about negative in numeric_std package

[move]

Hi all:

is negative in VHDL free of overflow?

like this

signal din : std_logic_vector(7 downto 0);

......

when i do

cout <= std_logic_vector(- signed(din));

for example when din equal b"10000000" in complient binary = -128,
after negate it, we got 128 : b"10000000" , is it a overflow? how to
get the right answer? what is the best way to do

a = - b ; ?

Thanks ALL in advance !!!

liubenyuan

[KJ]

On Aug 29, 1:23*pm, move <[email protected]> wrote:
> Hi all:
>
> is negative in VHDL free of overflow?
>
> like this
>
> signal din : std_logic_vector(7 downto 0);
>

std_logic_vectors do not have any numerical interpretation, they can
not 'overflow', they can not be 'positive' or 'negative'. To treat a
collection of bits as a signed or unsigned numeric thing, you should
put the following line of code to include use of the numeric_std
library.

use ieee.numeric_std

> .....
>
> when i do
>
> cout <= std_logic_vector(- signed(din));
>
> for example when din equal b"10000000" in complient binary = -128,
> after negate it, we got 128 : b"10000000" , is it a overflow? how to
> get the right answer?

With an 8 bit collection of bits interpreted as twos complement binary
as you've done, your range of operation is from -128 to +127. If
you're going to do something that may take you outside of that range
then you need another bit.

> what is the best way to do
>
> a = - b ; ?
>

'a' should have one more bit of precision than 'b'

signal b: signed(7 downto 0);
signal a: signed(8 downto 0);

KJ

[Mike Treseler]

move wrote:

> how to get the right answer?

That is an interesting question.

To synthesize any significant signed math,
I make an rtl-style sim model using reals to work
out all the register lengths.

To check the sims, I might also use this
www.python.org/download
as a functional calculator.

Once the real answers are right in simulation,
I convert the reals to integer ranges
for small numbers, or to numeric_std.signed
vectors for big numbers.

-- Mike Treseler

[Andy]

On Aug 29, 4:22*pm, Mike Treseler <[email protected]> wrote:
> move wrote:
> > how to get the right answer?
>
> That is an interesting question.
>
> To synthesize any significant signed math,
> I make an rtl-style sim model using reals to work
> out all the register lengths.
>
> To check the sims, I might also use thiswww.python.org/download
> as a functional calculator.
>
> Once the real answers are right in simulation,
> I convert the reals to integer ranges
> for small numbers, or to numeric_std.signed
> vectors for big numbers.
>
> * * * * -- Mike Treseler

There is a new standard package for fixed point, that does not
overflow or roll over, the result is large enough to handle the
largest possibility (n+1 bits in the case of adding or subtracting n
bit operands). Just declare a signed fixed point vector with zero
fractional bits, and you have mathematically accurate integer
arithmetic, but you have to keep up with the signal/variable widths
yourself (the compiler keeps you honest).

For quantities within integer'range, integer arithmetic automatically
handles overflow. The simulation keeps you honest, giving you
assertions if you overflow the range of a variable/signal.

Integer arithmetic also simulates many times faster than vector based
arithmetic.

Andy

[Mike Treseler]

Andy wrote:

> There is a new standard package for fixed point, that does not
> overflow or roll over, the result is large enough to handle the
> largest possibility (n+1 bits in the case of adding or subtracting n
> bit operands). Just declare a signed fixed point vector with zero
> fractional bits, and you have mathematically accurate integer
> arithmetic, but you have to keep up with the signal/variable widths
> yourself (the compiler keeps you honest).

I don't yet grok the difference from numeric_std.signed
for this application, but it sounds like it might be soup,
so I'll check it out.
Let's see
http://www.vhdl.org/vhdl-200x/vhdl-200x-ft/packages/files.html
Yes, Mr. Bishop has been busy.
This looks like a nice example:
http://www.vhdl.org/vhdl-200x/vhdl-200x-ft/packages/fixed_synth.vhdl
This one too
http://www.vhdl.org/vhdl-200x/vhdl-200x-ft/packages/real_tests.vhdl

Thanks Andy.

-- Mike Treseler

[move]

On Aug 30, 3:27*am, KJ <[email protected]> wrote:
> On Aug 29, 1:23*pm, move <[email protected]> wrote:

It seem there IS OVERFLOW, so when do negatate, one more bit is needed
to act as a protect bit;

i do a simple simulate :

library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;

entity test is
end entity test;

architecture rtl of test is
signal din : std_logic_vector(7 downto 0);
signal dout : std_logic_vector(7 downto 0);
begin
dout <= std_logic_vector(-signed(din));

pstim : process
begin
din <= b"00001111"; -- 15
wait for 100 ns; din <= b"11110001"; -- -15
wait for 100 ns; din <= b"10000000"; -- -128
wait;
end process pstim;
end architecture rtl;

so the results are:
11110001
00001111
10000000 -_-!

Thanks all!

