Skip to content

Add Texture-Row-Pitch-Readback.test #1167

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
EmilioLaiso wants to merge 2 commits into
base: main
Choose a base branch
from
+288 −0
Open

Add Texture-Row-Pitch-Readback.test #1167

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ac21fa2
add test for row pitch alignment in readbacks
EmilioLaiso May 11, 2026
7e487f1
Add Clang+Dx12 XFAIL
EmilioLaiso May 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
288 changes: 288 additions & 0 deletions test/Bugs/Texture-Row-Pitch-Readback.test
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,288 @@
#--- vertex.hlsl
struct PSInput {
float4 position : SV_POSITION;
};

PSInput main(float4 position : POSITION) {
PSInput o;
o.position = position;
return o;
}

#--- pixel.hlsl
struct PSInput {
float4 position : SV_POSITION;
};

// Emit a distinct color per pixel by encoding SV_POSITION.xy into the output.
// A readback that mis-strides the staging buffer would not match
// these expected per-pixel coordinates.
float4 main(PSInput input) : SV_TARGET {
return float4(input.position.x, input.position.y, 0.0, 1.0);
}

#--- pipeline-2x2.yaml
---
Shaders:
- Stage: Vertex
Entry: main
- Stage: Pixel
Entry: main
Buffers:
- Name: VertexData
Format: Float32
Stride: 12
Data: [ -1.0, -1.0, 0.0,
-1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, -1.0, 0.0,
-1.0, -1.0, 0.0 ]
- Name: Output
Format: Float32
Channels: 4
FillSize: 64 # 2x2 @ 16 B/pixel
OutputProps:
Width: 2
Height: 2
Depth: 1
- Name: Output_Expected
Format: Float32
Channels: 4
Data: [
# Row 0 (y = 0.5)
0.5, 0.5, 0.0, 1.0, 1.5, 0.5, 0.0, 1.0,
# Row 1 (y = 1.5)
0.5, 1.5, 0.0, 1.0, 1.5, 1.5, 0.0, 1.0,
]
Bindings:
VertexBuffer: VertexData
VertexAttributes:
- Format: Float32
Channels: 3
Offset: 0
Name: POSITION
RenderTarget: Output
DescriptorSets: []
Results:
- Result: ReadbackPixels
Rule: BufferExact
Actual: Output
Expected: Output_Expected
...

#--- pipeline-3x3.yaml
---
Shaders:
- Stage: Vertex
Entry: main
- Stage: Pixel
Entry: main
Buffers:
- Name: VertexData
Format: Float32
Stride: 12
Data: [ -1.0, -1.0, 0.0,
-1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, -1.0, 0.0,
-1.0, -1.0, 0.0 ]
- Name: Output
Format: Float32
Channels: 4
FillSize: 144 # 3x3 @ 16 B/pixel
OutputProps:
Width: 3
Height: 3
Depth: 1
- Name: Output_Expected
Format: Float32
Channels: 4
Data: [
# Row 0 (y = 0.5)
0.5, 0.5, 0.0, 1.0, 1.5, 0.5, 0.0, 1.0, 2.5, 0.5, 0.0, 1.0,
# Row 1 (y = 1.5)
0.5, 1.5, 0.0, 1.0, 1.5, 1.5, 0.0, 1.0, 2.5, 1.5, 0.0, 1.0,
# Row 2 (y = 2.5)
0.5, 2.5, 0.0, 1.0, 1.5, 2.5, 0.0, 1.0, 2.5, 2.5, 0.0, 1.0,
]
Bindings:
VertexBuffer: VertexData
VertexAttributes:
- Format: Float32
Channels: 3
Offset: 0
Name: POSITION
RenderTarget: Output
DescriptorSets: []
Results:
- Result: ReadbackPixels
Rule: BufferExact
Actual: Output
Expected: Output_Expected
...

