search

ponder-lab / Hybridize-Functions-Refactoring

Refactorings for optimizing imperative TensorFlow clients for greater efficiency.
Eclipse Public License 2.0
0 stars 0 forks source link

Importing a module from a different file with `from` results in missing call graph nodes #313

Closed khatchad closed 9 months ago

khatchad commented 10 months ago

$ diff edu.cuny.hunter.hybridize.tests/resources/HybridizeFunction/testPythonSideEffects8/in/A.py edu.cuny.hunter.hybridize.tests/resources/HybridizeFunction/testPythonSideEffects59/in/A.py


1c1
< from B import g
---
> import B
4c4
<     g()
---
>     B.g()

The LHS works, but the RHS doesn't.
khatchad commented 10 months ago

Relates to #311.

khatchad commented 10 months ago

For testPythonSideEffects8(), we have:

callees of node f : [g]

IR of node 4, context CallStringContext: [ script A.py.do()LRoot;@98 ]
<Code body of function Lscript A.py/f>
CFG:
BB0[-1..-2]
    -> BB1
BB1[0..1]
    -> BB2
BB2[-1..-2]
Instructions:
BB0
BB1
0   v4 = lexical:g@Lscript A.py              A.py [4:4] -> [4:5]
1   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v4 @1 exception:v5A.py [4:4] -> [4:7]
BB2

But for testPythonSideEffects59(), we have:

callees of node f : []

IR of node 4, context CallStringContext: [ script A.py.do()LRoot;@97 ]
<Code body of function Lscript A.py/f>
CFG:
BB0[-1..-2]
    -> BB1
BB1[0..2]
    -> BB2
BB2[-1..-2]
Instructions:
BB0
BB1
0   v5 = lexical:B@Lscript A.py              A.py [4:4] -> [4:5]
1   v3 = fieldref v5.v6:#g                   A.py [4:4] -> [4:7]
2   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v3 @2 exception:v7A.py [4:4] -> [4:9]
BB2
khatchad commented 10 months ago

Relevant diff lines:

