JIT: fold trivial redundant comparisons on loop induction variables

[EgorBo]

Closes #130206

static void Test()
{
    for (int i = 0; i < 100; i++)
    {
        if (i >= 0) // never false
        {
            Console.WriteLine("Hi");
        }
    }
}

 G_M24707_IG02:
        xor      ebx, ebx
        mov      rsi, 0x...      ; 'Hi'
-       jmp      SHORT G_M24707_IG05
-G_M24707_IG03:
-       inc      ebx
-G_M24707_IG04:
-       test     ebx, ebx
-       jl       SHORT G_M24707_IG03
-G_M24707_IG05:
+G_M24707_IG03:
        mov      rcx, rsi
        call     [System.Console:WriteLine(System.String)]
        inc      ebx
        cmp      ebx, 100
-       jl       SHORT G_M24707_IG04
+       jl       SHORT G_M24707_IG03

Some nice diffs, mostly in tests, though.

[dotnet-policy-service]

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

[Copilot]

Pull request overview

This PR adjusts assertion propagation’s SSA-based range query budget so that RangeCheck::TryGetRange can successfully compute ranges for common loop induction-variable patterns, enabling folding of trivially redundant comparisons.

Changes:

• Increase the default SSA walk budget used by assertionprop’s slow GetRange path from 64 to 256.
• Add an explanatory comment describing why TryGetRange needs a larger shared budget due to multiple internal SSA walks.

[EgorBo]

@MihuBot -nuget

[Copilot]

Copilot wasn't able to review any files in this pull request.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

[EgorBo]

PTAL @dotnet/jit-contrib simple change, just bumped the budget.

[JulieLeeMSFT]

SPMI diffs look good mostly, but LoopsIVWidened metric diff shows +0.06% to +10.23%. Is it expected?

[EgorBo]

SPMI diffs look good mostly, but LoopsIVWidened metric diff shows +0.06% to +10.23%. Is it expected?

That means more loops are IV widened. We generally don't pay attention to metrics unless we want to track something, so the numbers for metrics don't generally say that something improved/regressed.