use work.ppc_fx_insns.all;
entity execute1 is
+ generic (
+ EX1_BYPASS : boolean := true
+ );
port (
clk : in std_ulogic;
+ rst : in std_ulogic;
-- asynchronous
flush_out : out std_ulogic;
stall_out : out std_ulogic;
e_in : in Decode2ToExecute1Type;
+ l_in : in Loadstore1ToExecute1Type;
+
+ ext_irq_in : std_ulogic;
-- asynchronous
+ l_out : out Execute1ToLoadstore1Type;
f_out : out Execute1ToFetch1Type;
e_out : out Execute1ToWritebackType;
+ dbg_msr_out : out std_ulogic_vector(63 downto 0);
+
icache_inval : out std_ulogic;
terminate_out : out std_ulogic
);
e : Execute1ToWritebackType;
lr_update : std_ulogic;
next_lr : std_ulogic_vector(63 downto 0);
+ mul_in_progress : std_ulogic;
+ div_in_progress : std_ulogic;
+ cntz_in_progress : std_ulogic;
+ slow_op_dest : gpr_index_t;
+ slow_op_rc : std_ulogic;
+ slow_op_oe : std_ulogic;
+ slow_op_xerc : xer_common_t;
+ ldst_nia : std_ulogic_vector(63 downto 0);
end record;
+ constant reg_type_init : reg_type :=
+ (e => Execute1ToWritebackInit, lr_update => '0',
+ mul_in_progress => '0', div_in_progress => '0', cntz_in_progress => '0',
+ slow_op_rc => '0', slow_op_oe => '0', slow_op_xerc => xerc_init,
+ next_lr => (others => '0'), ldst_nia => (others => '0'), others => (others => '0'));
signal r, rin : reg_type;
- signal ctrl: ctrl_t := (others => (others => '0'));
- signal ctrl_tmp: ctrl_t := (others => (others => '0'));
+ signal a_in, b_in, c_in : std_ulogic_vector(63 downto 0);
+ signal ctrl: ctrl_t := (irq_state => WRITE_SRR0, others => (others => '0'));
+ signal ctrl_tmp: ctrl_t := (irq_state => WRITE_SRR0, others => (others => '0'));
signal right_shift, rot_clear_left, rot_clear_right: std_ulogic;
+ signal rot_sign_ext: std_ulogic;
signal rotator_result: std_ulogic_vector(63 downto 0);
signal rotator_carry: std_ulogic;
signal logical_result: std_ulogic_vector(63 downto 0);
signal countzero_result: std_ulogic_vector(63 downto 0);
+ signal popcnt_result: std_ulogic_vector(63 downto 0);
+ signal parity_result: std_ulogic_vector(63 downto 0);
+
+ -- multiply signals
+ signal x_to_multiply: Execute1ToMultiplyType;
+ signal multiply_to_x: MultiplyToExecute1Type;
+
+ -- divider signals
+ signal x_to_divider: Execute1ToDividerType;
+ signal divider_to_x: DividerToExecute1Type;
+
+ type privilege_level is (USER, SUPER);
+ type op_privilege_array is array(insn_type_t) of privilege_level;
+ constant op_privilege: op_privilege_array := (
+ OP_ATTN => SUPER,
+ OP_MFMSR => SUPER,
+ OP_MTMSRD => SUPER,
+ OP_RFID => SUPER,
+ OP_TLBIE => SUPER,
+ others => USER
+ );
+
+ function instr_is_privileged(op: insn_type_t; insn: std_ulogic_vector(31 downto 0))
+ return boolean is
+ begin
+ if op_privilege(op) = SUPER then
+ return true;
+ elsif op = OP_MFSPR or op = OP_MTSPR then
+ return insn(20) = '1';
+ else
+ return false;
+ end if;
+ end;
procedure set_carry(e: inout Execute1ToWritebackType;
carry32 : in std_ulogic;
end case;
end;
+ function msr_copy(msr: std_ulogic_vector(63 downto 0))
+ return std_ulogic_vector is
+ variable msr_out: std_ulogic_vector(63 downto 0);
+ begin
+ -- ISA says this:
+ -- Defined MSR bits are classified as either full func-
+ -- tion or partial function. Full function MSR bits are
+ -- saved in SRR1 or HSRR1 when an interrupt other
+ -- than a System Call Vectored interrupt occurs and
+ -- restored by rfscv, rfid, or hrfid, while partial func-
+ -- tion MSR bits are not saved or restored.
+ -- Full function MSR bits lie in the range 0:32, 37:41, and
+ -- 48:63, and partial function MSR bits lie in the range
+ -- 33:36 and 42:47. (Note this is IBM bit numbering).
