vhdl - VHDL Vivado 组合循环警报
问题描述
我正在尝试实现一个简单的 ALU：
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.NUMERIC_STD.ALL;
entity simple_alu is
Port (
clk : in std_logic;
rst : in std_logic;
op : in std_logic_vector (1 downto 0);
in0 : in std_logic_vector (31 downto 0);
in1 : in std_logic_vector (31 downto 0);
res : out std_logic_vector (31 downto 0);
done_flag : out std_logic
);
end simple_alu;
architecture Behavioral of simple_alu is
type state_type is (start, add, sub, mul, pow, done);
signal state, next_state : state_type := start;
signal status : std_logic := '0';
signal inter, shift_reg, zero : std_logic_vector (31 downto 0) := (others => '0');
begin
SYNC_PROC: process (clk, rst)
begin
if (rst = '0') then
if rising_edge(clk) then
state <= next_state;
res <= inter;
done_flag <= status;
end if;
else
state <= start;
res <= (others => '0');
done_flag <= '0';
end if;
end process;
--MEALY State-Machine - Outputs based on state and inputs
OUTPUT_DECODE: process (state, in0, in1, shift_reg)
variable result, temp : std_logic_vector (31 downto 0);
variable flag : std_logic := '0';
begin
shift_reg <= in1;
temp := temp;
flag := flag;
result := result;
case state is
when start =>
result := std_logic_vector(to_signed(1, 32));
temp := in0;
flag := '0';
when add => result := std_logic_vector(signed(in0) + signed(in1));
when sub => result := std_logic_vector(signed(in0) - signed(in1));
when mul => result := std_logic_vector(resize(signed(in0) * signed(in1), 32));
when pow =>
if (shift_reg(shift_reg'low) = '1') then
result := std_logic_vector(resize(signed(result) * signed(temp), 32));
else
result := result;
end if;
temp := std_logic_vector(resize(signed(temp) * signed(temp), 32));
shift_reg <= std_logic_vector(shift_right(signed(shift_reg), 1));
when done =>
result := result;
flag := '1';
when others =>
end case;
inter <= result;
status <= flag;
end process;
NEXT_STATE_DECODE: process (state, op, shift_reg, zero, rst) -- rst indicates that one input (op, in0 or in1) changed
begin
--declare default state for next_state to avoid latches
next_state <= state; --default is to stay in current state
case (state) is
when start =>
case (op) is
when "00" => next_state <= add;
when "01" => next_state <= sub;
when "10" => next_state <= mul;
when "11" => next_state <= pow;
when others => next_state <= done;
end case;
when add => next_state <= done;
when sub => next_state <= done;
when mul => next_state <= done;
when pow =>
if (shift_reg = zero) then
next_state <= done;
else
next_state <= pow;
end if;
when done =>
if (rst = '1') then
next_state <= start;
end if;
when others =>
end case;
end process;
end Behavioral;
这似乎有效，至少在这个测试平台中：
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.NUMERIC_STD.ALL;
entity simple_alu_tb is
end simple_alu_tb;
architecture Behavioral of simple_alu_tb is
component simple_alu is
Port (
clk : in std_logic;
rst : in std_logic;
op : in std_logic_vector (1 downto 0);
in0 : in std_logic_vector (31 downto 0);
in1 : in std_logic_vector (31 downto 0);
res : out std_logic_vector (31 downto 0);
done_flag : out std_logic
);
end component;
signal clk : std_logic := '0';
signal rst : std_logic := '0';
signal op : std_logic_vector (1 downto 0) := (others => '0');
signal in0 : std_logic_vector (31 downto 0) := (others => '0');
signal in1 : std_logic_vector (31 downto 0) := (others => '0');
signal res : std_logic_vector (31 downto 0) := (others => '0');
signal done_flag : std_logic := '0';
constant clk_period : time := 1 ns;
begin
--Instantiate the Unit Under Test (UUT)
uut: simple_alu Port Map (
clk => clk,
rst => rst,
op => op,
in0 => in0,
in1 => in1,
res => res,
done_flag => done_flag
);
-- Clock process definitions
