AMD 核心软件组织的 GitHub 全景:1000+ 仓库与 ROCm / NPU 技术路线
Published:
AMD 的 GitHub 不是一个组织,而是一张“核心计算 + 芯片平台(含 NPU)+ AI 模型 + 图形生态”的分层地图。本文以 2026-10-27 的 GitHub 公开数据、ROCm 10.1.0 与 Ryzen AI 官方文档为准:先把组织和仓库盘清,再用图拆开 ROCm 与 Ryzen AI NPU 的调用关系、软件栈和技术路线。(由 AI 辅助整理,数据和链接以官方来源为准)
图示已统一画布尺寸;部分复杂图拆成左右/三列/四列面板,便于对照阅读。
0. 统计口径
本文覆盖 AMD 的 9 个主组织:ROCm、amd、AMD-AGI、AMD-Ecosystem、AMDResearch、GPUOpen-LibrariesAndSDKs、GPUOpen-Effects、GPUOpen-Tools、GPUOpen-Drivers。
这里的“全景”指 AMD 核心软件组织的全景,不是所有 AMD GitHub 组织的穷举;Xilinx 等只在 NPU 上游处引用。
共 1035 个公开仓库:596 个活跃非 fork、366 个 fork(含 17 个已归档 fork)、90 个归档仓库。三列相加为 1052,比 1035 多 17,因为归档 fork 被同时计入 fork 和归档。另外附录还收录 9 个外围官方组织(AMD-HPC、AMDESE、pensando、AMDEPYC、AMDComputeLibraries、ROCm-LS、ROCm-DS、amd-enterprise-ai、AMD-AIM),共 144 个仓库;另有 3 个历史组织的 14 个归档仓库。
不包含:
AMD-OSX:社区的黑苹果项目,不是 AMD 官方软件栈。AMDlab:名称相近,但内容是与半导体无关的建筑设计工具。- 历史组织名
ROCmSoftwarePlatform、RadeonOpenCompute、ROCm-Developer-Tools:共 14 个归档仓库,已迁移到 ROCm,见附录 11.3。 AMD-IT、AMD-DC-GPU、AMD-OpenSIL、AMD-AI等空组织:当前没有公开仓库,暂不展开。- Xilinx:AMD 收购后的独立 GitHub 组织;
mlir-aie、mlir-air是 NPU 编译器上游,只在第 8 节引用,不计入 1035。
说明:仓库数量、Star 数和简介会随时间变化;本文是 2026-10-27 的快照。完整清单保留 GitHub 官方英文简介,避免二次翻译失真;核心仓库在正文中给出中文解读。
1. 一张图看组织:AMD 的 GitHub 是“三层结构”
flowchart TB
AMD["AMD GitHub 生态<br/>1000+ 公开仓库"]
subgraph CORE["核心 GPU 计算"]
ROCM["ROCm<br/>414 个仓库"]
end
subgraph PLATFORM["芯片、平台与 AI"]
AMDORG["amd<br/>126 个仓库"]
AGI["AMD-AGI<br/>72 个仓库"]
ECO["AMD-Ecosystem<br/>213 个仓库"]
RES["AMDResearch<br/>23 个仓库"]
end
subgraph GRAPHICS["图形、游戏与可视化"]
LIB["GPUOpen-Libraries<br/>85 个仓库"]
FX["GPUOpen-Effects<br/>22 个仓库"]
TOOLS["GPUOpen-Tools<br/>66 个仓库"]
DRV["GPUOpen-Drivers<br/>14 个仓库"]
end
AMD --> CORE
AMD --> PLATFORM
AMD --> GRAPHICS
图 1:AMD GitHub 组织地图。 图里是三层:AMD GitHub 生态 → 3 类组织 → 9 个组织;仓库数量标在组织节点上,不再单独画仓库层。ROCm 负责 GPU 通用计算;amd / AMD-AGI / AMD-Ecosystem / AMDResearch 负责芯片平台、AI 模型与生态移植;四个 GPUOpen 组织负责图形、游戏和可视化。
1.1 数字先看
| 组织 | 官方定位 | 仓库数 | 活跃非 fork | fork(含归档) | 归档 |
|---|---|---|---|---|---|
| ROCm | AMD ROCm 软件栈:编译器、运行时、数学库、通信库、工具 | 414 | 225 | 124 | 81 |
| amd | AMD 公司级组织:Ryzen AI、AOCL、EPYC、FPGA、Quark | 126 | 97 | 26 | 3 |
| AMD-AGI | AMD AI Group:开源模型、训练框架、推理优化、Agent | 72 | 62 | 10 | 0 |
| AMD-Ecosystem | 开源生态软件的统一入口:LLM、数据科学、生命科学、仿真、金融 | 213 | 23 | 190 | 0 |
| AMDResearch | AMD 研究院:NPU 教学、性能工具、HPC 研究原型 | 23 | 21 | 2 | 0 |
| GPUOpen-LibrariesAndSDKs | 图形/游戏 SDK 与库 | 85 | 83 | 1 | 1 |
| GPUOpen-Effects | 图形特效库:FSR、TressFX、FEMFX 等 | 22 | 21 | 1 | 0 |
| GPUOpen-Tools | 图形/计算性能分析与调试工具 | 66 | 58 | 8 | 0 |
| GPUOpen-Drivers | 开源图形驱动与组件 | 14 | 6 | 4 | 5 |
fork列含已归档 fork;归档列含其中 17 个(ROCm 16、GPUOpen-Drivers 1),所以这两行三列相加会多于仓库数。这 9 个主组织在附录中的 fork 标签数为 349。
1.2 三个关键结论
第一,ROCm 正在“收拢”。
2025—2026 年 ROCm 最大的变化不是新增某个库,而是仓库结构重组:
- 自 ROCm 7.14 起,ROCm 改用 TheRock 构建和发布;
- 数学/算子库集中到
ROCm/rocm-libraries超级仓库; - 运行时、通信、性能工具等集中到
ROCm/rocm-systems超级仓库; - 旧的单组件仓库大量标记为
[DEPRECATED] Moved to ...,部分已归档。
所以今天看 ROCm,不能只看 ROCm/rocBLAS、ROCm/rccl 这些旧仓库,而要先看 TheRock、rocm-libraries、rocm-systems 三个入口。
第二,AMD 的 GitHub 远不止 ROCm。
amd组织覆盖 Ryzen AI NPU、AOCL CPU 数学库、EPYC 系统管理、FPGA/Vitis;AMD-AGI是 AMD 的 AI 模型与训练/推理优化组织;AMD-Ecosystem是 CUDA 项目向 ROCm 移植的“生态池”;- NPU 仓库分散在
amd(RyzenAI-SW、xdna-driver、IRON、Triton-XDNA)、ROCm(FastFlowLM、FastFlow-SDK、HRX)和AMDResearch(Riallto、NPUEval、Ryzers),详见第 8 节; - 四个 GPUOpen 组织构成一条完整的 图形/游戏技术栈,与 ROCm 计算栈相对独立。
第三,AMD-Ecosystem 的 213 个仓库里约 190 个是 fork。
这说明它的角色不是“核心 SDK”,而是 生态移植与适配层:把 llama.cpp、SGLang、FlashInfer、LMCache、MONAI、OpenFold、MuJoCo、gsplat 等上游项目移植到 AMD GPU。真正的核心实现仍然在 ROCm、amd、AMD-AGI 三个组织里。
2. 组织逐个看
2.1 ROCm:核心 GPU 计算栈
ROCm 是 AMD 的开放 GPU 计算平台,覆盖编译器、运行时、数学库、通信库、性能工具和调试工具。官方文档对它的定义是:
ROCm is AMD’s open-source GPU computing platform: an end-to-end ecosystem of compilers, runtimes, and libraries for AI, HPC, and domain-specific workloads.
ROCm 组织有 414 个仓库,但其中大量是历史仓库、fork 和镜像。建议按“新入口 / 仍活跃 / 已迁移”三组看:
新入口(先看这三个)
| 仓库 | 一句话 |
|---|---|
| TheRock | ROCm 的统一构建与发布系统,也是 HIP/ROCm 的顶层入口;自 ROCm 7.14 起取代旧的单体发布流程。 |
| rocm-libraries | 数学与算子库超级仓库:CK、hipBLAS/Lt、rocBLAS、MIOpen、rocFFT、rocPRIM、rocWMMA 等都在这里。 |
| rocm-systems | 系统项目超级仓库:HIP/CLR、ROCr、RCCL、rocSHMEM、rocprofiler、AMD SMI、hipFile 等都在这里。 |
仍活跃的组件/工具
| 仓库 | 一句话 |
|---|---|
| hip | HIP 的 C++ 运行时 API 与内核语言,CUDA 风格、可移植到 AMD GPU。 |
| HIPIFY | 把 CUDA 源码翻译成可移植 HIP C++ 的工具(hipify-clang / hipify-perl)。 |
| llvm-project | AMD 维护的 LLVM 分支,包含 amdclang、AMDGPU 后端、COMGR、device-libs。 |
| clr | AMD Compute Language Runtime,包含 hipamd、rocclr 和 OpenCL runtime。 |
| aiter | AMD 高性能 AI 算子库,Triton/CK/ASM 多后端,是 vLLM/SGLang/ATOM 的默认 kernel 后端。 |
| ATOM | 基于 AITER 的轻量 vLLM 风格 LLM 推理引擎。 |
| rocSHMEM | GPU 发起的 intra-kernel 网络库,OpenSHMEM 风格对称堆。 |
| ROCgdb | 基于 GDB 的 ROCm 源码级调试器,支持 CPU/GPU 异构调试。 |
| amdsmi | AMD SMI 系统管理接口,控制/监控 GPU,已取代 ROCm SMI。 |
| FastFlowLM | 在 Ryzen AI NPU 上运行 LLM 的优化推理栈。 |
| FastFlow-SDK | FastFlowLM + HRX + iron-tools + iron-ops 的 NPU SDK,目标 XDNA2。 |
| hrx-system | HRX:HIP Runtime Extended,加载/调度 NPU 程序。 |
已迁移/旧仓库(只作参考)
| 仓库 | 一句话 |
|---|---|
| ROCR-Runtime | AMD 的 HSA 运行时实现;已迁入 rocm-systems。 |
| rocBLAS | AMD 原生 BLAS 实现;已迁入 rocm-libraries。 |
| hipBLASLt | AMD 高性能灵活 GEMM 实现;已迁入 rocm-libraries,不是 CUDA/rocBLASLt 封装。 |
| MIOpen | AMD 开源深度学习 primitives 库;已迁入 rocm-libraries。 |
| composable_kernel | 用 HIP C++ 组合高性能 ML kernel 的编程模型;已迁入 rocm-libraries。 |
| RCCL | 多 GPU/多节点集合通信库;已迁入 rocm-systems。 |
| rocprofiler-sdk | ROCm profiling/tracing 基础设施;已归档/迁入 rocm-systems。 |
“仍活跃”只表示 GitHub 仓库未归档/仍在使用;其中部分组件已在
rocm-libraries/rocm-systems里有镜像。写新代码优先从新入口开始。
ROCm 仓库状态提示:
- 看到
[DEPRECATED] Moved to ROCm/rocm-libraries:说明该库已迁入rocm-libraries; - 看到
[DEPRECATED] Moved to ROCm/rocm-systems:说明该库已迁入rocm-systems; - 看到
ARCHIVED:历史仓库,通常只保留只读镜像或旧版发布; - 看到
FORK:上游项目的 AMD 分支(如 PyTorch、LLVM、vLLM)。
注意:FastFlowLM / FastFlow-SDK / hrx-system 是 NPU 项目,托管在 ROCm 组织不等于它们属于 ROCm GPU 栈。详见第 8 节。
2.2 amd:芯片平台与 CPU/NPU 软件
amd 组织不是 ROCm 的重复,而是 AMD 公司层面的“芯片平台软件”:
- Ryzen AI / XDNA NPU:RyzenAI-SW、xdna-driver、IRON、Triton-XDNA、GAIA、Quark;
- AOCL / Zen CPU 数学库:BLIS、libFLAME、FFTW、LibM、Sparse、Crypto、Compression、ZenDNN;
- EPYC 系统管理与安全:ESMI、HSMP、APML、SEV、perfspect、node-scraper;
- FPGA / 嵌入式 / ISP:Vitis-AI、Kria-RoboticsAI、mini-isp、Linux ISP;
- 视频分析 / VP AI:VAIP、XIR、VART。
| 仓库 | 一句话 |
|---|---|
| RyzenAI-SW | Ryzen AI Software 的示例/教程子集;完整运行时和 Vitis AI EP 在官方安装包中。 |
| xdna-driver | AMD XDNA NPU 的 Linux 内核驱动(amdxdna.ko)与 XRT SHIM。 |
| IRON | 面向 AMD NPU 的 close-to-metal 编程库。 |
| Triton-XDNA | 在 Ryzen AI NPU 上运行 Triton kernel 的后端。 |
| gaia | 在 Ryzen AI PC 上构建本地 AI agent 的开源项目。 |
| Quark | AMD 深度学习模型量化工具链,支持 PyTorch/ONNX 和多种低精度格式。 |
| ZenDNN | 面向 Zen CPU 的 DNN 推理加速库。 |
| ZenDNN-pytorch-plugin | ZenDNN 的 PyTorch 插件(zentorch)。 |
| aocl | AMD Optimizing CPU Libraries 元包。 |
| aocl-sparse | AMD 优化稀疏线性代数库。 |
| AMD-PACE | 面向 AMD CPU 的 PyTorch LLM 推理优化扩展。 |
| amd_smi_exporter | 把 EPYC/GPU 指标导出到 Prometheus。 |
| sev-utils | AMD SEV 内存加密工具。 |
| Vitis-AI | Versal AI Edge 的 AI 推理栈。 |
| vaip | Vitis AI 的 AI 推理运行时(VAIP)。 |
2.3 AMD-AGI:开源模型、训练与推理优化
AMD-AGI 是 AMD 的 AI Group。它的仓库可以分成四类:
- 开源基础模型:Instella、Instella-MoE、Instella-Math、Instella-T2I、InstellaVL、AMD-LLM、Nitro、Hummingbird;
- 训练框架:Primus、Primus-Turbo、Primus-SaFE、torchtitan-amd、maxtext-slurm;
- 推理/算子优化:Hyperloom、GEAK、Apex、AgentKernelArena、AMDKernelVault、Infera;
- 算法研究:PARD、Gumiho、FLy、DUET-VLM、ReasonLite、Tyr-the-Pruner、LoKRA、ALTO。
注意:AMD 官方模型页有些链接仍指向旧组织
AMD-AIG-AIMA,但 GitHub 会把旧仓库地址重定向到AMD-AGI。可以认为AMD-AGI是当前组织名。
| 仓库 | 一句话 |
|---|---|
| Instella | AMD 全开源语言模型系列,权重、数据、训练代码开放。 |
| Instella-MoE | 全开源 MoE 语言模型,端到端在 AMD Instinct GPU 上训练。 |
| Instella-Math | AMD 全开源推理/数学模型,用长链式思维 RL 训练。 |
| AMD-LLM | AMD-135M 语言模型的训练代码与资源。 |
| Nitro-E | 高效训练的文生图扩散模型系列。 |
| Nitro-T | 一天内从零训练的文生图扩散模型。 |
| Hummingbird | 0.9B 高效文生视频扩散模型。 |
| Primus | YAML 驱动的大规模基础模型训练框架,支持 Megatron、TorchTitan、MaxText。 |
| Primus-Turbo | 面向 AMD GPU 的大规模训练高性能加速库。 |
| Hyperloom | 自动优化 AMD GPU 上 LLM 推理负载的 agentic 系统。 |
| GEAK | 多 agent GPU kernel 性能优化器,面向 CDNA。 |
| Apex | 用 LLM agent 优化 GPU kernel 的 agent/RL 环境与基准。 |
| AgentKernelArena | 评估和训练 GPU kernel 优化 agent 的平台。 |
| AMDKernelVault | HIP/Triton kernel 生成、优化与 LLM 训练的 playground。 |
另外还有 AMD-AIM(AMD AI Group, AIM)的 21 个仓库,聚焦 Radeon、Radeon Cloud 和 GPU MODE 参考 kernel,详见附录 11.2。
2.4 AMD-Ecosystem:CUDA → ROCm 的生态移植池
AMD-Ecosystem 的官方定位是:
The unified public home for open-source ecosystem software libraries, covering simulation, finance, LLM extension, life science, data science, and standalone utilities.
