Add ssd_update_row_addrs (pytorch#2953)

Summary:
Pull Request resolved: pytorch#2953

X-link: facebookresearch/FBGEMM#53

When pipeline prefetching is enabled, data in a scratch pad of the
current iteration can be moved to L1 or a scratch pad of the next
iteration during the prefetch step. `ssd_update_row_addrs` updates the
memory addresses of data that is relocated to the correct location.

Differential Revision: D60983150

fbgemm_gpu/src/ssd_split_embeddings_cache/ssd_split_embeddings_cache_cuda.cu

@@ -538,3 +538,112 @@ std::tuple<Tensor, Tensor> ssd_generate_row_addrs_cuda(
 
   return {ssd_row_addrs, post_bwd_evicted_indices};
 }
+
+__global__ __launch_bounds__(kMaxThreads) void ssd_update_row_addrs_kernel(
+    at::PackedTensorAccessor32<int64_t, 1, at::RestrictPtrTraits>
+        ssd_row_addrs_curr,
+    const at::PackedTensorAccessor32<int32_t, 1, at::RestrictPtrTraits>
+        ssd_curr_next_map,
+    const at::PackedTensorAccessor32<int32_t, 1, at::RestrictPtrTraits>
+        lxu_cache_locations_curr,
+    const at::PackedTensorAccessor32<int32_t, 1, at::RestrictPtrTraits>
+        linear_index_inverse_indices_curr,
+    const at::PackedTensorAccessor32<int32_t, 1, at::RestrictPtrTraits>
+        unique_indices_count_cumsum_curr,
+    const at::PackedTensorAccessor32<int32_t, 1, at::RestrictPtrTraits>
+        cache_set_inverse_indices_curr,
+    const uint64_t lxu_cache_weights_addr,
+    const uint64_t inserted_ssd_weights_addr_next,
+    const int* N_unique_curr,
+    const uint64_t cache_row_bytes // has to be 64 bits to prevent overflow
+) {
+  const auto n_curr = blockDim.y * blockIdx.x + threadIdx.y;
+  if (n_curr >= *N_unique_curr) {
+    return;
+  }
+
+  // Find mapping between n_curr and n_next
+  const auto n_next = ssd_curr_next_map[n_curr];
+
+  // Return if the row is not used in both previous and next iterations
+  if (n_next < 0) {
+    return;
+  }
+
+  // Find out if the row gets moved to the nextent iteration's scratch pad or
+  // L1 by checking the lxu_cache_locations_curr
+  const auto cache_set_id_curr = cache_set_inverse_indices_curr[n_curr];
+  const auto segment_start_curr =
+      unique_indices_count_cumsum_curr[cache_set_id_curr];
+  const auto segment_end_curr =
+      unique_indices_count_cumsum_curr[cache_set_id_curr + 1];
+  const auto cache_loc_curr = lxu_cache_locations_curr
+      [linear_index_inverse_indices_curr[segment_start_curr]];
+
+  const uint64_t ptr_addr = (cache_loc_curr == -1)
+      // The row is moved from the previous iteration's scratch pad to the
+      // next iteration's scratch pad
+      ? (inserted_ssd_weights_addr_next + (n_next * cache_row_bytes))
+      // The row is moved from the previous iteration's scratch pad to L1 cache
+      : (lxu_cache_weights_addr + (cache_loc_curr * cache_row_bytes));
+
+  // Set pointer address
+  for (auto l = segment_start_curr + threadIdx.x; l < segment_end_curr;
+       l += blockDim.x) {
+    auto dst = linear_index_inverse_indices_curr[l];
+    *reinterpret_cast<uint64_t*>(&ssd_row_addrs_curr[dst]) = ptr_addr;
+  }
+}
+
+void ssd_update_row_addrs_cuda(
+    const Tensor& ssd_row_addrs_curr,
+    const Tensor& ssd_curr_next_map,
+    const Tensor& lxu_cache_locations_curr,
+    const Tensor& linear_index_inverse_indices_curr,
+    const Tensor& unique_indices_count_cumsum_curr,
+    const Tensor& cache_set_inverse_indices_curr,
+    const Tensor& lxu_cache_weights,
+    const Tensor& inserted_ssd_weights_next,
+    const Tensor& unique_indices_length_curr) {
+  TENSORS_ON_SAME_CUDA_GPU_IF_NOT_OPTIONAL(
+      ssd_row_addrs_curr,
+      ssd_curr_next_map,
+      lxu_cache_locations_curr,
+      linear_index_inverse_indices_curr,
+      unique_indices_count_cumsum_curr,
+      cache_set_inverse_indices_curr,
+      lxu_cache_weights,
+      inserted_ssd_weights_next,
+      unique_indices_length_curr);
+
+  CUDA_DEVICE_GUARD(ssd_row_addrs_curr);
+
+  const auto lxu_cache_weights_addr =
+      reinterpret_cast<uint64_t>(lxu_cache_weights.data_ptr());
+  const auto inserted_ssd_weights_addr_next =
+      reinterpret_cast<uint64_t>(inserted_ssd_weights_next.data_ptr());
+  const auto cache_row_bytes =
+      lxu_cache_weights.size(1) * lxu_cache_weights.element_size();
+  constexpr auto kNumWarps = kMaxThreads / kWarpSize;
+
+  ssd_update_row_addrs_kernel<<<
+      div_round_up(ssd_row_addrs_curr.numel(), kNumWarps),
+      dim3(kWarpSize, kNumWarps),
+      0,
+      at::cuda::getCurrentCUDAStream()>>>(
+      ssd_row_addrs_curr.packed_accessor32<int64_t, 1, at::RestrictPtrTraits>(),
+      ssd_curr_next_map.packed_accessor32<int32_t, 1, at::RestrictPtrTraits>(),
+      lxu_cache_locations_curr
+          .packed_accessor32<int32_t, 1, at::RestrictPtrTraits>(),
+      linear_index_inverse_indices_curr
+          .packed_accessor32<int32_t, 1, at::RestrictPtrTraits>(),
+      unique_indices_count_cumsum_curr
+          .packed_accessor32<int32_t, 1, at::RestrictPtrTraits>(),
+      cache_set_inverse_indices_curr
+          .packed_accessor32<int32_t, 1, at::RestrictPtrTraits>(),
+      lxu_cache_weights_addr,
+      inserted_ssd_weights_addr_next,
+      unique_indices_length_curr.data_ptr<int32_t>(),
+      cache_row_bytes);
+  C10_CUDA_KERNEL_LAUNCH_CHECK();
+}

