VHDL and Latch

[Weng Tianxiang]

Hi,
I am very confused with latch generation in VHDL.

1. I have been using VHDL for 7 years and I have never met a situation
I need a latch.

2. I want to know why VHDL let VHDL programmers guess what is to be
generated in the following situation that I know is only case a latch
may be generated:

process(a, ...)
begin
-- signalA <= '0';
case state is
when A_S =>
if(a = "000001") then
signalA <= '1'; <--- a latch is generated, why not
generating an error ?
state_ns <= B_S;
else
state_ns <= A_S;
end if;

....
end process;

Because the first line " Latch_A <= '0'; " is easily missed when the
process content is big, signalA is generated by VHDL definition as a
latch.

Here are my questions:
1. Latch is rarily used throughout all VHDL, why doesn't VHDL
introduce a latch() statement to specially be used for this purpose
while generating an error in the above process().

2. Here is a latch function definition true table:
Enable = 1:
CLK = H, D = H, OUT = H
CLK = H, D = L, OUT = L
CLK = L, D = X, OUT = Q0 (latched data).

It means when clock is high, data is transparent. Input is directed
into output.
When clock is low, the data is latched on falling edge of clock.

In the above process(), there is no clock specified. Which clock will
be used if there are multiple clocks in the design ?

3. In the above example, condition: K = (state = A_S and a = "000001")
should be true to set signalA. What is K role? If K is used as the
latch enable signal, half clock the latch is transparent and its
stored data would be destroyed if there is a glitch with K and could
not keep signalA true value unchanged.

4. I don't know how to design the latch for signalA with K input?

If you know the answers, please help.

Thank you.

Weng

 

[KJ]

Hi,
I am very confused with latch generation in VHDL.

1. I have been using VHDL for 7 years and I have never met a situation
I need a latch.

2. I want to know why VHDL let VHDL programmers guess what is to be
generated in the following situation that I know is only case a latch
may be generated:

process(a, ...)
begin
-- signalA <= '0';
case state is
when A_S =>
if(a = "000001") then
signalA <= '1'; <--- a latch is generated, why not
generating an error ?
state_ns <= B_S;
else
state_ns <= A_S;
end if;

...
end process;

Because the first line " Latch_A <= '0'; " is easily missed when the
process content is big, signalA is generated by VHDL definition as a
latch.

Here are my questions:
1. Latch is rarily used throughout all VHDL, why doesn't VHDL
introduce a latch() statement to specially be used for this purpose
while generating an error in the above process().

2. Here is a latch function definition true table:
Enable = 1:
CLK = H, D = H, OUT = H
CLK = H, D = L, OUT = L
CLK = L, D = X, OUT = Q0 (latched data).

It means when clock is high, data is transparent. Input is directed
into output.
When clock is low, the data is latched on falling edge of clock.

In the above process(), there is no clock specified. Which clock will
be used if there are multiple clocks in the design ?

3. In the above example, condition: K = (state = A_S and a = "000001")
should be true to set signalA. What is K role? If K is used as the
latch enable signal, half clock the latch is transparent and its
stored data would be destroyed if there is a glitch with K and could
not keep signalA true value unchanged.

4. I don't know how to design the latch for signalA with K input?

If you know the answers, please help.

Thank you.

Weng

 

[KJ]

Hi,
I am very confused with latch generation in VHDL.

1. I have been using VHDL for 7 years and I have never met a situation
I need a latch.

In FPGAs and CPLDs there is not much need for a latch because a flip flop
will do just fine.

2. I want to know why VHDL let VHDL programmers guess what is to be
generated in the following situation that I know is only case a latch
may be generated:

There is no guessing involved. Certain coding styles infer a latch, just
like others infer a flip flop and still others infer a block of memory.

process(a, ...)
begin
-- signalA <= '0';
case state is
when A_S =>
if(a = "000001") then
signalA <= '1'; <--- a latch is generated, why not
generating an error ?
state_ns <= B_S;
else
state_ns <= A_S;
end if;

...
end process;

Because the first line " Latch_A <= '0'; " is easily missed

Yeah, I missed it too. I don't even see anything called "Latch_A".

when the
process content is big, signalA is generated by VHDL definition as a
latch.

That's why many people (myself included) discourage the use of processes
that are not synchronously clocked (i.e. the sensitivity consists of one
signal...'Clock'). Doing so completely avoids the above coding situation
which can (if you're not careful) infer an unintended latch.

Here are my questions:
1. Latch is rarily used throughout all VHDL, why doesn't VHDL
introduce a latch() statement to specially be used for this purpose
while generating an error in the above process().

Maybe you should read what you wrote again. "Latch is rarily used ..." and
"introduce a latch() statement to...". The obvious questions here are
- Why clutter the language for something you claim would rarely be used?
- For those that do use latches, are you suggesting that there code should
no longer be accepted, resulting in an error?

My guess is that there would not be much support for a 'rarely used'
addition that also breaks legacy code. If you want such a function then
VHDL allows you to write it yourself and incorporate it wherever you like
inside your own code.

