Weekly Org Report (2025-11-17..2025-11-23)

[github-actions]

Issue Metrics (2025-11-17..2025-11-23)

Summary

• Total issues/PRs: 9
• Closed issues/PRs: 9
• Average time to close: -575224 days, 09:47:17

Issues and Pull Requests

• [FEATURE] Collect all gpu macro definitions into one readme docs — eunomia-bpf/bpftime #509 (✅ Closed)
• Best practices for real-time streaming of eBPF SSL/TLS capture data in Kubernetes? — eunomia-bpf/bpf-developer-tutorial #187 (✅ Closed)
• Outdated README file — eunomia-bpf/bpftime #508 (✅ Closed)
• the collector program fails to load the eBPF object,error is R2 min value is negative — eunomia-bpf/agentsight Add workflow dispatch trigger page deploy #20 (✅ Closed)
• [FEATURE] Reduce the cost of cudaHostRegister to improve performance — eunomia-bpf/bpftime #506 (✅ Closed)
• [FEATURE] Add Per-thread runtime distribution example for bpftime GPU tracing — eunomia-bpf/bpftime #505 (✅ Closed)
• [FEATURE] Add SM / warp / lane mapping example for bpftime GPU tracing — eunomia-bpf/bpftime #504 (✅ Closed)
• [FEATURE] setTargetTriple API change for LLVM ≥15 — eunomia-bpf/llvmbpf Adaptation Issues on CentOS 7 #36 (✅ Closed)
• Weekly Org Report (2025-11-03..2025-11-09) — eunomia-bpf/eunomia.dev Weekly Org Report (2025-11-03..2025-11-09) #54 (✅ Closed)

Org Activity Addendum (eunomia-bpf, 2025-11-17..2025-11-23)

New Stars

• Total new stars: 61

By Repository:

• eunomia-bpf/eunomia-bpf: 3 new stars
• eunomia-bpf/wasm-bpf: 2 new stars
• eunomia-bpf/bpf-developer-tutorial: 14 new stars
• eunomia-bpf/GPTtrace: 2 new stars
• eunomia-bpf/libbpf-rs-starter-template: 2 new stars
• eunomia-bpf/cilium-ebpf-starter-template: 1 new stars
• eunomia-bpf/KEN: 1 new stars
• eunomia-bpf/bpftime: 14 new stars
• eunomia-bpf/eunomia.dev: 2 new stars
• eunomia-bpf/GPTtrace-web: 1 new stars
• eunomia-bpf/inline-hook-demo: 1 new stars
• eunomia-bpf/eGPU: 2 new stars
• eunomia-bpf/llvmbpf: 2 new stars
• eunomia-bpf/basic-cuda-tutorial: 1 new stars
• eunomia-bpf/cupti-tutorial: 1 new stars
• eunomia-bpf/xpu-perf: 1 new stars
• eunomia-bpf/schedcp: 1 new stars
• eunomia-bpf/agentsight: 4 new stars
• eunomia-bpf/MCPtrace: 4 new stars
• eunomia-bpf/gpu_ext_policy: 2 new stars

New Repositories

• eunomia-bpf/xsched — created 2025-11-17T05:00:09Z
• eunomia-bpf/faiss — created 2025-11-18T20:04:27Z
• eunomia-bpf/llama.cpp — created 2025-11-19T01:57:05Z
• eunomia-bpf/gpu_ext_policy — created 2025-11-20T06:16:17Z

Pull Requests Opened

• [Feature]: Add cutlass support — eunomia-bpf/bpftime #511
• [~]Fix: Change dynamic library loading to avoid glibc error — eunomia-bpf/bpftime #510
• Add option for LLVM preload workaround in CMake and update JIT context — eunomia-bpf/llvmbpf docs: Update README.md #38
• Update for "Reduce the cost of cudaHostRegister to improve performance" — eunomia-bpf/llvmbpf [BUG] docs/tutorials listed in .gitignore #37
• fix the issue where the eBPF verifier rejects the probe_SSL_write_ex_exit — eunomia-bpf/agentsight Update build instruction for bpftime #21
• Make tutorials build on a DGX Spark w/ CUDA 13.0 — eunomia-bpf/cupti-tutorial Add README #3
• [+]Feat: Reduce the cost of cudaHostRegister to improve performance — eunomia-bpf/bpftime #507
• Fix GPU-CPU memory synchronization and illegal memory access on NVIDI… — eunomia-bpf/bpftime #503

