Accelerate FPGA Prototyping with MATLAB and Simulink

S t b 21st 2010 September

Stephan van Beek Senior Application Engineer

From Idea to Implementation

Algorithm Development

MATLAB Simulink Stateflow

Need for FPGA prototyping:

Real Time Simulation Real-Time Modeling accuracy Interfacing with peripherals

Key Takeaways

Automation of manual steps in FPGA prototyping allowing shorter iteration cycles Integration of FPGA development tools enhances verification Automatic A t ti HDL Code C d generation ti can b be adapted d t d to meet your requirement
3

From Idea to Implementation (manual)

Algorithm Development

MATLAB Simulink Stateflow

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

MathWorks Solutions
DESIGN

Algorithm Development

MATLAB Si Simulink li k Stateflow TES ST & VERIFIC CATION

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

Fixed-Point

IMPLEMENTATION
C, C++ MCU DSP VHDL, Verilog FPGA ASIC SPICE Analog Hardware

INTEGRATION

MathWorks Solutions
DESIGN

Algorithm Development

MATLAB Si Simulink li k Stateflow TES ST & VERIFIC CATION

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

Fixed-Point

IMPLEMENTATION
C, C++ MCU DSP VHDL, Verilog FPGA ASIC SPICE Analog Hardware

INTEGRATION

How to Handle Fixed-Point Conversion? MATLAB Pi = 3.141592653589793 Simulink Pi = 3.141601562500000 Stateflow

Signed, 14-bits, 11 bits fraction Resources? Overflow? Precision? Error=8.9089e-6

FPGA

Pi = 01100100100010

MathWorks solution: Fi d P i t Tools Fixed-Point T l

Simulink Fixed-Point

Convert floating point to fixed point models

Bit-true models in the same environment

MathWorks Solutions
DESIGN

Algorithm Development

MATLAB Si Simulink li k Stateflow TES ST & VERIFIC CATION

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

Fixed-Point

IMPLEMENTATION
C, C++ MCU DSP VHDL, Verilog FPGA ASIC SPICE Analog Hardware

INTEGRATION

-- -------------------------------------------------------------- File Name: hdlsrc\Stage1 -- Created: 2010-06-22 15:42:41 -- Generated by MATLAB 7.11 and Simulink HDL Coder 2.0 --- --------------------------------------------------------------- -------------------------------------------------------------- Module: Stage1 -- Source Path: m03_ddc_hdlgen/ddc_hdl/Lowpass Filter/Stage1 --- ------------------------------------------------------------LIBRARY IEEE; USE IEEE.std_logic_1164.all; USE IEEE.numeric_std.ALL; IEEE numeric std ALL; ENTITY Stage1 IS PORT( clk : IN std_logic; enb_1_1_1 : IN std_logic; reset : IN std_logic; Stage1_in_re : IN std_logic_vector(17 Stage1_in_im : IN std_logic_vector(17 Stage1_out_re : OUT std_logic_vector(17 Stage1_out_im : OUT std_logic_vector(17 ); END Stage1;

BEGIN -- Block Statements ce_output : PROCESS (clk, reset) BEGIN IF reset = '1' THEN ring g_count <= to_unsigned(1, g 3); ELSIF clk'event AND clk = '1' THEN IF enb_1_1_1 = '1' THEN ring_count <= ring_count(0) & ring_count(2 DOWNTO 1); END IF; END IF; END PROCESS ce_output; phase_0 <= ring_count(0) phase_1 <= ring_count(1)
DOWNTO DOWNTO DOWNTO DOWNTO 0); 0); 0); 0) ----sfix18_En16 sfix18_En16 sfix18_En16 sfix18_En16