2. Here is a latch function definition true table:
Enable = 1:
CLK = H, D = H, OUT = H
CLK = H, D = L, OUT = L
CLK = L, D = X, OUT = Q0 (latched data).

It means when clock is high, data is transparent. Input is directed
into output.

If you say so, it leaves a bit to the imagination, like why is the magic
signal 'Q0' the latched version of 'OUT'....personally I find the following
to be much more descriptive and easy to read if I ever did want to infer a
transparent latch.
if (CLK = '1') then
Q <= D; -- Chose not to call it 'OUT' as you did so as to not use a
VHDL keyword
end if;

When clock is low, the data is latched on falling edge of clock.

The falling edge of a clock that is low? There's another logic oddity.

In the above process(), there is no clock specified. Which clock will
be used if there are multiple clocks in the design ?

VHDL doesn't have 'clocks', it has 'signals'. The logic is almost
completely specified by the plain text code that is written. The only thing
VHDL leaves to the imagination (i.e. it's in the LRM and you must learn
this) is that if you don't specify the value for a signal then it will
remain at it's current value. In the above mentioned latch process that I
wrote, if the 'if condition' is not met (i.e. "CLK = '1'" is not true),
since I haven't specified anything for the 'else' branch, the VHDL LRM says
that signal 'Q' would remain unchanged.

Much like the say all software languages are defined I might add so it's not
at all peculiar.

3. In the above example, condition: K = (state = A_S and a = "000001")
should be true to set signalA. What is K role? If K is used as the
latch enable signal, half clock the latch is transparent and its
stored data would be destroyed if there is a glitch with K and could
not keep signalA true value unchanged.

4. I don't know how to design the latch for signalA with K input?

If you know the answers, please help.

1. Don't use latches.
2. Don't use coding styles that can result in unintended latches.
3. Concentrate on the functionality and performance of your design while
remaining within whatever I/O, logic resource, power, etc.constraints that
your design might have.

Kevin Jennings

 

[Jim Lewis]

Weng,
First, 1076.6-2004 introduces an attribute named combinational.
For a compliant tool, when this attribute is applied to a
process label and the process creates a latch then a synthesis
tool shall generate an error.

So request that your vendor implement the standard (if
they have not already). Vendors will implement what their
user community wants. If you want this, you must make sure
you talk to your vendor.

2. I want to know why VHDL let VHDL programmers guess what is to be
generated in the following situation that I know is only case a latch
may be generated:

process(a, ...)
begin
-- signalA <= '0';
case state is
when A_S =>
if(a = "000001") then
signalA <= '1'; <--- a latch is generated, why not
generating an error ?
state_ns <= B_S;
else
state_ns <= A_S;
end if;

...
end process;

Because the first line " Latch_A <= '0'; " is easily missed when the
process content is big, signalA is generated by VHDL definition as a
latch.

OTOH, one simple rule for preventing latches:
* Initialize all outputs (except state_ns) to their
inactive value.
* Either fully specify state_ns or assign it to state_reg.

Or you can follow KJ's rule of always using clocked
processes, however, this is not my preference.


Cheers,
Jim

 

[comp.arch.fpga]

1. Latch is rarily used throughout all VHDL,

That is not true.
While latches are not supported in most modern FPGAs and where
discouraged to use in older families,
in ASIC design it is very common to split D-Flip-Flops into two
latches distributed through
the pipeline. Also, VHDL ist not only used for synthesis. It was
intended to model systems and there
clearly are systems that behave like a latch so there must be a way to
model them.

Most synthesis tools issue a warning in the cases that you describe
which is an adequate reaction to
the situation: The tool accountered legal code of which it knows that
it is often written unintentionally.

Kolja Sulimma

 

[KJ]

That is not true.
While latches are not supported in most modern FPGAs and where
discouraged to use in older families,
in ASIC design it is very common to split D-Flip-Flops into two
latches distributed through
the pipeline.

Kolja, good point.

To follow up on this point a bit for the other readers. The reason that
latches are discouraged FPGA/CPLDs but not so in ASICs is because in the
ASIC world they actually have a hard set of logic that gets inferred from
the code that generates a latch whereas in the typical FPGA/CPLD world you
do not. Instead the latch gets created by cobbling together the LUTs or
macrocells. The problem with the cobbling together approach is that you
have virtually no control over the timing inside the FPGA/CPLD and yet a
latch will have setup and hold timing requirements that you will have no way
to guarantee.

If FPGA/CPLDs cobbled together LUTs to create flip flops the same argument
could be made for not using flip flops. But since FPGA/CPLD/ASICs all have
flip flops implemented as hard logic you don't have this issue. Also, if
the target device does have a hard latch as a resource that can be used then
the use of latches is just fine also.

Kevin Jennings

 

[Ralf Hildebrandt]

Weng Tianxiang schrieb:

I am very confused with latch generation in VHDL.

1. I have been using VHDL for 7 years and I have never met a situation
I need a latch.

I do it every day. Latches are only half as big as flipflops and are not
triggered by a highly active clock. They only need their enable signal.
This makes them then 1st choice for small low-power circuits.