它最重要的特征不是“仓库多”,而是 fork 比例极高:213 个仓库中约 190 个是 fork。这说明它是一个 移植/适配层,把上游 CUDA 项目逐个搬到 ROCm/HIP。
| 仓库 | 一句话 |
|---|---|
| llama.cpp | AMD 维护的 llama.cpp,ROCm 上的 LLM 推理。 |
| sglang | SGLang 的 ROCm 分支,高性能 LLM 服务框架。 |
| flashinfer | FlashInfer 的 ROCm 移植。 |
| LMCache | LLM KV cache 层,ROCm 版。 |
| verl | 字节 verl 强化学习框架的 ROCm 版。 |
| gsplat-amd | 3D Gaussian Splatting 的 AMD 移植。 |
| hipDF | GPU DataFrame 库。 |
| hipMM | HIP Memory Manager(ROCm-DS)。 |
| hipGRAPH | 可移植 GPU 图算法库。 |
| hipRaft | RAPIDS RAFT 的 HIP 移植。 |
| hipCIM | cuCIM 的 HIP 移植。 |
| moat | 把 CUDA 项目逐个移植到 ROCm/HIP 的项目。 |
| MONAI | 医学影像 AI 工具包(ROCm 移植)。 |
| OpenFold3 | AlphaFold3 风格的开源生物分子结构预测(ROCm 移植)。 |
| mujoco | MuJoCo 物理仿真(ROCm 移植)。 |
| IsaacLab | 机器人学习框架(ROCm 移植)。 |
2.5 AMDResearch:研究与教学原型
AMDResearch 更像“前孵化器”:NPU 教学、GPU 性能工具、HPC 研究原型、AI4Science 配方。它的仓库不一定进入正式 ROCm 发行版,但能提前看到 AMD 在探索什么。
| 仓库 | 一句话 |
|---|---|
| Ryzers | Ryzen AI 的模块化 Dockerfile 集合,覆盖 LLM/VLM/VLA、机器人、视觉、NPU。 |
| Riallto | 面向 Ryzen AI NPU 的教学与实验环境。 |
| AUP Learning Cloud | 定制 JupyterHub 的 AMD AI 学习云。 |
| DAGEE | CPU/GPU 任务图执行引擎。 |
| IntelliKit | agent-first 的 AMD GPU 性能与验证工具集(Kerncap/Metrix/Linex/Nexus)。 |
| NPUEval | 面向 Ryzen AI NPU 的 LLM 代码生成评测数据集。 |
| intelliperf | 自动瓶颈检测与优化编排。 |
| omniprobe | 记录所有 kernel 时长的轻量工具。 |
| AMDGeniePIM | PIM 架构 GEMV 分析性能模型。 |
| AMD-DMA-Latte | 基于 DMA 引擎的 ML 集合通信实现与微基准。 |
2.6 GPUOpen:图形、游戏与可视化
四个 GPUOpen 组织构成一条独立的图形栈,和 ROCm 计算栈并行:
- GPUOpen-LibrariesAndSDKs:SDK 与库(FidelityFX SDK、AMF、Vulkan/D3D12 内存分配器、HIP RT、Radeon ProRender);
- GPUOpen-Effects:特效库(FSR、TressFX、FEMFX、SSSR、CACAO、Denoiser);
- GPUOpen-Tools:工具(RGP、RGA、RMV、RRA、OCAT、Compressonator、GPU Reshape);
- GPUOpen-Drivers:开源驱动(AMDVLK、PAL、XGL、LLPC、GPURT)。
| 仓库 | 一句话 |
|---|---|
| VulkanMemoryAllocator | 易集成的 Vulkan 内存分配库。 |
| D3D12MemoryAllocator | 易集成的 Direct3D 12 内存分配库。 |
| FidelityFX-SDK | FidelityFX 效果 SDK 主仓库。 |
| AMF | AMD 高级媒体框架 SDK,访问硬件多媒体处理。 |
| HIPRT | HIP 光线追踪库。 |
| RadeonRays_SDK | CPU/GPU 光线求交加速库。 |
| FidelityFX-FSR2 | FidelityFX Super Resolution 2。 |
| TressFX | 毛发/毛皮渲染库。 |
| FEMFX | 有限元软体/形变效果库。 |
| radeon_gpu_profiler | RGP,深度 GPU workload 分析。 |
| radeon_gpu_analyzer | RGA,离线编译器与代码分析工具。 |
| compressonator | 纹理/3D 模型压缩、优化与分析工具。 |
| AMDVLK | AMD 开源 Vulkan 驱动(Linux)。 |
| pal | 平台抽象库,封装 AMD 3D/计算驱动的硬件与 OS 细节。 |
| llpc | LLVM-Based Pipeline Compiler,把 shader 编译成 PAL 兼容 code object。 |
| gpurt | GPU 光线追踪库。 |
3. ROCm 软件栈:从应用到 GPU
3.1 分层视图
flowchart TB A["应用与框架<br/>PyTorch / JAX / TensorFlow<br/>vLLM / SGLang / ATOM<br/>ONNX Runtime / MIGraphX"] B["AI 算子与加速库<br/>AITER / CK / hipBLASLt<br/>MIOpen / AOTriton / tritonBLAS"] C["ROCm Core SDK<br/>BLAS / FFT / RAND / SPARSE / SOLVER<br/>rocPRIM / rocWMMA / RCCL / rocSHMEM"] A --> B --> C
flowchart TB D["编程模型与编译器<br/>HIP / HIPIFY / OpenCL / OpenMP<br/>Triton / Gluon / FlyDSL<br/>amdclang / LLVM AMDGPU / COMGR"] E["运行时与驱动<br/>CLR:hipamd + rocclr<br/>ROCr(Linux)/ PAL(Windows)<br/>amdkfd + amdgpu"] F["硬件<br/>CDNA / RDNA"] D --> E --> F
图 2:ROCm 软件栈分层。 左列:应用→算子→Core SDK;右列:编程模型/编译器→运行时/驱动→GPU。
XDNA NPU 不在 ROCm 支持范围;图 2 只画 GPU 栈,NPU 见第 8 节。
3.2 仓库级调用关系
flowchart TB APP["应用<br/>PyTorch / vLLM / SGLang / ATOM"] --> AITER["AITER<br/>CK / Triton / ASM / hipBLASLt"] AITER --> MATH["数学与 DNN<br/>rocBLAS / rocFFT / MIOpen / rocWMMA"]
flowchart TB HIP["HIP Runtime<br/>CLR:hipamd + rocclr<br/>编译:LLVM AMDGPU / COMGR"] --> RT["ROCr(Linux)/ PAL(Windows)<br/>amdkfd + amdgpu"] RT --> HW["CDNA / RDNA GPU"]
图 3:ROCm 仓库级调用关系。 左列:应用→AITER→数学库;右列:HIP Runtime→ROCr/PAL→GPU。MIGraphX/ONNX Runtime 走独立图执行路径,未画入这张 AITER 主链。
3.3 编译与运行调用链
flowchart TB CUDA["CUDA 源码"] --> IFY["HIPIFY"] --> S["HIP C++ / OpenCL C"] S --> F["Clang 前端<br/>amdclang++"] --> O["LLVM AMDGPU<br/>+ ROCm device-libs"] TRI["Triton / Gluon / FlyDSL"] --> MLIR["Triton MLIR / Fly MLIR"] --> O O --> CM["COMGR<br/>链接 / 检查 code object"] --> C["HSACO code object"]
flowchart TB C["HSACO code object"] --> R["HIP Runtime API<br/>CLR: hipamd + rocclr"] R --> H["ROCr(HSA)<br/>Linux"] R --> P["PAL<br/>Windows"] H --> T["libhsakmt / ROCt thunk"] --> K["amdkfd + amdgpu 内核驱动"] P --> K K --> G["AMD GPU"]
图 4:编译链(左)与运行链(右)。 HIP 源码由 amdclang 编译,经 LLVM AMDGPU 后端生成 HSACO;运行时 CLR 在 Linux 上走 ROCr,在 Windows 上走 PAL,最终通过 amdkfd/amdgpu 驱动提交到 GPU。
几个容易混淆的点:
- HIP 不是“另一个 CUDA”:HIP 是 C++ 运行时 API + kernel 语言,API 与 CUDA 高度相似,但后端是 AMD 的 CLR/ROCr。
- CLR 不是一个库:CLR 包含
hipamd、rocclr、opencl三部分;rocclr是虚拟设备接口,Linux 对接 ROCr,Windows 对接 PAL。 - ROCT thunk 已并入 ROCR-Runtime:旧的
ROCT-Thunk-Interface仓库已标记废弃,libhsakmt/ROCT 的工作并入 ROCR-Runtime。 - COMGR 是编译器侧的“胶水”:负责 code object 的链接、编译和检查,不是运行时调度器。
3.4 roc* 和 hip* 的区别
官方文档说得很清楚:
roc*前缀:AMD 原生、用 HIP 写的高性能实现,例如rocBLAS、rocFFT、rocSPARSE、rocSOLVER、rocPRIM;hip*前缀:通常是可移植的封装层,实现与 CUDA 对应的 API,让 CUDA 应用尽量少改代码就能跑在 AMD GPU 上,例如hipBLAS、hipFFT、hipRAND、hipSPARSE、hipSOLVER、hipCUB;例外是hipBLASLt,它本身是 AMD 的高性能灵活 GEMM 实现。
对应关系可以记成:
| 功能 | AMD 实现 | CUDA 兼容封装 |
|---|---|---|
| BLAS | rocBLAS | hipBLAS |
| 灵活 GEMM | hipBLASLt(AMD 实现) | — |
| FFT | rocFFT | hipFFT |
| 随机数 | rocRAND | hipRAND |
| 稀疏 | rocSPARSE | hipSPARSE |
| 求解器 | rocSOLVER | hipSOLVER |
| 并行原语 | rocPRIM | hipCUB |
| 并行算法 | rocThrust | — |
| DNN | MIOpen | hipDNN(beta) |
| 张量 | hipTensor | — |
| 线程 | hipThreads | — |
例外:
hipBLASLt不是 CUDA 封装,而是 AMD 自己的高性能灵活 GEMM 库;hip*不应一概理解为 wrapper。
选择原则:
- 新项目、追求峰值性能:优先用
roc*或 AITER/CK; - 从 CUDA 迁移、想少改代码:先用
hip*封装层; - 需要灵活 GEMM / 融合:用 hipBLASLt + CK/AITER;
- 需要图推理:用 MIGraphX 或 ONNX Runtime。
4. 技术路线:AMD GPU 编程的五条主路
flowchart TB
subgraph ROUTES["五条主要编程路线"]
H["HIP C++<br/>原生高性能路线"]
C["CUDA → HIPIFY → HIP<br/>迁移路线"]
OCL["OpenCL C<br/>跨厂商标准路线"]
OMP["OpenMP / Fortran<br/>HPC 路线"]
TRI["Triton / Gluon / FlyDSL<br/>AI Kernel DSL 路线"]
end
H --> LLVM["amdclang / LLVM AMDGPU"]
C --> LLVM
OCL --> LLVM
OMP --> LLVM
TRI --> MLIR["Triton MLIR / rocMLIR"]
MLIR --> LLVM
LLVM --> HSACO["HSACO / code object"]
HSACO --> RT["CLR / ROCr / PAL"]
RT --> GPU["AMD GPU"]
PY["Python / hip-python"] --> H
FORT["Fortran / hipfort"] --> OMP
图 5:五条编程路线。 所有路线最终都会落到 LLVM AMDGPU 后端、HSACO 和 HIP/ROCr 运行时。
4.1 路线一:CUDA → HIP 迁移
适合:已有 CUDA 代码、想快速跑在 AMD GPU 上。
路径:
CUDA 源码 → HIPIFY(hipify-clang / hipify-perl)→ HIP C++ → amdclang → HSACO → CLR/ROCr → GPU
关键仓库:
HIPIFY:CUDA 到 HIP 的源码翻译;hipify_torch:PyTorch 生态的 CUDA→HIP 翻译;hipother:在 NVIDIA 平台上把 HIP API 映射回 CUDA 的兼容层;hipBLAS/hipFFT/hipRAND/hipSPARSE/hipSOLVER/hipCUB:CUDA 对应库的封装。
现实提醒:HIPIFY 不能自动翻译所有 CUDA 代码。没有 HIP 等价库的 CUDA 库、以及针对 NVIDIA 架构做的性能优化,仍然需要人工重写和调优。官方 HIPIFY 文档明确建议:迁移后要做代码审查、替换不支持的库、重新调试和性能优化。
4.2 路线二:原生 HIP 高性能
适合:新写 kernel、追求峰值性能、需要精细控制。
路径:
HIP C++ → amdclang → LLVM AMDGPU → HSACO → CLR/ROCr → GPU
关键仓库:
hip:HIP 运行时 API 与 kernel 语言;llvm-project:amdclang、AMDGPU 后端、device-libs;rocPRIM/rocThrust/rocWMMA/hipTensor:并行原语与张量;rocprofiler-sdk/rocprofiler-compute/ROCgdb:性能分析与调试。
优化工具链:
rocprofiler-compute:kernel 级指标;rocprofiler-systems:CPU/GPU 系统级 trace;roc-optiq:profiler 可视化;ROCgdb+ROCdbgapi:源码级调试;rocprof-trace-decoder:thread trace 解码。
4.3 路线三:OpenCL 标准路线
适合:已有 OpenCL 代码、需要跨厂商可移植性。
路径:
OpenCL C → clang/SPIR-V → CLR OpenCL runtime → ROCr/PAL → GPU
关键仓库:
clr中的opencl实现;clang-ocl(历史仓库);ROCm-OpenCL-Runtime(已归档,旧版);OpenCL-CTS、OpenCL-Headers(Khronos 规范/测试)。
ROCm 10.x 仍提供 amdrocm-opencl 包(OpenCL runtime + ICD loader),但 OpenCL 不是 AMD 当前 AI/HPC 的主推路线。新项目通常优先 HIP 或 Triton。
4.4 路线四:OpenMP / Fortran HPC
适合:Fortran/C/C++ HPC 应用、已有 OpenMP 代码。
路径:
Fortran/C/C++ + OpenMP target → amdclang / flang + LLVM OpenMP → GPU
关键仓库:
aomp:AMD 的 OpenMP offload 编译器发行版;flang:Fortran 前端;hipfort:Fortran 到 HIP 的绑定;ROCm-Device-Libs、llvm-project:device 端运行时和编译器。
典型应用:GROMACS、OpenFOAM、SPECFEM3D、rocHPL、rocHPCG。
4.5 路线五:Triton / Gluon / FlyDSL
适合:AI kernel、想用 Python 写高性能 kernel、需要比 HIP 更高的抽象。
路径:
Python Triton / Gluon / FlyDSL → Triton MLIR / Fly MLIR → LLVM AMDGPU → HSACO → GPU
关键仓库:
triton:AMD 维护的 Triton 分支;aotriton:Triton 算子的 ahead-of-time 编译与分发,首个 kernel 是 FlashAttention;tritonBLAS:基于 Triton 的分析模型驱动 GEMM 库,不需要 autotune;iris:AMD RAD 的 Triton 多 GPU RMA/SHMEM 框架;FlyDSL:Python 前端 + Fly MLIR dialect,显式表达 layout/tiling/data movement;gfx950-gluon-tutorials:在 MI350/MI355 上用 Gluon 写高性能 GEMM/Attention 的教程。
三条子路线怎么选:
| 子路线 | 抽象层次 | 适合 |
|---|---|---|
| Triton | 高 | 可移植、开发快、大多数 AI kernel |
| Gluon | 中低 | 需要显式控制 layout、pipeline、register |
| FlyDSL | 低 | 需要 MLIR 级 layout 代数和深度定制 |
Gluon 的官方教程说得很直白:Triton 强在可移植和生产力,但最后 10%—20% 的性能需要显式控制;Gluon 是 Triton 的 dialect,用 block-level 编程模型补上这一段。
5. AI 推理与训练路线
5.1 AI 推理:vLLM / SGLang / ATOM + AITER
flowchart TB MODEL["模型<br/>Llama / DeepSeek / Qwen / Kimi"] --> SERVE["推理框架<br/>vLLM / SGLang / ATOM"] SERVE --> AITER["AITER<br/>Attention / MoE / GEMM / Norm"] AITER --> BACKENDS["算子后端<br/>CK / Triton / ASM / hipBLASLt"] BACKENDS --> HIP["HIP Runtime"] LLAMA["llama.cpp"] --> HIP HIP --> RT["ROCr / PAL + amdgpu"] RT --> GPU["Instinct MI300 / MI350"] SERVE --> DIST["分布式<br/>RCCL / MORI / rocSHMEM"] DIST --> NIC["RDMA NIC<br/>Pensando AINIC / Thor2 / ConnectX-7"] ONNX["ONNX 模型"] --> MIGX["MIGraphX"] --> HIP
图 6:AI 推理路线。 AITER 是 AMD 当前 LLM 推理的核心算子层;vLLM、SGLang、ATOM 都通过 AITER 调用 CK/Triton/ASM/hipBLASLt;llama.cpp 直接走 HIP/rocBLAS,不经过 AITER。
AITER 的定位(官方 README):
- AMD 高性能 AI 算子库;
- 多后端:Triton、Composable Kernel、手写 ASM;
- 覆盖推理和训练,不只是 serving kernel;
- 框架无关,可集成到 vLLM、SGLang 或自定义框架;
- 官方称其为 “default kernel backend for LLM inference on AMD GPUs”。
ATOM 是基于 AITER 的轻量 vLLM 风格实现,支持 OpenAI 兼容 API、TP/DP/EP、P/D 分离、KV cache 传输(Mooncake/MORI-IO)、FP8/MXFP4/INT8/INT4 量化、speculative decoding。
MIGraphX 是另一条路线:把 ONNX 模型编译成 AMD GPU 上的图执行引擎,适合非 Transformer 或需要图优化的场景。
5.2 AI 训练:Primus + Megatron/TorchTitan/MaxText
flowchart TB S["Primus-SaFE<br/>稳定性 / 故障容错"] -.-> P["Primus<br/>YAML 驱动训练编排"] P --> B["训练后端<br/>Megatron-LM / TorchTitan<br/>JAX MaxText / Megatron Bridge"] B --> PT["PyTorch / JAX"]
flowchart TB PT["PyTorch / JAX"] --> T["Primus-Turbo / AITER / hipBLASLt"] T --> C["RCCL / MORI / rocSHMEM"] C --> H["HIP Runtime → ROCr → amdgpu"] H --> G["MI300X / MI325X / MI355X 集群"]
图 7:训练编排(左)与算子/运行(右)。 Primus 是编排层,下面是 Megatron/TorchTitan/MaxText,再下面是 Primus-Turbo/AITER/hipBLASLt,最后是 RCCL/MORI 和 GPU。
Primus 的官方文档把它定义为“YAML 驱动的、面向 AMD GPU 的大规模基础模型训练框架”,支持:
- Megatron-LM;
- TorchTitan;
- JAX MaxText;
- Megatron Bridge;
- HummingbirdXT。
它的三层生态是:
Primus-SaFE(稳定性/集群管理)
↓
Primus-LM(训练编排)
↓
Primus-Turbo(算子/kernel)
5.3 HPC 路线
flowchart TB APP["GROMACS / OpenFOAM / SPECFEM3D / rocHPL"] --> LANG["MPI + OpenMP + Fortran/C++"] LANG --> COMP["amdclang / AOMP / Flang / hipfort"]
flowchart TB LIB["rocBLAS / rocFFT / rocSPARSE / rocSOLVER"] --> COMM["RCCL / rocSHMEM"] COMM --> RT["HIP Runtime → ROCr → amdgpu"] RT --> GPU["MI300A / MI355X / MI250X"]
图 8:应用与编译(左)、数学库与运行(右)。 HPC 应用通常走 MPI + OpenMP/Fortran,再调用 rocBLAS/rocFFT/rocSPARSE/rocSOLVER 和 RCCL/rocSHMEM。
HPC 关键仓库:
aomp、flang、hipfort:编译与 Fortran 绑定;rocBLAS、rocFFT、rocSPARSE、rocSOLVER、rocALUTION:数学库;RCCL、rocSHMEM:集合通信;rocHPL、rocHPCG、rocHPL-MxP:HPC 基准;Gromacs、OpenFOAM_HMM、SPECFEM3D:应用移植。
5.4 分布式通信路线
flowchart TB APP["训练 / 推理框架"] --> CCL["集合通信"] CCL --> RCCL["RCCL<br/>AllReduce / AllGather / ReduceScatter"] CCL --> SHMEM["rocSHMEM<br/>GPU 发起、对称堆"] CCL --> MORI["MORI<br/>MORI-EP / MORI-IO / MORI-CCL / MORI-UMBP"] RCCL --> XGMI["xGMI / PCIe"] RCCL --> RDMA["RDMA"] SHMEM --> RDMA MORI --> RDMA RDMA --> NIC["Pensando AINIC / Broadcom Thor2 / ConnectX-7"] NIC --> CLUSTER["多节点 GPU 集群"]
