Remove duplicate range check in VSB.Append(char) #82264
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label. |
|
Tagging subscribers to this area: @dotnet/area-system-runtime Issue DetailsDiffs from CoreLibExample improvement: https://www.diffchecker.com/m1G2G2z2
|
stephentoub
approved these changes
Feb 16, 2023
dakersnar
reviewed
Feb 16, 2023
Comment on lines
+173
to
+176
| Span<char> chars = _chars; | ||
| if ((uint)pos < (uint)chars.Length) | ||
| { | ||
| _chars[pos] = c; | ||
| chars[pos] = c; |
There was a problem hiding this comment.
Both methods are indexing on a Span<char>, correct? Why does this new method avoid a range check?
There was a problem hiding this comment.
If both the index and span are locals, and you index into those same locals, the JIT avoids emitting an extra range check and helper for the throw.
There was a problem hiding this comment.
Hmm I might have misunderstood. My original assumption was that the if ((uint)pos < (uint)chars.Length) allows the JIT to skip the range check in chars[pos] = c;, is that not the case?
There was a problem hiding this comment.
Right, that is the case after this change.
But that optimization only works on locals. Before, we were reading the field twice (once in the length check and once in the assignment), and by the time we got to the assignment, the JIT couldn't tell that it already checked the length, so it emitted another one.
There was a problem hiding this comment.
Awesome, that makes sense, thanks for the explanation.
Out of curiosity, do you know what prevents the JIT from optimizing the original case? Is it the potential for the global _chars to be accessed by another thread after the if statement? Or is optimizing globals just out of scope for the JIT for some simpler reason?
There was a problem hiding this comment.
In a general case where you have a field on a struct/class, a different thread could swap out the value from under you between the two reads, so you can't make the assumption that avoiding the second length check is safe.
In this case, this is a field on a ref struct, so it may be a valid assumption? I'll leave that up to smarter people like @EgorBo though.
ghost
locked as resolved and limited conversation to collaborators
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Diffs from CoreLib
@MihuBot
Example improvement: https://www.diffchecker.com/m1G2G2z2