAR`: scalar to which the elements of the array are compared to;
- **Function params**:
- `in1`: elements of the vector to be passed to the kernel.
- `out`: elements of the result of the operation (vector) to be passed from the kernel.
### Kernel name: lessOrEqualThanS
```c++
templat
xilinx vitis资源库 （2000个子文件） 
api_fcn_multiInstr.cpp 12KB host.cpp 11KB host.cpp 11KB host.cpp 11KB host.cpp 11KB host.cpp 11KB host.cpp 10KB host.cpp 10KB host.cpp 10KB host.cpp 10KB graph.cpp 9KB graph.cpp 9KB mulVS.cpp 9KB test_cgSolver_spmv_jacobi.cpp 8KB graph.cpp 8KB backward.cpp 7KB diffSV.cpp 6KB main.cpp 6KB main.cpp 6KB main.cpp 6KB krnl_gemv.cpp 5KB main.cpp 5KB fcnKernel.cpp 5KB diffVS.cpp 5KB test_cgSolver_gemv.cpp 5KB main.cpp 5KB host.cpp 5KB host.cpp 5KB host.cpp 5KB main.cpp 4KB main.cpp 4KB rtmbackward.cpp 4KB krnl_loadAval.cpp 4KB main.cpp 4KB rtmforward_hbc.cpp 4KB rtmforward.cpp 4KB main.cpp 4KB forward.cpp 4KB main.cpp 3KB krnl_storeApk.cpp 3KB rtmforward_rbc.cpp 3KB sumVS.cpp 3KB forward.cpp 3KB graph.cpp 3KB forward.cpp 3KB krnl_loadAidx.cpp 3KB main.cpp 3KB test_cgSolver_gemv_jacobi.cpp 3KB absV.cpp 3KB sumVV.cpp 2KB mulVV.cpp 2KB main.cpp 2KB main.cpp 2KB propagate.cpp 2KB divVS.cpp 2KB graph.cpp 2KB graph.cpp 2KB graph.cpp 2KB graph.cpp 2KB graph.cpp 2KB graph.cpp 2KB main.cpp 2KB graph.cpp 2KB main.cpp 2KB diffMV.cpp 2KB outer.cpp 2KB graph.cpp 2KB laplacian.cpp 2KB mulMM.cpp 2KB cosV.cpp 2KB sumMM.cpp 2KB krnl_update_xr.cpp 2KB laplacian.cpp 2KB krnl_update_rk_jacobi.cpp 2KB krnl_loadPkApar.cpp 2KB sign.cpp 2KB krnl_loadArbParam.cpp 1KB krnl_dupTk4.cpp 1KB lessOrEqualThanS.cpp 1KB equalS.cpp 1KB norm_axis_1.cpp 1KB krnl_update_rk.cpp 1KB krnl_update_pk.cpp 1KB squareV.cpp 1KB sqrtV.cpp 1KB reciprocalV.cpp 1KB top.cpp 1KB krnl_control.cpp 1KB sum_axis_1.cpp 1KB top.cpp 1KB dataMover.cpp 1KB krnl_update_xk.cpp 1KB tileV.cpp 1KB krnl_duplicate.cpp 1KB top.cpp 1KB ones.cpp 1KB streamTimer.cpp 909B krnl_timer.cpp 906B s2mm1.cpp 801B s2mm1.cpp 801B共 2000 条
- 1
- 2
- 3
- 4
- 5
- 6
- 20
资源评论
胡须渣
- 粉丝: 2
最新资源
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
