Initialize derivative pointers that are allocated through malloc or realloc #1124
Conversation
…derivative pointers
|
clang-tidy review says "All clean, LGTM! 👍" |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #1124 +/- ##
==========================================
+ Coverage 94.37% 94.39% +0.02%
==========================================
Files 50 50
Lines 8366 8713 +347
==========================================
+ Hits 7895 8225 +330
- Misses 471 488 +17
... and 30 files with indirect coverage changes
|
|
clang-tidy review says "All clean, LGTM! 👍" |
| @@ -2842,6 +2877,19 @@ Expr* getArraySizeExpr(const ArrayType* AT, ASTContext& context, | |||
| ComputeEffectiveDOperands(Ldiff, Rdiff, derivedL, derivedR); | |||
| addToCurrentBlock(BuildOp(opCode, derivedL, derivedR), | |||
| direction::forward); | |||
| if (opCode == BO_Assign && derivedR) | |||
There was a problem hiding this comment.
Why is this required? I thought this PR is only modifying the differentiation of malloc calls.
There was a problem hiding this comment.
I took it one step further and covered realloc as well where there's not a decl stmt, but this may be required for malloc functions also. For example:
double *ptr = nullptr;
ptr = (double *)malloc(10 * sizeof(double));
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
clang-tidy review says "All clean, LGTM! 👍" |
| @@ -1788,6 +1790,25 @@ Expr* getArraySizeExpr(const ArrayType* AT, ASTContext& context, | |||
| return StmtDiff(); | |||
| } | |||
|
|
|||
| // If all arguments are constant literals, then this does not contribute to | |||
| // the gradient. | |||
| // FIXME: revert this when this is integrated in the activity analysis pass. | |||
There was a problem hiding this comment.
@ovdiiuv, can you comment here?
There was a problem hiding this comment.
I thought about that recently. I don't think it has to be reverted, rather modified to track arguments that are elements of constant sets for example. This would resolve #682.
There was a problem hiding this comment.
But that should be part of the analysis not here, right? That is, it needs to be implemented on your end.
There was a problem hiding this comment.
No, I think it's the right place.
|
clang-tidy review says "All clean, LGTM! 👍" |
This PR identifies calls to
mallocorreallocand for the derivative pointers it:mallocwithcalloccalloccall afterrealloc