AND enb_1_1_1; AND enb_1_1_1; AND enb_1_1_1;

phase_2 <= ring_count(2)

input_typeconvert_re _ _ <= signed(Stage1_in_re); _ _ input_typeconvert_im <= signed(Stage1_in_im); Delay_Pipeline_Phase0_process : PROCESS (clk, reset) BEGIN IF reset = '1' THEN input_pipeline_phase0_re <= (OTHERS => '0'); input_pipeline_phase0_im <= (OTHERS => '0'); ELSIF clk'event AND clk = '1' THEN IF phase_0 = '1' THEN input_pipeline_phase0_re <= input_typeconvert_re; input_pipeline_phase0_im <= input_typeconvert_im; END IF; END IF; END PROCESS Delay_Pipeline_Phase0_process; Delay_Pipeline_Phase1_process : PROCESS (clk, reset) BEGIN IF reset = '1' THEN input_pipeline_phase1_re(0 TO 1) <= (OTHERS => (OTHERS => '0')) input_pipeline_phase1_im(0 TO 1) <= (OTHERS => (OTHERS => '0')) ELSIF clk'event AND clk = '1' THEN IF phase_1 = '1' THEN input_pipeline_phase1_re(0) <= input_typeconvert_re; input_pipeline_phase1_re(1) <= input_pipeline_phase1_re(0); input_pipeline_phase1_im(0) <= input_typeconvert_im; input_pipeline_phase1_im(1) <= input_pipeline_phase1_im(0); END IF; END IF; END PROCESS Delay_Pipeline_Phase1_process;

-----------------------------------------------------------------Module Architecture: Stage1 ---------------------------------------------------------------ARCHITECTURE rtl OF Stage1 IS -- Local Functions -- Type T D Definitions fi iti TYPE input_pipeline_type IS ARRAY (NATURAL range <>) OF signed(17 DOWNTO 0); -- sfix18_En16 -- Constants CONSTANT coeffphase1_1 : signed(17 DOWNTO 0) := to_signed(20991, 18); -- sfix18_En18 CONSTANT coeffphase1_2 : signed(17 DOWNTO 0) := to_signed(65368, 18); -- sfix18_En18 CONSTANT coeffphase2_1 : signed(17 DOWNTO 0) := to_signed(65368, 18); -- sfix18_En18 CONSTANT coeffphase2_2 : signed(17 DOWNTO 0) := to_signed(20991, 18); -- sfix18_En18 CONSTANT coeffphase3_1 : signed(17 DOWNTO 0) := to_signed(89431, 18); -- sfix18_En18 CONSTANT coeffphase3_2 : signed(17 DOWNTO 0) := to_signed(0, 18); -- sfix18_En18 -- Signals SIGNAL ring_count SIGNAL phase_0 SIGNAL phase_1 SIGNAL phase_2 SIGNAL input_typeconvert_re SIGNAL input_typeconvert_im SIGNAL input_pipeline_phase0_re SIGNAL input_pipeline_phase0_im SIGNAL input_pipeline_phase1_re SIGNAL input_pipeline_phase1_im SIGNAL input_pipeline_phase2_re SIGNAL input_pipeline_phase2_im SIGNAL product_phase0_1_re SIGNAL product_phase0_1_im SIGNAL mul_temp SIGNAL mul_temp_1 SIGNAL product_phase0_2_re SIGNAL product product_phase0_2_im phase0 2 im SIGNAL mul_temp_2 SIGNAL mul_temp_3 SIGNAL product_phase1_1_re SIGNAL product_phase1_1_im SIGNAL mul_temp_4 SIGNAL mul_temp_5 SIGNAL product_phase1_2_re SIGNAL product_phase1_2_im SIGNAL mul temp 6

: : : : : : : : : : : : : : : : : : : : : : : : : : :

