← sources

ggerganov/llama.cpp b8956: b8956

source: github (github:ggerganov/llama.cpp)published 2026-04-28 13:48ingested 2026-04-29 08:10

extract

<details open> CANN: add new ops, optimize existing ops (#21204) New operators: - GGML_OP_SET: implement via aclnnInplaceCopy on target region - GGML_OP_CUMSUM: implement via aclnnCumsum - GGML_OP_FILL: implement via aclnnInplaceFillScalar - GGML_OP_DIAG: implement via aclnnInplaceCopy on diagonal strides - GGML_OP_TRI (lower/lower_diag/upper_diag/upper): implement via aclnnTril(-1/0) and aclnnTriu(0/1) with appropriate diagonal offsets - GGML_OP_SOLVE_TRI: implement via aclnnTriangularSolve - GGML_UNARY_OP_SOFTPLUS: implement via aclnnSoftplus Optimizations: - GLU (SwiGLU/GeGLU/GeGLU_ERF/GeGLU_QUICK): fuse with aclnnSwiGlu / aclnnGeGluV3 when applicable; fallback conditions now checked inside each function rather than at the call site - CROSS_ENTROPY_LOSS: replace 5-kernel sequence (LogSoftmax→Mul→ ReduceSum×2→Muls) with single aclnnSoftmaxCrossEntropyWithLogits call - L2_NORM: fix in-place ClampMin on norm result (was clamping wrong tensor); add eps clamping before division to avoid divide-by-zero - PAD_REFLECT_1D: eliminate per-ne[3] loop; assert contiguity and call ReflectionPad1d once on the full 4-D view; remove redundant nb copies - GET_ROWS: replace IndexSelect with GatherV2 per batch slice; refactor helper into gather_batched lambda with batch loop inlined - SET_ROWS: replace IndexCopy with InplaceIndexCopy per batch slice; refactor helper into scatter_batched lambda with batch loop inlined - OUT_PROD: replace O(ne[3]*ne[2]*ne[1]) Ger+InplaceAdd loop with per-slice Matmul loop (src0 @ src1^T); handles strided-broadcast batch dims where ne02/ne03 may differ from ne2/ne3 - backend memset_tensor: implement via aclrtMemset (was NULL) Bug fixes: - COUNT_EQUAL: use non-inplace EqTensor into a same-type temporary buffer instead of InplaceEqTensor, avoiding corruption of src0 - ACL graph cache (USE_ACL_GRAPH): restore node_type and src_type[] fields in ggml_graph_node_properties; has_matching_properties() was missing type checks, causing F16 and BF16 tensors (same nb[0]=2) to incorrectly share cached graphs and produce wrong results (ERR≈679) - graph cache op_params matching: compare full GGML_MAX_OP_PARAMS bytes so that ops differing only in parameters are not incorrectly replayed from cache </details> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-macos-arm64.tar.gz) - [macOS Apple Silicon (arm64, KleidiAI enabled)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-macos-arm64-kleidiai.tar.gz) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-arm64.tar.gz) - [Ubuntu s390x (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-s390x.tar.gz) - [Ubuntu x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-vulkan-x64.tar.gz) - [Ubuntu arm64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-vulkan-arm64.tar.gz) - [Ubuntu x64 (ROCm 7.2)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-rocm-7.2-x64.tar.gz) - [Ubuntu x64 (OpenVINO)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-openvino-2026.0-x64.tar.gz) - [Ubuntu x64 (SYCL FP32)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-sycl-fp32-x64.tar.gz) - [Ubuntu x64 (SYCL FP16)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-ubuntu-sycl-fp16-x64.tar.gz) **Android:** - [Android arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-android-arm64.tar.gz) **Windows:** - [Windows x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-cpu-x64.zip) - [Windows arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-cpu-arm64.zip) - [Windows x64 (CUDA 12)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-cuda-12.4-x64.zip) - [CUDA 12.4 DLLs](https://github.com/ggml-org/llama.cpp/releases/download/b8956/cudart-llama-bin-win-cuda-12.4-x64.zip) - [Windows x64 (CUDA 13)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-cuda-13.1-x64.zip) - [CUDA 13.1 DLLs](https://github.com/ggml-org/llama.cpp/releases/download/b8956/cudart-llama-bin-win-cuda-13.1-x64.zip) - [Windows x64 (Vulkan)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-vulkan-x64.zip) - [Windows x64 (SYCL)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-sycl-x64.zip) - [Windows x64 (HIP)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-win-hip-radeon-x64.zip) **openEuler:** - [openEuler x86 (310p)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-310p-openEuler-x86.tar.gz) - [openEuler x86 (910b, ACL Graph)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-910b-openEuler-x86-aclgraph.tar.gz) - [openEuler aarch64 (310p)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-310p-openEuler-aarch64.tar.gz) - [openEuler aarch64 (910b, ACL Graph)](https://github.com/ggml-org/llama.cpp/releases/download/b8956/llama-b8956-bin-910b-openEuler-aarch64-aclgraph.tar.gz)

retained text

No document body was retained for this record — only its metadata was stored.