Unnecessary sign-extension for (nuint)span.Length

[xtqqczze]

@EgorBo In theory, the length in (nuint)span.Length should be recognized as never negative, and the sign-extension elided, but doesn't appear to be the case (at least in R2R), see https://csharp.godbolt.org/z/14oW68fqz.

Originally posted by @xtqqczze in #84524 (comment)

static nuint Span0(ReadOnlySpan<byte> span) => (nuint)span.Length;
static nuint Span1(ReadOnlySpan<byte> span) => (uint)span.Length;

// crossgen2 8.0.0-preview.4.23208.99+07f00dc2f624946c54dbf85fea79c600a690f634

C:Span0(System.ReadOnlySpan`1[ubyte]):ulong:
; Emitting BLENDED_CODE for X64 CPU with AVX - Windows
						;; size=0 bbWeight=1 PerfScore 0.00
       movsxd   rax, dword ptr [rcx+08H]
						;; size=4 bbWeight=1 PerfScore 4.00
       ret      
						;; size=1 bbWeight=1 PerfScore 1.00

C:Span1(System.ReadOnlySpan`1[ubyte]):ulong:
; Emitting BLENDED_CODE for X64 CPU with AVX - Windows
						;; size=0 bbWeight=1 PerfScore 0.00
       mov      eax, dword ptr [rcx+08H]
						;; size=3 bbWeight=1 PerfScore 2.00
       ret      
						;; size=1 bbWeight=1 PerfScore 1.00

[ghost]

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

Issue Details

---

@EgorBo In theory, the length in (nuint)span.Length should be recognized as never negative, and the sign-extension elided, but doesn't appear to be the case (at least in R2R), see https://csharp.godbolt.org/z/14oW68fqz.

Originally posted by @xtqqczze in #84524 (comment)

static nuint Span0(ReadOnlySpan<byte> span) => (nuint)span.Length;
static nuint Span1(ReadOnlySpan<byte> span) => (uint)span.Length;

// crossgen2 8.0.0-preview.4.23208.99+07f00dc2f624946c54dbf85fea79c600a690f634

C:Span0(System.ReadOnlySpan`1[ubyte]):ulong:
; Emitting BLENDED_CODE for X64 CPU with AVX - Windows
						;; size=0 bbWeight=1 PerfScore 0.00
       movsxd   rax, dword ptr [rcx+08H]
						;; size=4 bbWeight=1 PerfScore 4.00
       ret      
						;; size=1 bbWeight=1 PerfScore 1.00

C:Span1(System.ReadOnlySpan`1[ubyte]):ulong:
; Emitting BLENDED_CODE for X64 CPU with AVX - Windows
						;; size=0 bbWeight=1 PerfScore 0.00
       mov      eax, dword ptr [rcx+08H]
						;; size=3 bbWeight=1 PerfScore 2.00
       ret      
						;; size=1 bbWeight=1 PerfScore 1.00

Author:	xtqqczze
Assignees:	-
Labels:	area-CodeGen-coreclr, untriaged
Milestone:	-

[EgorBo]

Should be a simple fix since Span.Length can be recognized via IsNeverNegative. We just need to mark such GT_CAST as GTF_UNSIGNED if its target is IsNeverNegative and we're upcasting

[SingleAccretion]

We already do:

runtime/src/coreclr/jit/morph.cpp

Lines 10190 to 10194 in 00dbf84

	// Try and see if we can make this cast into a cheaper zero-extending version.
	if (genActualTypeIsInt(src) && cast->TypeIs(TYP_LONG) && srcRange.IsNonNegative())
	{
	cast->SetUnsigned();
	}

The snippets in the issue are cases of an un-promoted span, which should be rare in practice.