Skip to content

[QC][DirectX] WaveOps/WaveReadLaneFirst fp16, fp32, and fp64 tests fail on Qualcomm under D3D12 #627

New issue
New issue
Open
Bug
Open
[QC][DirectX] WaveOps/WaveReadLaneFirst fp16, fp32, and fp64 tests fail on Qualcomm under D3D12#627
Bug
Labels
driver-bugBugs that are likely or confirmed GPU driver bugs
@Icohedron

Description

@Icohedron
Icohedron
opened on Dec 16, 2025

The WaveOps/WaveReadLaneFirst fp16, fp32, and fp64 tests fail on Qualcomm under D3D12 with DXC. (Clang is XFAILed by these tests).

The fp64 failure is due to lack of support for 64-bit floating point. (To be fixed by adding REQUIRES: Double to the test).
The fp16 and fp32 failures are likely driver bugs.

╭───┬──────────────────────┬─────────────┬──────────────────────┬─────────────┬───────────────────────────────────────────╮
│ # │      timestamp       │   run-id    │       workflow       │   status    │                   test                    │
├───┼──────────────────────┼─────────────┼──────────────────────┼─────────────┼───────────────────────────────────────────┤
│ 0 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ UNSUPPORTED │ WaveOps/WaveReadLaneFirst.128Threads.test │
│ 1 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ FAIL        │ WaveOps/WaveReadLaneFirst.fp16.test       │
│ 2 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ FAIL        │ WaveOps/WaveReadLaneFirst.fp32.test       │
│ 3 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ FAIL        │ WaveOps/WaveReadLaneFirst.fp64.test       │
│ 4 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ PASS        │ WaveOps/WaveReadLaneFirst.int16.test      │
│ 5 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ PASS        │ WaveOps/WaveReadLaneFirst.int32.test      │
│ 6 │ 2025-12-16T14:08:13Z │ 20270756753 │ Windows D3D12 QC DXC │ PASS        │ WaveOps/WaveReadLaneFirst.int64.test      │
╰───┴──────────────────────┴─────────────┴──────────────────────┴─────────────┴───────────────────────────────────────────╯

Logs: https://github.com/llvm/offload-test-suite/actions/runs/20270756753/job/58205649320#step:12:788

FAIL: OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp16.test (299 of 305)
******************** TEST 'OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp16.test' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 316
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp16.test C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp16.test' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp'
# RUN: at line 317
"C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe" -enable-16bit-types -T cs_6_5 -Fo C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp.o C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp/source.hlsl
# executed command: 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe' -enable-16bit-types -T cs_6_5 -Fo 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp.o' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp/source.hlsl'
# RUN: at line 318
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe -debug-layer C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp/pipeline.yaml C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp.o
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe' -debug-layer 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp/pipeline.yaml' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp16.test.tmp.o'
# .---command stdout------------
# | Using DirectX API
# | Configuring execution on device: Qualcomm(R) Adreno(TM) X1-85 GPU
# | RootSignature created.
# | Descriptor heap created.
# | Command structures created.
# | Creating SRV: { Size = 256, Register = t0, Space = 0 }
# | Creating UAV: { Size = 64, Register = u1, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 128, Register = u2, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 256, Register = u3, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 256, Register = u4, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 8, Register = u5, Space = 0, HasCounter = 0 }
# | Creating SRV: { Size = 64, Register = t6, Space = 0 }
# |     EltSize = 8 NumElts = 32
# | SRV: HeapIdx = 0 EltSize = 8 NumElts = 32
# |     EltSize = 2 NumElts = 32
# | UAV: HeapIdx = 1 EltSize = 2 NumElts = 32 HasCounter = 0
# |     EltSize = 4 NumElts = 32
# | UAV: HeapIdx = 2 EltSize = 4 NumElts = 32 HasCounter = 0
# |     EltSize = 8 NumElts = 32
# | UAV: HeapIdx = 3 EltSize = 8 NumElts = 32 HasCounter = 0
# |     EltSize = 8 NumElts = 32
# | UAV: HeapIdx = 4 EltSize = 8 NumElts = 32 HasCounter = 0
# |     EltSize = 8 NumElts = 1
# | UAV: HeapIdx = 5 EltSize = 8 NumElts = 1 HasCounter = 0
# |     EltSize = 16 NumElts = 4
# | SRV: HeapIdx = 6 EltSize = 16 NumElts = 4
# | Buffers created.
# | Event prepared.
# | PSO created.
# | Compute command list created.
# | Compute commands executed.
# | Read data back.
# `-----------------------------
# .---command stderr------------
# | Test failed: ExpectedOut5
# | Comparison Rule: BufferExact
# | Expected:
# | ---
# | Name:            ExpectedOut5
# | Format:          Float16
# | Stride:          8
# | Data:            [ 0x3C00, 0x4000, 0x4200, 0x4400 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Got:
# | ---
# | Name:            Out5
# | Format:          Float16
# | Stride:          8
# | Data:            [ 0x3C00, 0x3C00, 0x3C00, 0x3C00 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Full Hex 64bit representation of Expected Buffer Values:
# | [ 0x3c00, 0x4000, 0x4200, 0x4400 ]
# | Full Hex 64bit representation of Actual Buffer Values:
# | [ 0x3c00, 0x3c00, 0x3c00, 0x3c00 ]
# | 
# `-----------------------------
# error: command failed with exit status: 1

