[TODO-List-Cleanup] Delete NumChildren, GetChild and gtGetChildPointer
#61875
Conversation
TryGetUse does the same thing and is used more often. Also some editorial renames to standardize on the "user/use" terminology.
In most situations these methods are the wrong thing to call - they are not very efficient by design. Not surprisingly, thus, they were mostly only used in debug code. The few product dependencies were removed and replaced with equivalent alternatives, deduplicating the code. In the process, BasicBlock::firstNode() was removed as it was not used anywhere.
ghost
added
the
community-contribution
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT Issue DetailsAll these functions represent what is essentially duplicated code.
|
|
@dotnet/jit-contrib |
|
@AndyAyersMS PTAL. |
src/coreclr/jit/gentree.cpp
Outdated
| @@ -4986,199 +4986,20 @@ unsigned GenTree::GetScaledIndex() | |||
| } | |||
|
|
|||
| //------------------------------------------------------------------------ | |||
| // gtGetChildPointer: If 'parent' is the parent of this node, return the pointer | |||
| // to the child node so that it can be modified; otherwise, return nullptr. | |||
| // TryGetUse: Get the def -> def edge for a child of this tree. | |||
There was a problem hiding this comment.
Nit: def seems like an odd name choice here, and def -> def likewise.
Maybe child would be better?
There was a problem hiding this comment.
Renamed to operand as that's what other related APIs use (ReplaceOperand, VisitOperands, Operands).
ghost
locked as resolved and limited conversation to collaborators
|
Did you look at JIT throughput with this change? Especially for Tier0 code and or Tier0 / w PGO? Techempower runs show a notable drop in full PGO time to first response on around 11/30-12/1 and from a quick scan, this seems to be the only change that might be relevant. But a drop of this magnitude also seems a bit hard to believe...
|
|
@AndyAyersMS I've just checked - it's not this commit (I've checked the commit prior this one) which is 541 (already improved result) |
All these functions represent what is essentially duplicated code.
No diffs.