Use StartsWith literal optimization for OrdinalIgnoreCase in RegexCompiler / source generator

[stephentoub]

#66095 is optimizing MemoryExtensions.StartsWith(span, "literal", OrdinalIgnoreCase) by auto-vectorizing cases where the literal string is all ASCII. In Regex, we can take advantage of this in the compiler and source generator by recognizing sequences of appropriate sets, e.g. "[Aa][Bb][Cc]", which our IgnoreCase option now produces (and will produce more robustly with #61048). Today we'll output individual checks for each set, e.g. (ch | 0x20) == 'a', but we can start outputting StartsWith calls with the corresponding multi-character literal string rather than performing individual comparisons against each char.

cc: @EgorBo, @joperezr

[ghost]

Tagging subscribers to this area: @dotnet/area-system-text-regularexpressions
See info in area-owners.md if you want to be subscribed.

Issue Details

---

#66095 is optimizing MemoryExtensions.StartsWith(span, "literal", OrdinalIgnoreCase) by auto-vectorizing cases where the literal string is all ASCII. In Regex, we can take advantage of this in the compiler and source generator by recognizing sequences of appropriate sets, e.g. "[Aa][Bb][Cc]", which our IgnoreCase option now produces (and will produce more robustly with #61048). Today we'll output individual checks for each set, e.g. (ch | 0x20) == 'a', but we can start outputting StartsWith calls with the corresponding multi-character literal string rather than performing individual comparisons against each char.

cc: @EgorBo, @joperezr

Author:	stephentoub
Assignees:	stephentoub
Labels:	area-System.Text.RegularExpressions, tenet-performance
Milestone:	7.0.0