图 9:分布式通信路线。 节点内走 xGMI/PCIe,节点间走 RDMA;RCCL 是通用集合通信,rocSHMEM 是 GPU 发起的细粒度通信,MORI 面向 MoE EP 和 KV cache 传输。
三条路线的关系:
- RCCL:最通用的集合通信库,类似 NCCL 的 AMD 对应物;
- rocSHMEM:OpenSHMEM 风格,适合 GPU 内核内部直接做远程内存访问;
- MORI:模块化 RDMA 框架,MORI-EP 做 MoE dispatch/combine,MORI-IO 做 KV cache 点对点传输,MORI-CCL 做定制集合通信。
MORI 官方 README 明确支持 AMD Pensando DSC、Broadcom Thor2、NVIDIA Mellanox ConnectX-7 等 NIC,并已被 SGLang、vLLM、DeepSpeed 等集成。
6. 图形与光追路线
flowchart TB APP["游戏 / 渲染 / 内容创作"] --> API["Vulkan / DX12 / OpenGL"] APP --> FX["FidelityFX SDK"] --> API API --> AMDVLK["AMDVLK"] --> XGL["XGL"]
flowchart TB XGL["XGL"] --> PAL["PAL / LLPC"] --> KMD["amdgpu 内核驱动"] --> GPU["Radeon GPU"] HIPRT["HIP RT"] --> HIP["HIP / ROCm"] --> KMD
图 10:图形 API/驱动前端(左)与平台/内核/光追(右)。 图形栈走 Vulkan/DX12/OpenGL → AMDVLK/XGL/LLPC/PAL;光追可以走 FidelityFX/HIP RT;HIP RT 把光追带到 HIP/ROCm。
AMDVLK 的架构(官方 README):
- AMDVLK:开源 Vulkan 驱动;
- PAL:平台抽象库,封装硬件和 OS 细节;
- XGL:Vulkan API 层;
- LLPC:LLVM-Based Pipeline Compiler,把 shader 编译成 PAL 兼容 code object;
- GPURT:GPU 光线追踪库;
- LLVM:shader 编译基础设施。
说明:
pal、xgl、llpc、gpurt仓库目前为归档状态,但 AMDVLK 的架构仍由这些组件构成;阅读源码时以 AMDVLK README 引用的组件关系为准。
GPUOpen 工具链:
- RGP:Radeon GPU Profiler,深度分析 GPU workload;
- RGA:Radeon GPU Analyzer,离线编译和代码分析;
- RMV:Radeon Memory Visualizer,显存使用分析;
- RRA:Radeon Raytracing Analyzer,BVH 可视化;
- OCAT:FPS overlay 与性能测量;
- Compressonator:纹理/模型压缩;
- GPU Reshape:指令级插桩与验证。
7. 硬件路线:CDNA、RDNA、XDNA
flowchart TB MI100["MI100<br/>CDNA1 gfx908"] --> MI250["MI250X<br/>CDNA2 gfx90a"] MI250 --> MI300["MI300X/A<br/>CDNA3 gfx942"] MI300 --> MI355["MI355X<br/>CDNA4 gfx950"]
flowchart TB RDNA3["RX 7000<br/>RDNA3 gfx110x"] --> RDNA35["Ryzen AI Max<br/>RDNA3.5 gfx115x"] RDNA35 --> RDNA4["RX 9000 / R9700<br/>RDNA4 gfx120x"]
flowchart TB XDNA["XDNA / XDNA2<br/>AIE2 / AIE2P"] --> IRON["IRON / Triton-XDNA"]
图 11:Instinct(左)、Radeon/Ryzen(中)、Ryzen AI NPU(右)。 CDNA 面向数据中心 Instinct,RDNA 面向 Radeon/Ryzen,XDNA 面向 Ryzen AI NPU;前两者走 ROCm,XDNA 走 Ryzen AI Software。
7.1 关键规格
| 产品 | 架构 | LLVM target | 显存 | CU | 典型用途 |
|---|---|---|---|---|---|
| MI355X | CDNA4 | gfx950 | 288 GiB HBM3e | 256 | 大模型训练/推理 |
| MI350X | CDNA4 | gfx950 | 288 GiB HBM3e | 256 | 大模型训练/推理 |
| MI350P | CDNA4 | gfx950 | 144 GiB | 128 | 推理/训练 |
| MI325X | CDNA3 | gfx942 | 256 GiB HBM3e | 304 | 大模型推理/训练 |
| MI300X | CDNA3 | gfx942 | 192 GiB HBM3 | 304 | 大模型推理/训练 |
| MI300A | CDNA3 | gfx942 | 128 GiB HBM3 | 228 | APU,CPU+GPU 统一内存 |
| MI250X | CDNA2 | gfx90a | 128 GiB HBM2e | 220 | HPC/AI |
| MI210 | CDNA2 | gfx90a | 64 GiB HBM2e | 104 | HPC/AI |
| MI100 | CDNA1 | gfx908 | 32 GiB HBM2 | 120 | 早期 CDNA |
| RX 9070 XT | RDNA4 | gfx1201 | 16 GiB | 64 | 消费级/工作站 |
| RX 7900 XTX | RDNA3 | gfx1100 | 24 GiB | 96 | 消费级/工作站 |
| Ryzen AI Max+ 395 | RDNA3.5 | gfx1151 | 共享内存 | — | APU/iGPU |
数据来自 ROCm 10.1.0 官方 GPU specifications 与兼容性矩阵;不同 SKU 的显存/规格可能不同。
NPU(XDNA1/XDNA2)的规格和软件栈见第 8 节,不在 ROCm 兼容性矩阵里。
7.2 ROCm 10.1 的支持范围
官方兼容性矩阵列出的 LLVM target 包括:
- CDNA4:gfx950(MI355X、MI350X、MI350P)
- CDNA3:gfx942(MI325X、MI300X、MI300A)
- CDNA2:gfx90a(MI250X、MI250、MI210)
- CDNA1:gfx908(MI100)
- RDNA4:gfx1200、gfx1201(RX 9060/9070、Radeon AI PRO R9700)
- RDNA3:gfx1100、gfx1101、gfx1102、gfx1103
- RDNA3.5:gfx1150、gfx1151、gfx1152、gfx1153(Ryzen AI 300/400、Ryzen AI Max)
XDNA NPU 不在 ROCm 的支持列表里。Ryzen AI NPU 走的是第 8 节的独立软件栈(Ryzen AI Software、IRON、Triton-XDNA、FastFlow-SDK)。ROCm 支持的是 Ryzen APU 的 iGPU(RDNA3.5),不是 NPU。
8. NPU:Ryzen AI 的软件栈与路线
这一节讲的是 Ryzen AI PC 里的 XDNA NPU,不是 ROCm 的 GPU,也不是 Versal AI Edge 的 Vitis AI。
8.1 硬件:XDNA1 → XDNA2
AMD 的 NPU 不是 GPU 的“小核”,而是一块独立的空间数据流(spatial dataflow)AI Engine 阵列。每个 AI Engine tile 包含:
- VLIW/SIMD 向量处理器;
- RISC 标量处理器;
- 本地程序存储器和数据存储器;
- 专用互连与 DMA。
它不靠大缓存,而是靠 tile 本地内存和显式数据流,所以功耗低、时延确定。
| 代际 | NPU 架构 | 代表产品 | AIE 阵列 | NPU TOPS |
|---|---|---|---|---|
| XDNA1 | AIE2 | Ryzen 7040(Phoenix)、Ryzen 8040(Hawk Point) | 20 tiles,4×5 | 10 / 16 |
| XDNA2 | AIE2P | Ryzen AI 300(Strix Point)、Ryzen AI Max(Strix Halo)、Ryzen AI 400(Gorgon Point)、Kraken | 32 tiles,4×8 | 50 / 55 / 60 |
| XDNA2 嵌入式 | AIE2P | Ryzen AI Embedded P100 / X100、P121i | XDNA2 | 最高 50;P121i 30 |
XDNA2 相对 XDNA1 的关键变化:每 tile 2× MAC、1.6× on-chip memory、block floating point,AI Engine 性能约 3×,NPU 达到 50 TOPS。不同 SKU 的 TOPS 不同:Ryzen AI 300 / Max 为 50,Ryzen AI PRO 300 为 55,Ryzen AI 400 / PRO 400 最高 60。
数字来自 AMD XDNA 架构页、IRON ISCA 2025 教程、Ryzen AI 400 CES 2026 发布页和 Ryzen AI Embedded 发布页。XDNA2 的 32 tiles/4×8 指计算 tile;不同 SKU/工具显示的拓扑可能不同。
7040/8040 的 10/16 TOPS、PRO 300 的 55 TOPS、P121i 的 30 TOPS 分别来自 AMD 对应产品页/发布资料,链接见第 12 节。
flowchart TB P["Phoenix / Hawk Point<br/>Ryzen 7040 / 8040"] --> X1["XDNA1<br/>AIE2 · 20 tiles · 4×5<br/>10 / 16 TOPS"]
flowchart TB S["Strix Point / Strix Halo<br/>Ryzen AI 300 / Max"] --> X2["XDNA2<br/>AIE2P · 32 tiles · 4×8<br/>50 TOPS"] K["Kraken / Gorgon<br/>Ryzen AI 400"] --> X2 X2 --> H["最高 60 TOPS<br/>Ryzen AI 400 / PRO 400"]
flowchart TB E["P100 / X100"] --> EX["XDNA2<br/>最高 50 TOPS"] P2["P121i"] --> EX2["XDNA2<br/>30 TOPS"]
图 12:XDNA1(左)、XDNA2(中)、嵌入式 XDNA2(右)。 XDNA1 是 20 tiles/4×5、10/16 TOPS;XDNA2 是 32 tiles/4×8、50–60 TOPS;嵌入式 P100/X100 最高 50 TOPS,P121i 为 30 TOPS。
8.2 软件栈:两条栈、五层
Ryzen AI NPU 的软件栈可以按“部署栈”和“编程栈”来理解。两条栈共用驱动和硬件,但入口、编译器和运行时不同。
| 层 | 部署栈(ONNX) | 编程栈(IRON / Triton / FastFlow) |
|---|---|---|
| 应用/模型 | PyTorch / TensorFlow → ONNX;CNN/Transformer;LLM | IRON Python、Triton kernel、FastFlowLM |
| 框架/API | ONNX Runtime + Vitis AI EP;OGA/llama.cpp;Lemonade | IRON API、Triton、FastFlowLM |
| 量化/编译 | Quark(INT8/BF16);Vitis AI EP 图切分/编译 | IRON ops、MLIR-AIE / MLIR-AIR;Peano → PDI/xclbin |
| 运行时 | XRT + XRT SHIM(Linux)/ Windows NPU 驱动栈 | HRX(FastFlow-SDK)或 XRT(IRON/Triton,Linux) |
| 驱动/硬件 | amdxdna(Linux)/ Windows NPU 驱动 → XDNA1/XDNA2 | 同上 |
OGA= ONNX Runtime GenAI;Ryzen AI 文档里llama.cpp只支持 iGPU,NPU-only 的 LLM 路线用 OGA/FastFlowLM。Lemonade是多厂商的本地 LLM SDK,FastFlowLM 已集成到 Lemonade。
flowchart TB A1["应用/模型<br/>ONNX:PyTorch / TF → ONNX<br/>IRON:Python / Triton / FastFlowLM"] --> A2["框架/API<br/>ONNX Runtime + Vitis AI EP<br/>IRON API / Triton / FastFlowLM"]
flowchart TB B1["量化/编译<br/>Quark / Vitis AI EP<br/>MLIR-AIE / MLIR-AIR + Peano"] --> B2["编译产物<br/>量化模型 / 编译图<br/>PDI / xclbin"]
flowchart TB C1["运行时<br/>XRT + XRT SHIM<br/>HRX"] --> C2["驱动/硬件<br/>amdxdna / Windows NPU 驱动<br/>XDNA1 / XDNA2"]
图 13:NPU 软件栈的三段压缩视图。 左:应用/模型 + 框架/API;中:量化/编译 + 产物;右:运行时 + 驱动/硬件。完整五层见上表。
8.3 四条路线
把上面的栈拆开,实际有四个入口:
- ONNX / Vitis AI EP:最像“部署”的路线,适合 CNN/Transformer;
- IRON:close-to-metal,适合写自定义算子/kernel;
- Triton-XDNA:把 Triton kernel 编译到 NPU,适合 Triton 用户/研究;
- FastFlowLM / FastFlow-SDK:NPU-first LLM 运行时,适合开箱即用跑 LLM/VLM/ASR。
另外,Vitis-AI / VAIP / XIR / VART 是 Versal AI Edge 的独立路线,不是 Ryzen AI NPU,见 8.6。
flowchart TB A1["PyTorch / TF"] --> A2["ONNX"] --> A3["Quark"] --> A4["ORT + VAI-EP"] --> A5["XRT / SHIM"] --> A6["NPU"]
flowchart TB B1["IRON"] --> B2["MLIR-AIE"] --> B3["Peano → PDI"] --> B4["HRX / XRT"] --> B5["NPU"]
flowchart TB C1["Triton"] --> C2["triton-shared"] --> C3["MLIR Transform"] --> C4["MLIR-AIR/AIE"] --> C5["XRT binary"] --> C6["NPU"]
flowchart TB D1["Lemonade / GAIA"] --> D2["FastFlowLM"] --> D3["IRON ops"] --> D4["HRX"] --> D5["NPU"]
图 14:四条 NPU 路线(从左到右)。 A:ONNX/Vitis AI EP;B:IRON;C:Triton-XDNA;D:FastFlowLM/FastFlow-SDK。
| 路线 | 入口 | 编译/运行时 | 适合 | 成熟度 |
|---|---|---|---|---|
| A. ONNX / Vitis AI EP | PyTorch/TF → ONNX | Quark → ONNX Runtime + Vitis AI EP → XRT/XRT SHIM | CNN/Transformer 部署,改动最小 | 官方主推;Windows/Linux |
| B. IRON | Python IRON | MLIR-AIE + Peano → PDI → HRX/XRT | 自定义算子/kernel、数据流 | 官方开源;XDNA1/2 |
| C. Triton-XDNA | Triton kernel | triton-shared → MLIR Transform → MLIR-AIR/AIE → XRT binary | Triton 用户/研究 | 实验性;AIE2/AIE2P |
| D. FastFlowLM | Lemonade/GAIA/CLI | FastFlowLM → IRON ops → HRX | 本地 LLM/VLM/ASR,开箱即用 | 早期访问;XDNA2 |
- A 是 Ryzen AI Software 的主推部署路线:Vitis AI EP 会切分图,把能跑 NPU 的部分交给 NPU;Windows 上不支持的算子可回退 CPU/iGPU,Linux 当前是 NPU-only。Linux 的运行时栈是 XRT + XRT SHIM,Windows 走 NPU 驱动栈。
- B 的 IRON 本身不提供 kernel,kernel 来自 MLIR-AIE 的
aie_kernels;@iron.jit把 Python 定义的工作负载编译成 NPU 程序,ObjectFifo 负责声明式数据搬运。 - C 目前是实验项目,官方 README 说 matmul 可达到手写 NPU 实现的性能,90% 测试配置达到 baseline 的 90% 以上;支持 matmul、elementwise、softmax、layer norm。
- D 的 FastFlow-SDK 是早期访问,包含 FastFlowLM + HRX + iron-tools + iron-ops,目标 XDNA2;FastFlowLM 支持 256k context,已被 Lemonade 集成。
8.4 NPU 仓库地图
NPU 相关仓库不在一个组织里,而是分散在 amd、ROCm、AMDResearch,以及上游 Xilinx。
核心 NPU 仓库
| 组织 | 仓库 | 角色 | 一句话 |
|---|---|---|---|
amd | RyzenAI-SW | 示例/教程 | Ryzen AI Software 的示例与教程子集;完整运行时和 Vitis AI EP 在官方安装包中。 |
amd | xdna-driver | Linux 驱动 | amdxdna.ko + XRT SHIM + 固件打包;需要 Linux 6.10+。 |
amd | IRON | 编程 API | close-to-metal Python API,基于 MLIR-AIE 和 Peano。 |
amd | Triton-XDNA | 编译器后端 | Triton → MLIR-AIR/AIE 的实验后端,产出 XRT binary。 |
amd | Quark | 量化 | PyTorch/ONNX 量化工具链,支持 INT8/BF16 等低精度。 |
amd | gaia | Agent/应用 | 本地 AI agent;NPU 后端走 FastFlowLM。 |
amd | onnxruntime | 框架 fork | AMD 的 ONNX Runtime 分支;Ryzen AI Software 通过 ONNX Runtime + Vitis AI EP 部署模型。 |
ROCm | FastFlowLM | LLM 运行时 | NPU-first LLM/VLM/ASR/Embedding 推理,支持 256k context。 |
ROCm | FastFlow-SDK | SDK | FastFlowLM + HRX + iron-tools + iron-ops;XDNA2 早期访问。 |
ROCm | hrx-system | 运行时 | HRX:HIP Runtime Extended,加载/调度 NPU 程序。 |
AMDResearch | Riallto | 教学 | Ryzen AI NPU 的教学与实验环境。 |
AMDResearch | NPUEval | 评测 | 面向 AIE kernel 代码生成的 LLM 评测集。 |
AMDResearch | Ryzers | 环境 | Ryzen AI Dockerfiles,覆盖 LLM/机器人/视觉/NPU。 |
相邻/上游仓库
| 组织 | 仓库 | 角色 | 一句话 |
|---|---|---|---|
amd | Vitis-AI | Versal | Versal AI Edge 的推理栈,不是 Ryzen AI NPU。 |
amd | vaip / xir / vart | Versal | Vitis AI 的运行时、图 IR 和 runtime 组件。 |
amd | LIRA | 归档 | Ryzen AI 300 NPU 上的 ASR 示例,已归档。 |
ROCm | hrx-solutions / hrx-loom-library | 示例/工具 | HRX 方案与 kernel 语料。 |
Xilinx | mlir-aie | 上游编译器 | MLIR-AIE dialect / AIE kernels;iron-tools 的前身。 |
Xilinx | mlir-air | 上游编译器 | MLIR-AIR;Triton-XDNA 的 lowering 目标。 |
RyzenAI-SW是示例/教程仓库,不是完整安装包;完整 Ryzen AI Software 需要从 AMD 下载页安装。ROCm/FastFlowLM、FastFlow-SDK、HRX是 NPU 项目,只是托管在 ROCm 组织,不代表 ROCm 支持 XDNA NPU。Xilinx/mlir-aie、mlir-air不在本文 1035 个仓库的统计内,但它们是 IRON/Triton-XDNA 的上游。
8.5 与 ROCm 的边界
| 维度 | ROCm GPU | Ryzen AI NPU |
|---|---|---|
| 硬件 | CDNA / RDNA | XDNA1 / XDNA2 |
| 计算单元 | CU | AIE tiles |
| 编译 target | gfx908 / gfx90a / gfx942 / gfx950 / gfx110x / gfx120x | aie2 / aie2p |
| 编程 | HIP / OpenCL / OpenMP / Triton | ONNX + Vitis AI EP / IRON / Triton-XDNA |
| 运行时 | CLR / ROCr / PAL | XRT / XRT SHIM / HRX |
| 驱动 | amdgpu + amdkfd | amdxdna(Linux)/ Windows NPU 驱动 |
| 发行 | ROCm Core SDK / TheRock | Ryzen AI Software / FastFlow-SDK |
| 典型硬件 | MI300X / MI355X / RX 7900 | Ryzen AI 300 / 400 / Max |
| 组织 | ROCm | amd / ROCm / AMDResearch |
一句话:ROCm 管 GPU,Ryzen AI Software 管 NPU。Ryzen AI APU 上两者都有:RDNA3.5 iGPU 走 ROCm,XDNA2 NPU 走 Ryzen AI Software / IRON / FastFlow。
8.6 不要混淆:Vitis AI EP、Vitis AI 和 Versal
- Vitis AI EP:ONNX Runtime 的 Execution Provider,跑在 Ryzen AI NPU 上;Ryzen AI Software 用它做图切分/编译。
- amd/Vitis-AI、vaip、xir、vart:Vitis AI 的 Versal AI Edge 推理栈,目标 FPGA/自适应 SoC,不是 Ryzen AI NPU。典型路径是 ONNX → XIR → VAIP → VART → Versal。
- XDNA NPU vs RDNA iGPU:Ryzen AI APU 同时有两者;ROCm 支持 iGPU,Ryzen AI Software/IRON 支持 NPU。
8.7 一句话总结
Ryzen AI NPU 不是 ROCm 的附属品;它有自己的驱动(amdxdna)、运行时(XRT / HRX)、编译器(Vitis AI EP / MLIR-AIE + Peano / MLIR-AIR)和 LLM 运行时(FastFlowLM)。按目标选:部署走 ONNX/Vitis AI EP,写 kernel 走 IRON,Triton 用户走 Triton-XDNA,跑 LLM 走 FastFlowLM。
9. 仓库重组:为什么旧仓库都写着 “Moved”
