aboutsummaryrefslogtreecommitdiffstats
path: root/testsuite/gna
diff options
context:
space:
mode:
Diffstat (limited to 'testsuite/gna')
-rw-r--r--testsuite/gna/issue2065/dual_port_ram.vhdl107
-rw-r--r--testsuite/gna/issue2065/fft.vhdl606
-rw-r--r--testsuite/gna/issue2065/repro.vhdl15
-rwxr-xr-xtestsuite/gna/issue2065/testsuite.sh14
4 files changed, 742 insertions, 0 deletions
diff --git a/testsuite/gna/issue2065/dual_port_ram.vhdl b/testsuite/gna/issue2065/dual_port_ram.vhdl
new file mode 100644
index 000000000..7039d5c8f
--- /dev/null
+++ b/testsuite/gna/issue2065/dual_port_ram.vhdl
@@ -0,0 +1,107 @@
+-- dual_port_ram.vhd
+-- This file is part of bladeRF-wiphy.
+--
+-- Copyright (C) 2021 Nuand, LLC.
+--
+-- This program is free software; you can redistribute it and/or modify
+-- it under the terms of the GNU General Public License as published by
+-- the Free Software Foundation; either version 2 of the License, or
+-- (at your option) any later version.
+--
+-- This program is distributed in the hope that it will be useful,
+-- but WITHOUT ANY WARRANTY; without even the implied warranty of
+-- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+-- GNU General Public License for more details.
+--
+-- You should have received a copy of the GNU General Public License along
+-- with this program; if not, write to the Free Software Foundation, Inc.,
+-- 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+library ieee;
+ use ieee.std_logic_1164.all;
+ use ieee.numeric_std.all;
+
+entity dual_port_ram is
+ generic(
+ ADDR_BITS : in natural := 6;
+ DATA_BITS : in natural := 32
+ );
+ port(
+ clock : in std_logic;
+ reset : in std_logic;
+
+ acc : in std_logic;
+ solo : in std_logic;
+ write : in std_logic;
+
+ addr_a : in std_logic_vector(ADDR_BITS-1 downto 0);
+ in_a : in std_logic_vector(DATA_BITS-1 downto 0);
+ data_a : out std_logic_vector(DATA_BITS-1 downto 0);
+
+ addr_b : in std_logic_vector(ADDR_BITS-1 downto 0);
+ in_b : in std_logic_vector(DATA_BITS-1 downto 0);
+ data_b : out std_logic_vector(DATA_BITS-1 downto 0)
+ );
+end entity;
+
+architecture arch of dual_port_ram is
+ type ram_t is array(natural range <>) of std_logic_vector(DATA_BITS-1 downto 0);
+
+ signal ram : ram_t((2**ADDR_BITS-1) downto 0);
+begin
+ sync : process(clock, reset)
+ variable add_a, add_b : integer;
+ begin
+ if (reset = '1') then
+ for i in ram'range loop
+ ram(i) <= ( others => '0' );
+ end loop;
+ elsif (rising_edge(clock)) then
+ if (acc = '1') then
+ add_a := to_integer(unsigned(addr_a));
+ add_b := to_integer(unsigned(addr_b));
+
+ if (write = '1') then
+ ram(add_a) <= in_a;
+ data_a <= in_a;
+ if (solo = '0') then
+ ram(add_b) <= in_b;
+ data_b <= in_b;
+ else
+ data_b <= ( others => '0' );
+ end if;
+ else
+ data_a <= ram(add_a);
+ if (solo = '0') then
+ data_b <= ram(add_b);
+ else
+ data_b <= ( others => '0' );
+ end if;
+ end if;
+ end if;
+
+ end if;
+ end process;
+end architecture;
+
+architecture synth of dual_port_ram is
+ type ram_t is array(natural range <>) of std_logic_vector(DATA_BITS-1 downto 0);
+ signal ram : ram_t((2**ADDR_BITS-1) downto 0);
+begin
+ sync : process(clock, reset)
+ variable addra : integer;
+ variable addrb : integer;
+ begin
+ if (rising_edge(clock)) then
+ addra := to_integer(unsigned(addr_a));
+ addrb := to_integer(unsigned(addr_b));
+ if (write = '1') then
+ ram(addra) <= in_a;
+ ram(addrb) <= in_a;
+ end if;
+ data_a <= ram(addra);
+ data_b <= ram(addrb);
+ end if;
+ end process;
+
+end architecture synth;
diff --git a/testsuite/gna/issue2065/fft.vhdl b/testsuite/gna/issue2065/fft.vhdl
new file mode 100644
index 000000000..857e42203
--- /dev/null
+++ b/testsuite/gna/issue2065/fft.vhdl
@@ -0,0 +1,606 @@
+-- fft.vhd
+-- This file is part of bladeRF-wiphy.
