Execution Tests: Long Vector WaveActiveSum

[alsepkow]

Adds the basic framework for WaveActiveOp tests and the test cases for WaveActiveSum.
This partially addresses #7472
WARP requires an update for this test to pass so this test will not run in automation for now (no priority set in TAEF metadata).

[damyanp]

WARP requires an update for this test to pass so this test will not run in automation for now

You should explicitly mark these tests as priority 2 to prevent them from running, since we'll be setting the default for all the tests in the class to pri 0.

[alsepkow]

Ah, yes! Much cleaner.

________________________________ From: Damyan Pepper ***@***.***> Sent: Wednesday, November 5, 2025 8:09 PM To: microsoft/DirectXShaderCompiler ***@***.***> Cc: Alex Sepkowski ***@***.***>; Author ***@***.***> Subject: Re: [microsoft/DirectXShaderCompiler] Execution Tests: Long Vector WaveActiveSum (PR #7878) @damyanp commented on this pull request.

________________________________ In tools/clang/unittests/HLSLExec/LongVectors.cpp<#7878 (comment)>:

+

+ MaxWaveSize = waveOpts.WaveLaneCountMax; + } + + DXASSERT_NOMSG(MaxWaveSize > 0); + DXASSERT((MaxWaveSize & (MaxWaveSize - 1)) == 0, "must be a power of 2"); + + dispatchWaveOpTest<T, OP>(D3DDevice, VerboseLogging, OverrideInputSize, + MaxWaveSize); + } + + template <typename T, OpType OP> void runTest() { + WEX::TestExecution::SetVerifyOutput verifySettings( + WEX::TestExecution::VerifyOutputSettings::LogOnlyFailures); + + CheckDevice(); Yup, I understand that. Since we create the device in TEST_CLASS_SETUP I'm suggesting that we check for / recover from a removed device in TEST_METHOD_SETUP rather than having to explicitly sprinkle calls to CheckDevice around. — Reply to this email directly, view it on GitHub<#7878 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABK4EW3DO745FD2EYFJARIT33LCXVAVCNFSM6AAAAACLIAQ7Q6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTIMRVHA2TMMBQGA>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[damyanp]

Some uncontroversial names that just need to be fixed and a suggestion for a simpler type for a string.

Otherwise LGTM.

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[joaosaffran]

Some merges need to be cleaned, other than that LGTM

[damyanp]

Basically I expect this'll LGTM once the compilation errors are fixed :)