sgl-project--sglang/python/sglang/jit_kernel/csrc/gemm/marlin/marlin.cuh

#pragma once

#include <sgl_kernel/utils.cuh>

#include <iostream>

namespace device::marlin {
// Marlin params

// 8 warps are a good choice since every SM has 4 schedulers and having more
// than 1 warp per schedule allows some more latency hiding. At the same time,
// we want relatively few warps to have many registers per warp and small tiles.
static constexpr int default_threads = 256;

static constexpr int pipe_stages = 4;  // 4 pipeline stages fit into shared memory

static constexpr int min_thread_n = 64;
static constexpr int min_thread_k = 64;
static constexpr int max_thread_n = 256;

static constexpr int tile_size = 16;
static constexpr int max_par = 16;

// Repack params
static constexpr int repack_stages = 8;

static constexpr int repack_threads = 256;

static constexpr int tile_k_size = tile_size;
static constexpr int tile_n_size = tile_k_size * 4;

// Helpers
template <typename T, int n>
struct Vec {
  T elems[n];
  __device__ T& operator[](int i) {
    return elems[i];
  }
};

using I4 = Vec<int, 4>;

#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
// No support for async
#else

__device__ inline void cp_async4_pred(void* smem_ptr, const void* glob_ptr, bool pred = true) {
  const int BYTES = 16;
  uint32_t smem = static_cast<uint32_t>(__cvta_generic_to_shared(smem_ptr));
  asm volatile(
      "{\n"
      "   .reg .pred p;\n"
      "   setp.ne.b32 p, %0, 0;\n"
      "   @p cp.async.cg.shared.global [%1], [%2], %3;\n"
      "}\n" ::"r"((int)pred),
      "r"(smem),
      "l"(glob_ptr),
      "n"(BYTES));
}

__device__ inline void cp_async4(void* smem_ptr, const void* glob_ptr) {
  const int BYTES = 16;
  uint32_t smem = static_cast<uint32_t>(__cvta_generic_to_shared(smem_ptr));
  asm volatile(
      "{\n"
      "   cp.async.cg.shared.global [%0], [%1], %2;\n"
      "}\n" ::"r"(smem),
      "l"(glob_ptr),
      "n"(BYTES));
}

__device__ inline void cp_async_fence() {
  asm volatile("cp.async.commit_group;\n" ::);
}

template <int n>
__device__ inline void cp_async_wait() {
  asm volatile("cp.async.wait_group %0;\n" ::"n"(n));
}

#endif

}  // namespace device::marlin