[PHI] Fix argsort big tensor bug (#72712)

Enigmatisms · web-flow · commit 6ef8d19ce5d9 · 2025-05-15T14:58:46.000+08:00
* [PHI] Fixed argsort big tensor bug

* [PHI] Fixed shape mismatch problem.
diff --git a/paddle/phi/kernels/gpu/argsort_kernel.cu b/paddle/phi/kernels/gpu/argsort_kernel.cu
@@ -96,52 +96,57 @@ static __global__ void FillIndex(T* indices, T num_rows, T num_cols) {
   }
 }
 
+#define CUB_ARGSORT_WRAPPER(func, ...)                                        \
+  {                                                                           \
+    size_t temp_storage_bytes = 0;                                            \
+    PADDLE_ENFORCE_GPU_SUCCESS(                                               \
+        func(nullptr, temp_storage_bytes, __VA_ARGS__));                      \
+    DenseTensor temp_storage;                                                 \
+    int64_t temp_size = static_cast<int64_t>(temp_storage_bytes);             \
+    PADDLE_ENFORCE_GT(                                                        \
+        temp_size,                                                            \
+        0,                                                                    \
+        common::errors::InvalidArgument(                                      \
+            "Argsort temp storage size is %d, but should be greater than 0.", \
+            temp_size));                                                      \
+    temp_storage.Resize({temp_size});                                         \
+    ctx.template Alloc<uint8_t>(&temp_storage);                               \
+    PADDLE_ENFORCE_GPU_SUCCESS(                                               \
+        func(temp_storage.data<uint8_t>(), temp_storage_bytes, __VA_ARGS__)); \
+  }
+
+#define PREDICATE_CUB_ARGSORT(predicate, if_func, else_func, ...) \
+  if (predicate)                                                  \
+    CUB_ARGSORT_WRAPPER(if_func, __VA_ARGS__)                     \
+  else                                                            \
+    CUB_ARGSORT_WRAPPER(else_func, __VA_ARGS__)
+
 // Sort by flag descending, True: descending. False: Ascending.
 // Default is false.
 template <typename T, typename IndType>
 void ArgFullSort(const phi::GPUContext& ctx,
                  const DenseTensor* input,
                  DenseTensor* output,
                  DenseTensor* indices,
-                 const IndType num_rows,
-                 const IndType num_cols,
+                 const int64_t num_rows,
+                 const int64_t num_cols,
                  const bool descending) {
   auto cu_stream = ctx.stream();
-  DenseTensor input_indices;
-  const std::vector<IndType> dims = {num_rows, num_cols};
-  auto dim = common::make_ddim(dims);
-  input_indices.Resize(dim);
-  ctx.template Alloc<IndType>(&input_indices);
-  size_t temp_storage_bytes = -1;
-
   auto ComputeBlockSize = [](IndType col) {
     if (col > 512)
       return 1024;
     else if (col > 256 && col <= 512)
       return 512;
     else if (col > 128 && col <= 256)
       return 256;
-    else if (col > 64 && col <= 128)
-      return 128;
     else
-      return 64;
+      return 128;
   };
+  const int block_size = ComputeBlockSize(num_cols);
+  const int64_t maxGridDimX = ctx.GetCUDAMaxGridDimSize()[0];
 
-  int block_size = ComputeBlockSize(num_cols);
-  int maxGridDimX = ctx.GetCUDAMaxGridDimSize()[0];
-  // actually, int num_rows < max_grid_size
-  int grid_size = num_rows < maxGridDimX ? num_rows : maxGridDimX;
-  // Init a index array
-  FillIndex<<<grid_size, block_size, 0, cu_stream>>>(
-      input_indices.data<IndType>(), num_rows, num_cols);
-
-  T* sorted_out_ptr;
-  IndType* sorted_indices_ptr;
   const T* inp = input->data<T>();
-  T* out = ctx.template Alloc<T>(output);
-  IndType* ind = ctx.template Alloc<IndType>(indices);
-  sorted_out_ptr = out;
-  sorted_indices_ptr = ind;
+  IndType* sorted_indices_ptr = indices->data<IndType>();
 