Using latches means thinking very well about the behavior of the
circuit. Hazards, the common muxed-latch-problem and other stuff has to
be considered.

2. I want to know why VHDL let VHDL programmers guess what is to be
generated in the following situation that I know is only case a latch
may be generated:

process(a, ...)
begin
-- signalA <= '0';
case state is
when A_S =>
if(a = "000001") then
signalA <= '1'; <--- a latch is generated, why not
generating an error ?
state_ns <= B_S;
else
state_ns <= A_S;
end if;

...
end process;

signalA is not assigned in the else-branch. Therefore you get a latch,
because this process is not clocked.
This is a bad way to code a latch (because you easily trap into the
muxed-latch pitfall) but why this should be an error? All synthesis
tools will report that signalA has inferred a latch.

Because the first line " Latch_A <= '0'; " is easily missed when the
process content is big, signalA is generated by VHDL definition as a
latch.

If for the 1st line after the begin the comment "--" is removed you will
get combinational logic for signalA. This is _totally_ different
behavior, but if you want to avoid a latch a very clean and easy way to
do it.

Ralf

 

[Weng Tianxiang]

Weng,
First, 1076.6-2004 introduces an attribute named combinational.
For a compliant tool, when this attribute is applied to a
process label and the process creates a latch then a synthesis
tool shall generate an error.

So request that your vendor implement the standard (if
they have not already). Vendors will implement what their
user community wants. If you want this, you must make sure
you talk to your vendor.









OTOH, one simple rule for preventing latches:
* Initialize all outputs (except state_ns) to their
inactive value.
* Either fully specify state_ns or assign it to state_reg.

Or you can follow KJ's rule of always using clocked
processes, however, this is not my preference.

Cheers,
Jim- Hide quoted text -

- Show quoted text -

Hi Jim,
Thank you for your good information.

If a compiler does have the attribute capability, it would greatly
easy the trouble a lot. For almost any state machine, I have been
always using 2-processe method, it is clear and very informative, the
only drawback is the missing of signal initial value assignment. With
your new definition, I am very encouraged that I expect Xilinx compier
will have the ability soon.

Weng

 

[Weng Tianxiang]

In FPGAs and CPLDs there is not much need for a latch because a flip flop
will do just fine.




There is no guessing involved. Certain coding styles infer a latch, just
like others infer a flip flop and still others infer a block of memory.










Yeah, I missed it too. I don't even see anything called "Latch_A".


That's why many people (myself included) discourage the use of processes
that are not synchronously clocked (i.e. the sensitivity consists of one
signal...'Clock'). Doing so completely avoids the above coding situation
which can (if you're not careful) infer an unintended latch.




Maybe you should read what you wrote again. "Latch is rarily used ..." and
"introduce a latch() statement to...". The obvious questions here are
- Why clutter the language for something you claim would rarely be used?
- For those that do use latches, are you suggesting that there code should
no longer be accepted, resulting in an error?

My guess is that there would not be much support for a 'rarely used'
addition that also breaks legacy code. If you want such a function then
VHDL allows you to write it yourself and incorporate it wherever you like
inside your own code.





If you say so, it leaves a bit to the imagination, like why is the magic
signal 'Q0' the latched version of 'OUT'....personally I find the following
to be much more descriptive and easy to read if I ever did want to infer a
transparent latch.
if (CLK = '1') then
Q <= D; -- Chose not to call it 'OUT' as you did so as to not use a
VHDL keyword
end if;


The falling edge of a clock that is low? There's another logic oddity.




VHDL doesn't have 'clocks', it has 'signals'. The logic is almost
completely specified by the plain text code that is written. The only thing
VHDL leaves to the imagination (i.e. it's in the LRM and you must learn
this) is that if you don't specify the value for a signal then it will
remain at it's current value. In the above mentioned latch process that I
wrote, if the 'if condition' is not met (i.e. "CLK = '1'" is not true),
since I haven't specified anything for the 'else' branch, the VHDL LRM says
that signal 'Q' would remain unchanged.

Much like the say all software languages are defined I might add so it's not
at all peculiar.






1. Don't use latches.
2. Don't use coding styles that can result in unintended latches.
3. Concentrate on the functionality and performance of your design while
remaining within whatever I/O, logic resource, power, etc.constraints that
your design might have.

Kevin Jennings- Hide quoted text -

- Show quoted text -

 

[Weng Tianxiang]

In FPGAs and CPLDs there is not much need for a latch because a flip flop
will do just fine.




There is no guessing involved. Certain coding styles infer a latch, just
like others infer a flip flop and still others infer a block of memory.










Yeah, I missed it too. I don't even see anything called "Latch_A".


That's why many people (myself included) discourage the use of processes
that are not synchronously clocked (i.e. the sensitivity consists of one
signal...'Clock'). Doing so completely avoids the above coding situation
which can (if you're not careful) infer an unintended latch.




Maybe you should read what you wrote again. "Latch is rarily used ..." and
"introduce a latch() statement to...". The obvious questions here are
- Why clutter the language for something you claim would rarely be used?
- For those that do use latches, are you suggesting that there code should
no longer be accepted, resulting in an error?

My guess is that there would not be much support for a 'rarely used'
addition that also breaks legacy code. If you want such a function then
VHDL allows you to write it yourself and incorporate it wherever you like
inside your own code.





If you say so, it leaves a bit to the imagination, like why is the magic
signal 'Q0' the latched version of 'OUT'....personally I find the following
to be much more descriptive and easy to read if I ever did want to infer a
transparent latch.
if (CLK = '1') then
Q <= D; -- Chose not to call it 'OUT' as you did so as to not use a
VHDL keyword
end if;


The falling edge of a clock that is low? There's another logic oddity.




VHDL doesn't have 'clocks', it has 'signals'. The logic is almost
completely specified by the plain text code that is written. The only thing
VHDL leaves to the imagination (i.e. it's in the LRM and you must learn
this) is that if you don't specify the value for a signal then it will
remain at it's current value. In the above mentioned latch process that I
wrote, if the 'if condition' is not met (i.e. "CLK = '1'" is not true),
since I haven't specified anything for the 'else' branch, the VHDL LRM says
that signal 'Q' would remain unchanged.

Much like the say all software languages are defined I might add so it's not
at all peculiar.






1. Don't use latches.
2. Don't use coding styles that can result in unintended latches.
3. Concentrate on the functionality and performance of your design while
remaining within whatever I/O, logic resource, power, etc.constraints that
your design might have.

Kevin Jennings- Hide quoted text -

- Show quoted text -

Hi Kevin,
In my first posting, there are 3 questions:
1. Why doesn't compiler generate an error for unintended latch
generation?
This question is satisfactorily answered: there is a new attribute to
give errors if an unintended latch is to generate. It says that some
other people had already observed the situation long time ago.

2. If the above error is given an error, VHDL may include a special
statement to generate a latch.
This is another big problem: I don't know how ASIC people generate a
real latch using VHDL? I think they may most likely use latch library
to generate special latch instead of using VHDL statements. I read an
article about asynchronous FIFO written by two engineers one of whom
is Pete Alfke of Xilinx (it is the best article I have read in my
life). In the paper they say that they fail to generate two or 3
latches in their design using Xilinx chip. If so, it seems to me that
there is no reliable statement in VHDL to generate a latch for a FPGA
chip.

3. If the example would generate a latch for signalA, how it is
generated?

KJ answered the question. If the equation KJ suggested is true, it
would like the following:
if (state = stateA_S and a = "000001") then
signalA <= '1';
end if;

Finally I realized KJ saying is correct.

Thank you, KJ and Lewis.

Weng

 

[KJ]

Hi Kevin,
In my first posting, there are 3 questions:
1. Why doesn't compiler generate an error for unintended latch
generation?

Simply put because it is not an 'error' to write code that infers a latch.
To presume that every latch is 'unintended' would be wrong.

This question is satisfactorily answered: there is a new attribute to
give errors if an unintended latch is to generate. It says that some
other people had already observed the situation long time ago.

If you choose to use that attribute and find a tool that supports it then by
all means use it. But keep in mind that if you forget to use the new
attribute (because people are human) that you can still end up in the same
situation. It's up to you if that's the coding style you prefer to use to
avoid unintended latches, personally I prefer other coding styles but the
attribute is at least an improvement and gives another method which is good.

No matter which of these two methods you use, you can still get an
'unintended latch' with the following type of code which infers a latch

Q <= (C and D) or (not(C) and Q);

This is obviously a very stripped down example, a 'real' design might be
much more complicated, but you have to train yourself to look at every
concurrent statement and make sure that you don't see the output on both
sides of the equation.

It gets worse though because the 'latch' could be a function of multiple
concurrent statements that happen to loop back. The synthesizer report that
it finds a combinatorial loop works to catch that but nothing in the
language prevents it (nor should it, since it may not be 'unintended').

2. If the above error is given an error, VHDL may include a special
statement to generate a latch.

If you want a 'special' statement that creates a latch, then write one and
instantiate it until your heart is content....but don't presume that
everyone else agrees with you and that the language itself should only
support your coding style for creating a latch and any other method should
cause an error.

This is another big problem: I don't know how ASIC people generate a
real latch using VHDL? I think they may most likely use latch library
to generate special latch instead of using VHDL statements.

That doesn't preclude them from inferring a latch in other ways.

I read an
article about asynchronous FIFO written by two engineers one of whom
is Pete Alfke of Xilinx (it is the best article I have read in my
life). In the paper they say that they fail to generate two or 3
latches in their design using Xilinx chip.

No idea what you mean by this...but reading on

If so, it seems to me that
there is no reliable statement in VHDL to generate a latch for a FPGA
chip.