+ msr_out := (others => '0');
+ msr_out(63 downto 31) := msr(63 downto 31);
+ msr_out(26 downto 22) := msr(26 downto 22);
+ msr_out(15 downto 0) := msr(15 downto 0);
+ return msr_out;
+ end;
+
begin
rotator_0: entity work.rotator
port map (
- rs => e_in.read_data3,
- ra => e_in.read_data1,
- shift => e_in.read_data2(6 downto 0),
+ rs => c_in,
+ ra => a_in,
+ shift => b_in(6 downto 0),
insn => e_in.insn,
is_32bit => e_in.is_32bit,
right_shift => right_shift,
arith => e_in.is_signed,
clear_left => rot_clear_left,
clear_right => rot_clear_right,
+ sign_ext_rs => rot_sign_ext,
result => rotator_result,
carry_out => rotator_carry
);
logical_0: entity work.logical
port map (
- rs => e_in.read_data3,
- rb => e_in.read_data2,
+ rs => c_in,
+ rb => b_in,
op => e_in.insn_type,
invert_in => e_in.invert_a,
invert_out => e_in.invert_out,
- result => logical_result
+ result => logical_result,
+ datalen => e_in.data_len,
+ popcnt => popcnt_result,
+ parity => parity_result
);
countzero_0: entity work.zero_counter
port map (
- rs => e_in.read_data3,
+ clk => clk,
+ rs => c_in,
count_right => e_in.insn(10),
is_32bit => e_in.is_32bit,
result => countzero_result
);
+ multiply_0: entity work.multiply
+ port map (
+ clk => clk,
+ m_in => x_to_multiply,
+ m_out => multiply_to_x
+ );
+
+ divider_0: entity work.divider
+ port map (
+ clk => clk,
+ rst => rst,
+ d_in => x_to_divider,
+ d_out => divider_to_x
+ );
+
+ dbg_msr_out <= ctrl.msr;
+
+ a_in <= r.e.write_data when EX1_BYPASS and e_in.bypass_data1 = '1' else e_in.read_data1;
+ b_in <= r.e.write_data when EX1_BYPASS and e_in.bypass_data2 = '1' else e_in.read_data2;
+ c_in <= r.e.write_data when EX1_BYPASS and e_in.bypass_data3 = '1' else e_in.read_data3;
+
execute1_0: process(clk)
begin
if rising_edge(clk) then
- r <= rin;
- ctrl <= ctrl_tmp;
- assert not (r.lr_update = '1' and e_in.valid = '1')
- report "LR update collision with valid in EX1"
- severity failure;
- if r.lr_update = '1' then
- report "LR update to " & to_hstring(r.next_lr);
- end if;
+ if rst = '1' then
+ r <= reg_type_init;
+ ctrl.msr <= (MSR_SF => '1', MSR_LE => '1', others => '0');
+ ctrl.irq_state <= WRITE_SRR0;
+ else
+ r <= rin;
+ ctrl <= ctrl_tmp;
+ assert not (r.lr_update = '1' and e_in.valid = '1')
+ report "LR update collision with valid in EX1"
+ severity failure;
+ if r.lr_update = '1' then
+ report "LR update to " & to_hstring(r.next_lr);
+ end if;
+ end if;
end if;
end process;
variable bo, bi : std_ulogic_vector(4 downto 0);
variable bf, bfa : std_ulogic_vector(2 downto 0);
variable cr_op : std_ulogic_vector(9 downto 0);
+ variable cr_operands : std_ulogic_vector(1 downto 0);
variable bt, ba, bb : std_ulogic_vector(4 downto 0);
variable btnum, banum, bbnum : integer range 0 to 31;
variable crresult : std_ulogic;
variable l : std_ulogic;
variable next_nia : std_ulogic_vector(63 downto 0);
variable carry_32, carry_64 : std_ulogic;
+ variable sign1, sign2 : std_ulogic;
+ variable abs1, abs2 : signed(63 downto 0);
+ variable overflow : std_ulogic;
+ variable negative : std_ulogic;
+ variable zerohi, zerolo : std_ulogic;
+ variable msb_a, msb_b : std_ulogic;
+ variable a_lt : std_ulogic;
+ variable lv : Execute1ToLoadstore1Type;
+ variable irq_valid : std_ulogic;
+ variable exception : std_ulogic;
+ variable exception_nextpc : std_ulogic;
+ variable trapval : std_ulogic_vector(4 downto 0);
+ variable illegal : std_ulogic;
begin
result := (others => '0');
result_with_carry := (others => '0');
v := r;
v.e := Execute1ToWritebackInit;
+ lv := Execute1ToLoadstore1Init;
-- XER forwarding. To avoid having to track XER hazards, we
-- use the previously latched value.
