sglang_v0.5.2/flashinfer_0.3.1/csrc/vllm_custom_all_reduce.cu

// flashinfer: adapted from sglang + vllm code
// refer to: https://github.com/vllm-project/vllm/blob/v0.8.2/csrc/custom_all_reduce.cu
#include "flashinfer/comm/vllm_custom_all_reduce.cuh"
#include "pytorch_extension_utils.h"

// Fake pointer type, must match fptr_t type in ops.h.
// We use this type alias to indicate when pointers are passed in as int64_t.
using fptr_t = int64_t;
static_assert(sizeof(void*) == sizeof(fptr_t));

fptr_t init_custom_ar(const std::vector<fptr_t>& fake_ipc_ptrs, at::Tensor& rank_data, int64_t rank,
                      bool full_nvlink) {
  int world_size = fake_ipc_ptrs.size();
  if (world_size > 8) throw std::invalid_argument("world size > 8 is not supported");
  if (world_size % 2 != 0) throw std::invalid_argument("Odd num gpus is not supported for now");
  if (rank < 0 || rank >= world_size) throw std::invalid_argument("invalid rank passed in");

  vllm::Signal* ipc_ptrs[8];
  for (int i = 0; i < world_size; i++) {
    ipc_ptrs[i] = reinterpret_cast<vllm::Signal*>(fake_ipc_ptrs[i]);
  }
  return (fptr_t) new vllm::CustomAllreduce(ipc_ptrs, rank_data.data_ptr(), rank_data.numel(), rank,
                                            world_size, full_nvlink);
}

/**
 * Make sure tensor t's data lies completely within ((char)t.data_ptr()) +
 * t.numel() * t.element_size(). This is slightly weaker than t.is_contiguous()
 * because it allows transpose of contiguous slice (i.e. slicing the first
 * dimension). Currently, we require this because stride information is not
 * passed into the kernels and we treat input tensors as flat.
 *
 * Examples
 * A = torch.zeros(3, 3, 3)
 * 1. A: OK
 * 2. A[1:]: OK
 * 3. A.permute(2, 0, 1): OK
 * 4. A[1:].permute(2, 0, 1): OK
 * 5. A[None].expand(2, -1, -1, -1): Not OK
 * 6. A[:, 1:, 1:]: Not OK
 */
bool _is_weak_contiguous(at::Tensor& t) {
  return t.is_contiguous() || (t.storage().nbytes() - t.storage_offset() * t.element_size() ==
                               t.numel() * t.element_size());
}

/**
 * Performs an out-of-place allreduce and stores result in out.
 *
 * If _reg_buffer is null, assumes inp.data_ptr() is already IPC-registered.
 * Otherwise, _reg_buffer is assumed to be IPC-registered and inp is first
 * copied into _reg_buffer.
 */
void all_reduce(fptr_t _fa, at::Tensor& inp, at::Tensor& out, fptr_t _reg_buffer,
                int64_t reg_buffer_sz_bytes, int64_t num_ctas) {
  auto fa = reinterpret_cast<vllm::CustomAllreduce*>(_fa);
  const at::cuda::OptionalCUDAGuard device_guard(inp.device());
  auto stream = c10::cuda::getCurrentCUDAStream().stream();

  TORCH_CHECK_EQ(inp.scalar_type(), out.scalar_type());
  TORCH_CHECK_EQ(inp.numel(), out.numel());
  TORCH_CHECK(_is_weak_contiguous(out));
  TORCH_CHECK(_is_weak_contiguous(inp));
  auto input_size = inp.numel() * inp.element_size();
  auto reg_buffer = reinterpret_cast<void*>(_reg_buffer);
  if (reg_buffer) {
    TORCH_CHECK_LE(input_size, reg_buffer_sz_bytes);
    auto status =
        cudaMemcpyAsync(reg_buffer, inp.data_ptr(), input_size, cudaMemcpyDeviceToDevice, stream);
    TORCH_CHECK(status == cudaSuccess);
  } else {
    reg_buffer = inp.data_ptr();
  }
  switch (out.scalar_type()) {
    case at::ScalarType::Float: {
      fa->allreduce<float>(stream, reinterpret_cast<float*>(reg_buffer),
                           reinterpret_cast<float*>(out.data_ptr()), out.numel(), num_ctas);
      break;
    }
    case at::ScalarType::Half: {
      fa->allreduce<half>(stream, reinterpret_cast<half*>(reg_buffer),
                          reinterpret_cast<half*>(out.data_ptr()), out.numel(), num_ctas);
      break;
    }
#if (__CUDA_ARCH__ >= 800 || !defined(__CUDA_ARCH__))
    case at::ScalarType::BFloat16: {
      fa->allreduce<nv_bfloat16>(stream, reinterpret_cast<nv_bfloat16*>(reg_buffer),
                                 reinterpret_cast<nv_bfloat16*>(out.data_ptr()), out.numel(),
                                 num_ctas);
      break;
    }
#endif
    default:
      throw std::runtime_error("custom allreduce only supports float32, float16 and bfloat16");
  }
}

