80525dfcde
Through the new `-foffload-via-llvm` flag, CUDA kernels can now be lowered to the LLVM/Offload API. On the Clang side, this is simply done by using the OpenMP offload toolchain and emitting calls to `llvm*` functions to orchestrate the kernel launch rather than `cuda*` functions. These `llvm*` functions are implemented on top of the existing LLVM/Offload API. As we are about to redefine the Offload API, this wil help us in the design process as a second offload language. We do not support any CUDA APIs yet, however, we could: https://www.osti.gov/servlets/purl/1892137 For proper host execution we need to resurrect/rebase https://tianshilei.me/wp-content/uploads/2021/12/llpp-2021.pdf (which was designed for debugging). ``` ❯❯❯ cat test.cu extern "C" { void *llvm_omp_target_alloc_shared(size_t Size, int DeviceNum); void llvm_omp_target_free_shared(void *DevicePtr, int DeviceNum); } __global__ void square(int *A) { *A = 42; } int main(int argc, char **argv) { int DevNo = 0; int *Ptr = reinterpret_cast<int *>(llvm_omp_target_alloc_shared(4, DevNo)); *Ptr = 7; printf("Ptr %p, *Ptr %i\n", Ptr, *Ptr); square<<<1, 1>>>(Ptr); printf("Ptr %p, *Ptr %i\n", Ptr, *Ptr); llvm_omp_target_free_shared(Ptr, DevNo); } ❯❯❯ clang++ test.cu -O3 -o test123 -foffload-via-llvm --offload-arch=native ❯❯❯ llvm-objdump --offloading test123 test123: file format elf64-x86-64 OFFLOADING IMAGE [0]: kind elf arch gfx90a triple amdgcn-amd-amdhsa producer openmp ❯❯❯ LIBOMPTARGET_INFO=16 ./test123 Ptr 0x155448ac8000, *Ptr 7 Ptr 0x155448ac8000, *Ptr 42 ```
42 lines
1.3 KiB
Plaintext
42 lines
1.3 KiB
Plaintext
// RUN: %clang++ -foffload-via-llvm --offload-arch=native %s -o %t
|
|
// RUN: %t | %fcheck-generic
|
|
// RUN: %clang++ -foffload-via-llvm --offload-arch=native %s -o %t -fopenmp
|
|
// RUN: %t | %fcheck-generic
|
|
|
|
// UNSUPPORTED: aarch64-unknown-linux-gnu
|
|
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
|
|
// UNSUPPORTED: x86_64-pc-linux-gnu
|
|
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
|
|
|
|
#include <stdio.h>
|
|
|
|
extern "C" {
|
|
void *llvm_omp_target_alloc_shared(size_t Size, int DeviceNum);
|
|
void llvm_omp_target_free_shared(void *DevicePtr, int DeviceNum);
|
|
}
|
|
|
|
__global__ void square(int *Dst, short Q, int *Src, short P) {
|
|
*Dst = (Src[0] + Src[1]) * (Q + P);
|
|
Src[0] = Q;
|
|
Src[1] = P;
|
|
}
|
|
|
|
int main(int argc, char **argv) {
|
|
int DevNo = 0;
|
|
int *Ptr = reinterpret_cast<int *>(llvm_omp_target_alloc_shared(4, DevNo));
|
|
int *Src = reinterpret_cast<int *>(llvm_omp_target_alloc_shared(8, DevNo));
|
|
*Ptr = 7;
|
|
Src[0] = -2;
|
|
Src[1] = 8;
|
|
printf("Ptr %p, *Ptr: %i\n", Ptr, *Ptr);
|
|
// CHECK: Ptr [[Ptr:0x.*]], *Ptr: 7
|
|
printf("Src: %i : %i\n", Src[0], Src[1]);
|
|
// CHECK: Src: -2 : 8
|
|
square<<<1, 1>>>(Ptr, 3, Src, 4);
|
|
printf("Ptr %p, *Ptr: %i\n", Ptr, *Ptr);
|
|
// CHECK: Ptr [[Ptr]], *Ptr: 42
|
|
printf("Src: %i : %i\n", Src[0], Src[1]);
|
|
// CHECK: Src: 3 : 4
|
|
llvm_omp_target_free_shared(Ptr, DevNo);
|
|
}
|