end if;
v.lr_update := '0';
+ v.mul_in_progress := '0';
+ v.div_in_progress := '0';
+ v.cntz_in_progress := '0';
+
+ -- signals to multiply unit
+ x_to_multiply <= Execute1ToMultiplyInit;
+ x_to_multiply.insn_type <= e_in.insn_type;
+ x_to_multiply.is_32bit <= e_in.is_32bit;
+
+ if e_in.is_32bit = '1' then
+ if e_in.is_signed = '1' then
+ x_to_multiply.data1 <= (others => a_in(31));
+ x_to_multiply.data1(31 downto 0) <= a_in(31 downto 0);
+ x_to_multiply.data2 <= (others => b_in(31));
+ x_to_multiply.data2(31 downto 0) <= b_in(31 downto 0);
+ else
+ x_to_multiply.data1 <= '0' & x"00000000" & a_in(31 downto 0);
+ x_to_multiply.data2 <= '0' & x"00000000" & b_in(31 downto 0);
+ end if;
+ else
+ if e_in.is_signed = '1' then
+ x_to_multiply.data1 <= a_in(63) & a_in;
+ x_to_multiply.data2 <= b_in(63) & b_in;
+ else
+ x_to_multiply.data1 <= '0' & a_in;
+ x_to_multiply.data2 <= '0' & b_in;
+ end if;
+ end if;
+
+ -- signals to divide unit
+ sign1 := '0';
+ sign2 := '0';
+ if e_in.is_signed = '1' then
+ if e_in.is_32bit = '1' then
+ sign1 := a_in(31);
+ sign2 := b_in(31);
+ else
+ sign1 := a_in(63);
+ sign2 := b_in(63);
+ end if;
+ end if;
+ -- take absolute values
+ if sign1 = '0' then
+ abs1 := signed(a_in);
+ else
+ abs1 := - signed(a_in);
+ end if;
+ if sign2 = '0' then
+ abs2 := signed(b_in);
+ else
+ abs2 := - signed(b_in);
+ end if;
+
+ x_to_divider <= Execute1ToDividerInit;
+ x_to_divider.is_signed <= e_in.is_signed;
+ x_to_divider.is_32bit <= e_in.is_32bit;
+ if e_in.insn_type = OP_MOD then
+ x_to_divider.is_modulus <= '1';
+ end if;
+ x_to_divider.neg_result <= sign1 xor (sign2 and not x_to_divider.is_modulus);
+ if e_in.is_32bit = '0' then
+ -- 64-bit forms
+ if e_in.insn_type = OP_DIVE then
+ x_to_divider.is_extended <= '1';
+ end if;
+ x_to_divider.dividend <= std_ulogic_vector(abs1);
+ x_to_divider.divisor <= std_ulogic_vector(abs2);
+ else
+ -- 32-bit forms
+ x_to_divider.is_extended <= '0';
+ if e_in.insn_type = OP_DIVE then -- extended forms
+ x_to_divider.dividend <= std_ulogic_vector(abs1(31 downto 0)) & x"00000000";
+ else
+ x_to_divider.dividend <= x"00000000" & std_ulogic_vector(abs1(31 downto 0));
+ end if;
+ x_to_divider.divisor <= x"00000000" & std_ulogic_vector(abs2(31 downto 0));
+ end if;
ctrl_tmp <= ctrl;
-- FIXME: run at 512MHz not core freq
ctrl_tmp.tb <= std_ulogic_vector(unsigned(ctrl.tb) + 1);
+ ctrl_tmp.dec <= std_ulogic_vector(unsigned(ctrl.dec) - 1);
+
+ irq_valid := '0';
+ if ctrl.msr(MSR_EE) = '1' then
+ if ctrl.dec(63) = '1' then
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#900#, 64));
+ report "IRQ valid: DEC";
+ irq_valid := '1';
+ elsif ext_irq_in = '1' then
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#500#, 64));
+ report "IRQ valid: External";
+ irq_valid := '1';
+ end if;
+ end if;
terminate_out <= '0';
icache_inval <= '0';
stall_out <= '0';
f_out <= Execute1ToFetch1TypeInit;
+ -- send MSR[IR] and ~MSR[PR] up to fetch1
+ f_out.virt_mode <= ctrl.msr(MSR_IR);
+ f_out.priv_mode <= not ctrl.msr(MSR_PR);
-- Next insn adder used in a couple of places
next_nia := std_ulogic_vector(unsigned(e_in.nia) + 4);
right_shift <= '1' when e_in.insn_type = OP_SHR else '0';
rot_clear_left <= '1' when e_in.insn_type = OP_RLC or e_in.insn_type = OP_RLCL else '0';
rot_clear_right <= '1' when e_in.insn_type = OP_RLC or e_in.insn_type = OP_RLCR else '0';
+ rot_sign_ext <= '1' when e_in.insn_type = OP_EXTSWSLI else '0';
+
+ ctrl_tmp.irq_state <= WRITE_SRR0;
+ exception := '0';
+ illegal := '0';
+ exception_nextpc := '0';
+ v.e.exc_write_enable := '0';
+ v.e.exc_write_reg := fast_spr_num(SPR_SRR0);
+ v.e.exc_write_data := e_in.nia;
+
+ if ctrl.irq_state = WRITE_SRR1 then
+ v.e.exc_write_reg := fast_spr_num(SPR_SRR1);
+ v.e.exc_write_data := ctrl.srr1;
+ v.e.exc_write_enable := '1';
+ ctrl_tmp.msr(MSR_SF) <= '1';
+ ctrl_tmp.msr(MSR_EE) <= '0';
+ ctrl_tmp.msr(MSR_PR) <= '0';
+ ctrl_tmp.msr(MSR_IR) <= '0';
+ ctrl_tmp.msr(MSR_DR) <= '0';
+ ctrl_tmp.msr(MSR_RI) <= '0';
+ ctrl_tmp.msr(MSR_LE) <= '1';
+ f_out.redirect <= '1';
+ f_out.virt_mode <= '0';
+ f_out.priv_mode <= '1';
+ f_out.redirect_nia <= ctrl.irq_nia;
+ v.e.valid := e_in.valid;
+ report "Writing SRR1: " & to_hstring(ctrl.srr1);
+
+ elsif irq_valid = '1' and e_in.valid = '1' then
+ -- we need two cycles to write srr0 and 1
+ -- will need more when we have to write HEIR
+ -- Don't deliver the interrupt until we have a valid instruction
+ -- coming in, so we have a valid NIA to put in SRR0.
