JIT: Slightly relax "const vector propagation" heuristics

[EgorBo]

Fixes #76781

By changing too conservative heuristic I introduced in #70378, also changed SPMI to print PerfScore (will revert before merge)

Snippet from #76781

static void Foo(ref byte src, nuint length, ref byte dest)
{
    nuint idx = 0;
    Vector128<byte> mask = Vector128.Create((byte)0x7F);
    if (length >= (uint)Vector128<byte>.Count)
    {
        nuint vecEnd = length - (uint)Vector128<byte>.Count;
        do
        {
            Vector128<byte> vec = Vector128.LoadUnsafe(ref src, idx);
            vec &= mask;
            vec.StoreUnsafe(ref dest, idx);
            idx += (uint)Vector128<byte>.Count;
        } while (idx <= vecEnd);
    }
}

; Method Program:Foo(byref,ulong,byref)
G_M5292_IG01:              
       C5F877               vzeroupper 
G_M5292_IG02:              
       33C0                 xor      eax, eax
+      C4E179100532000000   vmovupd  xmm0, xmmword ptr [reloc @RWD00]
       4883FA10             cmp      rdx, 16
       7223                 jb       SHORT G_M5292_IG05
G_M5292_IG03:              
       4883C2F0             add      rdx, -16
-      90                   align    [1 bytes for IG04]
+      0F1F840000000000     align    [8 bytes for IG04]
G_M5292_IG04:              
-      C5FA6F0401           vmovdqu  xmm0, xmmword ptr [rcx+rax]
-      C5F9DB0513000000     vpand    xmm0, xmm0, xmmword ptr [reloc @RWD00]
-      C4C17A7F0400         vmovdqu  xmmword ptr [r8+rax], xmm0
+      C5F9DB0C01           vpand    xmm1, xmm0, xmmword ptr [rcx+rax]
+      C4C17A7F0C00         vmovdqu  xmmword ptr [r8+rax], xmm1
       4883C010             add      rax, 16
       483BC2               cmp      rax, rdx
       76E4                 jbe      SHORT G_M5292_IG04
G_M5292_IG05:              
       C3                   ret      
RWD00  	dq	7F7F7F7F7F7F7F7Fh, 7F7F7F7F7F7F7F7Fh
-; Total bytes of code: 45
+; Total bytes of code: 53  ;; but better overall perfscore!

[ghost]

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

Issue Details

---

Fixes #76781

By changing too conservative heuristic I introduced in #70378, also changed SPMI to print PerfScore (will revert before merge)

Snippet from #76781

static void Foo(ref byte src, nuint length, ref byte dest)
{
    nuint idx = 0;
    Vector128<byte> mask = Vector128.Create((byte)0x7F);
    if (length >= (uint)Vector128<byte>.Count)
    {
        nuint vecEnd = length - (uint)Vector128<byte>.Count;
        do
        {
            Vector128<byte> vec = Vector128.LoadUnsafe(ref src, idx);
            vec &= mask;     // use of hoisted constant

            vec.StoreUnsafe(ref dest, idx);

            idx += (uint)Vector128<byte>.Count;
        } while (idx <= vecEnd);
    }
}

; Method Program:Foo(byref,ulong,byref)
G_M5292_IG01:              
       C5F877               vzeroupper 
G_M5292_IG02:              
       33C0                 xor      eax, eax
+      C4E179100532000000   vmovupd  xmm0, xmmword ptr [reloc @RWD00]
       4883FA10             cmp      rdx, 16
       7223                 jb       SHORT G_M5292_IG05
G_M5292_IG03:              
       4883C2F0             add      rdx, -16
-      90                   align    [1 bytes for IG04]
+      0F1F840000000000     align    [8 bytes for IG04]
G_M5292_IG04:              
-      C5FA6F0401           vmovdqu  xmm0, xmmword ptr [rcx+rax]
-      C5F9DB0513000000     vpand    xmm0, xmm0, xmmword ptr [reloc @RWD00]
-      C4C17A7F0400         vmovdqu  xmmword ptr [r8+rax], xmm0
+      C5F9DB0C01           vpand    xmm1, xmm0, xmmword ptr [rcx+rax]
+      C4C17A7F0C00         vmovdqu  xmmword ptr [r8+rax], xmm1
       4883C010             add      rax, 16
       483BC2               cmp      rax, rdx
       76E4                 jbe      SHORT G_M5292_IG04
G_M5292_IG05:              
       C3                   ret      
RWD00  	dq	7F7F7F7F7F7F7F7Fh, 7F7F7F7F7F7F7F7Fh
-; Total bytes of code: 45
+; Total bytes of code: 53

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

[EgorBo]

@jakobbotsch is this a correct way to switch to PerfScore for superpmi-diff job?

[jakobbotsch]

@jakobbotsch is this a correct way to switch to PerfScore for superpmi-diff job?

I think it should work depending on the size of the diffs. If there are massive number of diffs then you will hit the problem we always had before #76238, but if it manages to complete I think the jit-analyze output should show perfscore analysis.

[EgorBo]

This need more thinking judging by regressions 🙁