#--- pipeline-16x3.yaml
---
Shaders:
- Stage: Vertex
Entry: main
- Stage: Pixel
Entry: main
Buffers:
- Name: VertexData
Format: Float32
Stride: 12
Data: [ -1.0, -1.0, 0.0,
-1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, -1.0, 0.0,
-1.0, -1.0, 0.0 ]
- Name: Output
Format: Float32
Channels: 4
FillSize: 768 # 16x3 @ 16 B/pixel
OutputProps:
Width: 16
Height: 3
Depth: 1
- Name: Output_Expected
Format: Float32
Channels: 4
Data: [
# Row 0 (y = 0.5)
0.5, 0.5, 0.0, 1.0, 1.5, 0.5, 0.0, 1.0, 2.5, 0.5, 0.0, 1.0, 3.5, 0.5, 0.0, 1.0,
4.5, 0.5, 0.0, 1.0, 5.5, 0.5, 0.0, 1.0, 6.5, 0.5, 0.0, 1.0, 7.5, 0.5, 0.0, 1.0,
8.5, 0.5, 0.0, 1.0, 9.5, 0.5, 0.0, 1.0, 10.5, 0.5, 0.0, 1.0, 11.5, 0.5, 0.0, 1.0,
12.5, 0.5, 0.0, 1.0, 13.5, 0.5, 0.0, 1.0, 14.5, 0.5, 0.0, 1.0, 15.5, 0.5, 0.0, 1.0,
# Row 1 (y = 1.5)
0.5, 1.5, 0.0, 1.0, 1.5, 1.5, 0.0, 1.0, 2.5, 1.5, 0.0, 1.0, 3.5, 1.5, 0.0, 1.0,
4.5, 1.5, 0.0, 1.0, 5.5, 1.5, 0.0, 1.0, 6.5, 1.5, 0.0, 1.0, 7.5, 1.5, 0.0, 1.0,
8.5, 1.5, 0.0, 1.0, 9.5, 1.5, 0.0, 1.0, 10.5, 1.5, 0.0, 1.0, 11.5, 1.5, 0.0, 1.0,
12.5, 1.5, 0.0, 1.0, 13.5, 1.5, 0.0, 1.0, 14.5, 1.5, 0.0, 1.0, 15.5, 1.5, 0.0, 1.0,
# Row 2 (y = 2.5)
0.5, 2.5, 0.0, 1.0, 1.5, 2.5, 0.0, 1.0, 2.5, 2.5, 0.0, 1.0, 3.5, 2.5, 0.0, 1.0,
4.5, 2.5, 0.0, 1.0, 5.5, 2.5, 0.0, 1.0, 6.5, 2.5, 0.0, 1.0, 7.5, 2.5, 0.0, 1.0,
8.5, 2.5, 0.0, 1.0, 9.5, 2.5, 0.0, 1.0, 10.5, 2.5, 0.0, 1.0, 11.5, 2.5, 0.0, 1.0,
12.5, 2.5, 0.0, 1.0, 13.5, 2.5, 0.0, 1.0, 14.5, 2.5, 0.0, 1.0, 15.5, 2.5, 0.0, 1.0,
]
Bindings:
VertexBuffer: VertexData
VertexAttributes:
- Format: Float32
Channels: 3
Offset: 0
Name: POSITION
RenderTarget: Output
DescriptorSets: []
Results:
- Result: ReadbackPixels
Rule: BufferExact
Actual: Output
Expected: Output_Expected
...