flowchart TB L["旧版:单仓库 / 单组件发布<br/>ROCm/ROCm、ROCm/MIOpen、ROCm/rocBLAS ..."] --> S["超级仓库<br/>rocm-libraries / rocm-systems"] L -.-> A["归档 / 只读镜像"]
flowchart TB S["超级仓库"] --> T["TheRock<br/>统一构建、测试、发布"] --> P["ROCm Core SDK 10.x<br/>DEB/RPM / Tarball / Wheel / Runfile"] --> E["ROCm Extras / Standalone ONNX"]
图 15:旧版单仓库(左)与超级仓库/TheRock 发布(右)。 从“每个组件一个仓库”转向“超级仓库 + TheRock 统一发布”。
9.1 三个新入口
TheRock:
- 官方定义:The HIP Environment and ROCm Kit;
- 角色:ROCm 的顶层构建与发布系统;
- 自 ROCm 7.14 起,ROCm 通过 TheRock 构建和发布;
- 提供 nightly ROCm、PyTorch、JAX 构建;
- 支持 Linux 多发行版和原生 Windows;
- 构建从源码需要约 200 GB 磁盘空间。
rocm-libraries:
- 数学/算子库超级仓库;
projects/下包括:composablekernel、hipblas、hipblas-common、hipblaslt、hipcub、hipdnn、hipfft、hiprand、hipsolver、hipsparse、hipsparselt、hiptensor、miopen、rocblas、rocfft、rocprim、rocrand、rocsolver、rocsparse、rocthrust、rocwmma、rpp;shared/下包括:rocroller、tensile、mxdatagenerator、origami。
rocm-systems:
- 系统项目超级仓库;
projects/下包括:amdsmi、aqlprofile、clr、hip、hipfile、hipother、hip-tests、rccl、rdc、rocdbgapi、rocdecode、rocjpeg、rocm-core、rocminfo、rocmsmilib、rocprofiler、rocprofiler-compute、rocprofiler-register、rocprofiler-sdk、rocprofiler-systems、rocr-debug-agent、rocrruntime、rocshmem、roctracer。
9.2 包结构变化
ROCm 10.x 的包名从 rocm-* / roc* / hip* 统一改为 amdrocm-*:
| 新包 | 内容 |
|---|---|
amdrocm-llvm | amdclang++、hipcc、flang |
amdrocm-runtime | HIP、ROCR、runtime compilation、COMGR |
amdrocm-blas | rocBLAS、hipBLAS、hipBLASLt、hipSPARSELt |
amdrocm-dnn | hipDNN、MIOpen |
amdrocm-fft | rocFFT、hipFFT、hipFFTW |
amdrocm-sparse | rocSPARSE、hipSPARSE |
amdrocm-solver | rocSOLVER、hipSOLVER、rocALUTION |
amdrocm-profiler | rocprofiler-systems、rocprofiler-compute、rocprofiler-sdk、roctracer |
amdrocm-debugger | rocgdb、ROCdbgapi、ROCR Debug Agent |
amdrocm-opencl | OpenCL runtime 和 ICD loader |
amdrocm-rccl | RCCL |
amdrocm-rocshmem | rocSHMEM |
amdrocm-hipfile | hipFile |
amdrocm-ck | Composable Kernel |
安装目录也从 /opt/rocm-7.2/ 改为 /opt/rocm/core-10.1,并通过 update-alternatives 提供 /opt/rocm 兼容符号链接。
9.3 对开发者的实际影响
- 不要再从旧仓库起步:新项目看
rocm-libraries/rocm-systems/TheRock; - 旧博客/脚本可能失效:包名、安装路径、组件归属都变了;
- 用 TheRock 构建更可控:适合需要最新功能或自定义组合的开发者;
- 发行版安装优先用
amdrocm-*元包:避免单独安装组件导致依赖冲突。
10. 怎么选:按目标找路线
| 你的目标 | 首选路线 | 关键仓库 |
|---|---|---|
| 跑 PyTorch 训练/推理 | PyTorch ROCm wheels 或 TheRock | pytorch、TheRock、rocm-libraries |
| 部署 LLM 推理 | vLLM / SGLang / ATOM + AITER | vllm、sglang、ATOM、aiter |
| 做 LLM 训练 | Primus + Megatron/TorchTitan/MaxText | Primus、Primus-Turbo、Megatron-LM |
| 写高性能 kernel | HIP 或 Triton/Gluon/FlyDSL | hip、triton、aotriton、FlyDSL |
| 从 CUDA 迁移 | HIPIFY → HIP | HIPIFY、rocm-libraries(hipBLAS/hipCUB) |
| HPC / Fortran | AOMP/hipfort + rocm-libraries | aomp、hipfort、rocm-libraries |
| 多机多卡 | RCCL / rocSHMEM / MORI | rocm-systems(RCCL)、rocSHMEM、mori |
| 图推理 / ONNX | MIGraphX / ONNX Runtime | AMDMIGraphX、onnxruntime、hip-ep |
| 图形 / 光追 | GPUOpen / AMDVLK / HIP RT | AMDVLK、FidelityFX-SDK、HIPRT |
| Ryzen AI NPU | ONNX/Vitis AI EP、IRON、Triton-XDNA、FastFlowLM | RyzenAI-SW、xdna-driver、IRON、FastFlow-SDK |
标
rocm-libraries/rocm-systems的条目表示组件已迁入超级仓库,旧单仓库只作参考。
10.1 三条最常用的组合
组合 A:LLM 推理
模型 → vLLM/SGLang/ATOM → AITER → CK/Triton/ASM/hipBLASLt → HIP → ROCr → MI300X/MI355X
组合 B:LLM 训练
Primus → Megatron/TorchTitan/MaxText → PyTorch/JAX → RCCL/MORI → HIP → MI300X/MI355X 集群
组合 C:HPC
GROMACS/OpenFOAM → MPI + OpenMP/Fortran → AOMP/hipfort → rocBLAS/rocFFT/rocSPARSE → RCCL → HIP → MI250X/MI300A
11. 附录:本文统计范围内的全部仓库清单
以下清单按组织分组,按 Star 数从高到低排列。
状态列含义:活跃 = GitHub 上非 fork、非归档;fork = 上游项目的 AMD 分支;归档 = 只读历史仓库。简介保留 GitHub 官方英文原文。
活跃只表示 GitHub 状态,不代表项目仍推荐使用;简介含[DEPRECATED] Moved to ...的仓库,请以rocm-libraries/rocm-systems为准。
fork列含已归档的 fork;附录状态列只标一个主状态,所以 9 个主组织在附录中的 fork 标签数为 349,小于正文的 fork 总数 366。说明:
ROCm/legacy-rocm-build是旧ROCm/ROCm主页仓库的归档名;今天 ROCm 的入口是ROCm/TheRock。表中的—表示 GitHub 仓库未填写简介,这类仓库多为内部工具、第三方库镜像或历史占位仓库。
11.1 九个主组织
ROCm:414 个仓库(ROCm 核心 GPU 计算栈)
活跃非 fork 225 · fork 124 · 归档 81 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| legacy-rocm-build | 6,814 | 归档 | AMD ROCm™ Software - GitHub Home |
| hip | 4,421 | 活跃 | HIP: C++ Heterogeneous-Compute Interface for Portability |
| FastFlowLM | 1,995 | 活跃 | Run LLMs on AMD Ryzen™ AI NPUs in minutes; purpose-built and deeply optimized for the AMD NPUs. |
| TheRock | 1,408 | 活跃 | The HIP Environment and ROCm Kit - A lightweight open source build system for HIP and ROCm |
| MIOpen | 1,194 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| HIPIFY | 734 | 活跃 | HIPIFY: Convert CUDA to Portable C++ Code |
| tensorflow-upstream | 707 | fork | TensorFlow ROCm port |
| aiter | 598 | 活跃 | AI Tensor Engine for ROCm |
| composable_kernel | 548 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo. NOTE: develop branch is maintained as a read-only mirror |
| rocm-systems | 543 | 活跃 | super repo for rocm systems projects |
| ROCm-docker | 521 | 活跃 | Dockerfiles for the various software layers defined in the ROCm software platform |
| amdgpu | 470 | 活跃 | This repository is a fork of the upstream AMDGPU Driver, used by the ROCm project. To submit a patch, please follow the guidelines at https://docs.kernel.org/process/submitting-patches.html#submittingpatches . For rep… |
| rocm-libraries | 456 | 活跃 | super repo for rocm libraries |
| hcc | 444 | 归档 | HCC is an Open Source, Optimizing C++ Compiler for Heterogeneous Compute currently for the ROCm GPU Computing Platform |
| rccl | 420 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocBLAS | 403 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| k8s-device-plugin | 399 | 活跃 | Kubernetes (k8s) device plugin to enable registration of AMD GPU to a container cluster |
| omnitrace | 358 | 归档 | Omnitrace: Application Profiling, Tracing, and Analysis |
| AMDMIGraphX | 337 | 活跃 | AMD's graph optimization engine. |
| rocm-examples | 328 | 活跃 | A collection of examples for the ROCm software stack |
| FlyDSL | 304 | 活跃 | FlyDSL is the Python front‑end of the project: a Flexible Layout Python DSL for expressing tiling, partitioning, data movement, and kernel structure at a high level. |
| ROCR-Runtime | 278 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| Tensile | 262 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| pytorch | 261 | fork | Tensors and Dynamic neural networks in Python with strong GPU acceleration. To use PyTorch with latest ROCm releases, see: https://rocm.docs.amd.com/en/latest/rocm-for-ai/pytorch.html |
| aomp | 246 | 活跃 | AOMP is an open source Clang/LLVM based compiler with added support for the OpenMP® API on Radeon™ GPUs. Use this repository for releases, issues, documentation, packaging, and examples. |
| llvm-project | 244 | fork | This is the AMD-maintained fork of the LLVM git repository. This repository accepts pull requests and issues related to AMD fork-specific topics (amd/*). For all other issues/PRs, please submit upstream at https://git… |
| flash-attention | 241 | fork | Fast and memory-efficient exact attention |
| MIVisionX | 218 | 活跃 | AMD MIVisionX is a computer vision toolkit built around a highly optimized, conformant open-source implementation of the Khronos OpenVX™ 1.3.2 specification. As of the 4.0.0 release, MIVisionX ships three components: … |
| HIP-Examples | 214 | 归档 | Examples for HIP |
| iris | 203 | 活跃 | AMD RAD's multi-GPU Triton-based framework for seamless multi-GPU programming |
| rocFFT | 203 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| ATOM | 201 | 活跃 | AiTer Optimized Model |
| rocMLIR | 195 | 活跃 | — |
| mori | 188 | 活跃 | Modular RDMA Interface |
| clr | 185 | 活跃 | — |
| ROC-smi | 183 | 活跃 | ROC System Management Interface |
| rocPRIM | 177 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| ROCm-OpenCL-Runtime | 169 | 归档 | ROCm OpenOpenCL Runtime |
| gpufort | 168 | 归档 | GPUFORT: S2S translation tool for CUDA Fortran and Fortran+X in the spirit of hipify |
| rocprofiler-compute | 166 | 归档 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocm_smi_lib | 155 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| hipBLAS | 154 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| rocprofiler | 153 | 归档 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| triton | 146 | fork | Development repository for the Triton language and compiler |
| amd_matrix_instruction_calculator | 145 | 活跃 | A tool for generating information about the matrix multiplication instructions in AMD Radeon™ and AMD Instinct™ accelerators |
| rocSHMEM | 145 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocWMMA | 143 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| librocdxg | 142 | 活跃 | repo for AMD ROCDXG project |
| amdsmi | 136 | 活跃 | AMD SMI |
| rocSPARSE | 136 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| spur | 136 | 活跃 | spur your compute |
| HIP-CPU | 133 | 归档 | An implementation of HIP that works on CPUs, across OSes. |
| rocRAND | 130 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| hipCaffe | 125 | 归档 | (Deprecated) hipCaffe: the HIP port of Caffe |
| vllm | 124 | fork | A high-throughput and memory-efficient inference and serving engine for LLMs |
| rocThrust | 123 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| rocSOLVER | 118 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| hipBLASLt | 116 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| LLVM-AMDGPU-Assembler-Extra | 113 | 归档 | LLVM AMDGPU Assembler Helper Tools |
| ROCmValidationSuite | 110 | 活跃 | A system validation and diagnostics tool for monitoring, stress testing, detecting, and troubleshooting issues impacting AMD GPUs in high-performance computing environments |
| gpu-operator | 108 | 活跃 | — |
| gpuaidev | 106 | 活跃 | Repository to host ROCm Developer Hub Notebook Tutorials |
| rocALUTION | 104 | 活跃 | Next generation library for iterative sparse solvers for ROCm platform |
| aotriton | 102 | 活跃 | Ahead of Time (AOT) Triton Math Library |
| hipfort | 97 | 活跃 | Fortran interfaces for ROCm libraries |
| ROCm-Device-Libs | 97 | 归档 | ROCm Device Libraries |
| rccl-tests | 94 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| ROCT-Thunk-Interface | 93 | 活跃 | ROCm's Thunk Interface |
| rocm_bandwidth_test | 90 | 活跃 | Bandwidth test for ROCm |
| rocm-blogs | 85 | 活跃 | — |
| hipCUB | 83 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| roctracer | 83 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| ROCgdb | 80 | 活跃 | This is ROCgdb, the ROCm source-level debugger for Linux, based on GDB, the GNU source-level debugger. |
| TransferBench | 80 | 活跃 | TransferBench is a utility capable of benchmarking simultaneous copies between user-specified devices (CPUs/GPUs) |
| rocprof-compute-viewer | 79 | 活跃 | — |
| rocm-cmake | 78 | 活跃 | CMake modules used within the ROCm libraries |
| TransformerEngine | 77 | 活跃 | — |
| rocHPL | 75 | 活跃 | High Performance Linpack for Next-Generation AMD HPC Accelerators |
| rpp | 74 | 活跃 | AMD ROCm Performance Primitives (RPP) library is a comprehensive high-performance computer vision library for AMD processors with HIP/OpenCL/CPU back-ends. |
| FastFlowLM-internal | 72 | 归档 | FLM app mirror |
| rocm-xio | 72 | 活跃 | A ROCm library for GPU-Initiated IO. This provides support for initiating IO from a ROCm-capable GPU against a range of targets including RDMA NICs, NVMe SSDs and AMD GPU DMA engines (SDMAs). |
