myhdl-list

[myhdl-list] Function calls in combinatorial blocks

[Per K. <bas...@gm...>]

Hi!
Is there a particular reason that we do not support function calls in
combinatorial blocks?

In 0.7 I get:
  File "[...]_analyze.py", line 165, in _analyzeGens
    _isMem(obj) or _isTupleOfInts(obj)

In 0.8dev I get:
  File "[...]_analyze.py", line 168, in _analyzeGens
    raise ConversionError(_error.UnsupportedType, n, info)
myhdl.ConversionError: File file.py, line 11: Object type is not supported
in this context: f

Here is a test snippet:
#########################
from myhdl import *

def w(a, b, c, clk):
    def f(a):
        out = intbv(0)[len(a):]
        if a<3:
            out[:] = a
        else:
            out[:] = 0
        return out

    @always_comb
    def cm():
        b.next = a

    @always(clk.posedge)
    def ck():
        c.next = f(a)

    return instances()

def test():
    a = Signal(intbv(0)[10:])
    b = Signal(intbv(0)[10:])
    c = Signal(intbv(0)[10:])
    clk = Signal(bool(0))
    dut = toVerilog(w, a, b,c ,clk)
    return instances()

s = Simulation(test())
s.run()
##########################

-Cheers!
/Per

Re: [myhdl-list] Function calls in combinatorial blocks

[Christopher F. <chr...@gm...>]

On 3/1/2013 9:12 AM, Per Karlsson wrote:
> Hi!
> Is there a particular reason that we do not support function calls in
> combinatorial blocks?
>

I don't know if there is a reason, my guess is
not.  But it is currently *not* supported.  To
have the feature added it is ideal to create a
MEP describing the feature.  Can simply start
by describing the enhancement request to the
mailing-list.

My opinion, there are two levels: modeling and
conversion.  In some cases we might want to take
baby steps to implement, first just add for modeling
and then for conversion.

Regards,
Chris

Re: [myhdl-list] Function calls in combinatorial blocks

[Christopher F. <chr...@gm...>]

On 3/1/13 3:16 PM, Christopher Felton wrote:
> On 3/1/2013 9:12 AM, Per Karlsson wrote:
>> Hi!
>> Is there a particular reason that we do not support function calls in
>> combinatorial blocks?
>>
>
> I don't know if there is a reason, my guess is
> not.  But it is currently *not* supported.  To
> have the feature added it is ideal to create a
> MEP describing the feature.  Can simply start
> by describing the enhancement request to the
> mailing-list.
>
> My opinion, there are two levels: modeling and
> conversion.  In some cases we might want to take
> baby steps to implement, first just add for modeling
> and then for conversion.

I should clarify, this works for "modeling" you can
have a function on the right hand side, I was speaking
generally in the second paragraph.

Best of my knowledge, functions as described (on the RHS
of and assignment) are not convertible but I don't see
why we can't create a MEP and add it to the list of
future features.

Regards,
Chris

Re: [myhdl-list] Function calls in combinatorial blocks

[Christopher F. <chr...@gm...>]

On 3/2/2013 7:11 AM, Christopher Felton wrote:
> On 3/1/13 3:16 PM, Christopher Felton wrote:
>> On 3/1/2013 9:12 AM, Per Karlsson wrote:
>>> Hi!
>>> Is there a particular reason that we do not support function calls in
>>> combinatorial blocks?
>>>
>>
>> I don't know if there is a reason, my guess is
>> not.  But it is currently *not* supported.  To
>> have the feature added it is ideal to create a
>> MEP describing the feature.  Can simply start
>> by describing the enhancement request to the
>> mailing-list.
>>
>> My opinion, there are two levels: modeling and
>> conversion.  In some cases we might want to take
>> baby steps to implement, first just add for modeling
>> and then for conversion.
>
> I should clarify, this works for "modeling" you can
> have a function on the right hand side, I was speaking
> generally in the second paragraph.
>
> Best of my knowledge, functions as described (on the RHS
> of and assignment) are not convertible but I don't see
> why we can't create a MEP and add it to the list of
> future features.
>
> Regards,
> Chris
>
>

Hmmm, not sure what happened when I replied to this
thread awhile ago.  I (*thought*) I tested this and
saw the same phenomenon and thought it was not a
supported feature.  Yet again, I was *wrong*!
(a new post reminded me)

A function in an /always_comb/ should be supported.
I included an example below.  If you have an example
that is not working let us know and I will see if
I can help.  Sorry about any confusion.

