Test failure JIT/SIMD/VectorAdd_ro/VectorAdd_ro

[v-wenyuxu]

Run: runtime-coreclr outerloop 20230604.2

Failed test:

R2R-CG2 linux x64 Checked no_tiered_compilation @ (Alpine.315.Amd64.Open)Ubuntu.1804.Amd64.Open@mcr.microsoft.com/dotnet-buildtools/prereqs:alpine-3.15-helix-amd64
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

R2R-CG2 windows x86 Checked no_tiered_compilation @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

R2R-CG2 windows x64 Checked no_tiered_compilation @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

R2R-CG2 windows x86 Checked @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

R2R-CG2 linux x64 Checked no_tiered_compilation @ Ubuntu.1804.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

coreclr windows x86 Checked @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

Error message:

Assert.Equal() Failure
Expected: 100
Actual:   -1



Stack trace
   at Xunit.Assert.Equal[T](T expected, T actual, IEqualityComparer`1 comparer) in /_/src/xunit.assert/Asserts/EqualityAsserts.cs:line 96
   at Program.<<Main>$>g__TestExecutor58|0_59(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

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

Issue Details

---

Run: runtime-coreclr outerloop 20230604.2

Failed test:

R2R-CG2 linux x64 Checked no_tiered_compilation @ (Alpine.315.Amd64.Open)Ubuntu.1804.Amd64.Open@mcr.microsoft.com/dotnet-buildtools/prereqs:alpine-3.15-helix-amd64
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

R2R-CG2 windows x86 Checked no_tiered_compilation @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

R2R-CG2 windows x64 Checked no_tiered_compilation @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

R2R-CG2 windows x86 Checked @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

R2R-CG2 linux x64 Checked no_tiered_compilation @ Ubuntu.1804.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll

coreclr windows x86 Checked @ Windows.10.Amd64.Open
- JIT/SIMD/VectorAdd_ro/VectorAdd_ro.dll
- JIT/SIMD/VectorHWAccel_ro/VectorHWAccel_ro.dll
- JIT/SIMD/VectorHWAccel2_ro/VectorHWAccel2_ro.dll

Error message:

Assert.Equal() Failure
Expected: 100
Actual:   -1



Stack trace
   at Xunit.Assert.Equal[T](T expected, T actual, IEqualityComparer`1 comparer) in /_/src/xunit.assert/Asserts/EqualityAsserts.cs:line 96
   at Program.<<Main>$>g__TestExecutor58|0_59(StreamWriter tempLogSw, StreamWriter statsCsvSw, <>c__DisplayClass0_0&)

Author:	v-wenyuxu
Assignees:	-
Labels:	arch-x86, os-linux, os-windows, arch-x64, area-CodeGen-coreclr, untriaged, blocking-outerloop
Milestone:	-

[markples]

@tannergooding Is this caused by #84821?

[tannergooding]

Possibly. CC @dotnet/avx512-contrib

Will take a look.

[tannergooding]

Simplified repro is:

using System.Runtime.CompilerServices;
using System.Runtime.Intrinsics;

public class Program
{
    public static int Main()
    {
        return TryVectorAdd(1, 2, 1 + 2) ? 100 : 0;
    }

    [MethodImpl(MethodImplOptions.NoInlining)]
    public static bool TryVectorAdd(float a, float b, float c)
    {
        Vector128<float>A = Vector128.Create(a);
        Vector128<float>B = Vector128.Create(b);
        Vector128<float> C = A + B;

        return C == Vector128.Create(c);
    }
}

We emit the following and are missing the embedded broadcast for xmm1:

G_M39597_IG01:              ;; offset=0000H
 C5F877               vzeroupper
                                                ;; size=3 bbWeight=1 PerfScore 1.00
G_M39597_IG02:              ;; offset=0003H
 C4E27918C0           vbroadcastss xmm0, xmm0
 C5F858C1             vaddps   xmm0, xmm0, xmm1
 C4E27918CA           vbroadcastss xmm1, xmm2
 C5F8C2C100           vcmpps   xmm0, xmm0, xmm1, 0
 C5F850C0             vmovmskps rax, xmm0
 83F80F               cmp      eax, 15
 0F94C0               sete     al
; ............................... 32B boundary ...............................
 0FB6C0               movzx    rax, al
                                                ;; size=32 bbWeight=1 PerfScore 11.50
G_M39597_IG03:              ;; offset=0023H
 C3                   ret