decuda 0.1.1

CUDA to HIP, SYCL, OpenCL, and Rust GPU migration tool — automatic source-code translator for porting CUDA C++ kernels to AMD ROCm HIP, Intel oneAPI SYCL, Khronos OpenCL, and Rust GPU (cust / rust-gpu)
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

# decuda examples

This directory holds a comprehensive set of CUDA input files and the
real decuda output for every supported backend.

## Layout

```
examples/
├── cu/                          # CUDA input files (.cu) — single source of truth
│   ├── saxpy.cu                 # kernels, launches, malloc/free, shared mem
│   ├── histogram.cu             # atomics, shared bins, grid-stride, warp intrinsics
│   ├── transpose.cu             # 2D dim3 grid/block, shared-memory tile
│   ├── stream_pipeline.cu       # streams, events, async memcpy, smem+stream launches
│   ├── device_helpers.cu        # __device__ helpers, inline hints, __constant__, __launch_bounds__
│   ├── reduction.cu             # warp-shuffle reduction, __shfl_sync, tree reduction, atomicAdd
│   ├── stencil_3d.cu            # 3D 7-point stencil, shared-memory halo, 3D dim3
│   ├── device_management.cu     # multi-GPU, error handling, pinned memory, cudaDeviceSynchronize
│   ├── managed_memory.cu        # cudaMallocManaged, __managed__, unified memory
│   ├── warp_primitives.cu       # __shfl_sync, __ballot_sync, __any_sync, __all_sync, __activemask
│   ├── ptx_inline.cu            # inline PTX assembly (asm/asm volatile, flagged for manual review)
│   ├── rich.cu                  # atomics, shared mem, warp intrinsics, 2D launches, constant mem
│   ├── launch_in_comment.cu     # launches inside comments/strings (preprocessor invariant test)
│   ├── headers_only.cu          # every include-replacement path
│   └── empty.cu                 # no CUDA constructs (banner + verbatim source)
├── out/                         # decuda output (mirrors real <output>/<target>/ layout)
│   ├── hip/*.hip.cpp            # --target hip
│   ├── sycl/*.sycl.cpp          # --target sycl
│   ├── rust/*.rs                # --target rust  (cust / rust-gpu scaffold)
│   ├── opencl/*.cl              # --target opencl
│   └── migration-report.json.*  # structured per-file warnings report
├── inspect.rs                   # runnable example: programmatic IR inspection
├── migrate.rs                   # runnable example: programmatic in-memory migration
└── list_apis.rs                 # runnable example: enumerate the CUDA API database
```

## Regenerate the outputs

After changing decuda, regenerate the example outputs with:

```bash
cargo run -- migrate -i examples/cu -o examples/out --target all
```

This is exactly the command that produced the current `examples/out/`
tree. Each target writes into its own `out/<target>/` subfolder, and a
single timestamped `migration-report.json.<ts>` is written next to them.

## Run the runnable examples

```bash
cargo run --example inspect -- examples/cu/saxpy.cu
cargo run --example migrate  -- examples/cu/histogram.cu
cargo run --example list_apis -- hip
```