392,398c392,397
< 90   v17 = global:global script B.py         A.py [1:0] -> [7:3] [17=[g]]
< 91   v243 = fieldref v17.v242:#g             A.py [1:0] -> [7:3] [243=[g]17=[g]]
< 92   lexical:g@Lscript A.py = v243           A.py [1:0] -> [7:3] [243=[g]]
< 95   v245 = new <PythonLoader,Lscript A.py/f>@95<no information> [245=[f]]
< 96   global:global script A.py/f = v245      <no information> [245=[f]]
< 97   putfield v1.< PythonLoader, LRoot, f, <PythonLoader,LRoot> > = v245<no information> [245=[f]]
< 98   v247 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v245 @98 exception:v248A.py [7:0] -> [7:3] [245=[f]]
---
> 90   v242 = invokestatic < PythonLoader, LB, import()LB; > @90 exception:v243A.py [1:0] -> [1:8] [242=[B]]
> 91   lexical:B@Lscript A.py = v242           A.py [1:0] -> [1:0] [242=[B]]
> 94   v245 = new <PythonLoader,Lscript A.py/f>@94<no information> [245=[f]]
> 95   global:global script A.py/f = v245      <no information> [245=[f]]
> 96   putfield v1.< PythonLoader, LRoot, f, <PythonLoader,LRoot> > = v245<no information> [245=[f]]
> 97   v247 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v245 @97 exception:v248A.py [7:0] -> [7:3] [245=[f]]
511c510
< callees of node f : [g]
---
> callees of node f : []
524,542c523,525
< 0   v4 = lexical:g@Lscript A.py              A.py [4:4] -> [4:5]
< 1   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v4 @1 exception:v5A.py [4:4] -> [4:7]
< BB2
< 
< callees of node g : []
< 
< IR of node 5, context CallStringContext: [ script A.py.f.do()LRoot;@1 ]
< <Code body of function Lscript B.py/g>
< CFG:
< BB0[-1..-2]
<     -> BB1
< BB1[0..1]
<     -> BB2
< BB2[-1..-2]
< Instructions:
< BB0
< BB1
< 0   v4 = lexical:my_list@Lscript B.py        B.py [5:4] -> [5:11]
< 1   fieldref v4.v5:#0 = v2:#1 = v2:#1        B.py [5:4] -> [5:14]
---
> 0   v5 = lexical:B@Lscript A.py              A.py [4:4] -> [4:5]
> 1   v3 = fieldref v5.v6:#g                   A.py [4:4] -> [4:7]
> 2   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v3 @2 exception:v7A.py [4:4] -> [4:9]
558,576c541,556
< [Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ], v17] --> [SITE_IN_NODE{synthetic < PythonLoader, Lcom/ibm/wala/FakeRootClass, fakeRootMethod()V >:Lscript B.py in Everywhere}]
< [<field global script B.py>] --> [SITE_IN_NODE{synthetic < PythonLoader, Lcom/ibm/wala/FakeRootClass, fakeRootMethod()V >:Lscript B.py in Everywhere}]
< [Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ], v242] --> [[ConstantKey:g:<PythonLoader,Lstring>]]
< [upward:g:Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ]] --> [SMIK:SITE_IN_NODE{<Code body of function Lscript B.py>:Lscript B.py/g in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}@creator:Node: <Code body of function Lscript B.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]]
< [Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ], v243] --> [SMIK:SITE_IN_NODE{<Code body of function Lscript B.py>:Lscript B.py/g in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}@creator:Node: <Code body of function Lscript B.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]]
---
> [upward:B:Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ]] --> []
> [Node: <Code body of function Lscript A.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@2 ], v242] --> []
586,599c566,580
---
> [<field global script B.py>] --> [SITE_IN_NODE{synthetic < PythonLoader, Lcom/ibm/wala/FakeRootClass, fakeRootMethod()V >:Lscript B.py in Everywhere}]
603,605c584,586
< [catalog of SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}] --> [[ConstantKey:0:<PythonLoader,LRoot>], [ConstantKey:0:<PythonLoader,LRoot>]]
< [SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]},<field 0>] --> [[ConstantKey:10:<PythonLoader,LRoot>], [ConstantKey:1:<PythonLoader,LRoot>]]
< [SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}; [Core[Root]]] --> [[ConstantKey:10:<PythonLoader,LRoot>], [ConstantKey:1:<PythonLoader,LRoot>]]
---
> [catalog of SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}] --> [[ConstantKey:0:<PythonLoader,LRoot>]]
> [SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]},<field 0>] --> [[ConstantKey:10:<PythonLoader,LRoot>]]
> [SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}; [Core[Root]]] --> [[ConstantKey:10:<PythonLoader,LRoot>]]
611,619c592,596
< [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@98 ], v4] --> [SMIK:SITE_IN_NODE{<Code body of function Lscript B.py>:Lscript B.py/g in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}@creator:Node: <Code body of function Lscript B.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]]
< [Exc-Ret-V:Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@98 ]] --> [[Core[Exception]]]
< [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@98 ], v5] --> []
< [Node: <Code body of function Lscript B.py/g> Context: CallStringContext: [ script A.py.f.do()LRoot;@1 ], v1] --> [SMIK:SITE_IN_NODE{<Code body of function Lscript B.py>:Lscript B.py/g in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}@creator:Node: <Code body of function Lscript B.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]]
< [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@98 ], v2] --> []
< [Ret-V:Node: <Code body of function Lscript B.py/g> Context: CallStringContext: [ script A.py.f.do()LRoot;@1 ]] --> []
< [Node: <Code body of function Lscript B.py/g> Context: CallStringContext: [ script A.py.f.do()LRoot;@1 ], v4] --> [SITE_IN_NODE{<Code body of function Lscript B.py>:Llist in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}]
< [Node: <Code body of function Lscript B.py/g> Context: CallStringContext: [ script A.py.f.do()LRoot;@1 ], v2] --> [[ConstantKey:1:<PythonLoader,LRoot>]]
< [Node: <Code body of function Lscript B.py/g> Context: CallStringContext: [ script A.py.f.do()LRoot;@1 ], v5] --> [[ConstantKey:0:<PythonLoader,LRoot>]]
---
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v5] --> []
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v6] --> [[ConstantKey:g:<PythonLoader,Lstring>]]
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v3] --> []
> [Exc-Ret-V:Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ]] --> [[Core[Exception]]]
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v7] --> []
khatchad commented 10 months ago

The interesting stuff is in the PA.

511c510
< callees of node f : [g]
---
> callees of node f : []
524,542c523,525
< 0   v4 = lexical:g@Lscript A.py              A.py [4:4] -> [4:5]
< 1   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v4 @1 exception:v5A.py [4:4] -> [4:7]
< BB2
---
> 0   v5 = lexical:B@Lscript A.py              A.py [4:4] -> [4:5]
> 1   v3 = fieldref v5.v6:#g                   A.py [4:4] -> [4:7]
> 2   v2 = invokeFunction < PythonLoader, LCodeBody, do()LRoot; > v3 @2 exception:v7A.py [4:4] -> [4:9]

In 8, the function invoked is in v4, while in 59, it is in v3.

611,619c592,596
< [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@98 ], v4] --> [SMIK:SITE_IN_NODE{<Code body of function Lscript B.py>:Lscript B.py/g in CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]}@creator:Node: <Code body of function Lscript B.py> Context: CallStringContext: [ com.ibm.wala.FakeRootClass.fakeRootMethod()V@4 ]]
---
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v5] --> []
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v6] --> [[ConstantKey:g:<PythonLoader,Lstring>]]
> [Node: <Code body of function Lscript A.py/f> Context: CallStringContext: [ script A.py.do()LRoot;@97 ], v3] --> []

We can see that the points-to set for v3 is empty on the RHS.

khatchad commented 10 months ago

But, also interesting is the varying instructions, which makes me believe that the problem is (also?) with the CAst translation.

khatchad commented 9 months ago

I can't repro this on 2107422dde0bc40fefe1f918e2bde668f7a3b7b7.