fbgemm_gpu/src/ssd_split_embeddings_cache/ssd_split_table_batched_embeddings.cpp

@@ -148,6 +148,56 @@ std::tuple<Tensor, Tensor> ssd_generate_row_addrs_cuda(
     const Tensor& unique_indices_length,
     const Tensor& cache_set_sorted_unique_indices);
 
+/// @ingroup embedding-ssd
+///
+/// @brief Update memory addresses for SSD TBE data
+///
+/// When pipeline prefetching is enabled, data in a scratch pad of the
+/// current iteration can be moved to L1 or a scratch pad of the next
+/// iteration during the prefetch step. This operator updates the
+/// memory addresses of data that is relocated to the correct
+/// location.
+///
+/// @param ssd_row_addrs_curr The tensor that contains the row address
+///            of the current iteration
+/// @param inserted_ssd_weights_curr_next_map The tensor that contains
+///            mapping between the location of each index in the
+///            current iteration in the scratch pad of the next
+///            iteration. (-1 = the data has not been moved).
+///            inserted_ssd_weights_curr_next_map[i] is the location
+//             of index i in the next iteration's scratch pad.
+/// @param lxu_cache_locations_curr The tensor that contains cache
+///            slots where data is stored for the *full* list of
+///            indices for the current iteration. -1 is a sentinel
+///            value that indicates that data is not in cache.
+/// @param linear_index_inverse_indices_curr The tensor that contains
+///            the original position of linear indices before being
+///            sorted for the current iteration
+/// @param unique_indices_count_cumsum_curr The tensor that contains
+///            the the exclusive prefix sum results of the counts of
+///            unique indices for the current iteration
+/// @param cache_set_inverse_indices_curr The tensor that contains the
+///            original positions of cache sets before being sorted
+///            for the current iteration
+/// @param lxu_cache_weights The LXU cache tensor
+/// @param inserted_ssd_weights_next The scratch pad tensor for the
+///            next iteration
+/// @param unique_indices_length_curr The tensor that contains the
+///            number of unique indices (GPU tensor) for the current
+///            iteration
+///
+/// @return None
+void ssd_update_row_addrs_cuda(
+    const Tensor& ssd_row_addrs_curr,
+    const Tensor& inserted_ssd_weights_curr_next_map,
+    const Tensor& lxu_cache_locations_curr,
+    const Tensor& linear_index_inverse_indices_curr,
+    const Tensor& unique_indices_count_cumsum_curr,
+    const Tensor& cache_set_inverse_indices_curr,
+    const Tensor& lxu_cache_weights,
+    const Tensor& inserted_ssd_weights_next,
+    const Tensor& unique_indices_length_curr);
+
 namespace {
 class EmbeddingRocksDBWrapper : public torch::jit::CustomClassHolder {
  public:
@@ -315,5 +365,18 @@ TORCH_LIBRARY_FRAGMENT(fbgemm, m) {
       "    Tensor cache_set_sorted_unique_indices"
       ") -> (Tensor, Tensor)");
   DISPATCH_TO_CUDA("ssd_generate_row_addrs", ssd_generate_row_addrs_cuda);
+  m.def(
+      "ssd_update_row_addrs("
+      "    Tensor ssd_row_addrs_curr, "
+      "    Tensor inserted_ssd_weights_curr_next_map, "
+      "    Tensor lxu_cache_locations_curr, "
+      "    Tensor linear_index_inverse_indices_curr, "
+      "    Tensor unique_indices_count_cumsum_curr, "
+      "    Tensor cache_set_inverse_indices_curr, "
+      "    Tensor lxu_cache_weights, "
+      "    Tensor inserted_ssd_weights_next, "
+      "    Tensor unique_indices_length_curr"
+      ") -> ()");
+  DISPATCH_TO_CUDA("ssd_update_row_addrs", ssd_update_row_addrs_cuda);
 }
 } // namespace