clk_process :process
begin
clk <= '1';
wait for clk_period/2;
clk <= '0';
wait for clk_period/2;
end process;
stim_proc: process
begin
wait for 2*clk_period;
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 4*clk_period;
op <= "00"; -- add
in0 <= std_logic_vector(to_signed(12, 32));
in1 <= std_logic_vector(to_signed(3, 32));
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 2*clk_period;
assert (res = std_logic_vector(to_signed(15, 32))) report "addition failed" severity failure;
wait for 4*clk_period;
op <= "01"; -- sub
in0 <= std_logic_vector(to_signed(12, 32));
in1 <= std_logic_vector(to_signed(3, 32));
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 2*clk_period;
assert (res = std_logic_vector(to_signed(9, 32))) report "subtraction failed" severity failure;
wait for 4*clk_period;
op <= "10"; -- mul
in0 <= std_logic_vector(to_signed(12, 32));
in1 <= std_logic_vector(to_signed(3, 32));
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 2*clk_period;
assert (res = std_logic_vector(to_signed(36, 32))) report "multiplication failed" severity failure;
wait for 4*clk_period;
op <= "11"; -- pow
in0 <= std_logic_vector(to_signed(12, 32));
in1 <= std_logic_vector(to_signed(7, 32));
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 4*clk_period;
assert (res = std_logic_vector(to_signed(35831808, 32))) report "power failed" severity failure;
wait for 4*clk_period;
op <= "11"; -- pow
in0 <= std_logic_vector(to_signed(12, 32));
in1 <= std_logic_vector(to_signed(6, 32));
rst <= '1';
wait for clk_period;
rst <= '0';
wait for 4*clk_period;
assert (res = std_logic_vector(to_signed(2985984, 32))) report "power failed" severity failure;
wait;
end process;
end Behavioral;
我想将其实现为 AXI4-lite 组件。所以我生成包装器，调整写入过程并实例化我的模块，如下所示：
...
process (S_AXI_ACLK)
variable loc_addr :std_logic_vector(OPT_MEM_ADDR_BITS downto 0);
begin
if rising_edge(S_AXI_ACLK) then
if S_AXI_ARESETN = '0' then
-- command_reg <= (others => '0');
-- done_flag <= '0';
slv_reg1 <= (others => '0');
slv_reg2 <= (others => '0');
-- slv_reg3 <= (others => '0');
else
loc_addr := axi_awaddr(ADDR_LSB + OPT_MEM_ADDR_BITS downto ADDR_LSB);
if (slv_reg_wren = '1') then
case loc_addr is
when b"00" =>
for byte_index in 0 to (C_S_AXI_DATA_WIDTH/8-4) loop -- write to command register only
if ( S_AXI_WSTRB(byte_index) = '1' ) then
-- Respective byte enables are asserted as per write strobes
-- slave registor 0
command_reg(byte_index*8+7 downto byte_index*8) <= S_AXI_WDATA(byte_index*8+7 downto byte_index*8);
end if;
end loop;
when b"01" =>
for byte_index in 0 to (C_S_AXI_DATA_WIDTH/8-1) loop
if ( S_AXI_WSTRB(byte_index) = '1' ) then
-- Respective byte enables are asserted as per write strobes
-- slave registor 1
slv_reg1(byte_index*8+7 downto byte_index*8) <= S_AXI_WDATA(byte_index*8+7 downto byte_index*8);
end if;
end loop;
when b"10" =>
for byte_index in 0 to (C_S_AXI_DATA_WIDTH/8-1) loop
if ( S_AXI_WSTRB(byte_index) = '1' ) then
-- Respective byte enables are asserted as per write strobes
-- slave registor 2
slv_reg2(byte_index*8+7 downto byte_index*8) <= S_AXI_WDATA(byte_index*8+7 downto byte_index*8);
end if;
end loop;
-- when b"11" => -- do not write to reg3
-- for byte_index in 0 to (C_S_AXI_DATA_WIDTH/8-1) loop
-- if ( S_AXI_WSTRB(byte_index) = '1' ) then
-- -- Respective byte enables are asserted as per write strobes
-- -- slave registor 3
-- slv_reg3(byte_index*8+7 downto byte_index*8) <= S_AXI_WDATA(byte_index*8+7 downto byte_index*8);
-- end if;
-- end loop;
when others =>
command_reg <= command_reg;
done_flag <= done_flag;
slv_reg1 <= slv_reg1;
slv_reg2 <= slv_reg2;
-- slv_reg3 <= slv_reg3;
end case;
end if;
end if;
end if;
end process;
...