Pull Requests Merged

• [~]Fix: Change dynamic library loading to avoid glibc error — eunomia-bpf/bpftime #510
• Add option for LLVM preload workaround in CMake and update JIT context — eunomia-bpf/llvmbpf docs: Update README.md #38
• Update for "Reduce the cost of cudaHostRegister to improve performance" — eunomia-bpf/llvmbpf [BUG] docs/tutorials listed in .gitignore #37
• fix the issue where the eBPF verifier rejects the probe_SSL_write_ex_exit — eunomia-bpf/agentsight Update build instruction for bpftime #21
• [+]Feat: Reduce the cost of cudaHostRegister to improve performance — eunomia-bpf/bpftime #507
• [feature] Support for pytorch on GPU — eunomia-bpf/bpftime #501

Commits

• Remove vectoradd flamegraph SVG and folded stack files — eunomia-bpf/xpu-perf @aefc551 (2025-11-16T16:55:05-08:00)
• Add script to run all NVIDIA bpftrace tests and log results — eunomia-bpf/xpu-perf @3b6ada7 (2025-11-16T20:43:29-08:00)
• feat: Add untested NVIDIA function tests and update test runner — eunomia-bpf/xpu-perf @fdd954e (2025-11-16T22:44:38-08:00)
• Enhance NVIDIA bpftrace test runner with monolithic support and logging improvements — eunomia-bpf/xpu-perf @9a698c4 (2025-11-16T22:34:14-08:00)
• feat: Update command-line interface to support subcommands for profiling modes — eunomia-bpf/xpu-perf @275f2f2 (2025-11-16T19:48:01-08:00)
• feat: Add bpftrace scripts for prefetch hint generation and thrashing detection — eunomia-bpf/xpu-perf @f26acbb (2025-11-19T22:42:35-08:00)
• feat: Add bpftrace script for tracing uvm_va_block_select_residency function with detailed metrics — eunomia-bpf/xpu-perf @12ea28e (2025-11-19T14:44:33-08:00)
• docs: Add README entry for available filter functions in tracing — eunomia-bpf/xpu-perf @c6a3189 (2025-11-19T12:32:40-08:00)
• refactor: Update .gitignore and enhance README; modify kernel parameters for improved performance and usability — eunomia-bpf/co-processor-demo @06765ee (2025-11-16T21:37:23-08:00)
• Refactor and update UVM benchmark suite — eunomia-bpf/co-processor-demo @2345f95 (2025-11-16T20:57:35-08:00)
• refactor: Add Makefile and .gitignore for GEMM CUDA test application setup — eunomia-bpf/co-processor-demo @d36b73f (2025-11-17T14:56:10-08:00)
• Refactor Makefile and remove unused wrapper kernel to streamline build process; update gemm_policy_wrapper.h and gemm_test.cu for improved kernel launching with optional policy framework. — eunomia-bpf/co-processor-demo @352ef7d (2025-11-18T00:37:19-08:00)
• Enhance configuration loading by allowing model directory specification in load_config and build_tokenizer functions — eunomia-bpf/co-processor-demo @9cce0ac (2025-11-17T22:59:46-08:00)
• Add CUPTI-based kernel interception and policy enforcement for cuBLAS — eunomia-bpf/co-processor-demo @2796888 (2025-11-17T22:45:21-08:00)
• Remove --no-device-link flag from Makefiles and documentation across various projects to align with updated CUDA compilation practices. This change enhances compatibility with newer CUDA versions while maintaining performance optimizations. The modifications include adjustments in Makefiles for GPU examples, memory benchmarks, and scheduler components, as well as updates to README files to reflect the removal of the flag in example commands and instructions. — eunomia-bpf/co-processor-demo @48367f2 (2025-11-17T22:40:01-08:00)
• Refactor code structure for improved readability and maintainability — eunomia-bpf/co-processor-demo @6e8b5d5 (2025-11-17T22:16:40-08:00)
• Implement cuBLAS dynamic linking with runtime policy enforcement — eunomia-bpf/co-processor-demo @91a356c (2025-11-17T18:43:33-08:00)