   // create iter for counting input
   cub::CountingInputIterator<IndType> counting_iter(0);
@@ -151,78 +156,54 @@ void ArgFullSort(const phi::GPUContext& ctx,
                               cub::CountingInputIterator<IndType>>
       segment_offsets_t(counting_iter, SegmentOffsetIter(num_cols));
 
-  gpuError_t err;
-  if (descending) {
-    err = cub::DeviceSegmentedRadixSort::SortPairsDescending(
-        nullptr,
-        temp_storage_bytes,
-        inp,
-        sorted_out_ptr,
-        input_indices.data<IndType>(),
-        sorted_indices_ptr,
-        num_cols * num_rows,
-        num_rows,
-        segment_offsets_t,
-        segment_offsets_t + 1,
-        0,
-        sizeof(T) * 8,
-        cu_stream);
-  } else {
-    err =
-        cub::DeviceSegmentedRadixSort::SortPairs(nullptr,
-                                                 temp_storage_bytes,
-                                                 inp,
-                                                 sorted_out_ptr,
-                                                 input_indices.data<IndType>(),
-                                                 sorted_indices_ptr,
-                                                 num_cols * num_rows,
-                                                 num_rows,
-                                                 segment_offsets_t,
-                                                 segment_offsets_t + 1,
-                                                 0,
-                                                 sizeof(T) * 8,
-                                                 cu_stream);
-  }
-  PADDLE_ENFORCE_GPU_SUCCESS(err);
-
-  DenseTensor temp_storage;
-  int64_t temp_size = temp_storage_bytes;
-  temp_storage.Resize({temp_size});
-  ctx.template Alloc<uint8_t>(&temp_storage);
-
-  if (descending) {
-    err = cub::DeviceSegmentedRadixSort::SortPairsDescending(
-        temp_storage.data<uint8_t>(),
-        temp_storage_bytes,
-        inp,
-        sorted_out_ptr,
-        input_indices.data<IndType>(),
-        sorted_indices_ptr,
-        num_cols * num_rows,
-        num_rows,
-        segment_offsets_t,
-        segment_offsets_t + 1,
-        0,
-        sizeof(T) * 8,
-        cu_stream);
-  } else {
-    err =
-        cub::DeviceSegmentedRadixSort::SortPairs(temp_storage.data<uint8_t>(),
-                                                 temp_storage_bytes,
-                                                 inp,
-                                                 sorted_out_ptr,
-                                                 input_indices.data<IndType>(),
-                                                 sorted_indices_ptr,
-                                                 num_cols * num_rows,
-                                                 num_rows,
-                                                 segment_offsets_t,
-                                                 segment_offsets_t + 1,
-                                                 0,
-                                                 sizeof(T) * 8,
-                                                 cu_stream);
-  }
+  // num_rows is the total segments to be sorted
+  constexpr int64_t max_elements = 1 << 30;
+  const int64_t total_elements = num_cols * num_rows;
+  const int64_t segment_size = num_cols;
+  const int64_t element_per_call = std::min(max_elements, total_elements);
+  // make sure batch size is the multiple of segment_size
+  const int64_t batch_size = (element_per_call / segment_size) * segment_size;
+  int64_t offset = 0;
+  DenseTensor input_indices;
+
+  T* sorted_out_ptr = sorted_out_ptr = output->data<T>();
+  IndType* ind_ptr = nullptr;
 
