Fixing SpanHelpers.LastIndexOfAnyValueType to no longer create out of bounds GC refs #75857
Conversation
ghost
assigned 
|
Tagging subscribers to this area: @dotnet/area-system-memory Issue DetailsThis resolves #75792
|
There was a problem hiding this comment.
@tannergooding big thanks for your help! PTAL at my comments, I think I've found a bug
src/libraries/System.Private.CoreLib/src/System/SpanHelpers.T.cs
Outdated
Show resolved
Hide resolved
src/libraries/System.Private.CoreLib/src/System/SpanHelpers.T.cs
Outdated
Show resolved
Hide resolved
Co-authored-by: Adam Sitnik <adam.sitnik@gmail.com>
src/libraries/System.Private.CoreLib/src/System/SpanHelpers.T.cs
Outdated
Show resolved
Hide resolved
|
@stephentoub, would appreciate a secondary sign-off from you before merging. Great to have a second pair of eyes since we're backporting for .NET 7 |
| continue; | ||
| } | ||
|
|
||
| return ComputeLastIndex(ref searchSpace, ref currentSearchSpace, equals); | ||
| return ComputeLastIndex(offset, equals); |
There was a problem hiding this comment.
It doesn't affect this PR, but I wonder why it's written like this rather than:
if (equals != Vector128<TValue>.Zero)
{
return ComputeLastIndex(offset, equals);
}Is this just style, or we expect the not found case to be the most common and want the branch predictor to default to that?
There was a problem hiding this comment.
we expect the not found case to be the most common and want the branch predictor to default to that
That was my assumption when I was implementing it.
|
/backport to release/7.0-rc2 |
|
Started backporting to release/7.0-rc2: https://github.com/dotnet/runtime/actions/runs/3089699377 |
This resolves #75792