+ exception := '1';
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
- if e_in.valid = '1' then
+ elsif e_in.valid = '1' and ctrl.msr(MSR_PR) = '1' and
+ instr_is_privileged(e_in.insn_type, e_in.insn) then
+ -- generate a program interrupt
+ exception := '1';
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#700#, 64));
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
+ -- set bit 45 to indicate privileged instruction type interrupt
+ ctrl_tmp.srr1(63 - 45) <= '1';
+ report "privileged instruction";
+
+ elsif e_in.valid = '1' and e_in.unit = ALU then
+
+ report "execute nia " & to_hstring(e_in.nia);
v.e.valid := '1';
v.e.write_reg := e_in.write_reg;
- v.e.write_len := x"8";
- v.e.sign_extend := '0';
+ v.slow_op_dest := gspr_to_gpr(e_in.write_reg);
+ v.slow_op_rc := e_in.rc;
+ v.slow_op_oe := e_in.oe;
+ v.slow_op_xerc := v.e.xerc;
case_0: case e_in.insn_type is
when OP_ILLEGAL =>
- terminate_out <= '1';
- report "illegal";
+ -- we need two cycles to write srr0 and 1
+ -- will need more when we have to write HEIR
+ illegal := '1';
+ when OP_SC =>
+ -- check bit 1 of the instruction is 1 so we know this is sc;
+ -- 0 would mean scv, so generate an illegal instruction interrupt
+ -- we need two cycles to write srr0 and 1
+ if e_in.insn(1) = '1' then
+ exception := '1';
+ exception_nextpc := '1';
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#C00#, 64));
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
+ report "sc";
+ else
+ illegal := '1';
+ end if;
+ when OP_ATTN =>
+ -- check bits 1-10 of the instruction to make sure it's attn
+ -- if not then it is illegal
+ if e_in.insn(10 downto 1) = "0100000000" then
+ terminate_out <= '1';
+ report "ATTN";
+ else
+ illegal := '1';
+ end if;
when OP_NOP =>
-- Do nothing
- when OP_ADD =>
+ when OP_ADD | OP_CMP | OP_TRAP =>
if e_in.invert_a = '0' then
- a_inv := e_in.read_data1;
+ a_inv := a_in;
else
- a_inv := not e_in.read_data1;
+ a_inv := not a_in;
end if;
- result_with_carry := ppc_adde(a_inv, e_in.read_data2,
+ result_with_carry := ppc_adde(a_inv, b_in,
decode_input_carry(e_in.input_carry, v.e.xerc));
result := result_with_carry(63 downto 0);
- carry_32 := result(32) xor a_inv(32) xor e_in.read_data2(32);
+ carry_32 := result(32) xor a_inv(32) xor b_in(32);
carry_64 := result_with_carry(64);
- if e_in.output_carry = '1' then
- set_carry(v.e, carry_32, carry_64);
- end if;
- if e_in.oe = '1' then
- set_ov(v.e,
- calc_ov(a_inv(63), e_in.read_data2(63), carry_64, result_with_carry(63)),
- calc_ov(a_inv(31), e_in.read_data2(31), carry_32, result_with_carry(31)));
- end if;
- result_en := '1';
+ if e_in.insn_type = OP_ADD then
+ if e_in.output_carry = '1' then
+ set_carry(v.e, carry_32, carry_64);
+ end if;
+ if e_in.oe = '1' then
+ set_ov(v.e,
+ calc_ov(a_inv(63), b_in(63), carry_64, result_with_carry(63)),
+ calc_ov(a_inv(31), b_in(31), carry_32, result_with_carry(31)));
+ end if;
+ result_en := '1';
+ else
+ -- trap, CMP and CMPL instructions
+ -- Note, we have done RB - RA, not RA - RB
+ if e_in.insn_type = OP_CMP then
+ l := insn_l(e_in.insn);
+ else
+ l := not e_in.is_32bit;
+ end if;
+ zerolo := not (or (a_in(31 downto 0) xor b_in(31 downto 0)));
+ zerohi := not (or (a_in(63 downto 32) xor b_in(63 downto 32)));
+ if zerolo = '1' and (l = '0' or zerohi = '1') then
+ -- values are equal
+ trapval := "00100";
+ else
+ if l = '1' then
+ -- 64-bit comparison
+ msb_a := a_in(63);
+ msb_b := b_in(63);
+ else
+ -- 32-bit comparison
+ msb_a := a_in(31);
+ msb_b := b_in(31);
+ end if;
+ if msb_a /= msb_b then
+ -- Subtraction might overflow, but
+ -- comparison is clear from MSB difference.
+ -- for signed, 0 is greater; for unsigned, 1 is greater
+ trapval := msb_a & msb_b & '0' & msb_b & msb_a;
+ else
+ -- Subtraction cannot overflow since MSBs are equal.
+ -- carry = 1 indicates RA is smaller (signed or unsigned)
+ a_lt := (not l and carry_32) or (l and carry_64);
+ trapval := a_lt & not a_lt & '0' & a_lt & not a_lt;
+ end if;
+ end if;
+ if e_in.insn_type = OP_CMP then
+ if e_in.is_signed = '1' then
+ newcrf := trapval(4 downto 2) & v.e.xerc.so;
+ else
+ newcrf := trapval(1 downto 0) & trapval(2) & v.e.xerc.so;
+ end if;
+ bf := insn_bf(e_in.insn);
+ crnum := to_integer(unsigned(bf));
+ v.e.write_cr_enable := '1';
+ v.e.write_cr_mask := num_to_fxm(crnum);
+ for i in 0 to 7 loop
+ lo := i*4;
+ hi := lo + 3;
+ v.e.write_cr_data(hi downto lo) := newcrf;
+ end loop;
+ else
+ -- trap instructions (tw, twi, td, tdi)
+ if or (trapval and insn_to(e_in.insn)) = '1' then
+ -- generate trap-type program interrupt
+ exception := '1';
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#700#, 64));
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
+ -- set bit 46 to say trap occurred
+ ctrl_tmp.srr1(63 - 46) <= '1';
+ report "trap";
+ end if;
+ end if;
+ end if;
when OP_AND | OP_OR | OP_XOR =>
result := logical_result;
result_en := '1';
when OP_B =>
f_out.redirect <= '1';
if (insn_aa(e_in.insn)) then
- f_out.redirect_nia <= std_ulogic_vector(signed(e_in.read_data2));
+ f_out.redirect_nia <= b_in;
else
- f_out.redirect_nia <= std_ulogic_vector(signed(e_in.nia) + signed(e_in.read_data2));
+ f_out.redirect_nia <= std_ulogic_vector(signed(e_in.nia) + signed(b_in));
end if;
when OP_BC =>
-- read_data1 is CTR
bo := insn_bo(e_in.insn);
bi := insn_bi(e_in.insn);
if bo(4-2) = '0' then
- result := std_ulogic_vector(unsigned(e_in.read_data1) - 1);
+ result := std_ulogic_vector(unsigned(a_in) - 1);
result_en := '1';
v.e.write_reg := fast_spr_num(SPR_CTR);
end if;
- if ppc_bc_taken(bo, bi, e_in.cr, e_in.read_data1) = 1 then
+ if ppc_bc_taken(bo, bi, e_in.cr, a_in) = 1 then
f_out.redirect <= '1';
if (insn_aa(e_in.insn)) then
- f_out.redirect_nia <= std_ulogic_vector(signed(e_in.read_data2));
+ f_out.redirect_nia <= b_in;
else
- f_out.redirect_nia <= std_ulogic_vector(signed(e_in.nia) + signed(e_in.read_data2));
+ f_out.redirect_nia <= std_ulogic_vector(signed(e_in.nia) + signed(b_in));
end if;
end if;
when OP_BCREG =>
bo := insn_bo(e_in.insn);
bi := insn_bi(e_in.insn);
if bo(4-2) = '0' and e_in.insn(10) = '0' then
- result := std_ulogic_vector(unsigned(e_in.read_data1) - 1);
+ result := std_ulogic_vector(unsigned(a_in) - 1);
result_en := '1';
v.e.write_reg := fast_spr_num(SPR_CTR);
end if;
- if ppc_bc_taken(bo, bi, e_in.cr, e_in.read_data1) = 1 then
+ if ppc_bc_taken(bo, bi, e_in.cr, a_in) = 1 then
f_out.redirect <= '1';
- f_out.redirect_nia <= e_in.read_data2(63 downto 2) & "00";
+ f_out.redirect_nia <= b_in(63 downto 2) & "00";
end if;
+
+ when OP_RFID =>
+ f_out.redirect <= '1';
+ f_out.virt_mode <= b_in(MSR_IR) or b_in(MSR_PR);
+ f_out.priv_mode <= not b_in(MSR_PR);
+ f_out.redirect_nia <= a_in(63 downto 2) & "00"; -- srr0
+ -- Can't use msr_copy here because the partial function MSR
+ -- bits should be left unchanged, not zeroed.
