Definition of base causality order in PTX ISA memory consistency model

[admbbs]

I was reading the PTX ISA and got confused by the definition of base causality order.

An operation X precedes an operation Y in base causality order if:

1. X precedes Y in program order, or
2. X synchronizes with Y, or
3. For some operation Z,
   a. X precedes Z in program order and Z precedes Y in base causality order, or
   b. X precedes Z in base causality order and Z precedes Y in program order, or
   c. X precedes Z in base causality order and Z precedes Y in base causality order.

It seems to me that the 3.a and 3.b clauses are redundant. They both could be covered by the combination of 1, 2 and 3.c.

Then I read the ASPLOS_2019 PTX Memory Model paper and found a different definition (almost same except the first clause of program order)

An operation X precedes an operation Y in base causality order if:

1. X precedes Y in program order, or
2. X synchronizes with Y, or
3. For some operation Z,
   a. X precedes Z in program order and Z precedes Y in base causality order, or
   b. X precedes Z in base causality order and Z precedes Y in program order, or
   c. X precedes Z in base causality order and Z precedes Y in base causality order.

Finally, I went over the CUDA toolkit archive and found this change happened between PTX ISA 7.4 and PTX ISA 7.5.

So my question are

1. Is the current definition of base causality order redundant?
2. Why the definition changed in PTX ISA 7.5?

My appreciation in advance for any answers or clues!

---

Update Apr. 1.

I found the answer to the 2nd question, the change happend on the introduce of proxy. But my 1st question remains unanswered.