| bitsandbytes | 71 | fork | 8-bit CUDA functions for PyTorch |
| tritonBLAS | 69 | 活跃 | A lightweight triton-based General Matrix Multiplication (GEMM) library. |
| atmi | 68 | 归档 | Asynchronous Task and Memory Interface, or ATMI, is a runtime framework and programming model for heterogeneous CPU-GPU systems. It provides a consistent, declarative API to create task graphs on CPUs and GPUs (integr… |
| device-metrics-exporter | 68 | 活跃 | Device Metrics Exporter exports metrics from AMD devices (GPUs) to collectors like Prometheus. |
| hipSPARSE | 68 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| hipFFT | 65 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| hrx-system | 61 | 活跃 | HRX: Hip Runtime Extended |
| MIOpenGEMM | 61 | 归档 | MIOpenGEMM is now deprecated |
| hipDNN | 57 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| gfx950-gluon-tutorials | 55 | 活跃 | A practical guide to high-performance gluon kernel development on AMD GFX9 GPUs. |
| hipThreads | 51 | 活跃 | — |
| hipTensor | 50 | 活跃 | AMD’s C++ library for accelerating tensor primitives |
| rocm-cli | 49 | 活跃 | ROCm CLI is a command-line tool for setting up and running local AI on AMD GPUs, with a full-screen TUI dashboard for GPU telemetry, model serving, and chat. |
| container-toolkit | 47 | 活跃 | Offers tools that streamline the use of AMD GPUs with containers. |
| FLM_Q4NX_Converter | 46 | 活跃 | A converter for transferring gguf Q4_0, Q4_1 to FLM Q4NX |
| ROCm-CompilerSupport | 45 | 归档 | The compiler support repository provides various Lightning Compiler related services. |
| MAD | 44 | 活跃 | MAD (Model Automation and Dashboarding) |
| Megatron-LM | 44 | fork | Ongoing research training transformer models at scale |
| ROCclr | 43 | 归档 | — |
| rocminfo | 43 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| hip-python | 42 | 活跃 | HIP Python Low-level Bindings |
| hip-tests | 41 | 活跃 | — |
| HIPCC | 41 | 归档 | HIPCC: HIP compiler driver |
| rocHPCG | 41 | 活跃 | HPCG benchmark based on ROCm platform |
| ROCm-ComputeABI-Doc | 41 | 归档 | ROCm - AMDGPU Compute Application Binary Interface |
| rocm-install-on-linux | 40 | 活跃 | — |
| rocm-install-on-windows | 38 | 活跃 | — |
| Experimental_ROC | 36 | 归档 | Experimental and Intriguing Tools for ROCm |
| MISA | 36 | 活跃 | Machine Intelligence Shader Autogen. AMDGPU ML shader code generator. (previously iGEMMgen) |
| rocDecode | 35 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocr_debug_agent | 35 | 活跃 | The ROCdebug-agent is a library that can be loaded by ROCm Platform Runtime to provide some debugging functionality. |
| hipamd | 34 | 归档 | — |
| xformers | 34 | fork | Hackable and optimized Transformers building blocks, supporting a composable construction. |
| roc-optiq | 33 | 活跃 | A visualizer for the ROCm Profiler Tools |
| half | 32 | 活跃 | — |
| jax | 32 | fork | Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more |
| OpenFOAM_HMM | 32 | 活跃 | Refactoring OpenFOAM with OpenMP target offloading and use of HMM to offload work onto GPUs |
| ROCm.github.io | 32 | 归档 | ROCm Website |
| rocprofiler-sdk | 31 | 归档 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| k8s-gpu-dra-driver | 30 | 活跃 | K8s DRA driver for AMD GPUs |
| libhipcxx | 30 | 活跃 | The C++ Standard Library for your entire system. |
| rdc | 30 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocmProfileData | 30 | 活跃 | — |
| rocprofiler-systems | 30 | 归档 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| mxnet | 29 | fork | Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more |
| rocm-ernic | 29 | 活跃 | A vfio-user based emulated AMD Pensando RDMA NIC. |
| rocm-jax | 29 | 活跃 | — |
| hipify_torch | 28 | 活跃 | — |
| Thrust | 28 | 归档 | HIP back-end for Thrust that has been replaced by rocThrust |
| clang-ocl | 27 | 归档 | OpenCL compilation with clang compiler. |
| hipRAND | 27 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| hipSOLVER | 27 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| omnistat | 27 | 活跃 | Scale-out system monitoring |
| hipFile | 26 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocAL | 26 | 活跃 | The AMD rocAL is designed to efficiently decode and process images and videos from a variety of storage formats and modify them through a processing graph programmable by the user. |
| pyrsmi | 25 | 活跃 | python package of rocm-smi-lib |
| apex | 24 | fork | A PyTorch Extension: Tools for easy mixed precision and distributed training in Pytorch |
| ROCm-OpenCL-Driver | 24 | 归档 | ROCm OpenCL Compiler Tool Driver |
| clARMOR | 23 | 归档 | OpenCL tool to detect buffer overflows in GPU kernels |
| pytorch-micro-benchmarking | 23 | 活跃 | — |
| numba-hip | 22 | 归档 | HIP backend patch for Numba, the NumPy aware dynamic Python compiler using LLVM. |
| StatusMonitor | 22 | 归档 | OpenCL multiGPU sample monitoring system health |
| torch_migraphx | 22 | 活跃 | Torch-MIGraphX integrates AMD's graph inference engine with the PyTorch ecosystem. |
| fleet-chiplet-megakernel | 20 | 活跃 | — |
| roc-stdpar | 20 | 归档 | — |
| ROCdbgapi | 20 | 活跃 | The AMD Debugger API is a library that provides all the support necessary for a debugger and other tools to perform low level control of the execution and inspection of execution state of AMD's commercially available … |
| nnvm-rocm | 19 | 归档 | NNVM for ROCm Examples |
| rocm-aic | 19 | 活跃 | ROCm AMD Infinity Context - A distributed, shared, low-latency KV Cache solution for AI Inference.. |
| rocprof-trace-decoder | 19 | 活跃 | — |
| aws-ofi-rccl | 18 | 活跃 | — |
| cvs | 18 | 活跃 | — |
| hip-ep | 18 | 活跃 | ROCm Execution Provider for ONNX Runtime on AMD GPUs. Compiles ONNX models through MLIR to LLVM bitcode, JIT-loads them at session creation, and dispatches to hipDNN, MIOpen, hipBLASLt, and custom HIP kernels |
| tensorcast | 18 | 活跃 | — |
| rocBLAS-Examples | 17 | 归档 | Examples illustrating usage of the rocBLAS library |
| rocm-docs-core | 17 | 活跃 | ROCm Documentation Python package for ReadTheDocs build standardization |
| DeepEP | 15 | 活跃 | — |
| hipCOMP-core | 15 | 活跃 | hipCOMP is a library for fast lossless compression/decompression on the GPU. This repository contains the algorithms. |
| hipObject | 15 | 活跃 | A ROCm library for RDMA-based object storage with Direct-to-VRAM DMA support on AMD GPUs. |
| RIXL | 15 | 活跃 | DEPRECATED REPOSITORY. ROCm Inference Transfer Library (RIXL) is a port of the NIXL library for AMD GPUs. See README_rocm.md for AMD specific instructions for compiling RIXL |
| AMD_Hackathon | 14 | 活跃 | AMD_Robotics_Hackathon_2025 |
| FastFlow-SDK | 14 | 活跃 | FastFlow-SDK is a software stack for running and developing applications on AMD XDNA Neural Processing Engines (NPUs). |
| DirectGMA_CL | 13 | 归档 | Simple example showing how to use DGMA in OpenCL |
| HCC-Example-Application | 13 | 归档 | HCC Sample Applications |
| hipSPARSELt | 13 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| AITemplate | 12 | fork | AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference. |
| hcc-clang-upgrade | 12 | 归档 | stage the upgrade of hcc-clang to clang ToT |
| hipCollections | 12 | 活跃 | Header-only library of GPU-accelerated, concurrent data structures. |
| llvm | 12 | 归档 / fork | — |
| xla | 12 | fork | A machine learning compiler for GPUs, CPUs, and ML accelerators |
| aorta | 11 | 活跃 | — |
| gpu-cluster-networking | 11 | 活跃 | Cluster networking documentation for AMD Instinct accelerators |
| hipBench | 11 | 活跃 | HIP Kernel Benchmarking Library |
| rocm-amdgpu-bench | 11 | 归档 | — |
| hipDF | 10 | 归档 | hipDF - GPU DataFrame Library |
| madengine | 10 | 活跃 | madengine is a streamlined CLI tool for running and benchmarking AI models on ROCm GPUs, offering a production‑ready workflow for local single node or remote multi node execution with integrated performance monitoring. |
| onnxruntime | 10 | fork | ONNX Runtime: cross-platform, high performance scoring engine for ML models |
| rocm-core | 10 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| fio | 9 | fork | Flexible I/O Tester |
| Paddle | 9 | fork | PArallel Distributed Deep LEarning |
| rocmlirTriton | 9 | 活跃 | MLIR kernel generator for AMD GPUs (GEMM, conv, attention, fused), built on a Triton compilation backend. |
| criu | 8 | fork | Checkpoint/Restore tool |
| formal-models | 8 | 活跃 | — |
| hipother | 8 | 活跃 | — |
| instinct-docs | 8 | 活跃 | AMD Instinct Data Center GPU docs |
| MITuna | 8 | 活跃 | — |
| nccl-rccl-parser | 8 | fork | Tool to run rccl-tests/nccl-tests based on from an application |
| network-operator | 8 | 活跃 | Network Operator simplifies usage of AMD AINICs in a Kubernetes Environment |
| OSU_Microbenchmarks | 8 | 归档 | ROCm - UCX enabled OSU_Benchmarks |
| rbuild | 8 | 活跃 | Rocm build tool |
| rocJPEG | 8 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| rocm-kpack | 8 | 归档 | Prototype for a kernel packing library |
| rocPyDecode | 8 | 活跃 | rocPyDecode is a set of Python bindings to rocDecode C++ library which provides full HW acceleration for video decoding on AMD GPUs. |
| spur-cloud | 8 | 活跃 | GPU as a Service platform built on Spur |
| amdgpu-docs | 7 | 活跃 | AMD GPU Driver Documentation |
| axis | 7 | 活跃 | AXIS: Agent eXecution Isolation Substrate — secure sandbox runtime for AI agents |
| cupy | 7 | fork | A NumPy-compatible array library accelerated by CUDA |
| DeepSpeed | 7 | fork | DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective. |
| Gromacs | 7 | 活跃 | ROCm's implementation of Gromacs |
| rockrel | 7 | 活跃 | — |
| rtg_tracer | 7 | 活跃 | — |
| timemory | 7 | 活跃 | Modular C++ Toolkit for Performance Analysis and Logging. Profiling API and Tools for C, C++, CUDA, Fortran, and Python. The C++ template API is essentially a framework to creating tools: it is designed to provide a u… |
| amd-nhc | 6 | 活跃 | — |
| AMDiffusionBenchmark | 6 | 活跃 | Benchmarking training throughput of contemporary diffusion models on AMD's GPU hardware. |
| CTranslate2 | 6 | fork | Fast inference engine for Transformer models |
| embers | 6 | 活跃 | — |
| flang | 6 | 活跃 | Mirror of flang repo: The source repo is https://github.com/flang-compiler/flang . Once a day the master branch is updated from the upstream source repo and then locked. AOMP or ROCm developers may commit or create P… |
| hipBLAS-common | 6 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| jitify | 6 | 活跃 | A single-header C++ library for simplifying the use of HIP Runtime Compilation (HIPRTC). |
| repo-digest | 6 | 活跃 | Automated commit digests using Claude Code agents |
| rocHPL-MxP | 6 | 活跃 | — |
| rocm-llvm-python | 6 | 归档 | Low-level Cython and Python bindings to the (ROCm) LLVM and AMD COMGR C API. Also ships the official LLVM Clang bindings. |
| rocm-spack-pkgs | 6 | 归档 | Repository to host spack recipes for ROCm |
| transformers | 6 | fork | 🤗Transformers: State-of-the-art Natural Language Processing for Pytorch and TensorFlow 2.0. |
| ucx | 6 | fork | Unified Communication Library X |
| clang | 5 | 归档 / fork | — |
| DeepBench | 5 | fork | Benchmarking Deep Learning operations on different hardware |
| gpu-agent | 5 | 活跃 | GPU Agent provides programmable APIs to configure & monitor AMD Instinct GPUs |