+ ctrl_tmp.msr(63 downto 31) <= b_in(63 downto 31);
+ ctrl_tmp.msr(26 downto 22) <= b_in(26 downto 22);
+ ctrl_tmp.msr(15 downto 0) <= b_in(15 downto 0);
+ if b_in(MSR_PR) = '1' then
+ ctrl_tmp.msr(MSR_EE) <= '1';
+ ctrl_tmp.msr(MSR_IR) <= '1';
+ ctrl_tmp.msr(MSR_DR) <= '1';
+ end if;
+
when OP_CMPB =>
- result := ppc_cmpb(e_in.read_data3, e_in.read_data2);
+ result := ppc_cmpb(c_in, b_in);
result_en := '1';
- when OP_CMP =>
- bf := insn_bf(e_in.insn);
- l := insn_l(e_in.insn);
- v.e.write_cr_enable := '1';
- crnum := to_integer(unsigned(bf));
- v.e.write_cr_mask := num_to_fxm(crnum);
- for i in 0 to 7 loop
- lo := i*4;
- hi := lo + 3;
- v.e.write_cr_data(hi downto lo) := ppc_cmp(l, e_in.read_data1, e_in.read_data2, v.e.xerc.so);
- end loop;
- when OP_CMPL =>
- bf := insn_bf(e_in.insn);
- l := insn_l(e_in.insn);
- v.e.write_cr_enable := '1';
- crnum := to_integer(unsigned(bf));
- v.e.write_cr_mask := num_to_fxm(crnum);
- for i in 0 to 7 loop
- lo := i*4;
- hi := lo + 3;
- v.e.write_cr_data(hi downto lo) := ppc_cmpl(l, e_in.read_data1, e_in.read_data2, v.e.xerc.so);
- end loop;
- when OP_CNTZ =>
- result := countzero_result;
- result_en := '1';
- when OP_EXTS =>
- v.e.write_len := e_in.data_len;
- v.e.sign_extend := '1';
- result := e_in.read_data3;
+ when OP_CNTZ =>
+ v.e.valid := '0';
+ v.cntz_in_progress := '1';
+ stall_out <= '1';
+ when OP_EXTS =>
+ -- note data_len is a 1-hot encoding
+ negative := (e_in.data_len(0) and c_in(7)) or
+ (e_in.data_len(1) and c_in(15)) or
+ (e_in.data_len(2) and c_in(31));
+ result := (others => negative);
+ if e_in.data_len(2) = '1' then
+ result(31 downto 16) := c_in(31 downto 16);
+ end if;
+ if e_in.data_len(2) = '1' or e_in.data_len(1) = '1' then
+ result(15 downto 8) := c_in(15 downto 8);
+ end if;
+ result(7 downto 0) := c_in(7 downto 0);
result_en := '1';
when OP_ISEL =>
crbit := to_integer(unsigned(insn_bc(e_in.insn)));
if e_in.cr(31-crbit) = '1' then
- result := e_in.read_data1;
+ result := a_in;
else
- result := e_in.read_data2;
+ result := b_in;
end if;
result_en := '1';
- when OP_MCRF =>
+ when OP_CROP =>
cr_op := insn_cr(e_in.insn);
report "CR OP " & to_hstring(cr_op);
if cr_op(0) = '0' then -- MCRF
btnum := 31 - to_integer(unsigned(bt));
banum := 31 - to_integer(unsigned(ba));
bbnum := 31 - to_integer(unsigned(bb));
- case cr_op(8 downto 5) is
- when "1001" => -- CREQV
- crresult := not(e_in.cr(banum) xor e_in.cr(bbnum));
- when "0111" => -- CRNAND
- crresult := not(e_in.cr(banum) and e_in.cr(bbnum));
- when "0100" => -- CRANDC
- crresult := (e_in.cr(banum) and not e_in.cr(bbnum));
- when "1000" => -- CRAND
- crresult := (e_in.cr(banum) and e_in.cr(bbnum));
- when "0001" => -- CRNOR
- crresult := not(e_in.cr(banum) or e_in.cr(bbnum));
- when "1101" => -- CRORC
- crresult := (e_in.cr(banum) or not e_in.cr(bbnum));
- when "0110" => -- CRXOR
- crresult := (e_in.cr(banum) xor e_in.cr(bbnum));
- when "1110" => -- CROR
- crresult := (e_in.cr(banum) or e_in.cr(bbnum));
- when others =>
- report "BAD CR?";
- end case;
+ -- Bits 5-8 of cr_op give the truth table of the requested
+ -- logical operation
+ cr_operands := e_in.cr(banum) & e_in.cr(bbnum);
+ crresult := cr_op(5 + to_integer(unsigned(cr_operands)));
v.e.write_cr_mask := num_to_fxm((31-btnum) / 4);
for i in 0 to 31 loop
if i = btnum then
end if;
end loop;
end if;
+ when OP_MFMSR =>
+ result := ctrl.msr;
+ result_en := '1';
when OP_MFSPR =>
+ report "MFSPR to SPR " & integer'image(decode_spr_num(e_in.insn)) &
+ "=" & to_hstring(a_in);
+ result_en := '1';
if is_fast_spr(e_in.read_reg1) then
- result := e_in.read_data1;
+ result := a_in;
if decode_spr_num(e_in.insn) = SPR_XER then
+ -- bits 0:31 and 35:43 are treated as reserved and return 0s when read using mfxer
+ result(63 downto 32) := (others => '0');
result(63-32) := v.e.xerc.so;
result(63-33) := v.e.xerc.ov;
result(63-34) := v.e.xerc.ca;
case decode_spr_num(e_in.insn) is
when SPR_TB =>
result := ctrl.tb;
- when others =>
- result := (others => '0');
+ when SPR_DEC =>
+ result := ctrl.dec;
+ when others =>
+ -- mfspr from unimplemented SPRs should be a nop in
+ -- supervisor mode and a program interrupt for user mode
+ result := c_in;
+ if ctrl.msr(MSR_PR) = '1' then
+ illegal := '1';
+ end if;
end case;
end if;
- result_en := '1';
when OP_MFCR =>
if e_in.insn(20) = '0' then
-- mfcr
crnum := fxm_to_num(insn_fxm(e_in.insn));
v.e.write_cr_mask := num_to_fxm(crnum);
end if;
- v.e.write_cr_data := e_in.read_data3(31 downto 0);
+ v.e.write_cr_data := c_in(31 downto 0);
+ when OP_MTMSRD =>
+ if e_in.insn(16) = '1' then
+ -- just update EE and RI
