[bernardladenthin]

Summary

• The CI-built cpu-linux-s390x jar of s390x: IBM Z CPU variants z13 / z15 / z16 (patch 0023), 4-4.5x faster on a z16 #580 (llama-glibc-floor set of run 40475405391) on cfarm191 (LinuxONE z16, Debian 13, JDK 25): NativeLoadSmoke links b11901 with patches 0001-0023, ggml loads libggml-cpu-z16.so, ServerLauncher answers "Paris." on the big-endian Qwen3-0.6B Q4_K_M, with the same text as the static jar at temperature 0.
• Speed against the static 6.1.0-SNAPSHOT jar of run 40400361809, three alternating rounds (26 prompt / 64 generated tokens, t/s): 4 threads prompt 50.3-61.8 vs 14.4-16.0, generation 29.9-35.6 vs 9.9-10.0 (3.0-3.6x); 8 threads noisy on the shared host. TODO.md item ticked, one sentence in CLAUDE.md "Linux s390x".
• Ticks the Android AAR item: in the same run verify-aar-poms.py passes both poms, the consumer fixture's R8 assembleRelease passes for cpu and opencl, and the emulator's connectedDebugAndroidTest passes.

Test plan

• Docs-only change (TODO.md, CLAUDE.md)
• CI is green on this branch
• Docs updated

Related issues / PRs

Refs #580

Checklist

• I have read CONTRIBUTING.md and CODE_OF_CONDUCT.md
• My commits follow Conventional Commits
• No security-sensitive changes

🤖 Generated with Claude Code

https://claude.ai/code/session_01DuAkLtHrCTx81vP5Mt4jad

Summary by CodeRabbit

• Documentation
  • Added Linux s390x performance measurements: the modular jar runs at 29.9–35.6 tokens per second, compared with 9.9–10.0 for the static jar.
  • Recorded verification results for the s390x modular jar and Android AAR, including native loading, inference, build checks, and emulator testing.
  • Noted that measuring vector modules at native speed still requires real IBM Z hardware.

[coderabbitai]

No actionable comments were generated in the recent review.

Recent review info

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: a243de95-3332-43c2-b563-4308546b6ddf

Commits

Reviewing files that changed from the base of the PR and between 3457f40 and 40af8a6.

Files selected for processing: 2

• CLAUDE.md
• TODO.md

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.

Walkthrough

The PR replaces pending validation notes with completed results for the modular Linux s390x jar and Android AARs. It records s390x loading and inference measurements, plus Android POM, R8, release assembly, and emulator checks.

Changes: s390x jar validation

Layer / File(s)	Summary
Record s390x jar results TODO.md, CLAUDE.md	The notes record z16 CPU backend loading, inference results, and throughput comparisons between the modular jar and the static jar. The note about measuring vector-module speed on real IBM Z remains.

Android AAR validation

Layer / File(s)	Summary
Record Android AAR checks TODO.md	The note records POM dependency checks, CPU and OpenCL consumer fixture R8 and release assembly checks, and passing emulator instrumentation.

Priority: Low

Estimated code review effort: 1 (Trivial) | ~4 minutes

Change: Other

Possibly related PRs

• bernardladenthin/java-llama.cpp#580: Adds the modular Linux s390x artifact whose loading and inference results this PR records.
• bernardladenthin/java-llama.cpp#573: Adds s390x and Android checks whose results this PR records.
• bernardladenthin/java-llama.cpp#559: Adds Android consumer R8 builds and emulator instrumentation coverage, which this PR reports as passing.

Suggested reviewers: claude

Merge Risk: Minimal · up to 40af8

This PR records s390x and Android validation results without changing product behavior, so it is mergeable.

Architecture risk: Low · up to 40af8

The change affects 2 systems.

Changed systems: CLAUDE.md, TODO.md

Architecture concerns
No architecture-level concerns identified.

Review details

Systems and components

• observed — CLAUDE.md (service) was modified; 1 changed file maps to changed impact.
• observed — TODO.md (service) was modified; 1 changed file maps to changed impact.

Before / after behavior

• observed — Modified behavior in CLAUDE.md: Added CI measurements reporting the loaded z16 CPU module and generation throughput for the modular jar versus the static 6.1.0-SNAPSHOT jar.
• observed — Modified behavior in TODO.md: Replaces the pending plan to verify the modular s390x jar after a green Publish run with completed results: the jar loaded on cfarm191 with the z16 CPU backend, produced the same inference answer as the static jar, and recorded prompt and generation speed comparisons. The note that vector-module speed requires real IBM Z remains.
• observed — Modified behavior in TODO.md: Replaces the pending first-Publish-run check for Android AARs with the results from run 40475405391: both POMs mirror the core dependencies, CPU and OpenCL consumer fixtures pass R8 minification and release assembly without checker-qual, and emulator instrumentation passes.

Pre-merge checks: 5

Check name	Status	Explanation
Description Check		Check skipped - CodeRabbit’s high-level summary is enabled.
Title check		The title accurately identifies the documentation updates for s390x jar verification on cfarm191 and Android AAR validation results. It is specific and related to the main changes, although “run read”…
Docstring Coverage		No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Documentation-only edits and deletions within survi…
Linked Issues check		Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check		Check skipped because no linked issues were found for this pull request.

Full details: Title check

Explanation

The title accurately identifies the documentation updates for s390x jar verification on cfarm191 and Android AAR validation results. It is specific and related to the main changes, although “run read” is slightly unclear.

Full details: Docstring Coverage

Explanation

No eligible functions found in the changed files. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 0 functions across 0 files. (2 skipped: 2 unsupported.)

Finishing Touches: 💡 1

🛠️ Fix failing CI checks: 💡

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• 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.}

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud