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

// Generated by decuda.
// Host-side runtime calls are translated to the `cust` crate. Kernels
// are emitted as `TODO(decuda)` blocks: rust-gpu translation requires
// the kernel to be authored as a Rust fn. See `examples/` for a
// scaffolded SPIR-V kernel module you can flesh out.
//
// Add to your Cargo.toml:
//   [dependencies]
//   cust = "0.3"

// Fixture: only header includes, exercising every include-replacement path
// known to the parser (cuda_runtime.h, cuda.h, device_functions.h, and a
// generic cuda_*.h header).
// was: #include cuda_runtime.h  ->  cust::cuda_build_setup() /* TODO: import cust crate */
// was: #include cuda.h  ->  cust::cuda_build_setup()
#include <device_functions.h>
// was: #include cuda_fp16.h  ->  /* TODO: replace with target equivalent of cuda_fp16.h */

int main() { return 0; }