+ ctrl_tmp.msr(MSR_EE) <= c_in(MSR_EE);
+ ctrl_tmp.msr(MSR_RI) <= c_in(MSR_RI);
+ else
+ -- Architecture says to leave out bits 3 (HV), 51 (ME)
+ -- and 63 (LE) (IBM bit numbering)
+ ctrl_tmp.msr(63 downto 61) <= c_in(63 downto 61);
+ ctrl_tmp.msr(59 downto 13) <= c_in(59 downto 13);
+ ctrl_tmp.msr(11 downto 1) <= c_in(11 downto 1);
+ if c_in(MSR_PR) = '1' then
+ ctrl_tmp.msr(MSR_EE) <= '1';
+ ctrl_tmp.msr(MSR_IR) <= '1';
+ ctrl_tmp.msr(MSR_DR) <= '1';
+ end if;
+ end if;
when OP_MTSPR =>
report "MTSPR to SPR " & integer'image(decode_spr_num(e_in.insn)) &
- "=" & to_hstring(e_in.read_data3);
+ "=" & to_hstring(c_in);
if is_fast_spr(e_in.write_reg) then
- result := e_in.read_data3;
+ result := c_in;
result_en := '1';
if decode_spr_num(e_in.insn) = SPR_XER then
- v.e.xerc.so := e_in.read_data3(63-32);
- v.e.xerc.ov := e_in.read_data3(63-33);
- v.e.xerc.ca := e_in.read_data3(63-34);
- v.e.xerc.ov32 := e_in.read_data3(63-44);
- v.e.xerc.ca32 := e_in.read_data3(63-45);
+ v.e.xerc.so := c_in(63-32);
+ v.e.xerc.ov := c_in(63-33);
+ v.e.xerc.ca := c_in(63-34);
+ v.e.xerc.ov32 := c_in(63-44);
+ v.e.xerc.ca32 := c_in(63-45);
v.e.write_xerc_enable := '1';
end if;
else
--- TODO: Implement slow SPRs
--- case decode_spr_num(e_in.insn) is
--- when others =>
--- end case;
+ -- slow spr
+ case decode_spr_num(e_in.insn) is
+ when SPR_DEC =>
+ ctrl_tmp.dec <= c_in;
+ when others =>
+ -- mtspr to unimplemented SPRs should be a nop in
+ -- supervisor mode and a program interrupt for user mode
+ if ctrl.msr(MSR_PR) = '1' then
+ illegal := '1';
+ end if;
+ end case;
end if;
- when OP_POPCNTB =>
- result := ppc_popcntb(e_in.read_data3);
- result_en := '1';
- when OP_POPCNTW =>
- result := ppc_popcntw(e_in.read_data3);
+ when OP_POPCNT =>
+ result := popcnt_result;
result_en := '1';
- when OP_POPCNTD =>
- result := ppc_popcntd(e_in.read_data3);
+ when OP_PRTY =>
+ result := parity_result;
result_en := '1';
- when OP_PRTYD =>
- result := ppc_prtyd(e_in.read_data3);
- result_en := '1';
- when OP_PRTYW =>
- result := ppc_prtyw(e_in.read_data3);
- result_en := '1';
- when OP_RLC | OP_RLCL | OP_RLCR | OP_SHL | OP_SHR =>
+ when OP_RLC | OP_RLCL | OP_RLCR | OP_SHL | OP_SHR | OP_EXTSWSLI =>
result := rotator_result;
if e_in.output_carry = '1' then
set_carry(v.e, rotator_carry, rotator_carry);
end if;
result_en := '1';
- when OP_SIM_CONFIG =>
- -- bit 0 was used to select the microwatt console, which
- -- we no longer support.
- result := x"0000000000000000";
- result_en := '1';
-
- when OP_TDI =>
- -- Keep our test cases happy for now, ignore trap instructions
- report "OP_TDI FIXME";
when OP_ISYNC =>
f_out.redirect <= '1';
when OP_ICBI =>
icache_inval <= '1';
- when others =>
+ when OP_MUL_L64 | OP_MUL_H64 | OP_MUL_H32 =>
+ v.e.valid := '0';
+ v.mul_in_progress := '1';
+ stall_out <= '1';
+ x_to_multiply.valid <= '1';
+
+ when OP_DIV | OP_DIVE | OP_MOD =>
+ v.e.valid := '0';
+ v.div_in_progress := '1';
+ stall_out <= '1';
+ x_to_divider.valid <= '1';
+
+ when others =>
terminate_out <= '1';
report "illegal";
end case;
+ v.e.rc := e_in.rc and e_in.valid;
+
-- Update LR on the next cycle after a branch link
--
-- WARNING: The LR update isn't tracked by our hazard tracker. This
report "Delayed LR update to " & to_hstring(next_nia);
stall_out <= '1';
end if;
+
+ elsif e_in.valid = '1' then
+ -- instruction for other units, i.e. LDST
+ v.ldst_nia := e_in.nia;
+ v.e.valid := '0';
+ if e_in.unit = LDST then
+ lv.valid := '1';
+ end if;
+
elsif r.lr_update = '1' then
result_en := '1';
result := r.next_lr;
v.e.write_reg := fast_spr_num(SPR_LR);
- v.e.write_len := x"8";
- v.e.sign_extend := '0';
v.e.valid := '1';
+ elsif r.cntz_in_progress = '1' then
+ -- cnt[lt]z always takes two cycles
+ result := countzero_result;
+ result_en := '1';
+ v.e.write_reg := gpr_to_gspr(v.slow_op_dest);
+ v.e.rc := v.slow_op_rc;
+ v.e.xerc := v.slow_op_xerc;
+ v.e.valid := '1';
+ elsif r.mul_in_progress = '1' or r.div_in_progress = '1' then
+ if (r.mul_in_progress = '1' and multiply_to_x.valid = '1') or
+ (r.div_in_progress = '1' and divider_to_x.valid = '1') then
+ if r.mul_in_progress = '1' then
+ result := multiply_to_x.write_reg_data;
+ overflow := multiply_to_x.overflow;
+ else
+ result := divider_to_x.write_reg_data;
+ overflow := divider_to_x.overflow;
+ end if;
+ result_en := '1';
+ v.e.write_reg := gpr_to_gspr(v.slow_op_dest);
+ v.e.rc := v.slow_op_rc;
+ v.e.xerc := v.slow_op_xerc;
+ v.e.write_xerc_enable := v.slow_op_oe;
+ -- We must test oe because the RC update code in writeback
+ -- will use the xerc value to set CR0:SO so we must not clobber
+ -- xerc if OE wasn't set.