unsigned(2 DOWNTO 0); -- ufix3 std_logic; -- boolean std_logic; -- boolean std_logic; -- boolean signed(17 DOWNTO 0); -- sfix18_En16 signed(17 DOWNTO 0); -- sfix18_En16 signed(17 DOWNTO 0); -- sfix18_En16 signed(17 DOWNTO 0); -- sfix18_En16 input_pipeline_type(0 TO 1); -- sfix18_En16 input_pipeline_type(0 TO 1); -- sfix18_En16 signed(17 DOWNTO 0); -- sfix18_En16 signed(17 DOWNTO 0); -- sfix18_En16 signed(47 DOWNTO 0); -- sfix48_En48 signed(47 DOWNTO 0); -- sfix48_En48 signed(35 DOWNTO 0); -- sfix36_En34 signed(35 DOWNTO 0); -- sfix36_En34 signed(47 DOWNTO 0); -- sfix48_En48 signed(47 DOWNTO 0); -- sfix48_En48 sfix48 En48 signed(35 DOWNTO 0); -- sfix36_En34 signed(35 DOWNTO 0); -- sfix36_En34 signed(47 DOWNTO 0); -- sfix48_En48 signed(47 DOWNTO 0); -- sfix48_En48 signed(35 DOWNTO 0); -- sfix36_En34 signed(35 DOWNTO 0); -- sfix36_En34 signed(47 DOWNTO 0); -- sfix48_En48 signed(47 DOWNTO 0); -- sfix48_En48 signed(35 DOWNTO 0); -- sfix36 En34

Delay_Pipeline_Phase2_process : PROCESS (clk, reset) BEGIN IF reset = '1' THEN input_pipeline_phase2_re <= (OTHERS => '0');

10

Manual HDL Code Creation

Typical HDL designs contain many lines of code Days or maybe weeks to develop? How to implement Fixed Point in HDL? What if the specification changes?

MathWorks solution: A t Automatic ti HDL C Code d G Generation ti

Simulink HDL Coder

Simulink HDL Coder generates bit-true, cycle-accurate, synthesizable HDL code from Si li k models, Simulink d l Stateflow St t fl charts, h t and d MATLAB code d A Accelerates l generation i and d verification ifi i of f vendor d independent, readable RTL code

12

MathWorks Solutions
DESIGN

Algorithm Development

MATLAB Si Simulink li k Stateflow TES ST & VERIFIC CATION

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

Fixed-Point

IMPLEMENTATION
C, C++ MCU DSP VHDL, Verilog FPGA ASIC SPICE Analog Hardware

INTEGRATION

13

HDL Verification

Design the Test Bench twice

10 t 1 ratio 10-to-1 ti of fT Test t bench b h LOC to t Design D i LOC

Many stimuli-files from MATLAB These are ideal references which require pre- and post-processing How to analyze results?

MathWorks solution: R U S Re-Use System t L Level l Testbench T tb h

Co-Simulation with HDL Simulator

15

Co-Simulation with HDL Simulator

Re-use of MATLAB/Simulink testbench Extended analysis capabilities Dynamic Testbench (closed loop verification, multi domain) Automatic creation of co-simulation models Integrating handwritten HDL code
16

MathWorks Solutions
DESIGN

Algorithm Development

MATLAB Si Simulink li k Stateflow TES ST & VERIFIC CATION

Fixed Point Conversion HDL Code Creation HDL Verification FPGA Prototyping

Fixed-Point

IMPLEMENTATION
C, C++ MCU DSP VHDL, Verilog FPGA ASIC SPICE Analog Hardware

INTEGRATION

17

FPGA Prototyping

Building confidence T l flow Tool fl automation t ti Interfacing with peripherals

MathWorks solution: I t Integration ti with ith 3rd party t solutions l ti

FPGA-in-the-Loop Simulation

FPGA development board

FPGA Workflow Advisor

Automated workflow from model to FPGA prototyping

20

HDL Coder Only for Prototyping?

Folding to reduce area

Automatically A t ti ll add dd Pipelining to improve timing g

Summary

Automation of manual steps in FPGA prototyping allowing g shorter iteration cycles y

Assisted Fixed Point Conversion Automatic HDL Code Generation FPGA Turnkey Flow

Integration of FPGA development tools enhances verification

Improved analysis, closed loop verification, multi domain

Automatic HDL Code generation can be adapted to meet your requirements

Optimization for area and speed

Next steps ..

Visit www.mathworks.nl/fpga for more information

Request q a free guided g Simulink HDL Coder trial

Watch our Simulink HDL Coder webinars:

Questions??

23