Regards,
Chris


~~~ [Example with functions on RHS] ~~~~

     def cadd(a,b):
         return a + b

     def m_add_seq(clock,reset,a,b,c):
         @always_seq(clock.posedge, reset=reset)
         def hdl():
             c.next = cadd(a,b)
         return hdl

     def m_add_comb(a,b,c):
         @always_comb
         def hdl():
             c.next = cadd(a,b)
         return hdl

     def test_add():
         clock = Signal(bool(0))
         reset = ResetSignal(0, active=0, async=False)
         a,b = [Signal(intbv(0,min=-4,max=4)) for ii in (0,1)]
         cs = Signal(intbv(0,min=-8,max=8))
         cc = Signal(intbv(0,min=-8,max=8))

         @always(delay(3))
         def tb_clock():
             clock.next = not clock

         tb_dut_s = m_add_seq(clock,reset,a,b,cs)
         tb_dut_c = m_add_comb(a,b,cc)

         @instance
         def tb_stim():
             reset.next = False
             yield delay(10)
             reset.next = True
             yield delay(10)
             yield clock.posedge
             for aa in xrange(4):
                 for bb in xrange(4):
                     a.next = aa
                     b.next = bb
                     yield clock.posedge
                     yield clock.posedge
                     assert cs == (a+b)
                     assert cc == (a+b)
             raise StopSimulation

         Simulation((tb_clock,tb_dut_s,tb_dut_c,tb_stim)).run()
         toVerilog(m_add_seq,clock,reset,a,b,cs)
         toVerilog(m_add_comb,a,b,cc)

Re: [myhdl-list] Function calls in combinatorial blocks

[Per K. <bas...@gm...>]

There was an example in my original post. But of course there may be
something else that's wrong with that.
Especially since I just tried your example... and it works! :)
/Per


On Tue, Apr 2, 2013 at 3:42 PM, Christopher Felton
<chr...@gm...>wrote:

> On 3/2/2013 7:11 AM, Christopher Felton wrote:
> > On 3/1/13 3:16 PM, Christopher Felton wrote:
> >> On 3/1/2013 9:12 AM, Per Karlsson wrote:
> >>> Hi!
> >>> Is there a particular reason that we do not support function calls in
> >>> combinatorial blocks?
> >>>
> >>
> >> I don't know if there is a reason, my guess is
> >> not.  But it is currently *not* supported.  To
> >> have the feature added it is ideal to create a
> >> MEP describing the feature.  Can simply start
> >> by describing the enhancement request to the
> >> mailing-list.
> >>
> >> My opinion, there are two levels: modeling and
> >> conversion.  In some cases we might want to take
> >> baby steps to implement, first just add for modeling
> >> and then for conversion.
> >
> > I should clarify, this works for "modeling" you can
> > have a function on the right hand side, I was speaking
> > generally in the second paragraph.
> >
> > Best of my knowledge, functions as described (on the RHS
> > of and assignment) are not convertible but I don't see
> > why we can't create a MEP and add it to the list of
> > future features.
> >
> > Regards,
> > Chris
> >
> >
>
> Hmmm, not sure what happened when I replied to this
> thread awhile ago.  I (*thought*) I tested this and
> saw the same phenomenon and thought it was not a
> supported feature.  Yet again, I was *wrong*!
> (a new post reminded me)
>
> A function in an /always_comb/ should be supported.
> I included an example below.  If you have an example
> that is not working let us know and I will see if
> I can help.  Sorry about any confusion.
>
> Regards,
> Chris
>
>
> ~~~ [Example with functions on RHS] ~~~~
>
>      def cadd(a,b):
>          return a + b
>
>      def m_add_seq(clock,reset,a,b,c):
>          @always_seq(clock.posedge, reset=reset)
>          def hdl():
>              c.next = cadd(a,b)
>          return hdl
>
>      def m_add_comb(a,b,c):
>          @always_comb