+ if v.slow_op_oe = '1' then
+ v.e.xerc.ov := overflow;
+ v.e.xerc.ov32 := overflow;
+ v.e.xerc.so := v.slow_op_xerc.so or overflow;
+ end if;
+ v.e.valid := '1';
+ else
+ stall_out <= '1';
+ v.mul_in_progress := r.mul_in_progress;
+ v.div_in_progress := r.div_in_progress;
+ end if;
+ end if;
+
+ if illegal = '1' then
+ exception := '1';
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#700#, 64));
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
+ -- Since we aren't doing Hypervisor emulation assist (0xe40) we
+ -- set bit 44 to indicate we have an illegal
+ ctrl_tmp.srr1(63 - 44) <= '1';
+ report "illegal";
+ end if;
+ if exception = '1' then
+ v.e.exc_write_enable := '1';
+ if exception_nextpc = '1' then
+ v.e.exc_write_data := next_nia;
+ end if;
+ ctrl_tmp.irq_state <= WRITE_SRR1;
+ v.e.valid := '1';
end if;
v.e.write_data := result;
v.e.write_enable := result_en;
- v.e.rc := e_in.rc and e_in.valid;
+
+ -- generate DSI or DSegI for load/store exceptions
+ -- or ISI or ISegI for instruction fetch exceptions
+ if l_in.exception = '1' then
+ ctrl_tmp.srr1 <= msr_copy(ctrl.msr);
+ if l_in.instr_fault = '0' then
+ if l_in.segment_fault = '0' then
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#300#, 64));
+ else
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#380#, 64));
+ end if;
+ else
+ if l_in.segment_fault = '0' then
+ ctrl_tmp.srr1(63 - 33) <= l_in.invalid;
+ ctrl_tmp.srr1(63 - 35) <= l_in.perm_error; -- noexec fault
+ ctrl_tmp.srr1(63 - 44) <= l_in.badtree;
+ ctrl_tmp.srr1(63 - 45) <= l_in.rc_error;
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#400#, 64));
+ else
+ ctrl_tmp.irq_nia <= std_logic_vector(to_unsigned(16#480#, 64));
+ end if;
+ end if;
+ v.e.exc_write_enable := '1';
+ v.e.exc_write_reg := fast_spr_num(SPR_SRR0);
+ v.e.exc_write_data := r.ldst_nia;
+ report "ldst exception writing srr0=" & to_hstring(r.ldst_nia);
+ ctrl_tmp.irq_state <= WRITE_SRR1;
+ v.e.valid := '1'; -- complete the original load or store
+ end if;
+
+ -- Outputs to loadstore1 (async)
+ lv.op := e_in.insn_type;
+ lv.nia := e_in.nia;
+ lv.addr1 := a_in;
+ lv.addr2 := b_in;
+ lv.data := c_in;
+ lv.write_reg := gspr_to_gpr(e_in.write_reg);
+ lv.length := e_in.data_len;
+ lv.byte_reverse := e_in.byte_reverse;
+ lv.sign_extend := e_in.sign_extend;
+ lv.update := e_in.update;
+ lv.update_reg := gspr_to_gpr(e_in.read_reg1);
+ lv.xerc := v.e.xerc;
+ lv.reserve := e_in.reserve;
+ lv.rc := e_in.rc;
+ lv.insn := e_in.insn;
+ -- decode l*cix and st*cix instructions here
+ if e_in.insn(31 downto 26) = "011111" and e_in.insn(10 downto 9) = "11" and
+ e_in.insn(5 downto 1) = "10101" then
+ lv.ci := '1';
+ end if;
+ lv.virt_mode := ctrl.msr(MSR_DR);
+ lv.priv_mode := not ctrl.msr(MSR_PR);
-- Update registers
rin <= v;
-- update outputs
--f_out <= r.f;
+ l_out <= lv;
e_out <= r.e;
flush_out <= f_out.redirect;
end process;