[sgl-kernel] Opt per_token_quant_fp8 with warp reduce (sgl-project#8130)

Co-authored-by: luoyuan.luo <luoyuan.luo@antgroup.com>

Author: yuan-luo

sgl-kernel/csrc/gemm/per_token_quant_fp8.cu

@@ -1,18 +1,95 @@
 #include <ATen/cuda/CUDAContext.h>
 
 #include <cmath>
-#include <cub/block/block_reduce.cuh>
 #include <flashinfer/vec_dtypes.cuh>
 
 #include "utils.h"
 
-template <typename T>
+static constexpr int kWarpSize = 32;
+
+// ---------------------------------------------------------------------------
+// 1. Warp‑local, no shared memory
+//    • One warp handles one token.
+//    • Eight tokens per 256‑thread CTA.
+// ---------------------------------------------------------------------------
+template <typename T, int kTokensPerCTA = 8, int kVecSize = 16>
 __global__ void per_token_quant_fp8_kernel(
     const T* __restrict__ input,
     FP8_TYPE* __restrict__ output_q,
     float* __restrict__ output_s,
     const int64_t hidden_dim,
     const int64_t num_tokens) {
+  const int warp_id = threadIdx.x / kWarpSize;        // 0‑7  (8 warps)
+  const int lane_id = threadIdx.x & (kWarpSize - 1);  // 0‑31
+  const int token_id = blockIdx.x * kTokensPerCTA + warp_id;
+  if (token_id >= num_tokens) return;
+
+  // Global tensors for this token
+  const T* token_input = input + token_id * hidden_dim;
+  FP8_TYPE* token_output = output_q + token_id * hidden_dim;
+  float* token_scale = output_s + token_id;
+
+  //
+  // Pass-1: Perform a warp reduce to find the max_value of a token's hidden_dim
+  //
+  float max_value = 0.f;
+  using vec_t = flashinfer::vec_t<T, kVecSize>;
+  const int32_t num_vec_elems = hidden_dim / kVecSize;
+
+  for (int32_t i = lane_id; i < num_vec_elems; i += kWarpSize) {
+    vec_t input_vec;
+    input_vec.cast_load(token_input + i * kVecSize);
+
+#pragma unroll
+    for (uint32_t j = 0; j < kVecSize; ++j) {
+      max_value = fmaxf(max_value, fabsf(static_cast<float>(input_vec[j])));
+    }
+  }
+
+  float warp_max = warpReduceMax(max_value);
+
+  __shared__ float scale;
+  scale = warp_max / FP8_E4M3_MAX;
+  // Broadcast scale
+  if (lane_id == 0) {
+    token_scale[0] = scale;
+  }
+  float scale_inv = (scale == 0.f) ? 0.f : 1.0f / scale;
+
+  //
+  // Pass-2: quantize and write back
+  //
+  for (int i = lane_id; i < num_vec_elems; i += kWarpSize) {
+    vec_t input_vec;
+    input_vec.cast_load(token_input + i * kVecSize);
+    FP8_TYPE output_arr[kVecSize];
+#pragma unroll
+    for (uint32_t j = 0; j < kVecSize; ++j) {
+      float val = static_cast<float>(input_vec[j]) * scale_inv;
+      val = fmaxf(fminf(val, FP8_E4M3_MAX), -FP8_E4M3_MAX);
+
+#ifndef USE_ROCM
+      output_arr[j] = static_cast<FP8_TYPE>(val);
+#else
+      output_arr[j] = c10::Float8_e4m3fnuz(
+          __hip_cvt_float_to_fp8(val, fp8::fp8_type::__default_saturation, fp8::fp8_type::__default_interpret),
+          c10::Float8_e4m3fnuz::from_bits());
+#endif
+    }
+    *(uint4*)(token_output + i * kVecSize) = *(uint4*)output_arr;
+  }
+}
+
+// ---------------------------------------------------------------------------
+// 2.  Baseline kernel (1 token / CTA, CUB block reduce)
+// ---------------------------------------------------------------------------
+template <typename T>
+__global__ void per_token_quant_fp8_small_batch_kernel(
+    const T* __restrict__ input,
+    FP8_TYPE* __restrict__ output_q,
+    float* __restrict__ output_s,
+    const int64_t hidden_dim,
+    const int64_t num_tokens) {
   const int token_idx = blockIdx.x;
   if (token_idx >= num_tokens) return;
 
@@ -79,28 +156,41 @@ void sgl_per_token_quant_fp8(torch::Tensor input, torch::Tensor output_q, torch:
   CHECK_INPUT(input);
   CHECK_INPUT(output_q);
   CHECK_INPUT(output_s);
-
   const auto input_sizes = input.sizes();
   const int64_t num_tokens = input_sizes[0];
   const int64_t hidden_dim = input_sizes[1];
-
   TORCH_CHECK(hidden_dim % 16 == 0, "Hidden dimension must be divisible by 16, but got ", hidden_dim);
 
-  const int block_size = 256;
-  const int num_blocks = num_tokens;
-
-  dim3 grid(num_blocks);
-  dim3 block(block_size);
-
   cudaStream_t stream = at::cuda::getCurrentCUDAStream();
+  // Hard-code sm_count
+  int sm_count = 132;
+  constexpr int TOKENS_PER_CTA = 8;
+  const bool use_warp_kernel = (num_tokens >= sm_count * 2 * TOKENS_PER_CTA);
 
   DISPATCH_PYTORCH_DTYPE_TO_CTYPE_FLOAT_FP16(input.scalar_type(), scalar_t, [&] {
-    per_token_quant_fp8_kernel<scalar_t><<<grid, block, 0, stream>>>(
-        static_cast<scalar_t*>(input.data_ptr()),
-        static_cast<FP8_TYPE*>(output_q.data_ptr()),
-        static_cast<float*>(output_s.data_ptr()),
-        hidden_dim,
-        num_tokens);
+    if (use_warp_kernel) {
+      // -------- warp‑local ---------------------------------------------------
+      constexpr int THREADS = TOKENS_PER_CTA * kWarpSize;  // 256
+      dim3 grid((num_tokens + TOKENS_PER_CTA - 1) / TOKENS_PER_CTA);
+      dim3 block(THREADS);
+      per_token_quant_fp8_kernel<scalar_t, TOKENS_PER_CTA, 16><<<grid, block, 0, stream>>>(
+          static_cast<const scalar_t*>(input.data_ptr()),
+          static_cast<FP8_TYPE*>(output_q.data_ptr()),
+          static_cast<float*>(output_s.data_ptr()),
+          hidden_dim,
+          num_tokens);
+    } else {
+      // -------- baseline -----------------------------------------------------
+      constexpr int THREADS = 256;
+      dim3 grid(num_tokens);
+      dim3 block(THREADS);
+      per_token_quant_fp8_small_batch_kernel<scalar_t><<<grid, block, 0, stream>>>(
+          static_cast<const scalar_t*>(input.data_ptr()),
+          static_cast<FP8_TYPE*>(output_q.data_ptr()),
+          static_cast<float*>(output_s.data_ptr()),
+          hidden_dim,
+          num_tokens);
+    }
     return true;
   });
 }