#--- pipeline-17x3.yaml
---
Shaders:
- Stage: Vertex
Entry: main
- Stage: Pixel
Entry: main
Buffers:
- Name: VertexData
Format: Float32
Stride: 12
Data: [ -1.0, -1.0, 0.0,
-1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, 1.0, 0.0,
1.0, -1.0, 0.0,
-1.0, -1.0, 0.0 ]
- Name: Output
Format: Float32
Channels: 4
FillSize: 816 # 17x3 @ 16 B/pixel
OutputProps:
Width: 17
Height: 3
Depth: 1
- Name: Output_Expected
Format: Float32
Channels: 4
Data: [
# Row 0 (y = 0.5)
0.5, 0.5, 0.0, 1.0, 1.5, 0.5, 0.0, 1.0, 2.5, 0.5, 0.0, 1.0, 3.5, 0.5, 0.0, 1.0,
4.5, 0.5, 0.0, 1.0, 5.5, 0.5, 0.0, 1.0, 6.5, 0.5, 0.0, 1.0, 7.5, 0.5, 0.0, 1.0,
8.5, 0.5, 0.0, 1.0, 9.5, 0.5, 0.0, 1.0, 10.5, 0.5, 0.0, 1.0, 11.5, 0.5, 0.0, 1.0,
12.5, 0.5, 0.0, 1.0, 13.5, 0.5, 0.0, 1.0, 14.5, 0.5, 0.0, 1.0, 15.5, 0.5, 0.0, 1.0,
16.5, 0.5, 0.0, 1.0,
# Row 1 (y = 1.5)
0.5, 1.5, 0.0, 1.0, 1.5, 1.5, 0.0, 1.0, 2.5, 1.5, 0.0, 1.0, 3.5, 1.5, 0.0, 1.0,
4.5, 1.5, 0.0, 1.0, 5.5, 1.5, 0.0, 1.0, 6.5, 1.5, 0.0, 1.0, 7.5, 1.5, 0.0, 1.0,
8.5, 1.5, 0.0, 1.0, 9.5, 1.5, 0.0, 1.0, 10.5, 1.5, 0.0, 1.0, 11.5, 1.5, 0.0, 1.0,
12.5, 1.5, 0.0, 1.0, 13.5, 1.5, 0.0, 1.0, 14.5, 1.5, 0.0, 1.0, 15.5, 1.5, 0.0, 1.0,
16.5, 1.5, 0.0, 1.0,
# Row 2 (y = 2.5)
0.5, 2.5, 0.0, 1.0, 1.5, 2.5, 0.0, 1.0, 2.5, 2.5, 0.0, 1.0, 3.5, 2.5, 0.0, 1.0,
4.5, 2.5, 0.0, 1.0, 5.5, 2.5, 0.0, 1.0, 6.5, 2.5, 0.0, 1.0, 7.5, 2.5, 0.0, 1.0,
8.5, 2.5, 0.0, 1.0, 9.5, 2.5, 0.0, 1.0, 10.5, 2.5, 0.0, 1.0, 11.5, 2.5, 0.0, 1.0,
12.5, 2.5, 0.0, 1.0, 13.5, 2.5, 0.0, 1.0, 14.5, 2.5, 0.0, 1.0, 15.5, 2.5, 0.0, 1.0,
16.5, 2.5, 0.0, 1.0,
]
Bindings:
VertexBuffer: VertexData
VertexAttributes:
- Format: Float32
Channels: 3
Offset: 0
Name: POSITION
RenderTarget: Output
DescriptorSets: []
Results:
- Result: ReadbackPixels
Rule: BufferExact
Actual: Output
Expected: Output_Expected
...
#--- end

# Regression test for
# https://github.com/llvm/offload-test-suite/pull/1126 -
# D3D12 requires the placed-footprint RowPitch be a multiple of
# D3D12_TEXTURE_DATA_PITCH_ALIGNMENT (256 bytes). For render targets whose
# natural row size (Width * elementSize) is not already a multiple of 256,
# the readback path must walk the staging buffer at the aligned pitch, not
# the natural pitch. Before #1126 the buggy readback would surface garbage /
# zeroed pixels in row 1+ of any small or unaligned RT. Sub-256-byte rows on
# Vulkan and Metal are handled by their respective backends already; this
# test still runs everywhere so the invariant is enforced across the matrix.
#
# The four pipelines below share the same shaders (pixel output = (x+0.5,
# y+0.5, 0, 1) at each pixel center). The interesting variable is the RT
# dimensions:
#
# Size | Row bytes | Aligned? | Why include it
# 2 x 2 | 32 | no | Original repro from PR #1126.
# 3 x 3 | 48 | no | Square, multiple non-final rows.
# 16 x 3 | 256 | yes | Control: already-aligned fast path.
# 17 x 3 | 272 | no | Row > 256 but crosses to the 512 boundary,
# exercises different alignment arithmetic.
#
# Height > 1 is required everywhere -- a 1-row RT would only read the
# unpadded last row and silently sidestep the bug.

# Clang's HLSL -> DXIL lowering does not yet implement the graphics-stage
# signature intrinsics (llvm.dx.load.input / llvm.dx.store.output) needed
# for SV_POSITION / SV_TARGET plumbing; clang -> SPIR-V is unaffected.
# XFAIL: Clang && DirectX

# RUN: split-file %s %t
# RUN: %dxc_target -T vs_6_0 -Fo %t-vertex.o %t/vertex.hlsl
# RUN: %dxc_target -T ps_6_0 -Fo %t-pixel.o %t/pixel.hlsl
# RUN: %offloader %t/pipeline-2x2.yaml %t-vertex.o %t-pixel.o
# RUN: %offloader %t/pipeline-3x3.yaml %t-vertex.o %t-pixel.o
# RUN: %offloader %t/pipeline-16x3.yaml %t-vertex.o %t-pixel.o
# RUN: %offloader %t/pipeline-17x3.yaml %t-vertex.o %t-pixel.o
Loading