Unless the underlying hardware has a hard latch available as a resource to
use and most (if not all) FPGAs do not, then I would agree that there is no
reliable way to generate a latch inside an FPGA in any language, not just
VHDL. The lack of latches in FPGAs though has nothing at all to do with
VHDL or any language it is simply recognition by the FPGA suppliers that
there is very little demand for a hard latch inside the FPGA. A flip flop
would have the exact same coding issues, the only reason it's not a
'problem' is because there are hard flip flops inside the FPGA.

The 'problem' with inferring any storage element (latch, flip flop, big
hunk-o-memory, fifo, etc.) is when the underlying hardware doesn't directly
support the inferred element and it needs to be cobbled together from basic
logic blocks. In the FPGA/CPLD world, it is very difficult to meet internal
setup, hold time and 'no glitching' requirements between two arbitrary
signals.

Remember also, that you can build ANY logical element strictly from NAND
gates or NOR gates....so in theory you could write your code using only
these two building blocks. Were you to do so and target an FPGA, it would
most certainly fail if that design had any sort of storage.

3. If the example would generate a latch for signalA, how it is
generated?

Same as any other logical description....it gets mapped to whatever internal
resources exist in the device (LUTs, macrocells, flip flops, latches,
whatever the target device supports).

KJ answered the question. If the equation KJ suggested is true, it
would like the following:
if (state = stateA_S and a = "000001") then
signalA <= '1';
end if;

Finally I realized KJ saying is correct.

You're welcome.

Kevin Jennings

 

[Weng Tianxiang]

Simply put because it is not an 'error' to write code that infers a latch.
To presume that every latch is 'unintended' would be wrong.


If you choose to use that attribute and find a tool that supports it then by
all means use it. But keep in mind that if you forget to use the new
attribute (because people are human) that you can still end up in the same
situation. It's up to you if that's the coding style you prefer to use to
avoid unintended latches, personally I prefer other coding styles but the
attribute is at least an improvement and gives another method which is good.

No matter which of these two methods you use, you can still get an
'unintended latch' with the following type of code which infers a latch

Q <= (C and D) or (not(C) and Q);

This is obviously a very stripped down example, a 'real' design might be
much more complicated, but you have to train yourself to look at every
concurrent statement and make sure that you don't see the output on both
sides of the equation.

It gets worse though because the 'latch' could be a function of multiple
concurrent statements that happen to loop back. The synthesizer report that
it finds a combinatorial loop works to catch that but nothing in the
language prevents it (nor should it, since it may not be 'unintended').


If you want a 'special' statement that creates a latch, then write one and
instantiate it until your heart is content....but don't presume that
everyone else agrees with you and that the language itself should only
support your coding style for creating a latch and any other method should
cause an error.


That doesn't preclude them from inferring a latch in other ways.


No idea what you mean by this...but reading on


Unless the underlying hardware has a hard latch available as a resource to
use and most (if not all) FPGAs do not, then I would agree that there is no
reliable way to generate a latch inside an FPGA in any language, not just
VHDL. The lack of latches in FPGAs though has nothing at all to do with
VHDL or any language it is simply recognition by the FPGA suppliers that
there is very little demand for a hard latch inside the FPGA. A flip flop
would have the exact same coding issues, the only reason it's not a
'problem' is because there are hard flip flops inside the FPGA.

The 'problem' with inferring any storage element (latch, flip flop, big
hunk-o-memory, fifo, etc.) is when the underlying hardware doesn't directly
support the inferred element and it needs to be cobbled together from basic
logic blocks. In the FPGA/CPLD world, it is very difficult to meet internal
setup, hold time and 'no glitching' requirements between two arbitrary
signals.

Remember also, that you can build ANY logical element strictly from NAND
gates or NOR gates....so in theory you could write your code using only
these two building blocks. Were you to do so and target an FPGA, it would
most certainly fail if that design had any sort of storage.




Same as any other logical description....it gets mapped to whatever internal
resources exist in the device (LUTs, macrocells, flip flops, latches,
whatever the target device supports).





You're welcome.

Kevin Jennings

Hi KJ,
1. Q <= (C and D) or (not(C) and Q);
I have never seen such equation in my coding experiences and have no
idea how this equation would be written. The logic result is beyond a
reason. Could you please write it in equivalent latch equation in
informal VHDL?

2. FPGA of Xilinx chip really has latch primative and one may use it
using latch primative to call it. But it is hard to refer to it in
VHDL. I don't mean VHDL should have included latch statement, what I
mean is VHDL really lacks the statement element to refer to a latch in
a clear and reliable way and the lack can be easily corrected.

Thank you.

Weng

 

[Daniel S.]

Hi KJ,
1. Q <= (C and D) or (not(C) and Q);
I have never seen such equation in my coding experiences and have no
idea how this equation would be written. The logic result is beyond a
reason. Could you please write it in equivalent latch equation in
informal VHDL?

That equation is written exactly as-is and would work in theory.
Glitches on Q while C goes from 1 to 0 may break the feedback loop or
make it metastable in practice.