[David Bishop]

Andy wrote:
> On Aug 29, 4:22 pm, Mike Treseler <[email protected]> wrote:
>> move wrote:
>>> how to get the right answer?
>> That is an interesting question.
>>
>> To synthesize any significant signed math,
>> I make an rtl-style sim model using reals to work
>> out all the register lengths.
>>
>> To check the sims, I might also use thiswww.python.org/download
>> as a functional calculator.
>>
>> Once the real answers are right in simulation,
>> I convert the reals to integer ranges
>> for small numbers, or to numeric_std.signed
>> vectors for big numbers.
>>
>> -- Mike Treseler
>
> There is a new standard package for fixed point, that does not
> overflow or roll over, the result is large enough to handle the
> largest possibility (n+1 bits in the case of adding or subtracting n
> bit operands). Just declare a signed fixed point vector with zero
> fractional bits, and you have mathematically accurate integer
> arithmetic, but you have to keep up with the signal/variable widths
> yourself (the compiler keeps you honest).

This issue is exactly the reason that the fixed point package grows one
bit when it does an "abs" or a "-" on a number. When writing your code
it drives you nuts, but it keeps the precision correct.

You can get this package (downgraded for vhdl-93) at:
http://www.vhdl.org/fphdl/vhdl.html

[Mike Treseler]

move wrote:

> so the results are:
> 11110001
> 00001111
> 10000000 -_-!
>
> Thanks all!

Thanks for posting your results.

-- Mike Treseler

[Andy]

On Aug 30, 2:02*pm, David Bishop <[email protected]> wrote:
> Andy wrote:
> > On Aug 29, 4:22 pm, Mike Treseler <[email protected]> wrote:
> >> move wrote:
> >>> how to get the right answer?
> >> That is an interesting question.
>
> >> To synthesize any significant signed math,
> >> I make an rtl-style sim model using reals to work
> >> out all the register lengths.
>
> >> To check the sims, I might also use thiswww.python.org/download
> >> as a functional calculator.
>
> >> Once the real answers are right in simulation,
> >> I convert the reals to integer ranges
> >> for small numbers, or to numeric_std.signed
> >> vectors for big numbers.
>
> >> * * * * -- Mike Treseler
>
> > There is a new standard package for fixed point, that does not
> > overflow or roll over, the result is large enough to handle the
> > largest possibility (n+1 bits in the case of adding or subtracting n
> > bit operands). Just declare a signed fixed point vector with zero
> > fractional bits, and you have mathematically accurate integer
> > arithmetic, but you have to keep up with the signal/variable widths
> > yourself (the compiler keeps you honest).
>
> This issue is exactly the reason that the fixed point package grows one
> bit when it does an "abs" or a "-" on a number. * When writing your code
> it drives you nuts, but it keeps the precision correct.
>
> You can get this package (downgraded for vhdl-93) at:http://www.vhdl.org/fphdl/vhdl.html

Yes, declaring signals and variables of the correct size to handle the
results is a pain, but it keeps you aware of what is going on. The
functions help, but are still not very easy to use. This is (at
least) one place where I wish you could declare an unconstrained
variable, initialized to its appropriate width as follows:

variable a,b : sfixed(7 downto 0):= (others => 0);
variable ab_sum : sfixed := a + b;

You can do that with constants...

Andy

[Mike Treseler]

Andy wrote:

> Yes, declaring signals and variables of the correct size to handle the
> results is a pain, but it keeps you aware of what is going on.

.... if I know what the signed lengths ought to be in advance.
If I have to run some sims to figure this out,
I start with reals.

> The
> functions help, but are still not very easy to use. This is (at
> least) one place where I wish you could declare an unconstrained
> variable, initialized to its appropriate width as follows:
>
> variable a,b : sfixed(7 downto 0):= (others => 0);
> variable ab_sum : sfixed := a + b;
>
> You can do that with constants...

Great idea.
Yes, I do that with constants all the time.
I can't see any problem if the expression is static.
Consider an enhancement request to
http://www.eda.org/vasg/bugrep.htm

-- Mike Treseler

[Andy]

On Sep 2, 12:49*pm, Mike Treseler <[email protected]> wrote:
> Andy wrote:
> > Yes, declaring signals and variables of the correct size to handle the
> > results is a pain, but it keeps you aware of what is going on.
>
> ... if I know what the signed lengths ought to be in advance.
> If I have to run some sims to figure this out,
> I start with reals.
>
> > The
> > functions help, but are still not very easy to use. *This is (at
> > least) one place where I wish you could declare an unconstrained
> > variable, initialized to its appropriate width as follows:
>
> > variable a,b * *: sfixed(7 downto 0):= (others => 0);
> > variable ab_sum : sfixed * * * * * *:= a + b;
>
> > You can do that with constants...
>
> Great idea.
> Yes, I do that with constants all the time.
> I can't see any problem if the expression is static.
> Consider an enhancement request tohttp://www.eda.org/vasg/bugrep.htm
>
> * * * *-- Mike Treseler

I think therein lies the problem; the initialization expression is not
exactly static or even statically constrained for that matter. Perhaps
variable references inside a declarative region could be considered
static? Something like this could open up a big can of worms if not
carefully considered.

Andy

[Mike Treseler]

Andy wrote:

> I think therein lies the problem; the initialization expression is not
> exactly static or even statically constrained for that matter. Perhaps
> variable references inside a declarative region could be considered
> static? Something like this could open up a big can of worms if not
> carefully considered.

True, but such consideration is more likely
to occur at eda.org than it is here.
I think the idea has promise.

-- Mike Treseler