尝试编译 AMD ROCm 的 llvm-project

0，环境

ubuntu 22.04

gcc-11

x86_64 18cores/36threads

256GB RAM

rocm 6.0.2

Radeon VII

1，第一次尝试

构建命令：

cmake -G "Unix Makefiles" ../llvm \
-DLLVM_ENABLE_PROJECTS="clang;lld;lldb;mlir;openmp" \
-DLLVM_BUILD_EXAMPLES=ON \
-DLLVM_TARGETS_TO_BUILD="host;AMDGPU" \
-DCMAKE_BUILD_TYPE=Debug \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_RUNTIMES=all \
-DLLVM_BUILD_LLVM_DYLIB=ON \
-DCMAKE_INSTALL_PREFIX=../../local_amdgpu

make -j32

效果：

由于 ncurses 库没有搞定，所以失败了：

估计是lldb的layout src之类的命令会用到 ncurses,

对策：不编译 lldb项目

2，第二次尝试

 删掉了 lldb 和 openmp，调整了TARGET，加入了 NVPTX，

重来：

cmake -G "Unix Makefiles" ../llvm \
-DLLVM_ENABLE_PROJECTS="clang;lld;mlir" \
-DLLVM_BUILD_EXAMPLES=ON \
-DLLVM_TARGETS_TO_BUILD="Native;NVPTX;AMDGPU" \
-DCMAKE_BUILD_TYPE=Debug \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DLLVM_ENABLE_RUNTIMES=all \
-DLLVM_BUILD_LLVM_DYLIB=ON \
-DCMAKE_INSTALL_PREFIX=../../local_amdgpu

make -j34

编译成功，

安装 make install

效果：

3，测试编译rocm hip源码

3.1 源码示例

vectorAdd.hip

#include <stdio.h>
#include <hip/hip_runtime.h>

__global__ void vectorAdd(const float *A, const float *B, float *C) {
  int i = blockDim.x * blockIdx.x + threadIdx.x;

  C[i] = A[i] + B[i] + 0.0f;
}

int main(void) {
  hipError_t err = hipSuccess;
  int numElements = 512;
  size_t size = numElements * sizeof(float);
  printf("[Vector addition of %d elements]\n", numElements);

  float *h_A = (float *)malloc(size);
  float *h_B = (float *)malloc(size);
  float *h_C = (float *)malloc(size);

  if (h_A == NULL || h_B == NULL || h_C == NULL) {
    fprintf(stderr, "Failed to allocate host vectors!\n");
    exit(EXIT_FAILURE);
  }

  for (int i = 0; i < numElements; ++i) {
    h_A[i] = rand() / (float)RAND_MAX;
    h_B[i] = rand() / (float)RAND_MAX;
  }

  float *d_A = NULL;
  err = hipMalloc((void **)&d_A, size);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to allocate device vector A (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  float *d_B = NULL;
  err = hipMalloc((void **)&d_B, size);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to allocate device vector B (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  float *d_C = NULL;
  err = hipMalloc((void **)&d_C, size);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to allocate device vector C (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  printf("Copy input data from the host memory to the CUDA device\n");
  err = hipMemcpy(d_A, h_A, size, hipMemcpyHostToDevice);

  if (err != hipSuccess) {
    fprintf(stderr,
            "Failed to copy vector A from host to device (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  err = hipMemcpy(d_B, h_B, size, hipMemcpyHostToDevice);

  if (err != hipSuccess) {
    fprintf(stderr,
            "Failed to copy vector B from host to device (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  int threadsPerBlock = 256;
  int blocksPerGrid = (numElements + threadsPerBlock - 1) / threadsPerBlock;
  printf("CUDA kernel launch with %d blocks of %d threads\n", blocksPerGrid,
         threadsPerBlock);
  vectorAdd<<<blocksPerGrid, threadsPerBlock>>>(d_A, d_B, d_C);
  err = hipGetLastError();

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to launch vectorAdd kernel (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  printf("Copy output data from the CUDA device to the host memory\n");
  err = hipMemcpy(h_C, d_C, size, hipMemcpyDeviceToHost);

  if (err != hipSuccess) {
    fprintf(stderr,
            "Failed to copy vector C from device to host (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  for (int i = 0; i < numElements; ++i) {
    if (fabs(h_A[i] + h_B[i] - h_C[i]) > 1e-5) {
      fprintf(stderr, "Result verification failed at element %d!\n", i);
      exit(EXIT_FAILURE);
    }
  }

  printf("Test PASSED\n");

  err = hipFree(d_A);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to free device vector A (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  err = hipFree(d_B);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to free device vector B (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  err = hipFree(d_C);

  if (err != hipSuccess) {
    fprintf(stderr, "Failed to free device vector C (error code %s)!\n",
            hipGetErrorString(err));
    exit(EXIT_FAILURE);
  }

  free(h_A);
  free(h_B);
  free(h_C);

  printf("Done\n");
  return 0;
}

3.2 编译运行

$ ../../local_amdgpu/bin/clang++ hello_vectorAdd.hip  --rocm-path=/opt/rocm

这样简单的运行 clang++，还不太确定是不是仅仅起到了编译驱动的作用，后面进一步分析。

运行成功，

效果：

未完待续。。。