-- Add user logic here
-- byte0 byte1 byte2 byte3
slv_reg0 <= done_flag & "0000000" & "00000000" & "00000000" & command_reg;
alu : simple_alu
port map (
clk => S_AXI_ACLK,
rst => slv_reg_wren, -- reset on every write to a register, high active
op => command_reg(1 downto 0),
in0 => slv_reg1,
in1 => slv_reg2,
res => slv_reg3,
done_flag => done_flag
);
-- User logic ends
但是，当我尝试为我的封装设计（包括 Zync UltraScale+ MPSoC、AXI 互连、处理器系统重置和我的 AXI 外设）生成比特流时，我收到以下错误：
ERROR: [DRC LUTLP-1] Combinatorial Loop Alert: 1 LUT cells form a combinatorial loop. This can create a race condition. Timing analysis may not be accurate. The preferred resolution is to modify the design to remove combinatorial logic loops. If the loop is known and understood, this DRC can be bypassed by acknowledging the condition and setting the following XDC constraint on any one of the nets in the loop: 'set_property ALLOW_COMBINATORIAL_LOOPS TRUE [get_nets <myHier/myNet>]'. One net in the loop is design_1_i/simple_alu_0/U0/simple_alu_v1_0_S00_AXI_inst/alu/state[0]_i_2_n_0. Please evaluate your design. The cells in the loop are: design_1_i/simple_alu_0/U0/simple_alu_v1_0_S00_AXI_inst/alu/state[0]_i_2.
ERROR: [DRC LUTLP-1] Combinatorial Loop Alert: 1 LUT cells form a combinatorial loop. This can create a race condition. Timing analysis may not be accurate. The preferred resolution is to modify the design to remove combinatorial logic loops. If the loop is known and understood, this DRC can be bypassed by acknowledging the condition and setting the following XDC constraint on any one of the nets in the loop: 'set_property ALLOW_COMBINATORIAL_LOOPS TRUE [get_nets <myHier/myNet>]'. One net in the loop is design_1_i/simple_alu_0/U0/simple_alu_v1_0_S00_AXI_inst/alu/state[1]_i_3_n_0. Please evaluate your design. The cells in the loop are: design_1_i/simple_alu_0/U0/simple_alu_v1_0_S00_AXI_inst/alu/state[1]_i_3.