| hip-testsuite | 5 | 活跃 | — |
| hipMM | 5 | 归档 | HIP Memory Manager (ROCm-DS) |
| nvtop | 5 | fork | GPUs process monitoring for AMD, Intel and NVIDIA |
| rccl-rdma-sharp-plugins | 5 | 活跃 | — |
| rdma-perftest | 5 | 活跃 | — |
| rocm-docs | 5 | 活跃 | Source for the ROCm documentation website (rocm.docs.amd.com) |
| rocRoller | 5 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| rocsift | 5 | 活跃 | C99 AMDGPU Debug Library wrapping KFD sysfs, debugfs, and more. |
| SPIRV-LLVM-Translator | 5 | 活跃 | This is the AMD-maintained fork of the SPIRV-LLVM_Translator git repository. This repository accepts issues related to AMD fork-specific topics. For all other issues/PRs, please submit upstream at https://github.com/K… |
| amd-anp | 4 | 活跃 | AINIC Network Plugin |
| aomp-extras | 4 | 活跃 | hostcall services library, math library, and utilities |
| aqlprofile | 4 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| builder | 4 | fork | Continuous builder and binary build scripts for pytorch |
| cm | 4 | 活跃 | Frontend for configuring/building/testing CMake projects |
| dask-cuda-rocm | 4 | fork | Utilities for Dask and CUDA interactions |
| DeepLearningExamples | 4 | fork | State-of-the-Art Deep Learning scripts organized by models - easy to train and deploy with reproducible accuracy and performance on enterprise-grade infrastructure. |
| horovod | 4 | fork | Distributed training framework for TensorFlow, Keras, and PyTorch. |
| jax-aiter | 4 | 活跃 | — |
| mblas-bench | 4 | 活跃 | Basic BLAS benchmark for GEMMs |
| MIFin | 4 | 活跃 | Tuna centric MIOpen client |
| mxDataGenerator | 4 | 活跃 | [DEPRECATED] Moved to ROCm/rocm-libraries repo |
| rocCV | 4 | 活跃 | rocCV is an efficient GPU-accelerated library for image pre- and post-processing, powered by AMD's HIP platform |
| rocm-spack | 4 | fork | A flexible package manager that supports multiple versions, configurations, platforms, and compilers. |
| rocm-tests | 4 | 活跃 | A pytest based system test framework that validates the full AMD ROCm stack across diverse environments. |
| rocprofiler-systems-skills | 4 | 活跃 | A collection of skills used by the rocprof-sys team |
| AMD-ROCm-User-Experience- | 3 | 归档 | AMD ROCm User Experience, including Training, Support, and Technical Documentation. |
| amdgpu-windows-interop | 3 | 归档 | Support files for interfacing ROCm to the AMDGPU Windows Driver |
| bert | 3 | fork | TensorFlow code and pre-trained models for BERT |
| hip-fpsan | 3 | 活跃 | — |
| hrx-solutions | 3 | 活跃 | Solutions built on the HRX system |
| maxtext | 3 | fork | A simple, performant and scalable Jax LLM! |
| miopen_cudnn_ops | 3 | 归档 / fork | — |
| ompi | 3 | fork | Copy of the Open MPI repository |
| rocjitsu-test-corpus | 3 | 活跃 | — |
| rocm-papi-component | 3 | 归档 | PAPI integration in ROCm profiling and tracking tools |
| rocm-timeline-generator | 3 | fork | — |
| ROCmDS-cmake | 3 | 归档 | This is a collection of CMake modules that are useful for all ROCm-DS projects. By sharing the code in a single place it makes rolling out CMake fixes easier. |
| ucc | 3 | fork | Unified Communication Collectives Library |
| whisper | 3 | fork | Robust Speech Recognition via Large-Scale Weak Supervision |
| clang-tools-extra | 2 | 归档 / fork | Mirror of official clang-tools-extra git repository located at http://llvm.org/git/clang-tools-extra. Updated every five minutes. |
| cluster-networking | 2 | 活跃 | — |
| detectron2 | 2 | fork | Detectron2 is FAIR's next-generation research platform for object detection and segmentation. |
| device-config-manager | 2 | 活跃 | Device Config Manager allows configuring AMD devices (GPUs for now) using simple config JSON files |
| distributed_inference | 2 | 活跃 | InferenceX Distributed Inference MI355X Recipe |
| eigen | 2 | 归档 | This is the ROCmSoftwarePlatform fork of the Eigen library: https://gitlab.com/libeigen/eigen |
| FAMBench | 2 | fork | Benchmarks to capture important workloads. |
| gputt | 2 | fork | gpuTT: GPU Tensor Transpose library |
| hrx-loom-library | 2 | 活跃 | Example kernel corpus for the HRX loom compiler |
| lld | 2 | 归档 / fork | — |
| mesa-fork | 2 | 活跃 | A fork of https://gitlab.freedesktop.org/mesa/mesa - Mesa 3D graphics library |
| MIOpenExamples | 2 | 归档 | MIOpen examples |
| mshadow | 2 | fork | Matrix Shadow:Lightweight CPU/GPU Matrix and Tensor Template Library in C++/CUDA for (Deep) Machine Learning |
| pytorch-lightning | 2 | fork | The lightweight PyTorch wrapper for high-performance AI research. Scale your models, not the boilerplate. |
| rocGemmDriver | 2 | 活跃 | rocGemmDriver |
| rocm-installer-runfile | 2 | 活跃 | Code for the ROCm runfile installer which is one method users can install ROCm and the AMDGPU driver on Linux |
| rocprofiler-register | 2 | 归档 | [DEPRECATED] Moved to ROCm/rocm-systems repo |
| spur-toolkit | 2 | 活跃 | Tools to deploy and debug Spur. |
| tinygrad | 2 | fork | You like pytorch? You like micrograd? You love tinygrad! ❤️ |
| triton-inference-server-onnxruntime_backend-legacy | 2 | 活跃 | — |
| triton-inference-server-vllm_backend-legacy | 2 | 活跃 | — |
| vision | 2 | fork | Datasets, Transforms and Models specific to Computer Vision |
| .github | 1 | 活跃 | AMD ROCm™ Platform - GitHub Home |
| AFLplusplus | 1 | fork | The fuzzer afl++ is afl with community patches, qemu 5.1 upgrade, collision-free coverage, enhanced laf-intel & redqueen, AFLfast++ power schedules, MOpt mutators, unicorn_mode, and a lot more! |
| amd-rocm-optimization-guide | 1 | 活跃 | Performance optimization techniques and tutorials for AMD GPUs using HIP and ROCm |
| amd-rocm-programming-guide | 1 | 活跃 | AMD ROCm Programming Guide |
| benchmarks | 1 | fork | Benchmark code |
| CHROMA | 1 | 归档 | ROCm implementation of CHROMA |
| CNTK-1 | 1 | fork | Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit |
| compiler-rt | 1 | 归档 / fork | Mirror of official compiler-rt git repository located at http://llvm.org/git/compiler-rt. Updated every five minutes. |
| cublasgemm-benchmark | 1 | 归档 / fork | code for benchmarking GPU performance based on cublasSgemm and cublasHgemm |
| DeepSpeed-MII | 1 | fork | MII makes low-latency and high-throughput inference possible, powered by DeepSpeed. |
| DeepSpeedExamples | 1 | fork | Example models using DeepSpeed |
| dlrm | 1 | fork | An implementation of a deep learning recommendation model (DLRM) |
| dyninst | 1 | fork | DyninstAPI: Tools for binary instrumentation, analysis, and modification. |
| fastai | 1 | fork | The fastai deep learning library, plus lessons and tutorials |
| FBGEMM | 1 | fork | FB (Facebook) + GEMM (General Matrix-Matrix Multiplication) - https://code.fb.com/ml-applications/fbgemm/ |
| firdev-llvm-project | 1 | 归档 | THIS REPOSITORY IS DEPRECATED AND WILL BE DELETED SOON. The new repo is f18-llvm-project |
| fuzz-fill | 1 | 活跃 | Fuzzing to fill test suite gaps |
| ggml-staging-automation | 1 | 活跃 | Release management repo for AMD staging ggml project branches (prior to upstreaming). |
| hip-mpi-testsuite | 1 | 活跃 | — |
| hip-rdma-perftest | 1 | 归档 | Verbs/RDMA benchmarks with ROCM support |
| hip-verbs-acceptance | 1 | 活跃 | Acceptance suite for the HIP runtime and RDMA Verbs: functional and performance tests for GPU, NIC and GPU+NIC paths, gated by per-platform policies. |
| hipOMB | 1 | 归档 | OSU MPI benchmarks with ROCm support |
| InferenceMAX | 1 | fork | Open Source Continuous Inference Benchmarking - GB200 NVL72 vs MI355X vs B200 vs H200 vs MI325X & soon™ TPUv6e/v7/Trainium2/3/GB300 NVL72 - DeepSeek 670B MoE, GPTOSS |
| IRFuzzer | 1 | 归档 / fork | — |
| k8s-network-device-plugin | 1 | 活跃 | Network device plugin for Kubernetes |
| k8s-network-node-labeller | 1 | 活跃 | Discovers AMD AINIC's characteristics and labels Kubernetes nodes with the corresponding information |
| kineto | 1 | fork | A CPU+GPU Profiling library that provides access to timeline traces and hardware performance counters. |
| migraphx-benchmark | 1 | 活跃 | — |
| miopen-benchmark | 1 | 归档 / fork | benchmarking miopen |
| models | 1 | fork | Models and examples built with TensorFlow |
| mosaicml_benchmarks_internal | 1 | fork | Fast and flexible reference benchmarks |
| nixl | 1 | fork | AMD maintained fork of the NIXL repository |
| og | 1 | 归档 | OpenMP GCC support |
| OpenCL-CTS | 1 | fork | The OpenCL Conformance Tests |
| pyro | 1 | fork | Deep universal probabilistic programming with Python and PyTorch |
| pytest-abort | 1 | 活跃 | Pytest plugin + helpers for attributing hard crashes (SIGSEGV/SIGABRT) to the last running test and keeping HTML reports mergeable. |
| pytorch-examples | 1 | fork | A set of examples around pytorch in Vision, Text, Reinforcement Learning, etc. |
| Quartz | 1 | 活跃 | Central CI/CD Data Hub for the ROCm Ecosystem |
| rocgputreeshap | 1 | 活跃 | ROCm support for GPUTreeShap |
| rocm-blogs-sphinx | 1 | 活跃 | — |
| rocm-recipes | 1 | 活跃 | Recipes for rocm |
| SPECFEM3D | 1 | 归档 | ROCm implementation of SPECFEM3D |
| superbenchmark | 1 | fork | Provide hardware and software benchmarks for AI systems |
| system-acceptance-docs | 1 | 活跃 | — |
| taichi_examples | 1 | 活跃 | — |
| therock-ci-config | 1 | 活跃 | Shared CI runner and infrastructure config for ROCm projects |
| triton-inference-server-server-legacy | 1 | 活跃 | — |
| aomp-fork-loop | 0 | 归档 / fork | AOMP is an open source Clang/LLVM based compiler with added support for the OpenMP® API on Radeon™ GPUs. Use this repository for releases, issues, documentation, packaging, and examples,. |
| audio | 0 | fork | Data manipulation and transformation for audio signal processing, powered by PyTorch |
| Benchmark | 0 | 归档 / fork | — |
| blis | 0 | fork | BLAS-like Library Instantiation Software Framework |
| broadcom-roce-exporter | 0 | 活跃 | — |
| ByteMLPerf | 0 | fork | AI Accelerator Benchmark focuses on evaluating AI Accelerators from a practical production perspective, including the ease of use and versatility of software and hardware. |
| camunda-jenkins-shared-library | 0 | fork | Camunda community Jenkins Shared Library |
| ck-jit | 0 | 活跃 | — |
| ClassyVision | 0 | fork | An end-to-end PyTorch framework for image and video classification |
| ClientEnablement | 0 | 活跃 | — |
| common-infra-operator | 0 | 活跃 | Shared CI/CD workflows, pytest test framework, and utilities for ROCm GPU Operator components |
| D108850-lld-bug-reproduction | 0 | 归档 | Reproduction wrapper for the bug with LLVM D108850 |
| dask | 0 | fork | Parallel computing with task scheduling |
| dask-hip | 0 | 活跃 | Utilities for Dask and HIP interactions |
| distributed | 0 | 归档 / fork | A distributed task scheduler for Dask |
| dnn-benchmarking | 0 | 活跃 | Benchmarking and validation tool for hipDNN graphs. |
| f18-llvm-project | 0 | 归档 | Mirror of flang-compiler repo: The source repo is https://github.com/flang-compiler/f18-llvm-project . Once a day the main and fir-dev branches are updated from the upstream source repo and then locked. ROCm developer… |
| faiss | 0 | fork | A library for efficient similarity search and clustering of dense vectors. |
| fastai2 | 0 | fork | Temporary home for fastai v2 while it's being developed |
| fastbook | 0 | fork | The fastai book, published as Jupyter Notebooks |
