Vivado HLS introductory notes

 

video （B There are also ）：

Follow Xilinx SAE learn HLS

 

• Vivado HLS Explanation
  1. C/C++: HLS“ utilize C Write hardware directly ”, optimization algorithm + Guide the integrated tools →HDL(IP).
  2. The process
     1. Scheduling: What does the clock cycle do , Whether there are parallel operations ;
     2. Control logic extraction: Generating state machine （ State machine control algorithm process ）;
     3. Binding： Operation and resource mapping ;
  3. 
  4. It can't be integrated ： Dynamic memory allocation 、 Design operating system level ;
• Vivado HLS Practice demonstration
  1. The first thing to write is ：
     1. Source: The function of practice adder.c + adder.h  void add()
     2. Test bench: int main(){ test , call add()}
  2. Vivado HLS Create a project , add to Top Function(adder.c -> add()), add to test bench, Specify chip model .
  3. C simulation ( utilize Test bench: main() test )
  4. C synthesis( Get a comprehensive report )
  5. C/RTL Co-simulation Observe RTL Simulation results (Vivado simulator, Dump Trace=Port),  Open Wave Viewer
  6. Export RTL: Generate IP Core
• Vivado HLS Advanced practice
  1. New Solution (e.g. Pipeline…)
     1. stay solution – constrains – directives.tcl Deposit directives
     2. Use... Before cycling label Mark （directives Easy to find ）
        1. Myloop – insert directive (pipeline) (% It means that directives.tcl Within the constraints directive;# Represents a constraint in the source file (#pragma HLS))
     3. After synthesis ：compare reports – Add two... To compare solutions
  2. data type
     1. I won't support it char
     2. Any precision data type ： Declare in the header file #define W 18; typedef ap_int<W> data_t
     3. Return data type sizeof()( Of 1, 2, 4…bytes)
     4. Give Way Visual Studio Support any precision data type ：
     5. Numeric initialization ：
     6. Data type conversion assignment
        1. Assign a small range number to a large range number ： Signed number （ Extended symbol ）, An unsigned number （ Fill up 0）
        2. Assign a large range number to a small range number ： There may be data loss
        3. Addition of two numbers ： A wide +1; Multiply two numbers ： Sum of bit widths .