void dispose(fptr_t _fa) { delete reinterpret_cast<vllm::CustomAllreduce*>(_fa); }

int64_t meta_size() { return sizeof(vllm::Signal); }

void register_buffer(fptr_t _fa, const std::vector<fptr_t>& fake_ipc_ptrs) {
  auto fa = reinterpret_cast<vllm::CustomAllreduce*>(_fa);
  TORCH_CHECK(fake_ipc_ptrs.size() == fa->world_size_);
  void* ipc_ptrs[8];
  for (int i = 0; i < fake_ipc_ptrs.size(); i++) {
    ipc_ptrs[i] = reinterpret_cast<void*>(fake_ipc_ptrs[i]);
  }
  fa->register_buffer(ipc_ptrs);
}

// Use vector<int64_t> to represent byte data for python binding compatibility.
std::tuple<std::vector<int64_t>, std::vector<int64_t>> get_graph_buffer_ipc_meta(fptr_t _fa) {
  auto fa = reinterpret_cast<vllm::CustomAllreduce*>(_fa);
  auto [handle, offsets] = fa->get_graph_buffer_ipc_meta();
  std::vector<int64_t> bytes(handle.begin(), handle.end());
  return std::make_tuple(bytes, offsets);
}

// Use vector<int64_t> to represent byte data for python binding compatibility.
void register_graph_buffers(fptr_t _fa, const std::vector<std::vector<int64_t>>& handles,
                            const std::vector<std::vector<int64_t>>& offsets) {
  auto fa = reinterpret_cast<vllm::CustomAllreduce*>(_fa);
  std::vector<std::string> bytes;
  bytes.reserve(handles.size());
  for (int i = 0; i < handles.size(); i++) {
    bytes.emplace_back(handles[i].begin(), handles[i].end());
  }
  bytes.reserve(handles.size());
  fa->register_graph_buffers(bytes, offsets);
}

/*
void AllReduceSum(at::Tensor data, at::Tensor workspace, int64_t world_size, int64_t rank,
                  int64_t num_ctas
                  ) {
  printf("AllReduce called with num_ctas = %d\n", (int)num_ctas);

  float* workspace_ptr = workspace.data_ptr<float>();
  auto dtype = data.scalar_type();
  int hidden_size = data.size(-1);
  int token_num = data.numel() / hidden_size;
  auto fusion_op = tensorrt_llm::kernels::AllReduceFusionOp::NONE;
  if (fusion_op.has_value()) {
      auto fusion_op = fusion_op.value();
  } else {
      auto fusion_op = tensorrt_llm::kernels::AllReduceFusionOp::NONE;
  }
  auto stream = at::cuda::getCurrentCUDAStream();

  auto params = tensorrt_llm::kernels::AllReduceParams::deserialize(
      reinterpret_cast<int64_t*>(workspace_ptr), world_size, rank, dtype, token_num, hidden_size,
      fusion_op);

  auto strat_config = tensorrt_llm::kernels::AllReduceStrategyConfig::PUSH_MODE;
  auto strat_type = tensorrt_llm::kernels::AllReduceStrategyType::AUTO;

  customAllReduce(params, dtype, strat_type, strat_config, fusion_op, stream, num_ctas);
}
*/

TORCH_LIBRARY_FRAGMENT(TORCH_EXTENSION_NAME, m) {
  m.def("get_graph_buffer_ipc_meta", &get_graph_buffer_ipc_meta);
  m.def("register_graph_buffers", &register_graph_buffers);
  m.def("dispose", &dispose);
  m.def("meta_size", &meta_size);
  m.def("register_buffer", &register_buffer);
  m.def("init_custom_ar", &init_custom_ar);
  m.def("all_reduce", &all_reduce);
}