请原谅大量的代码，我找不到用更小的例子来显示错误的方法。
我尝试了这里提出的解决方案：
set_property SEVERITY {Warning} [get_drc_checks LUTLP-1]
但这无济于事。我还尝试设置set_property ALLOW_COMBINATORIAL_LOOPS TRUE两个网络，但这让我不确定电路的功能。我正在使用 Vivado v2018.3，我的目标是 Avnet 的 Ultra96。有什么线索吗？
编辑：我已经更新了代码以反映当前的实现，我收到有关result_reg,flag_reg和temp_reg. 我该如何解决这些问题？
解决方案
经过长期的奋斗，我终于想出了这个解决方案：
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.NUMERIC_STD.ALL;
entity simple_alu is
Port (
clk : in std_logic;
rst : in std_logic;
op : in std_logic_vector (1 downto 0);
in0 : in std_logic_vector (31 downto 0);
in1 : in std_logic_vector (31 downto 0);
res : out std_logic_vector (31 downto 0);
done_flag : out std_logic
);
end simple_alu;
architecture Behavioral of simple_alu is
type state_type is (start, add, sub, mul, pow, done);
signal state, next_state : state_type := start;
signal result, next_result, temp, next_temp, shift_reg, next_shift_reg, zero : std_logic_vector (31 downto 0) := (others => '0');
signal next_done_flag : std_logic := '0';
begin
SYNC_PROC: process (clk, rst)
begin
if rising_edge(clk) then
if (rst = '1') then
state <= start;
else
state <= next_state;
res <= next_result;
result <= next_result;
temp <= next_temp;
shift_reg <= next_shift_reg;
done_flag <= next_done_flag;
end if;
end if;
end process;
--MEALY State-Machine - Outputs based on state and inputs
OUTPUT_DECODE: process (state, result, in0, in1, temp, shift_reg)
begin
next_done_flag <= '0';
next_result <= result;
next_shift_reg <= shift_reg;
next_temp <= temp;
case state is
when start =>
next_result <= std_logic_vector(to_signed(1, 32));
next_temp <= in0;
next_shift_reg <= in1;
when add => next_result <= std_logic_vector(signed(in0) + signed(in1));
when sub => next_result <= std_logic_vector(signed(in0) - signed(in1));
when mul => next_result <= std_logic_vector(resize(signed(in0) * signed(in1), 32));
when pow =>
if (shift_reg(shift_reg'low) = '1') then
next_result <= std_logic_vector(resize(signed(result) * signed(temp), 32));
else
next_result <= result;
end if;
next_temp <= std_logic_vector(resize(signed(temp) * signed(temp), 32));
next_shift_reg <= std_logic_vector(shift_right(signed(shift_reg), 1));
when done => next_done_flag <= '1';
when others =>
end case;
end process;
NEXT_STATE_DECODE: process (state, op, shift_reg, zero)
begin
--declare default state for next_state to avoid latches
next_state <= state; --default is to stay in current state
case (state) is
when start =>
case (op) is
when "00" => next_state <= add;
when "01" => next_state <= sub;
when "10" => next_state <= mul;
when "11" => next_state <= pow;
when others => next_state <= done;
end case;
when add => next_state <= done;
when sub => next_state <= done;
when mul => next_state <= done;
when pow =>
if (shift_reg = zero) then
next_state <= done;
else
next_state <= pow;
end if;
when done =>
when others =>
end case;
end process;
end Behavioral;
问题是我不明白硬件描述是如何工作的，现在我知道了一点（至少我希望如此..）。尤其是时钟和非时钟进程是如何连接的（将中间结果保存在寄存器中）。我会留下这个问题，以防另一个初学者偶然发现同样的问题。如果您认为我应该删除它，请在评论中说明，我会这样做。
以下是一些对我有帮助的资源：
这个问题，特别是接受的答案
我在某处学到的一些规则：
- 不要从您写入的信号中读取。
- 有一个正确的敏感度列表（您阅读的所有信号都应该在敏感度列表中）
- 确保在每条路径中分配了写入的所有信号。（例如：在 if-else-statement 的每个分支中）
- 对于使用变量的进程，请确保在读取每个变量之前将其初始化为默认值（在另一个变量或信号中）。
推荐阅读
- isabelle - 如何将假设传递给语言环境的解释
- php - 如何使用php将一个数据库备份到另一个数据库
- python - python：txt文件中的索引超出范围
- java - MySQL插入意外的行
- c++ - 具有菱形继承结构的子类的大小似乎很奇怪
- curl - 通过 CURLing Gitlab API 将 webhook 添加到项目中
- html - 在 amp-lightbox 中使用 amp-video
- reactjs - 将引导站点转换为 React 的问题
- java - 如何在 Vaadin TextField 中呈现 html 内容
- javascript - 如何让 webpack treeshake 我的 ES6 模块？