• refactor: Enhance runtime dynamic linking by modifying Makefile and updating kernel invocation in wrapper — eunomia-bpf/co-processor-demo @a70d7e4 (2025-11-17T17:16:06-08:00)
• refactor: Update Makefile for dynamic linking and remove gemm_kernel.cu — eunomia-bpf/co-processor-demo @3a67c8b (2025-11-17T17:02:49-08:00)
• refactor: Implement runtime dynamic linking for GEMM kernel and policy integration — eunomia-bpf/co-processor-demo @e9b2391 (2025-11-17T16:53:25-08:00)
• refactor: Add RQ4-RQ6 sections for UVM prefetching, thrashing detection, and access counter migration experiments — eunomia-bpf/co-processor-demo @44abbb0 (2025-11-17T16:36:17-08:00)
• fix: Update baseline size factor reference in UVM slowdown calculation — eunomia-bpf/co-processor-demo @a89a1b3 (2025-11-17T15:52:30-08:00)
• Add Windows compatibility for memory management functions — eunomia-bpf/co-processor-demo @d1ed31a (2025-11-17T22:12:41-08:00)
• refactor: Consolidate size factors for kernel benchmarks to ensure consistent runtime across configurations — eunomia-bpf/co-processor-demo @cd3defc (2025-11-16T23:47:12-08:00)
• refactor: Enhance pointer chase kernel with GPU-based node initialization and multi-segment design for oversubscription testing — eunomia-bpf/co-processor-demo @f1d380e (2025-11-16T23:27:02-08:00)
• refactor: Add defensive checks for iterations and update input/output byte calculations in synthetic kernel — eunomia-bpf/co-processor-demo @af3b2d1 (2025-11-16T23:01:28-08:00)
• Add UVM memory management strategies and benchmark results for various kernels — eunomia-bpf/co-processor-demo @c2a0319 (2025-11-19T00:54:55-08:00)
• Refactor GEMM example to use nvJitLink policy framework — eunomia-bpf/co-processor-demo @71c04bb (2025-11-18T13:53:20-08:00)
• Add GEMM kernel implementation and policy framework integration; update .gitignore to include new test files. — eunomia-bpf/co-processor-demo @b354e2d (2025-11-18T13:44:36-08:00)
• Enhance README and policy framework with device function pointers; introduce generic wrapper kernel template for simplified kernel integration with policies. — eunomia-bpf/co-processor-demo @ea48f93 (2025-11-18T13:11:03-08:00)
• Refactor Makefile and source files to streamline policy application; enhance policy function pointer handling and update kernel execution logic. — eunomia-bpf/co-processor-demo @01de14f (2025-11-18T12:51:36-08:00)
• Refactor Makefile and source files to implement policy wrappers for extracted kernels; update policy application logic and enhance PTX extraction process. — eunomia-bpf/co-processor-demo @702a4a4 (2025-11-18T12:44:47-08:00)
• Add nvJitLink policy framework demo with user kernel integration and runtime policy application — eunomia-bpf/co-processor-demo @33d16ad (2025-11-18T12:25:25-08:00)
• Refactor Makefile and gemm_test.cu to link GEMM kernel with policy at runtime; remove user_kernel.cu as it is no longer needed. — eunomia-bpf/co-processor-demo @273e088 (2025-11-18T12:24:30-08:00)
• Add binary extraction and JIT rewriting framework — eunomia-bpf/co-processor-demo @685d17b (2025-11-18T12:12:33-08:00)
• Add demos for dynamic loading, cuGetProcAddress, nvJitLink, nvFatbin, and context-independent loading; enhance README files with detailed usage instructions and examples for each feature. — eunomia-bpf/co-processor-demo @3a7ba10 (2025-11-18T11:46:55-08:00)
• Add scheduling policies and GEMM benchmark for CUDA — eunomia-bpf/co-processor-demo @75482df (2025-11-18T19:33:49-08:00)
