Move row conversion code from cudf (#1838)

* Revert "Use cuda::proclaim_return_type on device lambdas (#1662)" This reverts commit 763406c. # Conflicts: # src/main/cpp/benchmarks/row_conversion.cpp # src/main/cpp/src/bloom_filter.cu # src/main/cpp/src/map_utils.cu # src/main/cpp/src/xxhash64.cu # thirdparty/cudf Signed-off-by: Nghia Truong <[email protected]> * Remove redundant header Signed-off-by: Nghia Truong <[email protected]> * Update copyright year Signed-off-by: Nghia Truong <[email protected]> --------- Signed-off-by: Nghia Truong <[email protected]>
NVIDIA · Mar 6, 2024 · 5a09147 · 5a09147
1 parent 99445a7
commit 5a09147
Show file tree

Hide file tree

Showing 7 changed files with 3,618 additions and 16 deletions.
diff --git a/src/main/cpp/CMakeLists.txt b/src/main/cpp/CMakeLists.txt
@@ -171,6 +171,7 @@ add_library(
   src/map_utils.cu
   src/murmur_hash.cu
   src/parse_uri.cu
+  src/row_conversion.cu
   src/timezones.cu
   src/utilities.cu
   src/xxhash64.cu

diff --git a/src/main/cpp/benchmarks/row_conversion.cpp b/src/main/cpp/benchmarks/row_conversion.cpp
@@ -48,15 +48,15 @@ void fixed_width(nvbench::state& state)
     bytes_per_row += cudf::size_of(t);
   }
 
-  auto rows = cudf::convert_to_rows_fixed_width_optimized(table->view());
+  auto rows = spark_rapids_jni::convert_to_rows_fixed_width_optimized(table->view());
 
   state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) {
     if (direction == "to row") {
-      auto _rows = cudf::convert_to_rows_fixed_width_optimized(table->view());
+      auto _rows = spark_rapids_jni::convert_to_rows_fixed_width_optimized(table->view());
     } else {
       for (auto const& r : rows) {
         cudf::lists_column_view const l(r->view());
-        auto out = cudf::convert_from_rows_fixed_width_optimized(l, schema);
+        auto out = spark_rapids_jni::convert_from_rows_fixed_width_optimized(l, schema);
       }
     }
   });
@@ -117,16 +117,16 @@ static void variable_or_fixed_width(nvbench::state& state)
     }
   }
 
-  auto rows = cudf::convert_to_rows(table->view());
+  auto rows = spark_rapids_jni::convert_to_rows(table->view());
 
   state.exec(nvbench::exec_tag::sync, [&](nvbench::launch& launch) {
-    auto new_rows = cudf::convert_to_rows(table->view());
+    auto new_rows = spark_rapids_jni::convert_to_rows(table->view());
     if (direction == "to row") {
-      auto _rows = cudf::convert_to_rows(table->view());
+      auto _rows = spark_rapids_jni::convert_to_rows(table->view());
     } else {
       for (auto const& r : rows) {
         cudf::lists_column_view const l(r->view());
-        auto out = cudf::convert_from_rows(l, schema);
+        auto out = spark_rapids_jni::convert_from_rows(l, schema);
       }
     }
   });

diff --git a/src/main/cpp/src/RowConversionJni.cpp b/src/main/cpp/src/RowConversionJni.cpp
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2022-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2022-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -16,8 +16,7 @@
 
 #include "cudf_jni_apis.hpp"
 #include "dtype_utils.hpp"
-
-#include <cudf/row_conversion.hpp>
+#include "row_conversion.hpp"
 
 extern "C" {
 
@@ -32,7 +31,7 @@ Java_com_nvidia_spark_rapids_jni_RowConversion_convertToRowsFixedWidthOptimized(
     cudf::jni::auto_set_device(env);
     cudf::table_view const* n_input_table = reinterpret_cast<cudf::table_view const*>(input_table);
     std::vector<std::unique_ptr<cudf::column>> cols =
-      cudf::convert_to_rows_fixed_width_optimized(*n_input_table);
+      spark_rapids_jni::convert_to_rows_fixed_width_optimized(*n_input_table);
     int const num_columns = cols.size();
     cudf::jni::native_jlongArray outcol_handles(env, num_columns);
     std::transform(cols.begin(), cols.end(), outcol_handles.begin(), [](auto& col) {
@@ -51,8 +50,9 @@ Java_com_nvidia_spark_rapids_jni_RowConversion_convertToRows(JNIEnv* env, jclass
   try {
     cudf::jni::auto_set_device(env);
     cudf::table_view const* n_input_table = reinterpret_cast<cudf::table_view const*>(input_table);
-    std::vector<std::unique_ptr<cudf::column>> cols = cudf::convert_to_rows(*n_input_table);
-    int const num_columns                           = cols.size();
+    std::vector<std::unique_ptr<cudf::column>> cols =
+      spark_rapids_jni::convert_to_rows(*n_input_table);
+    int const num_columns = cols.size();
     cudf::jni::native_jlongArray outcol_handles(env, num_columns);
     std::transform(cols.begin(), cols.end(), outcol_handles.begin(), [](auto& col) {
       return cudf::jni::release_as_jlong(col);
@@ -84,7 +84,7 @@ Java_com_nvidia_spark_rapids_jni_RowConversion_convertFromRowsFixedWidthOptimize
                    std::back_inserter(types_vec),
                    [](jint type, jint scale) { return cudf::jni::make_data_type(type, scale); });
     std::unique_ptr<cudf::table> result =
-      cudf::convert_from_rows_fixed_width_optimized(list_input, types_vec);
+      spark_rapids_jni::convert_from_rows_fixed_width_optimized(list_input, types_vec);
     return cudf::jni::convert_table_for_return(env, result);
   }
   CATCH_STD(env, 0);
@@ -110,7 +110,8 @@ JNIEXPORT jlongArray JNICALL Java_com_nvidia_spark_rapids_jni_RowConversion_conv
                    n_scale.begin(),
                    std::back_inserter(types_vec),
                    [](jint type, jint scale) { return cudf::jni::make_data_type(type, scale); });
-    std::unique_ptr<cudf::table> result = cudf::convert_from_rows(list_input, types_vec);
+    std::unique_ptr<cudf::table> result =
+      spark_rapids_jni::convert_from_rows(list_input, types_vec);
     return cudf::jni::convert_table_for_return(env, result);
   }
   CATCH_STD(env, 0);