Add select matrix tests - #1472
Conversation
| StructuredBuffer<float4> FloatCond : register(t1); | ||
| StructuredBuffer<float4> TrueVal : register(t2); | ||
| StructuredBuffer<float4> FalseVal : register(t3); |
There was a problem hiding this comment.
nit:
| StructuredBuffer<float4> FloatCond : register(t1); | |
| StructuredBuffer<float4> TrueVal : register(t2); | |
| StructuredBuffer<float4> FalseVal : register(t3); | |
| StructuredBuffer<uint4> FloatCond : register(t1); | |
| StructuredBuffer<uint4> TrueVal : register(t2); | |
| StructuredBuffer<uint4> FalseVal : register(t3); |
It might be more clear if a test fails when it is an integer as the dump of the hex representation would be clear. This is more of a thought aloud then an actual request for change
There was a problem hiding this comment.
Hmm that's a good point, though I would have to use int4 because I set FalseVal = -TrueVal, and about half the output values are negative which are just as unreadable in hex (e.g. -35 -> 0xffffffdd). Float32s also still get a readable decimal dump, so I think I'll keep this test as float. But will definitely keep this in mind for future tests
| # XFAIL: Clang | ||
|
|
||
| # RUN: split-file %s %t | ||
| # RUN: %dxc_target -HV 202x -T cs_6_5 -Fo %t.o %t/source.hlsl |
There was a problem hiding this comment.
Oh that was just copy and pasted from other tests. I think almost all other tests (matrix or not) use 6.5 - should we normally be using 6.0 instead? But yes I'll change it.
3 participants
Closes #792.
Adds matrix tests for
selecttestingfloat. Will remove the XFAIL once the Clang implementation is in.Only testing
floatbecauseselectbehaves the same at every element type and bit size, and we already did extensive testing on those in the scalar/vector tests.Assisted-by: Claude Opus 4.8