• Add cuBLAS demo with Makefile, README, and .gitignore — eunomia-bpf/co-processor-demo @6e5b3cd (2025-11-21T22:35:59-08:00)
• Remove QLoRA fine-tuning files and dependencies from the repository — eunomia-bpf/co-processor-demo @4cd2e6b (2025-11-21T23:13:26-08:00)
• refactor: Enhance random access pattern in rand_chunk_kernel and update size factors for oversubscription characterization — eunomia-bpf/co-processor-demo @453d1c6 (2025-11-17T00:48:49-08:00)
• feat: Add GEMM kernel implementation with Makefile and README documentation — eunomia-bpf/co-processor-demo @6dc3cd3 (2025-11-17T20:19:04-08:00)
• Add inline policy support for CLC scheduler with new policies and update Makefile — eunomia-bpf/co-processor-demo @9bdcb65 (2025-11-18T18:23:44-08:00)
• Add NoStealPolicy implementation and update Makefile and benchmark driver — eunomia-bpf/co-processor-demo @a0f72db (2025-11-18T17:40:56-08:00)
• Add CLC Policy Framework for Qwen3 CUDA Inference — eunomia-bpf/co-processor-demo @382b454 (2025-11-18T15:13:22-08:00)
• Add GEMM Benchmark System with comprehensive documentation and visualization — eunomia-bpf/co-processor-demo @49ca80e (2025-11-18T14:46:21-08:00)
• ds — eunomia-bpf/co-processor-demo @54266fa (2025-11-18T14:39:38-08:00)
• Add policy framework and GEMM kernel with policy support — eunomia-bpf/co-processor-demo @b640ce1 (2025-11-18T14:26:07-08:00)
• Add policy framework and GEMM kernel with nvJitLink support — eunomia-bpf/co-processor-demo @6457cae (2025-11-18T14:12:26-08:00)
• Add workload submodule for llama.cpp and enhance visualization scripts for dynamic subplot creation — eunomia-bpf/co-processor-demo @a1ec5e8 (2025-11-18T19:04:59-08:00)
• Add NVIDIA UVM Kernel Parameters Configuration Guide — eunomia-bpf/co-processor-demo @53ac820 (2025-11-19T22:13:21-08:00)
• Add gpu_ext_policy submodule for external GPU scheduling policies — eunomia-bpf/co-processor-demo @d3e8001 (2025-11-19T22:26:47-08:00)
• Add option for LLVM preload workaround in CMake and update JIT context (#38) — eunomia-bpf/llvmbpf @9ea0180 (2025-11-23T17:36:19+08:00)
• Update for buffer size & spin_lock setting (#37) — eunomia-bpf/llvmbpf @e54f5aa (2025-11-22T22:13:16-08:00)
• Refactor CUDA library detection and linking in CMakeLists.txt — eunomia-bpf/llvmbpf @f6c531d (2025-11-22T22:15:45-08:00)
• Add Spirv support (#35) — eunomia-bpf/llvmbpf @b8e4eea (2025-11-22T22:15:45-08:00)
• Update README and example PTX documentation to clarify GPU architecture detection and compilation options. Remove references to automatic compute capability detection for PTX generation, and note that the example code is hardcoded for a specific GPU architecture. Enhance clarity on build instructions and output examples. — eunomia-bpf/llvmbpf @4e36518 (2025-11-22T22:15:15-08:00)
• Published post and updated queue [skip ci] — eunomia-bpf/eunomia.dev @2f07894 (2025-11-18T07:12:50Z)
• Revise ideas.md with new project proposals and structure — eunomia-bpf/eunomia.dev @13249f1 (2025-11-17T12:02:30-08:00)
• Document APX-aware JIT backend project — eunomia-bpf/eunomia.dev @1b31ac5 (2025-11-17T11:46:32-06:00)
• ix the issue where the eBPF verifier rejects the probe_SSL_write_ex_exit (#21) — eunomia-bpf/agentsight @7d9de42 (2025-11-20T21:42:59-08:00)
• Add prefetch_always_max and prefetch_none BPF programs with necessary updates — eunomia-bpf/gpu_ext_policy @26e1420 (2025-11-23T01:41:37-08:00)
• Refactor code structure for improved readability and maintainability — eunomia-bpf/gpu_ext_policy @6f2d812 (2025-11-23T01:34:48-08:00)
• Move kfunc test for substring search from test_1 to test_3 — eunomia-bpf/gpu_ext_policy @c38b456 (2025-11-23T00:32:55-08:00)