Another VHDL way of putting it would be:

Q <= D when C = '1' else Q;

An equivalent VHDL process would be:

process(c,d)
begin
if(c='1') then
q <= d;
end if;
end process;

Exactly what each of these would actually infer is tool- and
hardware-specific. The only thing you can do is check your synthesis
reports to determine exactly what was synthesized. This applies to all
matters of getting HDL to map onto specific FPGA resources.

2. FPGA of Xilinx chip really has latch primative and one may use it
using latch primative to call it. But it is hard to refer to it in
VHDL. I don't mean VHDL should have included latch statement, what I
mean is VHDL really lacks the statement element to refer to a latch in
a clear and reliable way and the lack can be easily corrected.

Latches: "if somesig='1' then ... end if;"
Latches(2): see equations and process above

FFs: "if clk='1' and clk'event then ... end if;"
FFs(2): "if rising_edge(clk) then ... end if;"

That seems like plenty straight-forward VHDL to me. Latches are active
on levels, FFs are active on edges. Simple.

Xilinx devices do not have dedicated latches BUT they do allow you to
program a slice FF to bypass HALF of it so it can be used as a latch. As
far as power is concerned though, the whole FF is still connected to the
clock so you most likely won't be saving any power here. BTW, the
configuration logic that goes behind FPGA FFs uses more surface area
than the FF itself and the "latch bypass logic" probably uses nearly as
much power as a pure FF.

To reach high clock speeds, FPGAs and ASICs need their pipeline storage
elements to be able to forward values from D to Q with as little setup,
hold and settling time as possible to maximize timing margins between
clock edges. This is most effortlessly and reliably achieved with FFs,
hence the dominance of FFs in the world of high-speed digital systems
and the FF-oriented nature of FPGAs even though they cost a little more
surface area and power.

Simply put, FFs are the path of least unnecessary pain.

If you want lower power FPGAs, some flexibility has to be sacrificed. If
Xilinx stripped the latch option from slice FFs, it would reduce the
slice FF cells' power and size by 15-20% while making them a little bit
faster. I wonder how many would have a problem with this.

 

[KJ]

1. Q <= (C and D) or (not(C) and Q);
I have never seen such equation in my coding experiences and have no
idea how this equation would be written. The logic result is beyond a
reason. Could you please write it in equivalent latch equation in
informal VHDL?

It's the same old transparent latch that we've been talking about all along.
The following forms are exactly equivalent logic and will result in the
exact same synthesized result.
#1 -- Traditional form of writing a latch
process(C, D)
begin
if (C='1') then
Q <= D;
end if;
end process;

#2 -- Another traditional form of writing a latch.
Q <= D when (C = '1');

#3 -- Latch equations written as a sum of products
Q <= (C and D) or (not(C) and Q);

#4 -- Latch equation written in a sum of products form that includes a
'cover' term
Q <= (C and D) or (not(C) and Q) or (D and Q);

Try all four of the above out and run them through synthesis and you should
get the exact same fitted equations for all 4.

Of the four different forms, only #4 is written in a way that will not have
timing issues that are caused simply by the implementation (over which you
have no real control in an FPGA). However, the last term ".. or (D and Q)"
is redundant and any synthesizer worth a dime will 'optomize the logic' and
implement it as form #3. This 'optomization' though creates a logic hazard
when 'C' switches from '1' to '0'. This logic hazard will in turn show up
as a failure in the latch operation. Specifically, what you'll run into is
that even if you have 'D' set up oodles of time before 'C' switches from '1'
to '0', the output 'Q' can latch itself at either '1' or '0' (depending on
the targetted part) because of differences in prop delay in computing the
first two or terms or glitches in 'C'. That last "D and Q" term is
absolutely required in order to even have a shot at inferring a latch that
will actually work but since it is logically 'redundant' it will always get
optomized away. If you hark back to Boolean Logic 101, you may recall that
adding redundant terms is the method you employ to get rid of logic hazards.
The FPGA LUT implementation leaves even more to the imagination as you
wonder just what those RAM outputs are doing with this combinatorial loop
and whether 'Q' will glitch and cause even more havoc since it would feed
back to the address input of the LUT.

The bottom line is
- There are various forms of writing source code that infers a latch.
- Whether a latch is a 'good' thing or a 'bad' thing depends on the
targetted device itself and whether that device actually has a hard latch
and a way to guarantee the timing and glitch requirements of 'D' and 'C'.
- Inferring any storage element (flops, latches, memory) needs to be done in
such a way that the storage is not implemented as a collection of
LUTs/macrocells but instead uses hard storage resources of the device.

2. FPGA of Xilinx chip really has latch primative and one may use it
using latch primative to call it. But it is hard to refer to it in
VHDL.

You'll still have trouble meeting and guaranteeing the timing and glitch
requirements on 'D' relative to 'C' in order to get it to work correctly for
arbitrary signals 'D' and 'C'. If it does work, it likely won't scale well
(i.e. if you try to infer thousands of these as you would like to)

I don't mean VHDL should have included latch statement, what I
mean is VHDL really lacks the statement element to refer to a latch in
a clear and reliable way and the lack can be easily corrected.