+--
+-- Copyright (C) 2021 Nuand, LLC.
+--
+-- This program is free software; you can redistribute it and/or modify
+-- it under the terms of the GNU General Public License as published by
+-- the Free Software Foundation; either version 2 of the License, or
+-- (at your option) any later version.
+--
+-- This program is distributed in the hope that it will be useful,
+-- but WITHOUT ANY WARRANTY; without even the implied warranty of
+-- MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+-- GNU General Public License for more details.
+--
+-- You should have received a copy of the GNU General Public License along
+-- with this program; if not, write to the Free Software Foundation, Inc.,
+-- 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+library ieee;
+ use ieee.std_logic_1164.all;
+ use ieee.numeric_std.all;
+ use ieee.math_real.all;
+
+entity fft is
+ generic(
+ PARALLEL : in natural := 4;
+ N : in natural := 8;
+ BITS : in natural := 16
+ );
+ port(
+ clock : in std_logic;
+ reset : in std_logic;
+
+ inverse : in std_logic;
+ in_real : in std_logic_vector(BITS-1 downto 0);
+ in_imag : in std_logic_vector(BITS-1 downto 0);
+ in_valid : in std_logic;
+ in_sop : in std_logic;
+ in_eop : in std_logic;
+
+ out_real : out std_logic_vector(BITS-1 downto 0);
+ out_imag : out std_logic_vector(BITS-1 downto 0);
+ out_error : out std_logic;
+ out_valid : out std_logic;
+ out_sop : out std_logic;
+ out_eop : out std_logic
+ );
+end entity;
+
+architecture mult of fft is
+ type fft_out_t is record
+ out_real : std_logic_vector(BITS-1 downto 0);
+ out_imag : std_logic_vector(BITS-1 downto 0);
+ out_error : std_logic;
+ out_valid : std_logic;
+ out_sop : std_logic;
+ out_eop : std_logic;
+ end record;
+ type fft_out_arr_t is array(natural range <>) of fft_out_t;
+
+ signal fft_out : fft_out_arr_t(0 to PARALLEL-1);
+
+ signal in_idx : natural range 0 to PARALLEL;
+ signal out_idx : natural range 0 to PARALLEL;
+ signal in_mask : std_logic_vector(PARALLEL-1 downto 0);
+
+begin
+
+ sync : process(clock, reset)
+ variable tmp_idx : natural range 0 to PARALLEL;
+ begin
+ if (reset = '1') then
+ in_idx <= 0;
+ out_idx <= 0;
+ in_mask <= std_logic_vector(to_unsigned(1, PARALLEL));
+ elsif (rising_edge(clock)) then
+ if (in_eop = '1') then
+ if (in_idx = PARALLEL-1) then
+ tmp_idx := 0;
+ else
+ tmp_idx := tmp_idx + 1;
+ end if;
+ in_mask <= std_logic_vector(shift_left(to_unsigned(1, PARALLEL), tmp_idx));
+ in_idx <= tmp_idx;
+ end if;
+ if (out_eop = '1') then
+ if (out_idx = PARALLEL-1) then
+ out_idx <= 0;
+ else
+ out_idx <= out_idx + 1;
+ end if;
+ end if;
+ end if;
+ end process;
+
+ U_fft_gen: for i in 0 to PARALLEL-1 generate
+ U_fft_inst : entity work.fft(arch)
+ generic map(
+ N => N,
+ BITS => BITS
+ ) port map(
+ clock => clock,
+ reset => reset,
+ inverse => inverse,
+ in_real => in_real,
+ in_imag => in_imag,
+ in_valid => in_mask(i) and in_valid,
+ in_sop => in_mask(i) and in_sop,
+ in_eop => in_mask(i) and in_eop,
+ out_real => fft_out(i).out_real,
+ out_imag => fft_out(i).out_imag,
+ out_error => fft_out(i).out_error,
+ out_valid => fft_out(i).out_valid,
+ out_sop => fft_out(i).out_sop,
+ out_eop => fft_out(i).out_eop
+ );
+ end generate;
+
+ process(fft_out, out_idx)
+ begin
+ out_real <= fft_out(out_idx).out_real;
+ out_imag <= fft_out(out_idx).out_imag;
+ out_error <= fft_out(out_idx).out_error;
+ out_valid <= fft_out(out_idx).out_valid;
+ out_sop <= fft_out(out_idx).out_sop;
+ out_eop <= fft_out(out_idx).out_eop;
+ end process;
+
+end architecture mult;
+
+architecture arch of fft is
+ constant ADDR_BITS : integer := integer(ceil(log2(real(N))));
+ constant NUM_STAGES : integer := integer(ceil(log2(real(N))));
+ constant POSTBITS : integer := 0;
+ function PIPELINE_BITS return integer is
+ begin
+ return BITS + NUM_STAGES;
+ end function;
+
+ constant DATA_BITS : integer := PIPELINE_BITS*2;
+
+ type complex_sample_t is record
+ i : signed(PIPELINE_BITS-1 downto 0);
+ q : signed(PIPELINE_BITS-1 downto 0);
+ end record;
+
+ type complex_sample_arr_t is array(natural range <>) of complex_sample_t;
+ function NULL_COMPLEX_SAMPLE return complex_sample_t is
+ variable ret : complex_sample_t;
+ begin
+ ret.i := ( others => '0' );
+ ret.q := ( others => '0' );
+ return(ret);
+ end function;
+
+ type mem_bank_ctrl_t is record
+ acc : std_logic;
+ write : std_logic;
+ solo : std_logic;
+
+ addr_a : std_logic_vector(ADDR_BITS-1 downto 0);
+ in_a : std_logic_vector(DATA_BITS-1 downto 0);
+ data_a : std_logic_vector(DATA_BITS-1 downto 0);
+ addr_b : std_logic_vector(ADDR_BITS-1 downto 0);
+ in_b : std_logic_vector(DATA_BITS-1 downto 0);
+ data_b : std_logic_vector(DATA_BITS-1 downto 0);
+ end record;
+
+ function slv_to_cst(x : std_logic_vector) return complex_sample_t is
+ variable ret : complex_sample_t;
+ begin
+ ret.i := resize(signed(x(x'high-1 downto PIPELINE_BITS)), PIPELINE_BITS);
+ ret.q := resize(signed(x(PIPELINE_BITS-1 downto 0)), PIPELINE_BITS);
+ return(ret);
+ end function;
+
+ function reverse_bit_order(x : unsigned) return std_logic_vector is
+ variable ret : std_logic_vector(x'range);
+ begin
+ for i in x'range loop
+ ret(i) := x(x'high - i);
+ end loop;
+ return(ret);
+ end function;
+
+ function NULL_MEM_BANK_CTRL return mem_bank_ctrl_t is
+ variable ret : mem_bank_ctrl_t;
+ begin
+ ret.acc := '0';
+ ret.solo := '0';
+ ret.write := '0';
+ ret.addr_a := ( others => '0' );
+ ret.in_a := ( others => '0' );
+ ret.data_a := ( others => '0' );
+ ret.addr_b := ( others => '0' );
+ ret.in_b := ( others => '0' );
+ ret.data_b := ( others => '0' );
+ return(ret);
+ end function;
+
+ type fsm_t is (IDLE, LOAD, FIRST_STAGE, RUN_STAGE, WAIT_STAGE, READ_OUT, STOP, RESET_STAGE);
+ type r_fsm_t is (IDLE, PASSTHROUGH, MEM_READ);
+
+ type mem_bank_ctrl_arr_t is array(natural range <>) of mem_bank_ctrl_t;
+ type state_t is record
+ fsm : fsm_t;
+ rfsm : r_fsm_t;
+ count : integer range 0 to N+1;
+ bf_ready : std_logic;
+ iter : integer range 0 to N+2;
+
+ mbc : mem_bank_ctrl_arr_t(1 downto 0);
+ buffer_idx : std_logic;
+ write_idx : unsigned(ADDR_BITS-1 downto 0);
+
