Faster Vector128/64 compare on arm64

[EgorBo]

Apply @TamarChristinaArm's suggestions for faster vector comparison in #75849

bool Test1(Vector128<int> a, Vector128<int> b) => a == b;
bool Test2(Vector64<float> a, Vector64<float> b) => a != b;

Now emits:

; Method Tests:Test1
G_M48391_IG01:              
        A9BF7BFD          stp     fp, lr, [sp, #-0x10]!
        910003FD          mov     fp, sp
G_M48391_IG02:              
        6EA18C10          cmeq    v16.4s, v0.4s, v1.4s
-       6E31AA10          uminv   b16, v16.16b
-       0E013E00          umov    w0, v16.b[0]
-       7100001F          cmp     w0, #0
-       9A9F07E0          cset    x0, ne
+       6EB0AE10          uminp   v16.4s, v16.4s, v16.4s
+       4E083E00          umov    x0, v16.d[0]
+       B100041F          cmn     x0, #1
+       9A9F17E0          cset    x0, eq
G_M48391_IG03:              
        A8C17BFD          ldp     fp, lr, [sp], #0x10
        D65F03C0          ret     lr
; Total bytes of code: 36


; Method Tests:Test2
G_M64388_IG01:              
        A9BF7BFD          stp     fp, lr, [sp, #-0x10]!
        910003FD          mov     fp, sp
						;; size=8 bbWeight=1    PerfScore 1.50
G_M64388_IG02:              
        0E21E410          fcmeq   v16.2s, v0.2s, v1.2s
-       2E31AA10          uminv   b16, v16.8b
-       0E013E00          umov    w0, v16.b[0]
-       7100001F          cmp     w0, #0
-       9A9F17E0          cset    x0, eq
+       4E083E00          umov    x0, v16.d[0]
+       B100041F          cmn     x0, #1
+       9A9F07E0          cset    x0, ne
G_M64388_IG03:              
        A8C17BFD          ldp     fp, lr, [sp], #0x10
        D65F03C0          ret     lr
-; Total bytes of code: 36
+; Total bytes of code: 32

[ghost]

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

Issue Details

---

Apply @TamarChristinaArm's suggestions for faster vector comparison in #75849

bool Test1(Vector128<int> a, Vector128<int> b) => a == b;

Now emits:

; Method Test1
G_M3164_IG01:              
        A9BF7BFD          stp     fp, lr, [sp, #-0x10]!
        910003FD          mov     fp, sp
G_M3164_IG02:             
        6EA18C10          cmeq    v16.4s, v0.4s, v1.4s
        6EB0AE10          uminp   v16.4s, v16.4s, v16.4s
        4E083E00          umov    x0, v16.d[0]
        B100041F          cmn     x0, #1
        9A9F17E0          cset    x0, eq
G_M3164_IG03:              
        A8C17BFD          ldp     fp, lr, [sp], #0x10
        D65F03C0          ret     lr
; Total bytes of code: 36

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

[TamarChristinaArm]

G_M64388_IG02:              
        0E21E410          fcmeq   v16.2s, v0.2s, v1.2s
        0E043E00          umov    w0, v16.s[0]
        12800001          movn    w1, #0
        6B01001F          cmp     w0, w1
        9A9F07E0          cset    x0, ne

That doesn't look right, The 64-bit case should be transferring the entire register, so I'm expecting the same d[0] transfer here to an x register and the same compare as the 128-bit case, just without the compression step. Looks like this has ignored the top 32-bits.

[EgorBo]

G_M64388_IG02:              
        0E21E410          fcmeq   v16.2s, v0.2s, v1.2s
        0E043E00          umov    w0, v16.s[0]
        12800001          movn    w1, #0
        6B01001F          cmp     w0, w1
        9A9F07E0          cset    x0, ne

That doesn't look right, The 64-bit case should be transferring the entire register, so I'm expecting the same d[0] transfer here to an x register and the same compare as the 128-bit case, just without the compression step. Looks like this has ignored the top 32-bits.

Ah, so I thought, let me fix it 🙂

[TamarChristinaArm]

I assume the diff for the 128-bit one is reversed? but otherwise that looks good to me now.

[EgorBo]

I assume the diff for the 128-bit one is reversed? but otherwise that looks good to me now.

Yes, should be correct now, thanks!

[EgorBo]

@kunalspathak PTAL this one too please

[kunalspathak]

LGTM

[EgorBo]

@TamarChristinaArm it's too early to estimate the whole impact but I already see nice improvements from this, e.g.:

And even more here:

[TamarChristinaArm]

Naise!

[DrewScoggins]

Improvements: dotnet/perf-autofiling-issues#8660