I've shown above four different forms for inferring the exact same latch
(i.e. synthesizes to the exact same thing). Which form you prefer (or if
you prefer to make up your own) to use is up to you. Making up a new method
though will not prohibit the earlier forms from being used and outlawing the
other forms gets in the way of people who do use latches and are doing so in
a way that they can control the various issues so you're not helping them by
outlawing their method.

Kevin Jennings

 

[Weng Tianxiang]

It's the same old transparent latch that we've been talking about all along.
The following forms are exactly equivalent logic and will result in the
exact same synthesized result.
#1 -- Traditional form of writing a latch
process(C, D)
begin
if (C='1') then
Q <= D;
end if;
end process;

#2 -- Another traditional form of writing a latch.
Q <= D when (C = '1');

#3 -- Latch equations written as a sum of products
Q <= (C and D) or (not(C) and Q);

#4 -- Latch equation written in a sum of products form that includes a
'cover' term
Q <= (C and D) or (not(C) and Q) or (D and Q);

Try all four of the above out and run them through synthesis and you should
get the exact same fitted equations for all 4.

Of the four different forms, only #4 is written in a way that will not have
timing issues that are caused simply by the implementation (over which you
have no real control in an FPGA). However, the last term ".. or (D and Q)"
is redundant and any synthesizer worth a dime will 'optomize the logic' and
implement it as form #3. This 'optomization' though creates a logic hazard
when 'C' switches from '1' to '0'. This logic hazard will in turn show up
as a failure in the latch operation. Specifically, what you'll run into is
that even if you have 'D' set up oodles of time before 'C' switches from '1'
to '0', the output 'Q' can latch itself at either '1' or '0' (depending on
the targetted part) because of differences in prop delay in computing the
first two or terms or glitches in 'C'. That last "D and Q" term is
absolutely required in order to even have a shot at inferring a latch that
will actually work but since it is logically 'redundant' it will always get
optomized away. If you hark back to Boolean Logic 101, you may recall that
adding redundant terms is the method you employ to get rid of logic hazards.
The FPGA LUT implementation leaves even more to the imagination as you
wonder just what those RAM outputs are doing with this combinatorial loop
and whether 'Q' will glitch and cause even more havoc since it would feed
back to the address input of the LUT.

The bottom line is
- There are various forms of writing source code that infers a latch.
- Whether a latch is a 'good' thing or a 'bad' thing depends on the
targetted device itself and whether that device actually has a hard latch
and a way to guarantee the timing and glitch requirements of 'D' and 'C'.
- Inferring any storage element (flops, latches, memory) needs to be done in
such a way that the storage is not implemented as a collection of
LUTs/macrocells but instead uses hard storage resources of the device.




You'll still have trouble meeting and guaranteeing the timing and glitch
requirements on 'D' relative to 'C' in order to get it to work correctly for
arbitrary signals 'D' and 'C'. If it does work, it likely won't scale well
(i.e. if you try to infer thousands of these as you would like to)


I've shown above four different forms for inferring the exact same latch
(i.e. synthesizes to the exact same thing). Which form you prefer (or if
you prefer to make up your own) to use is up to you. Making up a new method
though will not prohibit the earlier forms from being used and outlawing the
other forms gets in the way of people who do use latches and are doing so in
a way that they can control the various issues so you're not helping them by
outlawing their method.

Kevin Jennings

Hi KJ,
Thank you very much for your points.

Your answer really opens my eye and expand my knowledge range in FPGA
and VHDL !!!

Weng

 

[Weng Tianxiang]

It's the same old transparent latch that we've been talking about all along.
The following forms are exactly equivalent logic and will result in the
exact same synthesized result.
#1 -- Traditional form of writing a latch
process(C, D)
begin
if (C='1') then
Q <= D;
end if;
end process;

#2 -- Another traditional form of writing a latch.
Q <= D when (C = '1');

#3 -- Latch equations written as a sum of products
Q <= (C and D) or (not(C) and Q);

#4 -- Latch equation written in a sum of products form that includes a
'cover' term
Q <= (C and D) or (not(C) and Q) or (D and Q);

Try all four of the above out and run them through synthesis and you should
get the exact same fitted equations for all 4.

Of the four different forms, only #4 is written in a way that will not have
timing issues that are caused simply by the implementation (over which you
have no real control in an FPGA). However, the last term ".. or (D and Q)"
is redundant and any synthesizer worth a dime will 'optomize the logic' and
implement it as form #3. This 'optomization' though creates a logic hazard
when 'C' switches from '1' to '0'. This logic hazard will in turn show up
as a failure in the latch operation. Specifically, what you'll run into is
that even if you have 'D' set up oodles of time before 'C' switches from '1'
to '0', the output 'Q' can latch itself at either '1' or '0' (depending on
the targetted part) because of differences in prop delay in computing the
first two or terms or glitches in 'C'. That last "D and Q" term is
absolutely required in order to even have a shot at inferring a latch that
will actually work but since it is logically 'redundant' it will always get
optomized away. If you hark back to Boolean Logic 101, you may recall that
adding redundant terms is the method you employ to get rid of logic hazards.
The FPGA LUT implementation leaves even more to the imagination as you
wonder just what those RAM outputs are doing with this combinatorial loop
and whether 'Q' will glitch and cause even more havoc since it would feed
back to the address input of the LUT.

