Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[DNR] Add multi-source reading to JSON reader benchmarks #17688

Draft
wants to merge 3 commits into
base: branch-25.02
Choose a base branch
from
Draft

[DNR] Add multi-source reading to JSON reader benchmarks #17688

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5a6ac07
adding multi-source to benchmarks
shrshi Jan 7, 2025
4f6c804
removing null columns only to evaluate multi source perf
shrshi Jan 7, 2025
2b578f5
cleanup
shrshi Jan 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions cpp/benchmarks/common/generate_input.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
/*
* Copyright (c) 2020-2024, NVIDIA CORPORATION.
* Copyright (c) 2020-2025, NVIDIA CORPORATION.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
Expand Down Expand Up @@ -236,10 +236,10 @@ class data_profile {
{cudf::type_id::INT32, cudf::type_id::FLOAT32, cudf::type_id::STRING}, 2};
std::map<cudf::type_id, distribution_params<numeric::decimal128>> decimal_params;

double bool_probability_true = 0.5;
std::optional<double> null_probability = 0.01;
cudf::size_type cardinality = 2000;
cudf::size_type avg_run_length = 4;
double bool_probability_true = 0.5;
std::optional<double> null_probability;
cudf::size_type cardinality = 2000;
cudf::size_type avg_run_length = 4;

public:
template <typename T,
Expand Down
49 changes: 32 additions & 17 deletions cpp/benchmarks/io/json/json_reader_input.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
/*
* Copyright (c) 2024, NVIDIA CORPORATION.
* Copyright (c) 2024-2025, NVIDIA CORPORATION.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
Expand Down Expand Up @@ -29,14 +29,15 @@
constexpr size_t default_data_size = 512 << 20;
constexpr cudf::size_type num_cols = 64;

void json_read_common(cuio_source_sink_pair& source_sink,
void json_read_common(cudf::io::source_info const& source,
size_t source_size,
cudf::size_type num_rows_to_read,
nvbench::state& state,
cudf::io::compression_type comptype = cudf::io::compression_type::NONE,
size_t data_size = default_data_size)
{
cudf::io::json_reader_options read_opts =
cudf::io::json_reader_options::builder(source_sink.make_source_info()).compression(comptype);
cudf::io::json_reader_options::builder(source).compression(comptype).lines(true);

auto mem_stats_logger = cudf::memory_stats_logger();
state.set_cuda_stream(nvbench::make_cuda_stream_view(cudf::get_default_stream().value()));
Expand All @@ -56,7 +57,7 @@ void json_read_common(cuio_source_sink_pair& source_sink,
state.add_element_count(static_cast<double>(data_size) / time, "bytes_per_second");
state.add_buffer_size(
mem_stats_logger.peak_memory_usage(), "peak_memory_usage", "peak_memory_usage");
state.add_buffer_size(source_sink.size(), "encoded_file_size", "encoded_file_size");
state.add_buffer_size(source_size, "encoded_file_size", "encoded_file_size");
}

cudf::size_type json_write_bm_data(
Expand All @@ -73,6 +74,7 @@ cudf::size_type json_write_bm_data(
cudf::io::json_writer_options::builder(sink, view)
.na_rep("null")
.rows_per_chunk(100'000)
.lines(true)
.compression(comptype);
cudf::io::write_json(write_opts);
return view.num_rows();
Expand All @@ -92,15 +94,16 @@ void BM_json_read_io(nvbench::state& state, nvbench::type_list<nvbench::enum_typ
static_cast<int32_t>(data_type::STRUCT)});
auto const num_rows = json_write_bm_data(source_sink.make_sink_info(), d_type);

json_read_common(source_sink, num_rows, state);
json_read_common(source_sink.make_source_info(), source_sink.size(), num_rows, state);
}

template <cudf::io::compression_type comptype, io_type IO>
void BM_json_read_compressed_io(
nvbench::state& state, nvbench::type_list<nvbench::enum_type<comptype>, nvbench::enum_type<IO>>)
template <cudf::io::compression_type comptype>
void BM_json_read_compressed_io(nvbench::state& state,
nvbench::type_list<nvbench::enum_type<comptype>>)
{
size_t const data_size = state.get_int64("data_size");
cuio_source_sink_pair source_sink(IO);
size_t const data_size = state.get_int64("data_size");
size_t const num_sources = state.get_int64("num_sources");

auto const d_type = get_type_or_group({static_cast<int32_t>(data_type::INTEGRAL),
static_cast<int32_t>(data_type::FLOAT),
static_cast<int32_t>(data_type::DECIMAL),
Expand All @@ -109,10 +112,22 @@ void BM_json_read_compressed_io(
static_cast<int32_t>(data_type::STRING),
static_cast<int32_t>(data_type::LIST),
static_cast<int32_t>(data_type::STRUCT)});
auto const num_rows =
json_write_bm_data(source_sink.make_sink_info(), d_type, comptype, data_size);

json_read_common(source_sink, num_rows, state, comptype, data_size);
std::vector<char> out_buffer;
auto sink = cudf::io::sink_info(&out_buffer);
auto const num_rows = json_write_bm_data(sink, d_type, comptype, data_size);

std::vector<cudf::host_span<char const>> hostbufs(
num_sources,
cudf::host_span<char const>(reinterpret_cast<char const*>(out_buffer.data()),
out_buffer.size()));
json_read_common(cudf::io::source_info{cudf::host_span<cudf::host_span<char const>>(
hostbufs.data(), hostbufs.size())},
out_buffer.size() * num_sources,
num_rows * num_sources,
state,
comptype,
data_size * num_sources);
}

template <data_type DataType, io_type IO>
Expand All @@ -123,7 +138,7 @@ void BM_json_read_data_type(
auto const d_type = get_type_or_group(static_cast<int32_t>(DataType));
auto const num_rows = json_write_bm_data(source_sink.make_sink_info(), d_type);

json_read_common(source_sink, num_rows, state);
json_read_common(source_sink.make_source_info(), source_sink.size(), num_rows, state);
}

using d_type_list = nvbench::enum_type_list<data_type::INTEGRAL,
Expand Down Expand Up @@ -153,9 +168,9 @@ NVBENCH_BENCH_TYPES(BM_json_read_io, NVBENCH_TYPE_AXES(io_list))
.set_type_axes_names({"io"})
.set_min_samples(4);

NVBENCH_BENCH_TYPES(BM_json_read_compressed_io,
NVBENCH_TYPE_AXES(compression_list, nvbench::enum_type_list<io_type::FILEPATH>))
NVBENCH_BENCH_TYPES(BM_json_read_compressed_io, NVBENCH_TYPE_AXES(compression_list))
.set_name("json_read_compressed_io")
.set_type_axes_names({"compression_type", "io"})
.set_type_axes_names({"compression_type"})
.add_int64_power_of_two_axis("data_size", nvbench::range(20, 29, 1))
.add_int64_axis("num_sources", nvbench::range(1, 5, 1))
.set_min_samples(4);
Loading