-  PADDLE_ENFORCE_GPU_SUCCESS(err);
+  while (offset < total_elements) {
+    const int64_t n_elements = std::min(batch_size, total_elements - offset);
+    const int64_t n_segments = n_elements / segment_size;
+
+    // allocate a temporary storage for input indices, with shape:
+    // [num_segments = n_elements / segment_size, segment_size]
+    // will be de-allocated once the sort is done, to save memory and
+    // avoid repeated allocation and deallocation
+    if (input_indices.initialized()) {
+      ind_ptr = input_indices.data<IndType>();
+    } else {
+      input_indices.Resize({n_segments, segment_size});
+      ind_ptr = ctx.template Alloc<IndType>(&input_indices);
+    }
+    const int64_t grid_size = std::min(n_segments, maxGridDimX);
+    // Init a index array
+    FillIndex<<<grid_size, block_size, 0, cu_stream>>>(
+        ind_ptr, n_segments, segment_size);
+
+    PREDICATE_CUB_ARGSORT(descending,
+                          cub::DeviceSegmentedRadixSort::SortPairsDescending,
+                          cub::DeviceSegmentedRadixSort::SortPairs,
+                          inp + offset,
+                          sorted_out_ptr,
+                          ind_ptr,
+                          sorted_indices_ptr + offset,
+                          n_elements,
+                          n_segments,
+                          segment_offsets_t,
+                          segment_offsets_t + 1,
+                          0,
+                          sizeof(T) * 8,
+                          cu_stream);
+    offset += n_elements;
+  }
 }
 
 template <typename T, typename Context>
@@ -247,10 +228,10 @@ void ArgsortKernel(const Context& dev_ctx,
   axis = (axis < 0) ? (in_dims.size() + axis) : axis;
   const T* in_data = input.data<T>();
   auto size = input.numel();
-  T* out_data = dev_ctx.template Alloc<T>(output);
-  int64_t* ids_data = dev_ctx.template Alloc<int64_t>(indices);
 
   if (rank == 0) {
+    dev_ctx.template Alloc<T>(output);
+    dev_ctx.template Alloc<int64_t>(indices);
     phi::Copy<Context>(dev_ctx, input, dev_ctx.GetPlace(), false, output);
     phi::funcs::set_constant(dev_ctx, indices, static_cast<int64_t>(0));
     return;
@@ -261,6 +242,8 @@ void ArgsortKernel(const Context& dev_ctx,
   // Compared to the following 'Special case for full sort', ascending sort is
   // 34 times faster and descending sort is 31 times faster.
   if (size == in_dims[axis]) {
+    T* out_data = dev_ctx.template Alloc<T>(output);
+    int64_t* ids_data = dev_ctx.template Alloc<int64_t>(indices);
 #ifdef PADDLE_WITH_CUDA
     const auto& exec_policy = thrust::cuda::par.on(dev_ctx.stream());
 #else
@@ -297,6 +280,8 @@ void ArgsortKernel(const Context& dev_ctx,
     const int64_t input_height =
         common::product(common::slice_ddim(in_dims, 0, in_dims.size() - 1));
     const int64_t input_width = in_dims[in_dims.size() - 1];
+    dev_ctx.template Alloc<int64_t>(indices);
+    dev_ctx.template Alloc<T>(output);
     ArgFullSort<T, int64_t>(dev_ctx,
                             &input,
                             output,
@@ -338,7 +323,6 @@ void ArgsortKernel(const Context& dev_ctx,
     // temp indices for sorting
     tmp_indices.Resize(trans_dims);
     dev_ctx.template Alloc<int64_t>(&tmp_indices);
-    dev_ctx.template Alloc<int64_t>(indices);
 
     ArgFullSort<T, int64_t>(dev_ctx,
                             &trans_inp,
@@ -347,10 +331,13 @@ void ArgsortKernel(const Context& dev_ctx,
                             input_height,
                             input_width,
                             descending);
-
-    TransposeKernel<int64_t, Context>(dev_ctx, tmp_indices, trans, indices);
+    // delay output allocation until after transpose, to avoid
+    // allocating too much memory
+    dev_ctx.template Alloc<T>(output);
+    dev_ctx.template Alloc<int64_t>(indices);
     // transpose back
     TransposeKernel<T, Context>(dev_ctx, tmp_out, trans, output);
+    TransposeKernel<int64_t, Context>(dev_ctx, tmp_indices, trans, indices);
     return;
   }
 }