The bottom line is
- There are various forms of writing source code that infers a latch.
- Whether a latch is a 'good' thing or a 'bad' thing depends on the
targetted device itself and whether that device actually has a hard latch
and a way to guarantee the timing and glitch requirements of 'D' and 'C'.
- Inferring any storage element (flops, latches, memory) needs to be done in
such a way that the storage is not implemented as a collection of
LUTs/macrocells but instead uses hard storage resources of the device.




You'll still have trouble meeting and guaranteeing the timing and glitch
requirements on 'D' relative to 'C' in order to get it to work correctly for
arbitrary signals 'D' and 'C'. If it does work, it likely won't scale well
(i.e. if you try to infer thousands of these as you would like to)


I've shown above four different forms for inferring the exact same latch
(i.e. synthesizes to the exact same thing). Which form you prefer (or if
you prefer to make up your own) to use is up to you. Making up a new method
though will not prohibit the earlier forms from being used and outlawing the
other forms gets in the way of people who do use latches and are doing so in
a way that they can control the various issues so you're not helping them by
outlawing their method.

Kevin Jennings

Hi KJ,
I have another question for you.

1. Is any combinational equation with target signal in both right part
and left part of the equation a latch like your equations 3 or 4 show?

2. Is there any general algorithm to change such an equation to a
latch equation?

Thank you.

Weng

 

[Jim Lewis]

Weng

Hi KJ,
I have another question for you.

1. Is any combinational equation with target signal in both right part
and left part of the equation a latch like your equations 3 or 4 show?

A latch is the most benign form.
There are oscillators (however when they occur by
accident, they are not necessarily this obvious):
Q <= not Q ;

There is also other strange latch-like behavior.
I don't have an equation for you, but I do remember
analyzing one - it was very interesting.

Cheers,
Jim

 

[Peter Alfke]

I think we will agree that a latch needs positive feedback from the
output to at least one of the inputs.
If the feedback is negative (Qbar to D, for example) you end up with
an oscillator.
Peter Alfke

 

[KJ]

Hi KJ,
I have another question for you.

1. Is any combinational equation with target signal in both right part
and left part of the equation a latch like your equations 3 or 4 show?

While Jim and Peter can debate whether or not it's a latch or an
oscillator, suffice it to say that any time you have something of the
form of equations #3 and #4 where you have the same signal on both
sides of the '<=' in a concurrent statement you have some form of
storage or memory being inferred that will cause you the exact same
timing/glitch/hazard issues to deal with as you would if you have
something that matches exactly #3 and #4.

The important thing to take away is not whether it's a latch or an
oscillator or anything else, just that red flags should go off in your
head when you see the output being used on the right side of the
equation. Whether this is a problem or not depends on what the target
device is. In ASICs this will not be a problem, in FPGAs you have to
be very careful or it will be a problem.

2. Is there any general algorithm to change such an equation to a
latch equation?

Not sure what you mean....all four forms that I presented are exactly
equivalent 'latch equations', they will produce the same synthesized
result. Equations #1 and #2 involve less typing then #3 and #4.

Also keep in mind, that latches do not need to fit into a single
equation, any time that you form a combinatorial loop you have to be
concerned about how this will get implemented. A combinatorial loop
happens when you have multiple equations, none of which inherently
shows any feedback but taken together the whole set does. A simple
example of this would be

#5
Q <= (C and X) or (not(C) and Y); -- Eqn. 5a
Y <= Q and Z; -- Eqn. 5b

Equation 5a, on the surface appears to be a 2->1 multiplexer where 'C'
is simply used to select between two signals 'X' and 'Y'.

Equation 5b, on the surface appears to be a simple anding of two
signals 'Q' and 'Z'.

But when you put them together the two form a combinatorial loop
because 'Q' depends on 'Y' and 'Y' depends on 'Q'. When you run
through synthesis, these things generally get flagged as warnings
somewhere. The nice thing is that it will find all these loops no
matter how complex the equations or how spread out they are as long as
they are totally self contained within the entire design. By that I
mean, lets say that Equation 5a was implemented in some entity
somewhere in some file and Equation 5b in a totally different entity
in some other file. By themselves, each entity/file will synthesize
just fine. But when you add the code for the entity that ties these
two blocks together that new design and synthesize that new block,
you'll have now formed the loop and the synthesizer should flag the
warning for you. Take those warning seriously if you're designing in
the FPGA/CPLD world.

Kevin Jennings

 

[Jim Lewis]

Peter,
No argument from me as you seem to be saying the same
thing I said.

I interpreted Weng's question as being, when a combinational
signal is on both the right and left side of an equation,
is the only hardware solution a latch.

The answer which we both observed is no and the simple
case is an oscillator.

Cheers,
Jim