|
|
|
library ieee;
|
|
|
|
use ieee.std_logic_1164.all;
|
|
|
|
use ieee.numeric_std.all;
|
|
|
|
|
|
|
|
library work;
|
|
|
|
use work.common.all;
|
|
|
|
use work.crhelpers.all;
|
|
|
|
|
|
|
|
entity writeback is
|
|
|
|
port (
|
|
|
|
clk : in std_ulogic;
|
|
|
|
|
|
|
|
e_in : in Execute1ToWritebackType;
|
|
|
|
l_in : in Loadstore1ToWritebackType;
|
|
|
|
|
|
|
|
w_out : out WritebackToRegisterFileType;
|
|
|
|
c_out : out WritebackToCrFileType;
|
|
|
|
|
|
|
|
complete_out : out std_ulogic
|
|
|
|
);
|
|
|
|
end entity writeback;
|
|
|
|
|
|
|
|
architecture behaviour of writeback is
|
|
|
|
begin
|
|
|
|
writeback_1: process(all)
|
|
|
|
variable x : std_ulogic_vector(0 downto 0);
|
|
|
|
variable y : std_ulogic_vector(0 downto 0);
|
|
|
|
variable w : std_ulogic_vector(0 downto 0);
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
variable cf: std_ulogic_vector(3 downto 0);
|
|
|
|
variable zero : std_ulogic;
|
|
|
|
variable sign : std_ulogic;
|
|
|
|
variable scf : std_ulogic_vector(3 downto 0);
|
|
|
|
begin
|
|
|
|
x(0) := e_in.valid;
|
|
|
|
y(0) := l_in.valid;
|
|
|
|
assert (to_integer(unsigned(x)) + to_integer(unsigned(y))) <= 1 severity failure;
|
|
|
|
|
|
|
|
x(0) := e_in.write_enable;
|
|
|
|
y(0) := l_in.write_enable;
|
|
|
|
assert (to_integer(unsigned(x)) + to_integer(unsigned(y))) <= 1 severity failure;
|
|
|
|
|
|
|
|
w(0) := e_in.write_cr_enable;
|
|
|
|
x(0) := (e_in.write_enable and e_in.rc);
|
|
|
|
assert (to_integer(unsigned(w)) + to_integer(unsigned(x))) <= 1 severity failure;
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
|
|
|
|
w_out <= WritebackToRegisterFileInit;
|
|
|
|
c_out <= WritebackToCrFileInit;
|
|
|
|
|
|
|
|
complete_out <= '0';
|
|
|
|
if e_in.valid = '1' or l_in.valid = '1' then
|
|
|
|
complete_out <= '1';
|
|
|
|
end if;
|
|
|
|
|
|
|
|
if e_in.write_enable = '1' then
|
|
|
|
w_out.write_reg <= e_in.write_reg;
|
|
|
|
w_out.write_data <= e_in.write_data;
|
|
|
|
w_out.write_enable <= '1';
|
|
|
|
end if;
|
|
|
|
|
|
|
|
if e_in.write_cr_enable = '1' then
|
|
|
|
c_out.write_cr_enable <= '1';
|
|
|
|
c_out.write_cr_mask <= e_in.write_cr_mask;
|
|
|
|
c_out.write_cr_data <= e_in.write_cr_data;
|
|
|
|
end if;
|
|
|
|
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
if e_in.write_xerc_enable = '1' then
|
|
|
|
c_out.write_xerc_enable <= '1';
|
|
|
|
c_out.write_xerc_data <= e_in.xerc;
|
|
|
|
end if;
|
|
|
|
|
|
|
|
if l_in.write_enable = '1' then
|
|
|
|
w_out.write_reg <= gpr_to_gspr(l_in.write_reg);
|
|
|
|
w_out.write_data <= l_in.write_data;
|
|
|
|
w_out.write_enable <= '1';
|
|
|
|
end if;
|
|
|
|
|
|
|
|
if l_in.rc = '1' then
|
|
|
|
-- st*cx. instructions
|
|
|
|
scf(3) := '0';
|
|
|
|
scf(2) := '0';
|
|
|
|
scf(1) := l_in.store_done;
|
|
|
|
scf(0) := l_in.xerc.so;
|
|
|
|
c_out.write_cr_enable <= '1';
|
|
|
|
c_out.write_cr_mask <= num_to_fxm(0);
|
|
|
|
c_out.write_cr_data(31 downto 28) <= scf;
|
|
|
|
end if;
|
|
|
|
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
-- Perform CR0 update for RC forms
|
|
|
|
-- Note that loads never have a form with an RC bit, therefore this can test e_in.write_data
|
|
|
|
if e_in.rc = '1' and e_in.write_enable = '1' then
|
|
|
|
sign := e_in.write_data(63);
|
|
|
|
zero := not (or e_in.write_data);
|
|
|
|
c_out.write_cr_enable <= '1';
|
|
|
|
c_out.write_cr_mask <= num_to_fxm(0);
|
|
|
|
cf(3) := sign;
|
|
|
|
cf(2) := not sign and not zero;
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
cf(1) := zero;
|
|
|
|
cf(0) := e_in.xerc.so;
|
Add basic XER support
The carry is currently internal to execute1. We don't handle any of
the other XER fields.
This creates type called "xer_common_t" that contains the commonly
used XER bits (CA, CA32, SO, OV, OV32).
The value is stored in the CR file (though it could be a separate
module). The rest of the bits will be implemented as a separate
SPR and the two parts reconciled in mfspr/mtspr in latter commits.
We always read XER in decode2 (there is little point not to)
and send it down all pipeline branches as it will be needed in
writeback for all type of instructions when CR0:SO needs to be
updated (such forms exist for all pipeline branches even if we don't
yet implement them).
To avoid having to track XER hazards, we forward it back in EX1. This
assumes that other pipeline branches that can modify it (mult and div)
are running single issue for now.
One additional hazard to beware of is an XER:SO modifying instruction
in EX1 followed immediately by a store conditional. Due to our writeback
latency, the store will go down the LSU with the previous XER value,
thus the stcx. will set CR0:SO using an obsolete SO value.
I doubt there exist any code relying on this behaviour being correct
but we should account for it regardless, possibly by ensuring that
stcx. remain single issue initially, or later by adding some minimal
tracking or moving the LSU into the same pipeline as execute.
Missing some obscure XER affecting instructions like addex or mcrxrx.
[paulus@ozlabs.org - fix CA32 and OV32 for OP_ADD, fix order of
arguments to set_ov]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
5 years ago
|
|
|
c_out.write_cr_data(31 downto 28) <= cf;
|
|
|
|
end if;
|
|
|
|
end process;
|
|
|
|
end;
|