Welcome to pyvsc Discussions!

[mballance]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[eramamoe]

Hi @mballance,
First, Great work on creating "PyVSC"!!
I am trying to use PyVSC to port "uvm_mem_mam" UVM Memory allocation manager to Python. While porting it, I had difficulties to implement the following System Verilog (SV) constraint using PyVSC. I do see examples in PyVSC documentation, where the "vsc.foreach" has been used to constraint the random variable (iterator) of "list" data type. But, in the following constraint, I need to constrain a non-list variable based on variable of type list, which seems to be disallowed currently in PyVSC.

Is this usage not-supported by PyVSC at this point? Please let me know, if you need more information. I greatly appreciate your feedback.

constraint uvm_mem_mam_policy_no_overlap { foreach (in_use[i]) { !(start_offset <= in_use[i].Xend_offsetX and start_offset + len - 1 >= in_use[i].Xstart_offsetX) } }

[mballance]

Hi @eramamoe,
Glad to hear that you're trying out PyVSC! Despite using UVM for many years, I've actually never used the memory allocator. I had a very quick look, and it appears that you can use a vsc.list_t with a class-type element (uvm_mem_region) to model this. I checked the docs and realized that the section on lists doesn't describe how to create a list with class-type elements. For now, have a look at this test:

pyvsc/ve/unit/test_list_object.py

Lines 12 to 34 in 00ce001

	def test_smoke(self):
	@vsc.randobj
	class item_c(object):
	def __init__(self):
	self.a = vsc.rand_uint8_t()
	self.b = vsc.rand_uint8_t()
	@vsc.randobj
	class container_c(object):
	def __init__(self):
	self.l = vsc.rand_list_t(item_c())
	for i in range(10):
	self.l.append(item_c())
	c = container_c()
	c.randomize()
	for i,it in enumerate(c.l):
	print("Item[" + str(i) + "] a=" + str(it.a) + " b=" + str(it.b))

I'll try to spend some time this evening updating the docs. Let me know if you face issues using this modeling approach and I'll be happy to have a look.

Best Regards,
Matthew

[eramamoe]

Hi @mballance ,
Thanks for your response. Sorry for a late reply from my end.

That is the implementation I had first, where I have defined "in_use" as a list of "uvm_mem_region" (tried both - vsc.list_t and vsc.rand_list_t). But, what is new about my implementation, which was not clearly shown in the "constraint" documentation/examples is, in almost all of the documentation involving "list", the constraint was applied on the list, whereas what this implementation warrants is constraining a third variable "start_offset" based on the values in the list "in_use". I ran into problems in this use case.

Here's a simple test case I tried:

@vsc.randobj
class my_sub_s(object):
    def __init__(self):
        self.a = vsc.rand_uint8_t()
        self.b = vsc.rand_uint8_t()

@vsc.randobj
class my_item_c(object):
    def __init__(self):
      self.test = vsc.rand_uint32_t()
      self.my_l = vsc.rand_list_t(my_sub_s())

      for i in range(5):
          inst = my_sub_s()
          inst.a = i*10
          inst.b = i*10 + 5
          self.my_l.append(inst)

    @vsc.constraint
    def test_c(self):
      with vsc.foreach(self.my_l, it=True, idx=True) as (i,it):
        #self.test > it.a
        #self.test < it.b
        self.test in vsc.rangelist(vsc.rng(it.a,it.b))

my_item = my_item_c()
for i in range(len(my_item.my_l)):
  print(str(my_item.my_l[i].a) + "  " + str(my_item.my_l[i].b))

for i in range(5):
  my_item.randomize()
  print(my_item.test)

In the above code, I am expecting "my_item.test" be always inside one of list elements "my_item.my_l[x].a" and "my_item.my_l[x].b".

When, I define self.my_l = vsc.list_t(my_sub_s()), I get errors during randomization as follows:

Problem Constraints:
Constraint 1:
(test(1) > <unknown>.<unknown>[4].a(40));
(test > <unknown>.<unknown>[4].a);
Constraint 2:
(test(1) > <unknown>.<unknown>[2].a(20));
(test > <unknown>.<unknown>[2].a);
Constraint 3:
(test(1) > <unknown>.<unknown>[3].a(30));
(test > <unknown>.<unknown>[3].a);
Constraint 4:
(test(1) > <unknown>.<unknown>[1].a(10));
(test > <unknown>.<unknown>[1].a);

And, when I define self.my_l = vsc.rand_list_t(my_sub_s()), the randomization returns "my_list.test", outside of constraint ranges.

Do you see any issue in the above constraint scheme? Appreciate your insight. Thanks!

[mballance]

Hi @eramamoe,
I believe 'list_t' is what you want. A 'rand_list_t' will treat each element as randomizable, so the 'a' and 'b' fields of each array element will be randomized.
I think the issue you're facing here is that a foreach constraint is unrolled to form the AND of each element. So, you have something like:
test in [0..5] AND test in [10..15] AND test in [20..25] etc.

I'll admit this isn't terribly clear from the solve failure message. I'll have to look at that...

Since you're working with memory allocation, I think you want to select a value of 'test' that *does not* fall within any of the already-allocated ranges. So, you'll want a constraint that looks like this:

            @vsc.constraint
            def test_c(self):
                with vsc.foreach(self.my_l, it=True, idx=True) as (i,it):
                    self.test.not_inside(vsc.rangelist([it.a,it.b]))

Hope this helps!

Best Regards,
Matthew

[eramamoe]

Hi @mballance ,
Got it. I agree with your comments. The constraint in UVM MAM is now working as expected. Thanks!