gui: ensure inst pins are from register insts

[AcKoucher]

The filter was not excluding instances that are not registers. So we might end displaying data of a path from/to e.g., a buffer.

This should have some impact on performance as the number of total endpoints will be generally smaller. E. g., ngt45/black-parrot:

Start Points
    Instance Pins 38501
    Register Pins 38477
End Points
    Instance Pins 111839
    Register Pins 76486

[github-actions[bot]]

clang-tidy review says "All clean, LGTM! :+1:"

[maliberty]

Do these non-reg end points go into some bucket? Do you have a sense of what sort of end points are being excluded in bp?

[AcKoucher]

@maliberty We're not distinguishing registers from other types of instances when using the filter. So currently the filter types are actually instance to instance, instance to IO, IO to instance.

As for the types of endpoints being excluded in bp:

StartPoints
    Instance Pins 38501
    Register Pins 38477
        Non-Register Instance Types
            Macro
EndPoints
    Instance Pins 111839
    Register Pins 76486
        Non-Register Instance Types
            Buffer/inverter
            Clock buffer/inverter
            Macro
            Multi-Input combinational cell

[maliberty]

I'm curious why so many non-sequential elements are endpoints. The SDC is quite simple so its not from that. What does one of these paths look like?

If we separate non-registers then we will have many more groups (eg reg2non-reg, input2non-reg, etc) which will be awkward.

[AcKoucher]

Well it looks like for the Macro endpoints we get paths:

• Macro to Macro:

  

• IO to Macro:

  

As for the other non-register endpoints, it seems there are no paths, even though they're considered endpoints by sta.

[maliberty]

If there is no path then are they excluded from the histogram (I assume they have no slack either)?

[maliberty]

I don't feel we need to split registers and macros into separate categories. We could change reg to internal or non-IO but I don't have a strong opinion. @oharboe any preference?

[oharboe]

We want the definition of OpenSTA all_registers when we filter for "register to register".

all_registers include endpoint registers within macros.

[AcKoucher]

If there is no path then are they excluded from the histogram (I assume they have no slack either)?

Yes the buckets only contain Sequential Cells and Macros. I think this PR is not needed then.