UPSTREAM PR #18467: vulkan: support buffer_from_host_ptr

[loci-dev]

Mirrored from ggml-org/llama.cpp#18467

For ggml-org/llama.cpp#18166 (comment).

Draft, the model loader change needs testing/polishing and to be merged with the other pending change.

[loci-review]

Explore the complete analysis inside the Version Insights

I've successfully generated the summary report for your project. The analysis shows significant performance regressions in PR #747 for the llama.cpp repository.

Key Highlights:

⚠️ Critical Issues Found:

• All top 10 functions show performance degradation (4.9% to 165.3% increase in response time)
• Throughput decreased across all functions (9.5% to 311.5% degradation)
• The most severely impacted function is std::_Rb_tree::_S_key with a 165.3% response time increase

Most Affected Components:

• STL Red-Black Tree operations
• STL Vector operations
• STL Hash table operations
• Standard algorithms and smart pointers

Recommendation: The report suggests not merging PR #747 in its current state due to severe performance regressions.

Would you like me to help you with:

1. Generating flame graphs for deeper analysis?
2. Retrieving Control Flow Graphs for specific functions?
3. Getting more detailed insights on particular functions?

[loci-review]

Explore the complete analysis inside the Version Insights

Perfect! I've generated a summary report for your project. Here are the key findings:

Summary Report for llama.cpp PR #747

Performance Analysis Result: ✅ PASSED

The analysis comparing the base version to the target version shows:

• No significant performance regressions detected
• All modified functions show performance changes of less than 2% for both:
  • Response Time (execution time per function)
  • Throughput Time (total time including nested calls)

Conclusion:
Pull Request #747 maintains performance stability and is safe to merge from a performance perspective. The changes introduced have minimal to no negative impact on the codebase performance.

Would you like me to provide more detailed information about specific functions or any other aspect of this analysis?

[loci-review]

Explore the complete analysis inside the Version Insights

Perfect! I've generated the summary report for your project. Here's what the analysis shows:

Key Findings:

Project: llama.cpp (auroralabs-loci) - Pull Request #747

Performance Impact: ✅ No significant changes detected

The comparison between the base version and target version shows:

• Response Time: No modified functions with changes > 2%
• Throughput Time: No modified functions with changes > 2%

Conclusion: This pull request maintains performance stability with no notable regressions or improvements exceeding the 2% measurement threshold. From a performance perspective, it appears safe to merge.