+ stage : integer range 0 to N;
+ twiddle_idx : unsigned(ADDR_BITS-2 downto 0);
+ tw : complex_sample_t;
+
+ sop : std_logic;
+ eop : std_logic;
+ N2_sample : complex_sample_t;
+ N2_sample_r : complex_sample_t;
+ out_sample : complex_sample_t;
+ valid : std_logic;
+ end record;
+
+ type butter_fly_t is record
+ A, B, TW : complex_sample_t;
+ addr_a : std_logic_vector(ADDR_BITS-1 downto 0);
+ addr_b : std_logic_vector(ADDR_BITS-1 downto 0);
+ valid : std_logic;
+ end record;
+
+ type butter_fly_arr_t is array(natural range <>) of butter_fly_t;
+ signal bf_pl : butter_fly_arr_t(0 to 3);
+
+ function NULL_BF_T return butter_fly_t is
+ variable ret : butter_fly_t;
+ begin
+ ret.A := NULL_COMPLEX_SAMPLE;
+ ret.B := NULL_COMPLEX_SAMPLE;
+ ret.TW := NULL_COMPLEX_SAMPLE;
+ ret.addr_a := ( others => '0' );
+ ret.addr_b := ( others => '0' );
+ ret.valid := '0';
+ return(ret);
+ end function;
+
+ function shift_sample(x : complex_sample_t ; enable : std_logic) return complex_sample_t is
+ variable ret : complex_sample_t;
+ begin
+ if (enable = '0') then
+ ret.i := shift_right(x.i, POSTBITS*NUM_STAGES);
+ ret.q := shift_right(x.q, POSTBITS*NUM_STAGES);
+ else
+ ret.i := shift_right(x.i, NUM_STAGES+POSTBITS*NUM_STAGES);
+ ret.q := shift_right(x.q, NUM_STAGES+POSTBITS*NUM_STAGES);
+ end if;
+ return(ret);
+ end function;
+
+ function NULL_STATE_T return state_t is
+ variable ret : state_t;
+ begin
+ ret.fsm := IDLE;
+ ret.rfsm := IDLE;
+ for i in ret.mbc'range loop
+ ret.mbc(i) := NULL_MEM_BANK_CTRL;
+ end loop;
+
+ ret.count := 0;
+
+ ret.iter := 0;
+ ret.bf_ready := '0';
+
+ ret.buffer_idx := '0';
+ ret.write_idx := ( others => '0' );
+
+ ret.stage := 0;
+
+ ret.twiddle_idx := ( others => '0' );
+
+ ret.tw.i := ( others => '0' );
+ ret.tw.q := ( others => '0' );
+
+ ret.sop := '0';
+ ret.eop := '0';
+ ret.valid := '0';
+ ret.N2_sample := NULL_COMPLEX_SAMPLE;
+ ret.N2_sample_r := NULL_COMPLEX_SAMPLE;
+ ret.out_sample := NULL_COMPLEX_SAMPLE;
+ return(ret);
+ end function;
+
+ function rc_func(x : real) return real is
+ begin
+ if (x < 0.0) then
+ return(ceil(x));
+ else
+ return(floor(x));
+ end if;
+ end function;
+
+ function gen_roots_of_unity return complex_sample_arr_t is
+ variable t_s, t_c : real := 0.0;
+ variable ret : complex_sample_arr_t(((N/2)-1) downto 0);
+ begin
+ for i in 0 to (N/2)-1 loop
+ t_c := rc_func(cos(real(MATH_2_PI * real(i) / real(N))) * real(2**(BITS-1) - 1));
+ t_s := rc_func(sin(real(MATH_2_PI * real(i) / real(N))) * real(2**(BITS-1) - 1));
+ ret(i).i := to_signed(integer(t_c), PIPELINE_BITS);
+ ret(i).q := to_signed(integer(t_s), PIPELINE_BITS);
+ --report integer'image(i) & " = " & integer'image(integer(t_c)) &
+ -- " , " & integer'image(integer(t_s)) ;
+ end loop;
+
+ return(ret);
+ end function;
+
+ constant TLUT : complex_sample_arr_t(((N/2)-1) downto 0) := gen_roots_of_unity;
+
+ signal current, future : state_t := NULL_STATE_T;
+
+ signal muxed_mbc : mem_bank_ctrl_arr_t(1 downto 0);
+
+ signal data_mbc : mem_bank_ctrl_arr_t(1 downto 0);
+ signal curr_data : mem_bank_ctrl_t;
+
+ signal mix : complex_sample_t;
+ signal T_A, T_B : complex_sample_t;
+
+ signal comp_mbc : mem_bank_ctrl_t;
+begin
+ U_mem_banks: for i in 0 to 1 generate
+ U_mem_bank: entity work.dual_port_ram(synth)
+ generic map(
+ ADDR_BITS => ADDR_BITS,
+ DATA_BITS => DATA_BITS
+ )
+ port map(
+ clock => clock,
+ reset => reset,
+
+ acc => muxed_mbc(i).acc,
+ solo => muxed_mbc(i).solo,
+ write => muxed_mbc(i).write,
+
+ addr_a => muxed_mbc(i).addr_a,
+ in_a => muxed_mbc(i).in_a,
+ data_a => data_mbc(i).data_a,
+
+ addr_b => muxed_mbc(i).addr_b,
+ in_b => muxed_mbc(i).in_b,
+ data_b => data_mbc(i).data_b
+ );
+ end generate;
+
+ comp_mbc.addr_a <= bf_pl(3).addr_a;
+ comp_mbc.in_a <= std_logic_vector(T_A.i) & std_logic_vector(T_A.q);
+ comp_mbc.addr_b <= bf_pl(3).addr_b;
+ comp_mbc.in_b <= std_logic_vector(T_B.i) & std_logic_vector(T_B.q);
+ comp_mbc.acc <= bf_pl(3).valid;
+ comp_mbc.write <= bf_pl(3).valid;
+ comp_mbc.solo <= '0';
+
+ sync : process(clock, reset)
+ begin
+ if (reset = '1') then
+ current <= NULL_STATE_T;
+ bf_pl(1).addr_a <= ( others => '0' );
+ bf_pl(1).addr_b <= ( others => '0' );
+ bf_pl(2) <= NULL_BF_T;
+ bf_pl(3) <= NULL_BF_T;
+ elsif (rising_edge(clock)) then
+ current <= future;
+
+ bf_pl(1).valid <= current.bf_ready;
+ bf_pl(1).addr_a <= current.mbc(0).addr_a;
+ bf_pl(1).addr_b <= current.mbc(0).addr_b;
+ bf_pl(2) <= bf_pl(1);
+ bf_pl(3) <= bf_pl(2);
+ end if;
+ end process;
+
+ butterfly : process(clock, reset)
+ begin
+ if (rising_edge(clock)) then
+ mix.i <= resize(shift_right(bf_pl(1).B.i * bf_pl(1).TW.i - bf_pl(1).B.q * bf_pl(1).TW.q, BITS-1-POSTBITS), PIPELINE_BITS);
+ mix.q <= resize(shift_right(bf_pl(1).B.i * bf_pl(1).TW.q + bf_pl(1).B.q * bf_pl(1).TW.i, BITS-1-POSTBITS), PIPELINE_BITS);
+ T_A.i <= shift_left(bf_pl(2).A.i, POSTBITS) + mix.i;
+ T_A.q <= shift_left(bf_pl(2).A.q, POSTBITS) + mix.q;
+ T_B.i <= shift_left(bf_pl(2).A.i, POSTBITS) - mix.i;
+ T_B.q <= shift_left(bf_pl(2).A.q, POSTBITS) - mix.q;
+ end if;
+ end process;
+
+ out_sop <= current.sop;
+ out_valid <= current.valid;
+ out_eop <= current.eop;
+ out_error <= '1' when current.fsm = STOP else '0';
+
+ out_real <= std_logic_vector(resize(current.out_sample.i, BITS));
+ out_imag <= std_logic_vector(resize(current.out_sample.q, BITS));
+
+ comb : process(all)
+ variable tmp_addr_a, tmp_addr_b : unsigned(ADDR_BITS-1 downto 0);
+ variable ones_reg : unsigned(ADDR_BITS-2 downto 0);
+ variable tmp_tw : complex_sample_t;
+ begin
+ tmp_tw := current.tw;
+ if (inverse = '1' ) then
+ bf_pl(1).TW <= tmp_tw;
+ else
+ bf_pl(1).TW.i <= tmp_tw.i;
+ bf_pl(1).TW.q <= -tmp_tw.q;
+ end if;
+ bf_pl(1).A <= slv_to_cst(curr_data.data_a);
+ if (current.fsm = FIRST_STAGE or (current.fsm = WAIT_STAGE and current.stage = 0)) then
+ bf_pl(1).B <= current.N2_sample_r;
+ else
+ bf_pl(1).B <= slv_to_cst(curr_data.data_b);
+ end if;
+ if (current.buffer_idx = '0') then
+ muxed_mbc(0) <= current.mbc(0); -- during RUN_STAGES: READ
+ curr_data <= data_mbc(0);
+
+ muxed_mbc(1) <= comp_mbc; -- during RUN_STAGES: WRITE
+ else
+ muxed_mbc(0) <= comp_mbc; -- during RUN_STAGES: WRITE
+
+ muxed_mbc(1) <= current.mbc(0); -- during RUN_STAGES: READ
+ curr_data <= data_mbc(1);
+ end if;
+
+ future <= current;
+
+ for i in future.mbc'range loop
+ future.mbc(i) <= NULL_MEM_BANK_CTRL;
+ end loop;
+ future.bf_ready <= '0';
+ future.sop <= '0';
+ future.eop <= '0';
+ future.valid <= '0';
+
+ ones_reg := ( others => '1' );
+
+ -- note, this updates on the next cycle
+ if (current.fsm = FIRST_STAGE or current.fsm = RUN_STAGE or current.fsm = WAIT_STAGE) then
+ tmp_tw := TLUT(to_integer(current.twiddle_idx));
+ future.tw <= tmp_tw;
+ future.twiddle_idx <= to_unsigned(current.iter, ones_reg'high+1)
+ and shift_left(ones_reg, NUM_STAGES-1-current.stage);
+ end if;
+
+ future.N2_sample_r <= current.N2_sample;
+
+ case current.fsm is
+ when IDLE =>
+ if (in_sop = '1') then
+ future.fsm <= LOAD;
+ if (in_valid = '1') then
+ future.mbc(0).addr_b <= std_logic_vector(to_unsigned(1, ADDR_BITS));
+ future.mbc(0).addr_a <= reverse_bit_order(current.write_idx);
+ future.mbc(0).in_a <= std_logic_vector(resize(signed(in_real), PIPELINE_BITS) & resize(signed(in_imag), PIPELINE_BITS));
+ future.mbc(0).acc <= '1';
+ future.mbc(0).solo <= '1';
+ future.mbc(0).write <= '1';
+ future.write_idx <= current.write_idx + 1;
+ future.count <= 1;
+ end if;
+ end if;
+ when LOAD =>
+ if (in_valid = '1') then
+ future.write_idx <= current.write_idx + 1;
+ future.count <= current.count + 1;
+ if (current.write_idx = (N/2)) then
+ future.mbc(0).addr_a <= reverse_bit_order(current.write_idx-32);
+ future.mbc(0).addr_b <= reverse_bit_order(current.write_idx);
+ future.N2_sample.i <= resize(signed(in_real), PIPELINE_BITS);
+ future.N2_sample.q <= resize(signed(in_imag), PIPELINE_BITS);
+ future.bf_ready <= '1';
+ future.mbc(0).acc <= '1';
+ future.fsm <= FIRST_STAGE;
+ else
+ future.mbc(0).addr_b <= std_logic_vector(to_unsigned(1, ADDR_BITS));
+ future.mbc(0).addr_a <= reverse_bit_order(current.write_idx);
+ future.mbc(0).in_a <= std_logic_vector(resize(signed(in_real), PIPELINE_BITS) & resize(signed(in_imag), PIPELINE_BITS));
+ future.mbc(0).acc <= '1';
+ future.mbc(0).solo <= '1';
+ future.mbc(0).write <= '1';
+ end if;
+ end if;
+ if (in_eop = '1') then
+ future.fsm <= STOP;
+ end if;
+ when FIRST_STAGE =>
+ if (in_valid = '1') then
+ future.count <= current.count + 1;
+ future.write_idx <= current.write_idx + 1;
+ future.bf_ready <= '1';
+ future.mbc(0).addr_a <= reverse_bit_order(current.write_idx-32);
+ future.mbc(0).addr_b <= reverse_bit_order(current.write_idx);
+ future.mbc(0).acc <= '1';
+ future.N2_sample.i <= resize(signed(in_real), PIPELINE_BITS);
+ future.N2_sample.q <= resize(signed(in_imag), PIPELINE_BITS);
+ if (current.write_idx = N-1) then
+ future.iter <= 3;
+ future.fsm <= WAIT_STAGE;
+ if (in_eop = '0') then
+ future.fsm <= STOP;
+ end if;
+ else
+ if (in_eop = '1') then
+ future.fsm <= STOP;
+ end if;
+ end if;
+ end if;
+ when RUN_STAGE =>
+ future.mbc(0).acc <= '1';
+ future.bf_ready <= '1';
+ tmp_addr_a := rotate_left(to_unsigned(current.iter*2, ADDR_BITS), current.stage);
+ tmp_addr_b := rotate_left(to_unsigned(current.iter*2+1, ADDR_BITS), current.stage);
+
+ future.mbc(0).addr_a <= std_logic_vector(tmp_addr_a);
+ future.mbc(0).addr_b <= std_logic_vector(tmp_addr_b);
+ if (current.iter = (N/2)-1) then
+ future.iter <= 3;
+ future.fsm <= WAIT_STAGE;
+ else
+ future.iter <= current.iter + 1;
+ end if;
+
+ when WAIT_STAGE =>
+ if (current.iter = 0) then
+ future.buffer_idx <= not current.buffer_idx;
+ if (current.stage < NUM_STAGES-1) then
+ future.stage <= current.stage + 1;
+ future.fsm <= RUN_STAGE;
+ future.iter <= 0;
+ else
+ future.fsm <= READ_OUT;
+ future.iter <= N/2 + 2;
+ future.mbc(0).addr_a <= std_logic_vector(to_unsigned(N/2+1, ADDR_BITS));
+ future.mbc(0).acc <= '1';
+ future.mbc(0).solo <= '1';
+ end if;
+ else
+ future.iter <= current.iter - 1;
+ end if;
+ when READ_OUT =>
+ if (current.iter = N+1) then
+ future.fsm <= RESET_STAGE;
+ future.eop <= '1';
+ end if;
+ if (current.iter < N) then
+ future.mbc(0).addr_a <= std_logic_vector(to_unsigned(current.iter, ADDR_BITS));
+ end if;
+ future.iter <= current.iter + 1;
+ future.mbc(0).acc <= '1';
+ future.mbc(0).solo <= '1';
+
+ when others =>
+ future <= NULL_STATE_T;
+ end case;
+
+ case current.rfsm is
+ when IDLE =>
+ if (current.fsm = RUN_STAGE and current.stage = NUM_STAGES - 1) then
+ future.rfsm <= PASSTHROUGH;
+ end if;
+ when PASSTHROUGH =>
+ if (current.fsm = RUN_STAGE) then
+ if (current.iter = 4) then
+ future.N2_sample <= T_B;
+ future.sop <= '1';
+ end if;
+ end if;
+
+ if (current.iter > 3 or current.fsm = WAIT_STAGE) then
+ if (current.iter = (N/2)+2) then
+ future.out_sample <= shift_sample(current.N2_sample, inverse);
+ else
+ future.out_sample <= shift_sample(T_A, inverse);
+ end if;
+ future.valid <= '1';
+ end if;
+
+ if (current.fsm = READ_OUT) then
+ future.rfsm <= MEM_READ;
+ end if;
+ when MEM_READ =>
+ if (current.iter = N+1) then
+ future.rfsm <= IDLE;
+ end if;
+ future.out_sample <= shift_sample(slv_to_cst(curr_data.data_a), inverse);
+ future.valid <= '1';
+ when others =>
+ future <= NULL_STATE_T;
+ end case;
+
+ end process;
+
+
+end architecture;
diff --git a/testsuite/gna/issue2065/repro.vhdl b/testsuite/gna/issue2065/repro.vhdl
new file mode 100644
index 000000000..a6dc56d26
--- /dev/null
+++ b/testsuite/gna/issue2065/repro.vhdl
@@ -0,0 +1,15 @@
+entity repro is
+ generic (depth : natural := 5);
+ port (inp : bit := '0');
+end entity;
+
+architecture mult of repro is
+ signal s : bit;
+begin
+ gen: if depth > 0 generate
+ inst : entity work.repro
+ generic map (depth => depth - 1)
+ port map(inp => s and inp);
+ end generate;
+end architecture mult;
+
diff --git a/testsuite/gna/issue2065/testsuite.sh b/testsuite/gna/issue2065/testsuite.sh
new file mode 100755
index 000000000..8cc43fcb1
--- /dev/null
+++ b/testsuite/gna/issue2065/testsuite.sh
@@ -0,0 +1,14 @@
+#! /bin/sh
+
+. ../../testenv.sh
+
+export GHDL_STD_FLAGS=--std=08
+analyze repro.vhdl
+elab_simulate repro
+
+analyze dual_port_ram.vhdl
+analyze fft.vhdl
+
+clean
+
+echo "Test successful"