JIT: Redundant bounds check is not eliminated

[EgorBo]

A tracking issue to link all cases currently not optimized by RyuJIT around bounds checking. Contributes to #94941

Checks inside loops

• Safe vectorization is a lot slower than unsafe #111309
• JIT does not hoist bound check outside a loop #69194
• Optimization of array bounds checks involving bitwise AND could be improved #65550
• Loops aren't cloned for Spans #82946
  • Closed by Loop cloning for Span #113575
• JIT doesn't elide bounds checking when iterating over a span and an array of known length #110986
  • Closed by Loop cloning for Span #113575
• Redundant bounds check #104890
• Bounds checks are no longer elided when using nint for indexing. #44415
• Elide bounds checks for partial array/span iteration #31638
• JIT: Index use can't be completely optimized away #11870
• [JIT] Elide bounds check for reverse iteration over arrays and spans #9505
• JIT: Bounds check not elided for inlined function with safe arguments on Windows #112019
  • Closed by Loop cloning for Span #113575
• JIT: Missed i >= 0 assertion #113809
  • Closed by [Bounds Checks] Make union/intersect operations more precise #113862
• JIT: Redundant bounds check for when loopUpperBound==arrayLength #114744
  • Closed by Fold "new T[size].Length" into "size" #114845

Multidimensional arrays

• JIT: Unnecessary bounds checks for MD array accesses #103321
• Loop cloning can't eliminate bounds check when assigning to multi-dim jagged array #54074

Miscellaneous

• JIT: Bounds check for "i + cns < arrlen" bound #112523
• JIT: Block clonning for subsequent bounds checks #112524
  • Closed by [Experiment] Clone blocks with bounds checks #112595 and Handle overlapped groups of bounds checks #112660
• Codegen: Bound check not elided in ref struct Span field index access #72004
• JIT: Improve rangecheck for certain binops #79257
• C# value discard leads to range check for happy path #99943
• Manual bounds checks are less efficient #80256
• Extra bound checks are not eliminated for pinning #35748
• [JIT] Redundant Span<> bounds check not elided for fields in ref structs #10645
• Use stelm.ref helper calls towards array bounds check assertions and range check elimination #8096
• Redundant bounds check for arr[X >> CNS] pattern #109899
  • Closed by JIT: (x u>> cns) -> [0..(x's max value >> cns)] #109900
• JIT: Redundant bound check #111027
• JIT: Redundant bounds check depending on branch order #113054
  • Closed by Bounds checks: make MergeEdgeAssertions more precise #113233

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[neon-sunset]

Not sure if the issue #109043 belongs here too since it does not go through normal bounds check but otherwise seems related.

[MichalPetryka]

This should also include reverse loops, iterating over multiple collections at once and non 1 increments.

[EgorBo]

Not sure if the issue #109043 belongs here too since it does not go through normal bounds check but otherwise seems related.

I think it indeed duplicates #80256
We're going to close/merge some of these issues as duplicates

This should also include reverse loops, iterating over multiple collections at once and non 1 increments.

It's in the list already - #9505