Move double-precision tests to DoublePrecision class

[Copilot]

Addresses mmerecki's comment on PR #8134: several tests operating on double-precision values remained in DxilConf_SM69_Vectorized_Core but belong in DxilConf_SM69_Vectorized_Double.

Changes

Moved from DxilConf_SM69_Vectorized_Core to DxilConf_SM69_Vectorized_Double:

• CastToFloat64_* (9 variants: HLSLBool_t, HLSLHalf_t, float, uint16_t, uint32_t, uint64_t, int16_t, int32_t, int64_t)
• AsDouble_uint32_t
• AsUint_SplitDouble_double

Technical change:

• Removed tests from DxilConf_SM69_Vectorized_Core class
• Added same tests to DxilConf_SM69_Vectorized_Double class using HLK_TEST_DOUBLE macro
• Tests are now properly associated with Device.Graphics.D3D12.DXILCore.ShaderModel69.DoublePrecision specification
• Tests organized in sections: "Explicit Cast to Double (from various types)" and "AsType (double precision)"

Test names and OpType template parameters remain unchanged.

Original prompt

Create a follow-up PR in microsoft/DirectXShaderCompiler to address mmerecki's comment on PR #8134 (#8134 (comment)).

Problem:
PR #8134 split long vector execution tests into separate classes based on HLK requirements, introducing DxilConf_SM69_Vectorized_Core and DxilConf_SM69_Vectorized_Double test classes in tools/clang/unittests/HLSLExec/LongVectors.cpp. mmerecki noted that several tests still live under DxilConf_SM69_Vectorized_Core but appear to be double-precision related and should be moved to DxilConf_SM69_Vectorized_Double.

The tests listed by mmerecki:

• DxilConf_SM69_Vectorized_Core::CastToFloat64_HLSLBool_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_HLSLHalf_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_float
• DxilConf_SM69_Vectorized_Core::CastToFloat64_uint16_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_uint32_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_uint64_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_int16_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_int32_t
• DxilConf_SM69_Vectorized_Core::CastToFloat64_int64_t
• DxilConf_SM69_Vectorized_Core::AsDouble_uint32_t
• DxilConf_SM69_Vectorized_Core::AsUint_SplitDouble_double

Expected behavior / fix:

• Update tools/clang/unittests/HLSLExec/LongVectors.cpp so that these tests are associated with the DoublePrecision HLK specification.
• Concretely, remove the above tests from DxilConf_SM69_Vectorized_Core and add equivalent tests to DxilConf_SM69_Vectorized_Double.
  • Prefer using HLK_TEST_DOUBLE for the tests being moved so they pick up the "Device.Graphics.D3D12.DXILCore.ShaderModel69.DoublePrecision" specification.
  • Ensure the resulting test method names remain stable/consistent and compilation passes.
• Validate that moved tests still call the correct OpType values (OpType::CastToFloat64, OpType::AsDouble, OpType::AsUint_SplitDouble) with appropriate template type arguments.

Notes / pointers:

• Relevant file: tools/clang/unittests/HLSLExec/LongVectors.cpp (see locations around DxilConf_SM69_Vectorized_Core and DxilConf_SM69_Vectorized_Double classes).
• DxilConf_SM69_Vectorized_Double class defines HLK_TEST_DOUBLE and already contains many double tests.

Deliverables:

• A new PR with changes only necessary to re-home these tests from Core to Double.
• Include a concise PR description referencing mmerecki's comment and why the tests belong in the DoublePrecision group.

This pull request was created from Copilot chat.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[alsepkow]

@copilot We don't need to group the tests like you did. You can leave them in the original location. Just update the macro call.

[Copilot]

@copilot We don't need to group the tests like you did. You can leave them in the original location. Just update the macro call.

Done in fee4f4a. Changed HLK_TEST to HLK_TEST_DOUBLE for the 11 tests in their original locations without any reorganization.

[alsepkow]

@copilot I was wrong. You can revert that last commit.

[Copilot]

@copilot I was wrong. You can revert that last commit.

Reverted in 13ebe08. The tests are now moved to the Double class as in the original approach.

[tex3d]

LGTM.