JIT: Run optRedundantBranches twice and try to compute more accurate doms

[EgorBo]

Closes #70480

void Test(object o)
{
    if (o is int n)
        Console.WriteLine(n);
}

Asm diff for ^: https://www.diffchecker.com/8j7VGRuz

BB04 and BB07 are unreachable, because of "dead" BB04, BB06 reports BB01 as a dominator instead of BB02

Some nice spmi diffs

[ghost]

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

Issue Details

---

Closes #70480

void Test(object o)
{
    if (o is int n)
        Console.WriteLine(n);
}

Asm diff for ^: https://www.diffchecker.com/8j7VGRuz

Author:	EgorBo
Assignees:	-
Labels:	area-CodeGen-coreclr
Milestone:	-

[EgorBo]

PTAL @AndyAyersMS @dotnet/jit-contrib
Diffs up to

Total bytes of delta: -54496 (-0.08 % of base)

for libraries.pmi.windows.arm64.checked.mch

[EgorBo]

/azp run runtime-coreclr outerloop, fuzzlyn

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[AndyAyersMS]

I like the idea, but it seems kind of costly.

[jakobbotsch]

Antigen has better coverage for loops than Fuzzlyn, so you may also want to consider running that (although I do see Fuzzlyn found something).

[EgorBo]

/azp run fuzzlyn, antigen

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[AndyAyersMS]

Are the fuzzlyn/antigen failures real?

[EgorBo]

Are the fuzzlyn/antigen failures real?

Yes, I plan to address them tomorrow

[EgorBo]

/azp run fuzzlyn, antigen

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[EgorBo]

@AndyAyersMS PTAL, throughput regression is lowered from 1% to 0.01-0.1%, diff is slightly smaller but the pattern in question is handled. I plan a few more experiments here in future

Antigen/Fuzzlyn failures are unrelated

[AndyAyersMS]

Happy that you found a very surgical way to check what might be affected.