[PointKernel]

Description

Closes #10528; depends on #24731.

This PR adds peak-memory tracking to the remaining non-join NVBench benchmarks. It also routes device pack, list scatter, and Thrust iterator allocations through the tracked resource, and excludes string replacement and repetition setup from memory tracking.

This also corrects two inherited benchmark inconsistencies: the Thrust iterator reductions now honor the configured RMM resource, and replace with api=multi constructs its target/replacement columns outside the timed callback. These changes alter existing runtime baselines, so lower timings across this PR should not be interpreted as libcudf algorithm speedups. Treat this commit as a baseline change for those cases; the before/after measurements and allocator controls quantify the effect.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[PointKernel]

/ok to test

[coderabbitai]

Important

Review skipped

Review was skipped as selected files did not have any reviewable changes.

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cudf
• Branch: main
• File: .coderabbit.yaml
• Commit: dfb7633
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: eb8c3626-0cb2-4c78-8fe1-344683fc2405

📥 Commits

Reviewing files that changed from the base of the PR and between 3429a66 and d4b41f6.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Source mode: Auto
• Repository: NVIDIA/cudf
• Branch: main
• File: .coderabbit.yaml
• Commit: 35a12aa
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 3198493d-3599-48c6-858e-78868370bcf9

📥 Commits

Reviewing files that changed from the base of the PR and between 35a12aa and 3429a66.

📒 Files selected for processing: 19

• cpp/benchmarks/ast/transform_fp64.cpp
• cpp/benchmarks/binaryop/compiled_binaryop.cpp
• cpp/benchmarks/bitmask/mask_to_bools.cpp
• cpp/benchmarks/copying/pack.cpp
• cpp/benchmarks/io/fst.cu
• cpp/benchmarks/io/parquet/experimental/hybrid_scan/hybrid_scan_metadata.cpp
• cpp/benchmarks/iterator/iterator.cu
• cpp/benchmarks/json/json.cu
• cpp/benchmarks/lists/copying/scatter_lists.cu
• cpp/benchmarks/string/experimental/stringview_compare.cu
• cpp/benchmarks/string/make_strings_column.cu
• cpp/benchmarks/string/repeat_strings.cpp
• cpp/benchmarks/string/replace.cpp
• cpp/benchmarks/string/url_decode.cu
• cpp/benchmarks/transform/is_null.cpp
• cpp/benchmarks/transform/one_hot_encode.cpp
• cpp/benchmarks/transform/polynomials_concurrent.cpp
• cpp/benchmarks/transform/transform_dispatch.cpp
• cpp/benchmarks/type_dispatcher/type_dispatcher.cu

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 10 remain after this review.

📝 Summary

Summary by CodeRabbit

• Benchmarking
  • Added peak memory usage reporting across a range of benchmarks, making memory consumption visible alongside existing benchmark results.
  • Updated iterator benchmark execution to use a no-sync policy on the default stream.

The pull request adds peak-memory measurements to benchmarks across AST, binary operations, I/O, iterators, strings, and transforms. It also changes memory-resource selection in the pack benchmark and the execution policy used by iterator Thrust reductions.

Changes: Benchmark memory metrics

Layer / File(s)	Summary
Peak memory metrics for single-operation benchmarks cpp/benchmarks/ast/transform_fp64.cpp, cpp/benchmarks/binaryop/compiled_binaryop.cpp, cpp/benchmarks/bitmask/mask_to_bools.cpp, cpp/benchmarks/io/parquet/.../hybrid_scan_metadata.cpp, cpp/benchmarks/json/json.cu, cpp/benchmarks/string/make_strings_column.cu, cpp/benchmarks/string/url_decode.cu, cpp/benchmarks/transform/*, cpp/benchmarks/type_dispatcher/type_dispatcher.cu	These benchmarks create memory statistics loggers and add peak memory usage as an NVBench buffer-size metric.
Peak memory metrics for benchmark variants cpp/benchmarks/io/fst.cu, cpp/benchmarks/string/experimental/stringview_compare.cu, cpp/benchmarks/string/repeat_strings.cpp, cpp/benchmarks/string/replace.cpp	The FST JSON and string benchmark variants report peak memory usage. The replace multi branch now constructs target and replacement columns before its execution callback.
Memory resources and execution policies cpp/benchmarks/copying/pack.cpp, cpp/benchmarks/iterator/iterator.cu, cpp/benchmarks/lists/copying/scatter_lists.cu	The pack benchmark selects its output resource using a boolean. Iterator Thrust reductions use RMM’s no-sync execution policy on the default stream. The iterator and scatter benchmarks also report peak memory usage.

Priority: ⬇️ Low

Estimated code review effort: 3 (Moderate) | ~20 minutes

Change: Feature

Possibly related PRs

• NVIDIA/cudf#22548: Adds the same peak-memory metric to an earlier set of benchmarks.

Suggested labels: improvement

Suggested reviewers: vyasr, bdice, paul-aiyedun

Merge Risk: ⚪ Minimal · up to 3429a

No identified issue blocks merging the benchmark memory-tracking changes after normal checks.

Pre-merge checks: 3 · 1 · 1

Failed checks: 1 warning · 1 inconclusive

Check name	Status	Explanation	Resolution
Docstring Coverage		Docstring coverage is 29.41% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 34 functions across 19 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Linked Issues check		Issue #10528 requires peak_memory_usage for all NVBench benchmarks. This PR adds the logger and metric to the listed remaining non-join benchmarks, including branch-local tracking in replace.cpp a…	Provide reviewable evidence that every benchmark covered by #10528 has peak_memory_usage at the reviewed head, or identify the remaining benchmarks that require instrumentation.

Passed checks: 3 passed

Check name	Status	Explanation
Title check		The title concisely and accurately summarizes the main change: tracking peak memory in the remaining benchmarks.
Out of Scope Changes check		The changed resource selection in pack.cpp and scatter_lists.cu, the no-sync Thrust policy in iterator.cu, and the setup exclusions in the string benchmarks directly support accurate peak-memory…
Description check		The description directly matches the changeset. It explains peak-memory tracking, tracked allocation resources, and benchmark behavior corrections.

Full details: Out of Scope Changes check

Explanation

The changed resource selection in pack.cpp and scatter_lists.cu, the no-sync Thrust policy in iterator.cu, and the setup exclusions in the string benchmarks directly support accurate peak-memory tracking. The remaining changes add the logger and peak_memory_usage metric to benchmark executions. No unrelated change is shown.

Full details: Linked Issues check

Explanation

Issue #10528 requires peak_memory_usage for all NVBench benchmarks. This PR adds the logger and metric to the listed remaining non-join benchmarks, including branch-local tracking in replace.cpp and repeat_strings.cpp. The diff also routes relevant allocations through the current tracked resource. The supplied evidence does not establish whether all other benchmarks, including join benchmarks, already provide this metric at the reviewed head. Dependency #24731 alone does not establish that code coverage.

Resolution

Provide reviewable evidence that every benchmark covered by #10528 has peak_memory_usage at the reviewed head, or identify the remaining benchmarks that require instrumentation.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[mattgara]

This looks good to me, I just have one concern about this potentially creating a shift in existing benchmarks, it would be good to confirm that it doesn't.

[mattgara]

Thanks for updating the PR description.

[PointKernel]

/merge