• Add always_max BPF program and cleanup tool for struct_ops instances — eunomia-bpf/gpu_ext_policy @24b699a (2025-11-23T00:42:52-08:00)
• Refactor struct_ops to implement uvm_bpf_test_trigger_kfunc and update cleanup logic for struct_ops map — eunomia-bpf/gpu_ext_policy @50d8e86 (2025-11-23T00:39:29-08:00)
• Refactor code structure for improved readability and maintainability — eunomia-bpf/gpu_ext_policy @4fa280a (2025-11-23T00:58:52-08:00)
• Refactor struct_ops/test_1 to include kfunc test for substring search — eunomia-bpf/gpu_ext_policy @aa81dbe (2025-11-22T18:59:32-08:00)
• Enhance adaptive prefetch policy with NVML integration for PCIe throughput monitoring and update threshold dynamically — eunomia-bpf/gpu_ext_policy @41b5b75 (2025-11-23T11:53:46-08:00)
• Add adaptive threshold prefetch policy implementation and update Makefile — eunomia-bpf/gpu_ext_policy @49d6dcb (2025-11-23T11:14:11-08:00)
• Initial commit — eunomia-bpf/gpu_ext_policy @517aeb7 (2025-11-19T22:16:19-08:00)
• Add cleanup detection for old struct_ops instances in userspace program — eunomia-bpf/gpu_ext_policy @0e88d03 (2025-11-21T16:11:37-08:00)
• Remove bootstrap BPF program and associated files; add new kernel module with struct_ops support — eunomia-bpf/gpu_ext_policy @52decf6 (2025-11-19T22:26:16-08:00)
• Enhance GCN Benchmark with UVM Support and Add Testing Script — eunomia-bpf/schedcp @9442dbb (2025-11-22T11:53:48-08:00)
• feat: Optimize memory usage in PyTorch benchmarks with chunked index_add and add GraphSAGE training benchmark — eunomia-bpf/schedcp @df7cbbd (2025-11-22T01:25:51-08:00)
• feat: Add CPU benchmark script for performance evaluation — eunomia-bpf/schedcp @c7dfd01 (2025-11-21T17:41:27-08:00)
• feat: Refactor GPU index creation to unify UVM and non-UVM logic — eunomia-bpf/schedcp @0202331 (2025-11-21T17:30:07-08:00)
• feat: Enhance UVM support by skipping CPU aggregation and index writing in UVM mode — eunomia-bpf/schedcp @ae994a9 (2025-11-21T17:22:20-08:00)
• feat: Add support for CUDA Unified Virtual Memory in FAISS benchmark script — eunomia-bpf/schedcp @cb86f6f (2025-11-21T17:10:12-08:00)
• Add FAISS UVM tests and benchmark results — eunomia-bpf/schedcp @8b61186 (2025-11-21T16:49:42-08:00)
• feat: Add FAISS benchmark runner script and README for setup instructions — eunomia-bpf/schedcp @2b38d0e (2025-11-21T16:07:10-08:00)
• Enhance CUDA support in llama.cpp with UVM for large model handling — eunomia-bpf/schedcp @0875f6a (2025-11-18T23:37:02-08:00)
• Remove roofline analysis scripts and related files for DeepSeek-R1 — eunomia-bpf/schedcp @e67ac99 (2025-11-18T20:14:54-08:00)
• feat: Replace llama.cpp submodule with eunomia-bpf fork — eunomia-bpf/schedcp @0be3ef9 (2025-11-18T19:29:18-08:00)
• Update README.md for testing under 5090 platform and add benchmarks for ncmoe configurations — eunomia-bpf/schedcp @cf77a82 (2025-11-19T00:04:52-08:00)
• feat: Add PyTorch benchmark workloads and testing scripts — eunomia-bpf/schedcp @ef90ff9 (2025-11-21T23:12:51-08:00)
• feat: Implement PyTorch UVM allocator with memory statistics and add GNN training benchmark — eunomia-bpf/schedcp @8a8dccd (2025-11-22T01:08:03-08:00)
• [feature] Support for pytorch on GPU (#501) — eunomia-bpf/bpftime @6aa1989 (2025-11-23T18:26:19+08:00)
• [+]Feat: Reduce the cost of cudaHostRegister to improve performance (#507) — eunomia-bpf/bpftime @6c37dd0 (2025-11-23T21:44:34+08:00)
• [~]Fix: Change dynamic library loading to avoid glibc error (#510) — eunomia-bpf/bpftime @2dbab79 (2025-11-23T23:22:26+08:00)