--

********************
FAIL: OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp32.test (300 of 305)
******************** TEST 'OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp32.test' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 313
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp32.test C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp32.test' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp'
# RUN: at line 314
"C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe" -T cs_6_5 -Fo C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp.o C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp/source.hlsl
# executed command: 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe' -T cs_6_5 -Fo 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp.o' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp/source.hlsl'
# RUN: at line 315
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe -debug-layer C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp/pipeline.yaml C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp.o
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe' -debug-layer 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp/pipeline.yaml' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp32.test.tmp.o'
# .---command stdout------------
# | Using DirectX API
# | Configuring execution on device: Qualcomm(R) Adreno(TM) X1-85 GPU
# | RootSignature created.
# | Descriptor heap created.
# | Command structures created.
# | Creating SRV: { Size = 512, Register = t0, Space = 0 }
# | Creating UAV: { Size = 128, Register = u1, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 256, Register = u2, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 512, Register = u3, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 512, Register = u4, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 16, Register = u5, Space = 0, HasCounter = 0 }
# | Creating SRV: { Size = 64, Register = t6, Space = 0 }
# |     EltSize = 16 NumElts = 32
# | SRV: HeapIdx = 0 EltSize = 16 NumElts = 32
# |     EltSize = 4 NumElts = 32
# | UAV: HeapIdx = 1 EltSize = 4 NumElts = 32 HasCounter = 0
# |     EltSize = 8 NumElts = 32
# | UAV: HeapIdx = 2 EltSize = 8 NumElts = 32 HasCounter = 0
# |     EltSize = 16 NumElts = 32
# | UAV: HeapIdx = 3 EltSize = 16 NumElts = 32 HasCounter = 0
# |     EltSize = 16 NumElts = 32
# | UAV: HeapIdx = 4 EltSize = 16 NumElts = 32 HasCounter = 0
# |     EltSize = 16 NumElts = 1
# | UAV: HeapIdx = 5 EltSize = 16 NumElts = 1 HasCounter = 0
# |     EltSize = 16 NumElts = 4
# | SRV: HeapIdx = 6 EltSize = 16 NumElts = 4
# | Buffers created.
# | Event prepared.
# | PSO created.
# | Compute command list created.
# | Compute commands executed.
# | Read data back.
# `-----------------------------
# .---command stderr------------
# | Test failed: ExpectedOut5
# | Comparison Rule: BufferExact
# | Expected:
# | ---
# | Name:            ExpectedOut5
# | Format:          Float32
# | Stride:          8
# | Data:            [ 1.5, 2.5, 3.5, 4.5 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Got:
# | ---
# | Name:            Out5
# | Format:          Float32
# | Stride:          16
# | Data:            [ 1.5, 1.5, 1.5, 1.5 ]
# | OutputProps:
# |   Height:          0
# |   Width:           0
# |   Depth:           0
# | ...
# | Full Hex 64bit representation of Expected Buffer Values:
# | [ 0x1.5, 0x2.5, 0x3.5, 0x4.5 ]
# | Full Hex 64bit representation of Actual Buffer Values:
# | [ 0x1.5, 0x1.5, 0x1.5, 0x1.5 ]
# | 
# `-----------------------------
# error: command failed with exit status: 1

