Add pool memory resource to libcudf basic example

cpp/examples/README.md

@@ -5,4 +5,5 @@ libcudf examples.
 
 Current examples:
 
-- Basic: example that demonstrates basic use case with libcudf and building a custom application with libcudf.
+- Basic: demonstrates a basic use case with libcudf and building a custom application with libcudf
+- Strings: demonstrates using libcudf for accessing and creating strings columns and for building custom kernels for strings

cpp/examples/basic/README.md

@@ -15,7 +15,7 @@ cmake -S . -B build/
 # Build
 cmake --build build/ --parallel $PARALLEL_LEVEL
 # Execute
-build/libcudf_example
+build/basic_example
 ```
 
 If your machine does not come with a pre-built libcudf binary, expect the

cpp/examples/basic/src/process_csv.cpp

@@ -19,6 +19,10 @@
 #include <cudf/io/csv.hpp>
 #include <cudf/table/table.hpp>
 
+#include <rmm/mr/device/cuda_memory_resource.hpp>
+#include <rmm/mr/device/device_memory_resource.hpp>
+#include <rmm/mr/device/pool_memory_resource.hpp>
+
 #include <memory>
 #include <string>
 #include <utility>
@@ -72,6 +76,21 @@ std::unique_ptr<cudf::table> average_closing_price(cudf::table_view stock_info_t
 
 int main(int argc, char** argv)
 {
+  // Construct a CUDA memory resource using RAPIDS Memory Manager (RMM)
+  // This is the default memory resource for libcudf for allocating device memory.
+  rmm::mr::cuda_memory_resource cuda_mr{};
+  // Construct a memory pool using the CUDA memory resource
+  // Using a memory pool for device memory allocations is important for good performance in libcudf.
+  // The pool defaults to allocating half of the available GPU memory.
+  rmm::mr::pool_memory_resource mr{&cuda_mr};
+
+  // Set the pool resource to be used by default for all device memory allocations
+  // Note: It is the user's responsibility to ensure the `mr` object stays alive for the duration of
+  // it being set as the default
+  // Also, call this before the first libcudf API call to ensure all data is allocated by the same
+  // memory resource.
+  rmm::mr::set_current_device_resource(&mr);
+
   // Read data
   auto stock_table_with_metadata = read_csv("4stock_5day.csv");
 

cpp/examples/strings/README.md

@@ -0,0 +1,37 @@
+# libcudf C++ examples using strings columns
+
+This C++ example demonstrates using libcudf APIs to access and create
+strings columns.
+
+The example source code loads a csv file and produces a redacted strings
+column from the names column using the values from the visibilities column.
+
+Four examples are included:
+1. Using libcudf APIs to build the output
+2. Using a simple custom kernel with dynamic memory
+3. Using a custom kernel with pre-allocated device memory
+4. Using a two-pass approach to improve performance
+
+These examples are described in more detail in
+https://developer.nvidia.com/blog/mastering-string-transformations-in-rapids-libcudf/
+
+## Compile and execute
+
+```bash
+# Configure project
+cmake -S . -B build/
+# Build
+cmake --build build/ --parallel $PARALLEL_LEVEL
+# Execute
+build/libcudf_apis names.csv
+--OR--
+build/custom_with_malloc names.csv
+--OR--
+build/custom_prealloc names.csv
+--OR--
+build/custom_optimized names.csv
+```
+
+If your machine does not come with a pre-built libcudf binary, expect the
+first build to take some time, as it would build libcudf on the host machine.
+It may be sped up by configuring the proper `PARALLEL_LEVEL` number.