Timing results without synthesis?

[Sergey Katsev]

Hi all,

As can be seen by my "VHDL Fixed Point Package" thread problems, I'm
having some synthesis problems with that package.

It appears as though the code *is* synthesizeable, just not by ISE,
right now.

That being said, are there any existing methodologies to get even
approximate timing results (both number of clock cycles and
propagations) for my design without synthesizing it? The design is
basically a couple of inter-connected systolic arrays which do
intermediate mathematical operations...

Thanks,

Sergey

 

[radarman]

Sergey Katsev wrote:

Hi all,

As can be seen by my "VHDL Fixed Point Package" thread problems, I'm
having some synthesis problems with that package.

It appears as though the code *is* synthesizeable, just not by ISE,
right now.

That being said, are there any existing methodologies to get even
approximate timing results (both number of clock cycles and
propagations) for my design without synthesizing it? The design is
basically a couple of inter-connected systolic arrays which do
intermediate mathematical operations...

Thanks,

Sergey

Actual timing will depend on the part, speed grade, routing, and amount
of combinational logic between registers among other things. Most of
those factors are component specific. So, while you could estimate
based on timing figures in the data sheet, they really won't be all
that useful.

A better use of time would probably be determining why ISE can't
compile your code. I haven't seen very much "clean" code fail to
compile properly under ISE (though I have seen some - so it's not out
of the question). Also, if you have access to Modelsim, run your code
through it with the "check for synthesis" option turned on. You might
be surprised at what it flags.

 

[Mike Treseler]

Sergey Katsev wrote:

As can be seen by my "VHDL Fixed Point Package" thread problems, I'm
having some synthesis problems with that package.

It appears as though the code *is* synthesizeable, just not by ISE,
right now.

Get the free version on altera quartus, and try it there.
Quartus is very compliant ieee standards and might work for you.

-- Mike Treseler

 

[Mike Treseler]

radarman wrote:

A better use of time would probably be determining why ISE can't
compile your code. I haven't seen very much "clean" code fail to
compile properly under ISE (though I have seen some - so it's not out
of the question).

Other posters have noted that one problem is
negative array indexes. Legal VHDL but not supported yet by ISE

Also, if you have access to Modelsim, run your code
through it with the "check for synthesis" option turned on. You might
be surprised at what it flags.

All I've ever found is sensitivity list warnings
that make no sense for synchronous processes.

-- Mike Treseler