--

********************
FAIL: OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp64.test (301 of 305)
******************** TEST 'OffloadTest-d3d12 :: WaveOps/WaveReadLaneFirst.fp64.test' FAILED ********************
Exit Code: 2170

Command Output (stdout):
--
# RUN: at line 316
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp64.test C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\split-file.exe' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\OffloadTest\test\WaveOps\WaveReadLaneFirst.fp64.test' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp'
# RUN: at line 317
"C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe" -T cs_6_5 -Fo C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp.o C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp/source.hlsl
# executed command: 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\DXC\build\bin\dxc.exe' -T cs_6_5 -Fo 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp.o' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp/source.hlsl'
# RUN: at line 318
c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe -debug-layer C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp/pipeline.yaml C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp.o
# executed command: 'c:\users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\bin\offloader.exe' -debug-layer 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp/pipeline.yaml' 'C:\Users\github\dev\_work\offload-test-suite\offload-test-suite\llvm-project\build\tools\OffloadTest\test\d3d12\WaveOps\Output\WaveReadLaneFirst.fp64.test.tmp.o'
# .---command stdout------------
# | Using DirectX API
# | Configuring execution on device: Qualcomm(R) Adreno(TM) X1-85 GPU
# | RootSignature created.
# | Descriptor heap created.
# | Command structures created.
# | Creating SRV: { Size = 1024, Register = t0, Space = 0 }
# | Creating UAV: { Size = 256, Register = u1, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 512, Register = u2, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 1024, Register = u3, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 1024, Register = u4, Space = 0, HasCounter = 0 }
# | Creating UAV: { Size = 32, Register = u5, Space = 0, HasCounter = 0 }
# | Creating SRV: { Size = 64, Register = t6, Space = 0 }
# |     EltSize = 32 NumElts = 32
# | SRV: HeapIdx = 0 EltSize = 32 NumElts = 32
# |     EltSize = 4 NumElts = 64
# | UAV: HeapIdx = 1 EltSize = 4 NumElts = 64 HasCounter = 0
# |     EltSize = 8 NumElts = 64
# | UAV: HeapIdx = 2 EltSize = 8 NumElts = 64 HasCounter = 0
# |     EltSize = 16 NumElts = 64
# | UAV: HeapIdx = 3 EltSize = 16 NumElts = 64 HasCounter = 0
# |     EltSize = 16 NumElts = 64
# | UAV: HeapIdx = 4 EltSize = 16 NumElts = 64 HasCounter = 0
# |     EltSize = 32 NumElts = 1
# | UAV: HeapIdx = 5 EltSize = 32 NumElts = 1 HasCounter = 0
# |     EltSize = 16 NumElts = 4
# | SRV: HeapIdx = 6 EltSize = 16 NumElts = 4
# | Buffers created.
# | Event prepared.
# `-----------------------------
# .---command stderr------------
# | PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace and instructions to reproduce the bug.
# | Stack dump:
# | 0.	Program arguments: c:\\users\\github\\dev\\_work\\offload-test-suite\\offload-test-suite\\llvm-project\\build\\bin\\offloader.exe -debug-layer C:\\Users\\github\\dev\\_work\\offload-test-suite\\offload-test-suite\\llvm-project\\build\\tools\\OffloadTest\\test\\d3d12\\WaveOps\\Output\\WaveReadLaneFirst.fp64.test.tmp/pipeline.yaml C:\\Users\\github\\dev\\_work\\offload-test-suite\\offload-test-suite\\llvm-project\\build\\tools\\OffloadTest\\test\\d3d12\\WaveOps\\Output\\WaveReadLaneFirst.fp64.test.tmp.o
# | D3D: ID3D12Device::CreateComputeShader: Shader uses double precision float ops which are not supported on the current device.
# | Exception Code: 0x0000087A
# | #0 0x00007ffece6050d4 (C:\Windows\System32\KERNELBASE.dll+0x1250d4)
# | #1 0x0e487ffea31862e8
# `-----------------------------
# error: command failed with exit status: 0x87a

--

Metadata

Metadata

Assignees

No one assigned

    Labels

    driver-bugBugs that are likely or confirmed GPU driver bugs

    Type

    Bug

    Projects

    HLSL Support

    Status

    No status

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions