Hi,
My problem is I'm trying to learn about the implementation of assertion
languages like PSL/Sugar in digital logic simulators.
The assertion language allows you to more easily say "if this signal
combination/sequence follows that signal combination/sequence then ..."
I feel that you should be able to think of this as the simulator
providing a stream of signal values over simulator time that could be
translated into a string of characters interleaved with clock cycle
markers. Then the temporal assertion could be translated into a regular
expression acting on the string.
When I make the mental analogy above, then I think of the practicle
problems of such an implementation. The first is, if you have a finite
bounded re - e.g. no infinite matches i.e. no plus or star characters in
the RE, but you have a simulator churning out a huge sequence of
characters, how do you do the RE match efficiently?
I am thinking of a program skeleton that looks something like this
psl_re = PSL_subset_to_R E(PSL)
simulator.start (psl_re.clock, psl_re.signals_ to_capture)
for cycles_psl_sign als in simulator.next_ cycle():
if psl_re.optimise d_search(cycles _psl_signals):
raise PSL_PRE_CONDITI ON_FOUND
It is the psl_re.optimise d_search() routine that I am after.
Is there any way of computing how big a string buffer needs to be (call
it B), so that you could then maybe keep a sliding buffer of a minimum
of just the last 2*B characters, sliding along and matching in B
character increments and guarantee to find the matches?
Help, or alternative algorithms would be appreciated.
Thanks, Paddy.
My problem is I'm trying to learn about the implementation of assertion
languages like PSL/Sugar in digital logic simulators.
The assertion language allows you to more easily say "if this signal
combination/sequence follows that signal combination/sequence then ..."
I feel that you should be able to think of this as the simulator
providing a stream of signal values over simulator time that could be
translated into a string of characters interleaved with clock cycle
markers. Then the temporal assertion could be translated into a regular
expression acting on the string.
When I make the mental analogy above, then I think of the practicle
problems of such an implementation. The first is, if you have a finite
bounded re - e.g. no infinite matches i.e. no plus or star characters in
the RE, but you have a simulator churning out a huge sequence of
characters, how do you do the RE match efficiently?
I am thinking of a program skeleton that looks something like this
psl_re = PSL_subset_to_R E(PSL)
simulator.start (psl_re.clock, psl_re.signals_ to_capture)
for cycles_psl_sign als in simulator.next_ cycle():
if psl_re.optimise d_search(cycles _psl_signals):
raise PSL_PRE_CONDITI ON_FOUND
It is the psl_re.optimise d_search() routine that I am after.
Is there any way of computing how big a string buffer needs to be (call
it B), so that you could then maybe keep a sliding buffer of a minimum
of just the last 2*B characters, sliding along and matching in B
character increments and guarantee to find the matches?
Help, or alternative algorithms would be appreciated.
Thanks, Paddy.