>          def hdl():
>              c.next = cadd(a,b)
>          return hdl
>
>      def test_add():
>          clock = Signal(bool(0))
>          reset = ResetSignal(0, active=0, async=False)
>          a,b = [Signal(intbv(0,min=-4,max=4)) for ii in (0,1)]
>          cs = Signal(intbv(0,min=-8,max=8))
>          cc = Signal(intbv(0,min=-8,max=8))
>
>          @always(delay(3))
>          def tb_clock():
>              clock.next = not clock
>
>          tb_dut_s = m_add_seq(clock,reset,a,b,cs)
>          tb_dut_c = m_add_comb(a,b,cc)
>
>          @instance
>          def tb_stim():
>              reset.next = False
>              yield delay(10)
>              reset.next = True
>              yield delay(10)
>              yield clock.posedge
>              for aa in xrange(4):
>                  for bb in xrange(4):
>                      a.next = aa
>                      b.next = bb
>                      yield clock.posedge
>                      yield clock.posedge
>                      assert cs == (a+b)
>                      assert cc == (a+b)
>              raise StopSimulation
>
>          Simulation((tb_clock,tb_dut_s,tb_dut_c,tb_stim)).run()
>          toVerilog(m_add_seq,clock,reset,a,b,cs)
>          toVerilog(m_add_comb,a,b,cc)
>
>
>
>
> ------------------------------------------------------------------------------
> Own the Future-Intel(R) Level Up Game Demo Contest 2013
> Rise to greatness in Intel's independent game demo contest. Compete
> for recognition, cash, and the chance to get your game on Steam.
> $5K grand prize plus 10 genre and skill prizes. Submit your demo
> by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2
> _______________________________________________
> myhdl-list mailing list
> myh...@li...
> https://lists.sourceforge.net/lists/listinfo/myhdl-list
>

Re: [myhdl-list] Function calls in combinatorial blocks

[Christopher F. <chr...@gm...>]

On 4/2/2013 10:17 AM, Per Karlsson wrote:
> There was an example in my original post. But of course there may be
> something else that's wrong with that.
> Especially since I just tried your example... and it works! :)
> /Per
>
>

Doh, missed the original example, will give it another
look and see I see anything.

.chris

Re: [myhdl-list] Function calls in combinatorial blocks

[Christopher F. <chr...@gm...>]

On 4/2/2013 10:40 AM, Christopher Felton wrote:
> On 4/2/2013 10:17 AM, Per Karlsson wrote:
>> There was an example in my original post. But of course there may be
>> something else that's wrong with that.
>> Especially since I just tried your example... and it works! :)
>> /Per
>>
>>
>
> Doh, missed the original example, will give it another
> look and see I see anything.
>
> .chris
>

The difference between my example and the one you
provided is that the function is in the local scope
(in the original example).  If I modify your example
slightly it works in the /@always_comb/.

I don't know if a local function working in the
sequential just happens to work or if the local
function not working in the comb is a bug.  I will
investigate (at some point :) ).

Regards,
Chris

~~~~ [modified example] ~~~~
     from myhdl import *

     def f(a):
         out = intbv(0)[len(a):]
         if a<3:
             out[:] = a
         else:
             out[:] = 0
         return out

     def w(a, b, c, clk):
         @always_comb
         def cm():
             b.next = f(a)

         @always(clk.posedge)
         def ck():
             c.next = f(a)

         return instances()

     def test():
         a = Signal(intbv(0)[10:])
         b = Signal(intbv(0)[10:])
         c = Signal(intbv(0)[10:])
         clk = Signal(bool(0))
         dut = toVerilog(w, a, b,c ,clk)
         return instances()

     s = Simulation(test())
     s.run()