[Intel] Add a means to check which intel gpu generation for vulkan and directX

[farzonl]

This change fixes #709

[farzonl]

I don't know what the family prefix is for Battlemage, so this is unlikely to work on CI. will need help figuring this out.

[bogner]

I guess based on these docs BattleImage would be E2

[farzonl]

Suggested change

	case 0x0B00: // Ponte Vecchio (Xe-HPC / Data Center Max)
	case 0x0B00: // Ponte Vecchio (Xe-HPC / Data Center Max)
	case 0xE200: // Battlemage Discrete
	case 0x6400: // Lunar Lake Integrated

[bogner]

I feel like these names might bite us at some point in the future. Should we bikeshed them a little?

[farzonl]

sure bikeshed away. I just want a tag to xfail on.

[bogner]

@V-FEXrt suggested backwards looking ranges, so Intel-Gen-10 instead of Intel-Legacy-GPU and Intel-Gen-X or Intel-Gen-Current for the "modern" one. This also allows, for example, Intel-Gen-6 if we needed older GPUS (we won't...) and makes it fairly obvious how to update this if another generation pops up.

[bogner]

LGTM. I'll let you use your discretion on the naming, and obviously we should wait for the intel bots to clear.

[farzonl]

hlsl-clang-vk failure is not related this looks like a spirv validator issue on res-of-matrix-in-struct.test.

# .---command stderr------------
# | error: line 72: In Logical addressing, variables can only allocate a storage buffer pointer if the VariablePointersStorageBuffer capability is declared
# |   %a1 = OpVariable %_ptr_Function_struct_A Function %35
# | 
# | clang-dxc: error: spirv-val command failed with exit code 1 (use -v to see invocation)
# `-----------------------------
# error: command failed with exit status: 1

[farzonl]

Looks like we need to file 3 new test failures against DXC. The transpose one we likely have a bug for already.

Exec-Tests-Windows (windows-intel, check-hlsl-vk)

Failed Tests (4):
  OffloadTest-vk :: Feature/CBuffer/Matrix/LayoutKeyword/transpose.test
  OffloadTest-vk :: Feature/PushConstant/types_16bit.test
  OffloadTest-vk :: Feature/Textures/Texture2D.SRVToUAV.array.test.yaml
  OffloadTest-vk :: Feature/Textures/Texture2D.SRVToUAV.test.yaml

transpose.test

Test failed: OutRM2x3
# | Comparison Rule: BufferExact
# | Expected:
# | ---
# | Name:            ExpectedOutRM2x3
# | Format:          Float32
# | Data:            [ 1, 4, 2, 5, 3, 6 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Got:
# | ---
# | Name:            OutRM2x3
# | Format:          Float32
# | Data:            [ 0, 0, 0, 0, 0, 0 ]

types_16bit.test Looks like results are off by 8.

# CHECK: 65000
# |          ^
# | <stdin>:27:18: note: scanning from here
# |  Data: [ 1.5, -5, 64992 ]

Texture2D.SRVToUAV.array.test.yaml

Name:            ExpectedOut
# | Format:          Float32
# | Channels:        4
# | ArraySize:       2
# | Data:
# |   - [ 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1 ]
# |   - [ 0, 0, 2, 2, 0, 2, 0, 2, 2, 0, 0, 2, 0, 2, 2, 2 ]
# | OutputProps:
# |   Height:          2
# |   Width:           2
# |   Depth:           1
# | ...
# | Got:
# | ---
# | Name:            Out
# | Format:          Float32
# | Channels:        4
# | ArraySize:       2
# | Data:
# |   - [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
# |   - [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]

Texture2D.SRVToUAV.test.yaml

Name:            ExpectedOut
# | Format:          Float32
# | Channels:        4
# | Data:            [ 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1 ]
# | OutputProps:
# |   Height:          2
# |   Width:           2
# |   Depth:           1
# | ...
# | Got:
# | ---
# | Name:            Out
# | Format:          Float32
# | Channels:        4
# | Data:            [ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
# | OutputProps:
# |   Height:          2
# |   Width:           2
# |   Depth:           1

[farzonl]

Looks like we need to file a bug for all of WaveOps against Xe intel gpus.

Exec-Tests-Windows (windows-intel, check-hlsl-clang-d3d12)

Failed Tests (20):
  OffloadTest-clang-d3d12 :: WaveOps/ComponentAccumulationDataRace.test
  OffloadTest-clang-d3d12 :: WaveOps/ComponentDataRace.test
  OffloadTest-clang-d3d12 :: WaveOps/GroupSharedMatrixEltComponentDataRace.test
  OffloadTest-clang-d3d12 :: WaveOps/GroupSharedMatrixRowComponentDataRace.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.fp16.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.fp32.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.fp64.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.int16.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.int32.test
  OffloadTest-clang-d3d12 :: WaveOps/WaveActiveSum.int64.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixProduct.32.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixProduct.fp16.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixProduct.fp64.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixProduct.int16.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixProduct.int64.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixSum.32.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixSum.fp16.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixSum.fp64.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixSum.int16.test
  OffloadTest-clang-d3d12 :: WaveOps/WavePrefixSum.int64.test

#1293