Popular repositories Loading
-
-
-
cuda-sgemm-kernels
cuda-sgemm-kernels PublicFour successive CUDA SGEMM kernels (naive to 1D block-tiling) plus a fused warp-shuffle softmax, benchmarked against cuBLAS on a Tesla T4. 27.6x over naive, 42.3% of cuBLAS.
Cuda
-
tensor-dsl-compiler
tensor-dsl-compiler PublicA tiny tensor DSL compiler in C++ emitting textual LLVM IR, with a loop-fusion pass: 5 loops to 1, 4 buffers eliminated, IR instructions 69 to 19 (3.63x). 6/6 tests verified against a NumPy oracle.
C++
-
packet-ring-buffer-c
packet-ring-buffer-c PublicBounded producer/consumer queue in C: mutex+condvar vs lock-free SPSC, benchmarked on real IPv4/TCP parsing. Valgrind + ThreadSanitizer clean.
C
-
adversarial-review-fraud-detection
adversarial-review-fraud-detection PublicThree-agent LangGraph benchmark: LLM fraud detection collapses from 100% to 0% recall as an adversarial curriculum escalates. Deterministic evidence-grounding gate.
Python
If the problem persists, check the GitHub status page or contact support.



