Missing CI test coverage for Vector512

[MihaZupan]

The following test passed CI without any errors in #88166:

[Fact]
[MethodImpl(MethodImplOptions.AggressiveOptimization)]
public static void SanityCheckAO()
{
    if (Vector512.IsHardwareAccelerated && Avx512BW.IsSupported)
        throw new Exception("Yay?");
}

Dumping the CpuId, it looks like all processors in CI that report Avx512BW.IsSupported == true also report isGenuineIntel=True familyID=6 extendedModelID=5 model=5.

runtime/src/coreclr/vm/codeman.cpp

Lines 1882 to 1892 in ddf47f0

	if (xarchCpuInfo.FamilyId == 0x06)
	{
	if (xarchCpuInfo.ExtendedModelId == 0x05)
	{
	if (xarchCpuInfo.Model == 0x05)
	{
	// * Skylake (Server)
	// * Cascade Lake
	// * Cooper Lake
	CPUCompileFlags.Set(CORJIT_FLAGS::CORJIT_FLAG_VECTOR512_THROTTLING);

With #86655, those no longer report Vector512.IsHardwareAccelerated, therefore we have no CI coverage of such code paths?

Tagging subscribers to this area: @dotnet/area-system-runtime-intrinsics
See info in area-owners.md if you want to be subscribed.

Issue Details

---

The following test passed CI without any errors in #88166:

[Fact]
[MethodImpl(MethodImplOptions.AggressiveOptimization)]
public static void SanityCheckAO()
{
    if (Vector512.IsHardwareAccelerated && Avx512BW.IsSupported)
        throw new Exception("Yay?");
}

Dumping the CpuId, it looks like all processors in CI that report Avx512BW.IsSupported == true also report isGenuineIntel=True familyID=6 extendedModelID=5 model=5.

runtime/src/coreclr/vm/codeman.cpp

Lines 1882 to 1892 in ddf47f0

	if (xarchCpuInfo.FamilyId == 0x06)
	{
	if (xarchCpuInfo.ExtendedModelId == 0x05)
	{
	if (xarchCpuInfo.Model == 0x05)
	{
	// * Skylake (Server)
	// * Cascade Lake
	// * Cooper Lake
	CPUCompileFlags.Set(CORJIT_FLAGS::CORJIT_FLAG_VECTOR512_THROTTLING);

With #86655, those no longer report Vector512.IsHardwareAccelerated, therefore we have no CI coverage of such code paths?

Author:	MihaZupan
Assignees:	-
Labels:	area-System.Runtime.Intrinsics
Milestone:	-

[tannergooding]

We should add an outerloop leg to the AVX-512 run that sets DOTNET_PreferredVectorBitWidth=512

We are getting some stress coverage still, just randomly based on compStressCompile(STRESS_GENERIC_VARN, 20) and noting this only impacts Vector512.IsHardwareAccelerated lightup. It doesn't impact direct usage or validation of the System.Runtime.Intrinsics.X86 code paths.