| frugally-deep | 0 | fork | Header-only library for using Keras (TensorFlow) models in C++. |
| FunctionalPlus | 0 | fork | Functional Programming Library for C++. Write concise and readable C++ code. |
| gloo | 0 | fork | Collective communications library with various primitives for multi-machine training. |
| GOTCHA | 0 | fork | GOTCHA is a library for wrapping function calls in shared libraries |
| grover | 0 | 归档 / fork | Code for Defending Against Neural Fake News, https://rowanzellers.com/grover/ |
| hsa-class | 0 | 归档 | — |
| instinct-ai-examples | 0 | 活跃 | — |
| jax-lab | 0 | 活跃 | A JAX benchmarking framework for reproducible and comparable evaluation of configurable workloads on AMD GPUs. |
| jax-triton | 0 | fork | Development fork of `jax-triton`: integrations between JAX and OpenAI Triton |
| jax-triton.old | 0 | 活跃 | ROCm fork of jax-triton |
| jaxpp | 0 | 活跃 | — |
| kernel-module-management | 0 | 活跃 | — |
| kernl | 0 | fork | Kernl lets you run PyTorch transformer models several times faster on GPU with a single line of code, and is designed to be easily hackable. |
| libfabric | 0 | fork | Open Fabric Interfaces |
| libflame | 0 | fork | High-performance object-based library for DLA computations |
| llama-cpp-temp-fork | 0 | fork | This is a temporary fork of llama.cpp to support AMD development efforts while we are migrating GitHub organizations. Final location: https://github.com/AMD-Ecosystem/llama.cpp |
| migraphx_userguide | 0 | 归档 | — |
| miopenDockerBuilder | 0 | 归档 | Scripts and |
| MONAI-Core | 0 | 归档 / fork | AI Toolkit for Healthcare Imaging |
| multi-gpu-programming-models | 0 | fork | Examples demonstrating available options to program multiple GPUs in a single node or a cluster |
| onnxruntime-training-examples | 0 | fork | Examples for using ONNX Runtime for model training. |
| OpenCL-Headers | 0 | fork | Khronos OpenCL-Headers |
| OpenLLM | 0 | fork | Operating LLMs in production |
| openwebtext | 0 | fork | An open clone of the GPT-2 WebText dataset by OpenAI. Still WIP. |
| pollara-roce-exporter | 0 | 活跃 | — |
| pytorch.github.io | 0 | fork | The website for PyTorch |
| pytorch_scatter | 0 | fork | PyTorch Extension Library of Optimized Scatter Operations |
| QoLA | 0 | 活跃 | Quality-of-Life AITER -- a wrapper library that sits atop AITER to provide a disentangled build solution for AITER consumers. |
| recommenders-addons | 0 | fork | Additional utils and helpers to extend TensorFlow when build recommendation systems, contributed and maintained by SIG Recommenders. |
| res_embed | 0 | fork | Embed resources into CMake targets in a cross-platform way |
| rocALmodels | 0 | 归档 / fork | TensorFlow models repository with RALI support |
| rocComposer | 0 | 归档 | AMD composer for High Performance Deep Learning Kernels and Libraries |
| ROCgdb-docs | 0 | 活跃 | This repository contains the documentation for AMD ROCm Debugger (ROCgdb) on Linux, AMD’s source-level debugger based on the GNU Debugger (GDB). ROCgdb enables heterogeneous debugging across ROCm software stacks that … |
| rocGemmDriver-1 | 0 | 归档 / fork | rocGemmDriver |
| rockbuilder | 0 | 活跃 | — |
| rocLibUtils | 0 | 活跃 | Shared code for the math and communication libraries |
| rocm-cmake-build-tools | 0 | 活跃 | A collection of CMake modules/functions used internally within the ROCm libraries |
| rocm-docs-style | 0 | 活跃 | — |
| rocm-fortran-generator | 0 | 活跃 | Generates the Fortran bindings of HIP, ROCTx and the ROCm math libraries from their C headers |
| rocm-framework-tests | 0 | 活跃 | — |
| rocm-handbook | 0 | 活跃 | ROCm handbooks target developers who require a unified, book-style reference for ROCm and HIP |
| rocm-security-gh | 0 | 活跃 | This Repo provides Security baseline files, scan tools, branch protection rules etc., to use by the ROCm org repositories |
| rocm-spack-packages | 0 | fork | Spack's community package recipes |
| rocprofiler-systems-report | 0 | 归档 | Workspace for generating reports assisting maintenance and engineering oversight |
| sccache | 0 | fork | sccache with ROCm and CUDA compilation (NVCC) support |
| shap | 0 | fork | A game theoretic approach to explain the output of any machine learning model. |
| SPIRV-Headers | 0 | fork | SPIRV-Headers |
| sukha-tools | 0 | fork | Profiling tools intended for Nvidia or AMD GPUs |
| tensorflow-addons | 0 | fork | Useful extra functionality for TensorFlow 2.x maintained by SIG-addons |
| tensorflow-build | 0 | fork | Build-related tools for TensorFlow |
| tensorflow-runtime | 0 | fork | A performant and modular runtime for TensorFlow |
| tensorflow-serving | 0 | 活跃 | TensorFlow Serving dockerfiles for ROCM |
| tensorflow_fashionmnist | 0 | fork | — |
| tensorpipe | 0 | fork | A tensor-aware point-to-point communication primitive for machine learning |
| text-generation-inference | 0 | fork | Large Language Model Text Generation Inference |
| TheRockMigrationTest | 0 | fork | The HIP Environment and ROCm Kit - A lightweight open source build system for HIP and ROCm |
| tiny-rocm-cxx | 0 | 归档 | — |
| torchrec | 0 | fork | Pytorch domain library for recommendation systems |
| torchtitan | 0 | fork | A PyTorch native platform for training generative AI models |
| translate | 0 | fork | Translate - a PyTorch Language Library |
| triton-inference-server-backend-legacy | 0 | 活跃 | — |
| triton-inference-server-core-legacy | 0 | 活跃 | — |
| triton-inference-server-pytorch_backend-legacy | 0 | 归档 | — |
| triton-inference-server-third_party-legacy | 0 | 活跃 | — |
| triton-kernels | 0 | fork | — |
| ucx-py-rocm | 0 | fork | Python bindings for UCX |
| utk-magma | 0 | fork | — |
| vcpkg | 0 | fork | C++ Library Manager for Windows, Linux, and MacOS |
amd:126 个仓库(AMD 芯片平台、Ryzen AI、AOCL、EPYC、FPGA)
活跃非 fork 97 · fork 26 · 归档 3 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| gaia | 1,598 | 活跃 | Build AI agents for your PC |
| RyzenAI-SW | 894 | 活跃 | AMD Ryzen™ AI Software includes the tools and runtime libraries for optimizing and deploying AI inference on AMD Ryzen™ AI powered PCs. |
| xdna-driver | 623 | 活跃 | AMD XDNA Driver (amdxdna.ko) for Linux and XRT SHIM library |
| OpenCL-caffe | 529 | 活跃 | This is a Experimental version of OpenCL by AMD Research, we now recommend you to use The official BVLC Caffe OpenCL branch is over at Caffe branch now at https://github.com/BVLC/caffe/tree/opencl |
| skills | 466 | 活跃 | Official AMD catalog of AI agent skills. Empower your AI agents with AMD's optimized SW stack. |
| HPCTrainingExamples | 220 | 活跃 | — |
| MxGPU-Virtualization | 210 | 活跃 | — |
| Quark | 186 | 活跃 | — |
| blis | 171 | fork | BLAS-like Library Instantiation Software Framework |
| ZenDNN | 160 | 活跃 | — |
| IRON | 151 | 活跃 | Close-to-metal programming for AMD NPUs |
| mini-isp | 146 | 活跃 | A minimal, open-source Image Signal Processor (ISP) for AMD FPGA, implemented in Verilog. |
| aocl-libm-ose | 136 | 活跃 | AOCL-LibM |
| playbooks | 119 | 活跃 | Official repository of AMD Playbooks |
| amd-lab-notes | 118 | 活跃 | AMD lab notes with code examples to demonstrate use of AMD GPUs |
| AMD-ASPFW | 114 | 活跃 | — |
| ryzen-ai-documentation | 108 | 活跃 | Onboarding documentation source for the AMD Ryzen™ AI Software Platform. The AMD Ryzen™ AI Software Platform enables developers to take pretrained machine learning models in popular frameworks and run them on laptop… |
| Chromium-WebCL | 95 | 活跃 | WebCL implementation for Chromium |
| furious.js | 82 | 活跃 | scientific computing package for JavaScript - inspired by NumPy |
| amd_smi_exporter | 67 | 活跃 | The AMD SMI Exporter exports AMD EPYC CPU & Datacenter GPU metrics to the Prometheus server. |
| firmware_binaries | 65 | 活跃 | — |
| amd-fftw | 64 | 活跃 | FFTW code optimized for AMD based processors |
| UIF | 62 | 活跃 | — |
| Triton-XDNA | 58 | 活跃 | — |
| amd_energy | 55 | 活跃 | — |
| fuzzyHSA | 53 | 活跃 | — |
| esmi_ib_library | 51 | 活跃 | E-SMI: EPYC™ System management Interface In-band Library |
| libflame | 50 | fork | High-performance object-based library for DLA computations |
| aocl-compression | 45 | 活跃 | A software library of lossless data compression methods tuned and optimized for AMD “Zen”-based CPUs |
| apml_modules | 44 | 活跃 | AMD APML modules, apml_sbtsi and apml_sbrmi registers to hwmon framework provding temperature and power metrics of the socket. They are extended by registering a misc_device, which provides ioctl interface to BMC admins. |
| amd_hsmp | 43 | 活跃 | AMD HSMP module to provide user interface to system management features. |
| InfinityHub-CI | 39 | 活跃 | — |
| aocl-libmem | 38 | 活跃 | A library of AMD optimized memory and string functions |
| aocl-sparse | 38 | 活跃 | AMD optimized Sparse Linear Algebra library |
| HPCTrainingDock | 37 | 活跃 | — |
| ZenDNN-pytorch-plugin | 36 | 活跃 | — |
| aocl-crypto | 33 | 活跃 | — |
| aocl | 30 | 活跃 | — |
| LIRA | 29 | 归档 | This tool helps you easily deploy ASR models on NPUs on AMD's Ryzen AI 300 series laptops |
| esmi_oob_library | 28 | 活跃 | for hosting E-SMI Out-of-band code |
| win-libm | 27 | 归档 | Core Math functions for MS Windows |
| sd-sandbox | 26 | 活跃 | Open source Stable Diffusion examples and workflows for Ryzen AI software |
| virtio-npu | 23 | 活跃 | — |
| openmm-hip | 21 | 活跃 | — |
| aocl-utils | 20 | 活跃 | AOCL-Utils library to get CPU architecture, Cache information and CPU features flags etc. |
| Kria-RoboticsAI | 20 | 活跃 | — |
| AMD-PACE | 16 | 活跃 | AMD PACE is a PyTorch extension for high-performance LLM inference on AMD CPUs. It provides a framework to develop and test novel optimizations, accelerating real-time deployment. |
| go_amd_smi | 16 | 活跃 | — |
| whisper.cpp | 16 | fork | Port of OpenAI's Whisper model in C/C++ |
| Vitis-AI | 15 | 活跃 | AMD’s stack for AI inference on Versal AI Edge and Versal AI Edge Gen 2. Host tools, examples, and reference designs organized per device series. |
| ZenDNN-onnxruntime | 14 | 活跃 | — |
| aocl-fftz | 12 | 活跃 | AMD's High-performance Fast Fourier Transform (FFT) library optimized for Zen-based processors |
| node-scraper | 12 | 活跃 | Automated data collection and analysis for the purposes of system debug. |
| sev-utils | 12 | 活跃 | Utilities and tools for AMD SEV memory encryption technologies |
| ama-sdk | 11 | 活跃 | — |
| aocl-dlp | 11 | 活跃 | A library of AMD Optimized Deep Learning Primitives |
| ZenDNN-tensorflow-plugin | 11 | 活跃 | — |
| ADA | 10 | 活跃 | AMD Deployment Assistant |
| AOCL.jl | 10 | 活跃 | — |
| Open-Field-Health-Check | 10 | 活跃 | — |
| scalapack | 10 | 活跃 | DEPRECATED. This Scalapck repository is deprecated. The last version in this repository is 3.0. Refer to "aocl-scalapack" repository under the same "amd" organization for AOCL Scalapack 3.1 release onwards. https://gi… |
| aocl-data-analytics | 9 | 活跃 | AMD Optimized Data Analytics Library |
| resource-optimizer | 9 | 活跃 | — |
| ZenDNN-pytorch | 9 | 活跃 | — |
| GenAI-contest | 8 | 活跃 | — |
| mumps-build | 8 | 活跃 | — |
| spack | 8 | fork | A flexible package manager that supports multiple versions, configurations, platforms, and compilers. |
| onnxruntime | 7 | fork | ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator |
| skillscope | 7 | 活跃 | AMD's testing harness for AI Agent Skills |
| ImageTilingNPU | 6 | 活跃 | — |
| quark-documentation | 6 | 归档 | Quark (new quantizer) documentation. https://quark.docs.amd.com/latest/index.html# |
| aocl-scalapack | 5 | fork | AMD optimized ScaLAPACK development repository. Please refer build and test documentation from AOCL User Guide at AMD Developer portal https://developer.amd.com/amd-aocl/ |
| dotnet-runtime | 5 | fork | .NET is a cross-platform runtime for cloud, mobile, desktop, and IoT apps. |
| vaip | 5 | 活跃 | vaip open source |
| Linux_ISP_Kernel | 4 | 活跃 | — |
| pensando-elk | 4 | 活跃 | — |
| xir | 4 | 活跃 | vaip open source |
| agentsim | 3 | 活跃 | Developer tool for visualizing, debugging, and understanding agent traces |
| aocl-spack | 3 | 活跃 | DEPRECATED. This Spack repo is deprecated. Refer to "spack" repository under the same "amd" organization for AOCL Spack packages. https://github.com/amd/spack |
| convnet-benchmarks | 3 | fork | Easy benchmarking of all public open-source implementations of convnets |
| DynamicDispatch | 3 | 活跃 | vaip open source |
| embedded-x86-ai | 3 | 活跃 | — |
| Linux_ISP_libcamera | 3 | 活跃 | — |
| logslop | 3 | 活跃 | LogsLop is a log summarization tool that removes repeated messages from log files, enabling readers and agents to home in quickly on key message events. |
| node-unzip | 3 | fork | node.js cross-platform unzip using streams |
| rocALUTION | 3 | fork | Next generation library for iterative sparse solvers for ROCm platform |
| ryzenai-mcp-server | 3 | 活跃 | Sharing code that allows for a fully managed MCP server that provides up-to-date access to AMD Ryzen AI documentation, code samples, and official examples (docs, code samples, and examples are all pulled from public i… |
| ZenDNN-tensorflow | 3 | 活跃 | — |
| apollo | 2 | fork | An open autonomous driving platform |
| Gyanam-GPU-Observability-Assistant | 2 | 活跃 | Fleet Intelligence for Realtime GPU observability |
| lz4 | 2 | fork | Extremely Fast Compression algorithm |
| opencv | 2 | fork | Open Source Computer Vision Library |
| OPTIMA-Architecture | 2 | 活跃 | OPTIMA — Data Intelligence Platform AI and IT Analytics. A self-hosted Docker Compose data intelligence platform combining Apache Iceberg, PuppyGraph, Dremio, Dagster, and an AI assistant for natural language queries… |
| transformers | 2 | 活跃 | vaip open source |
| zstd | 2 | fork | Zstandard - Fast real-time compression algorithm |
| AOCL_jll.jl | 1 | 活跃 | — |
| existdb-node | 1 | fork | A node.js client library for the eXist-db Native XML Database. |
| flashrom-amd | 1 | 活跃 | Publish AMD's changes to the opensource flashrom project |
| git-time-lapse | 1 | 活跃 | — |
| graph_engine | 1 | 活跃 | vaip open source |
| oneDNN-Zen | 1 | 活跃 | For upstreaming AMD CPU optimization in OneDNN, which is widely used in all framework like PyTorch, TF, vLLM |
| tock | 1 | fork | A secure embedded operating system for microcontrollers |
| tools-and-sdks | 1 | 活跃 | For distributing software via Spack tool |
| vart | 1 | 活跃 | vaip open source |
| VVAS | 1 | 活跃 | For the VAI 6.2 release, we are planning to migrate the VVAS GHES source code to github.com/amd/VVAS |
| acoll | 0 | 活跃 | — |
| aocl-rng | 0 | 活跃 | AOCL-RNG 4.2 Library Documentation. For latest API information, refer to link in README.md |
| astc-encoder | 0 | fork | The Arm ASTC Encoder, a compressor for the Adaptive Scalable Texture Compression data format. |
| avm | 0 | fork | AVM (AOM Video Model) is the reference software for AV2 codec from Alliance for Open Media (https://aomedia.org/). |
| ClickHouse | 0 | fork | ClickHouse® is a real-time analytics database management system |
| LMCache_zen | 0 | 活跃 | — |
| lzbench | 0 | fork | lzbench is an in-memory benchmark of open-source compressors |
| lzbench-old | 0 | 活跃 | — |
| mlcflow | 0 | fork | MLCFlow: Simplifying MLPerf Automations |
| mlperf-automations | 0 | fork | This repository contains automation scripts designed to run MLPerf Inference benchmarks. Originally developed for the Collective Mind (CM) automation framework, these scripts have been adapted to leverage the MLC auto… |
| opencv_extra | 0 | fork | OpenCV extra data |
| openrng | 0 | 活跃 | — |
| openssl | 0 | fork | TLS/SSL and crypto library |
| opus | 0 | fork | Modern audio compression for the internet. |
| perfspect | 0 | 活跃 | — |
| processwatch | 0 | 活跃 | — |
| quark-examples | 0 | 活跃 | — |
| target_factory | 0 | 活跃 | vaip open source |
| trace_logging | 0 | 活跃 | vaip open source |
| unilog | 0 | 活跃 | vaip open source |
| Yggdrasil | 0 | fork | Collection of builder repositories for BinaryBuilder.jl |
AMD-AGI:72 个仓库(AMD AI Group:模型、训练、推理优化)
活跃非 fork 62 · fork 10 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| Instella | 334 | 活跃 | Fully Open Language Models with Stellar Performance |
| Hyperloom | 256 | 活跃 | An agentic system that auto-optimizes LLM workloads on AMD GPUs. |
| GEAK | 196 | 活跃 | Generating Efficient AI-Centric Kernels |
| AMD-LLM | 195 | 活跃 | Training code and resources for AMD-135M language models on AMD GPUs. |
| Primus | 136 | 活跃 | A flexible and high-performance training framework designed for large-scale foundation model training on AMD GPUs |
| Nitro-E | 129 | 活跃 | Nitro-E is a family of text-to-image diffusion models focused on highly efficient training. |
| Nitro-1 | 126 | 活跃 | This repo provides a working re-implementation of Latent Adversarial Diffusion Distillation by AMD |
| AgentKernelArena | 123 | 活跃 | AgentKernelArena is a platform for evaluating and training AI agents for GPU kernel optimization, combining realistic tasks, correctness validation, and performance-based rewards for reinforcement learning. |
| TraceLens | 100 | 活跃 | Automating analysis from trace files |
| Magpie | 85 | 活跃 | A lightweight, general-purpose framework for evaluating GPU kernel and benchmark. |
| Instella-MoE | 80 | 活跃 | Instella MoE is a fully open state-of-the-art Mixture-of-Experts language model, trained end-to-end on AMD Instinct™ GPUs. |
| Apex | 78 | 活跃 | Agents, and RL environment, for optimizing GPU kernels on AMD ROCm using LLM agents. Benchmarks LLM serving workloads end-to-end, profiles bottleneck kernels, optimizes them via Claude Code or Codex, and scores on com… |
| Primus-Turbo | 71 | 活跃 | A high-performance acceleration library dedicated to large-scale model training on AMD GPUs |
| Micro-World | 70 | 活跃 | For world model code developing and releasing. |
| Primus-SaFE | 59 | 活跃 | Primus-SaFE(Stability and Fault Endurance) |
| Hummingbird | 47 | 活跃 | AMD 0.9B efficient text to video diffusion model |
| FLy | 45 | 活跃 | FLy: Training-Free Loosely Speculative Decoding: Accepting Semantically Correct Drafts Beyond Exact Match (ICLR 2026) |
| Nitro-T | 42 | 活跃 | Nitro-T is a family of text-to-image diffusion models focused on highly efficient training. |
| PARD | 37 | 活跃 | PARD: Accelerating LLM Inference with Low-Cost PARallel Draft Model Adaptation (ICLR 26) |
| Gumiho | 33 | 活跃 | Official Implementation of "Gumiho: A Hybrid Architecture to Prioritize Early Tokens in Speculative Decoding" (ICML'25) |
| DoSSR | 30 | 活跃 | [NeurIPS 2024]Taming Diffusion Prior for Image Super-Resolution with Domain Shift SDEs |
| maxtext-slurm | 30 | 活跃 | Toolkit for launching and observing MaxText training on Slurm-managed GPU clusters |
| DUET-VLM | 26 | 活跃 | DUET-VLM: Dual stage Unified Efficient Token reduction for VLM Training and Inference |
| Infera | 26 | 活跃 | More token goodput from frontier models. A distributed, SLA-aware serving mesh — disaggregated prefill/decode, KV-aware routing, and cache offload, tuned to your production SLA. |
| Instella-T2I | 24 | 活跃 | Instella-T2I: Pushing the Limits of 1D Discrete Latent Space Image Generation |
| InstellaVL | 24 | 活跃 | Official code repo for instellavl-1b |
| ReNeg | 19 | 活跃 | ReNeg: Learning Negative Embedding with Reward Guidance |
| HummingbirdXT | 17 | 活跃 | This repository presents an efficient acceleration pipeline for Diffusion Transformer (DiT) based video generation models, optimized for AMD client-grade GPUs, including Navi48 dGPUs and Strix Halo iGPUs. |
| torchtitan-amd | 17 | fork | A PyTorch native platform for training generative AI models |
| GEAK-eval | 15 | 活跃 | — |
| Neurips2025-GPU-kernels-Tutorial | 15 | 活跃 | Repo containing artifacts for Neurips 2025 tutorial- How to Build Agents to Generate Kernels for Faster LLMs (and Other Models!) |
| AMD-Hybrid-Models | 14 | 活跃 | Official repo for AMD hybrid models training and inference workflow |
| ReasonLite | 12 | 活跃 | Efficient Reasoning Model |
| AMDKernelVault | 10 | 活跃 | AMDKernelVault A research and engineering playground for HIP and Triton kernel generation, optimization, and LLM training, including continued pretraining, SFT, and RL (RLHF/RLVR) pipelines. |
| Instella-Math | 9 | 活跃 | Training code for Instella-Math model |
| pytorch-training-benchmark | 8 | 活跃 | DEPRECATED: This is a benchmark script only for archiving purpose |
| gpt-fast | 7 | 活跃 | The GPT-Fast for Multimodal Models on AMD GPUs |
| AMD_QTViT | 6 | 活跃 | QT-ViT: Improving Linear Attention in ViT with Quadratic Taylor Expansion |
| FarSkip-Collective | 6 | 活跃 | Training and inference implementation of FarSkip-Collective models enabling communication-computation overlap |
| Tyr-the-Pruner | 6 | 活跃 | Týr-the-Pruner: Structural Pruning LLMs via Global Sparsity Distribution Optimization (NeurIPS25) |
| AMD_FDViT | 5 | 活跃 | FDViT: Improve the Hierarchical Architecture of Vision Transformer (ICCV 2023) |
| AMD_IFFN | 5 | 活跃 | Enhancing Vision Transformer: Amplifying Non-Linearity in Feedforward Network Module (ICML 2024) |
| awesome-rocm-autodrive | 5 | 活跃 | Examples of training autodrive models in ROCm |
| PrimusClaw | 5 | 活跃 | LLM agent orchestration on Kubernetes: autonomous coding-agent sessions in per-session sandboxes. |
| DynamicChunkingDiT | 3 | 活跃 | Official code for DC-DiT: Adaptive Compute and Elastic Inference for Visual Generation via Dynamic Chunking |
| m3d_rocm | 3 | 活跃 | This project is an optimized version of Matrix3D. It has better compatibility with ROCm ecosystem. |
| AMD-Spark | 2 | 活跃 | SparK: Query-Aware Unstructured Sparsity with Recoverable KV Cache Channel Pruning |
| Athena-PRM | 2 | 活跃 | Athena-PRM: Enhancing Large Reasoning Model with Data-efficient Process Reward Models |
| HF_PEFT_GPT_OSS | 2 | 活跃 | Hugging Face PEFT for GPT_OSS |
| Primus-DLRM | 2 | 活跃 | DLRM implementation for Primus |
| sand-pipeline | 2 | 活跃 | Synthetic data generation pipeline, finetuning and evaluation scripts. |
| AMD-EfficientVLM | 1 | 活跃 | Introducing the AMD-EfficientVLM (Vision-Language Model) |
| axlearn | 1 | fork | An Extensible Deep Learning Library |
| nixl | 1 | fork | NVIDIA Inference Xfer Library (NIXL) |
| ALTO | 0 | 活跃 | ALTO: Advanced Low-precision Training and Optimization |
| AMD_MSWA | 0 | 活跃 | MSWA: Refining Local Attention with Multi-ScaleWindow Attention |
| AMDLongContextServing | 0 | 活跃 | Kimi-Linear long-context FP8 benchmark tooling for vLLM on AMD MI355X (AITER MLA head-padding + patched decode/prefill kernels). |
| composer | 0 | fork | Supercharge Your Model Training |
| diffusion-models-inference | 0 | 活跃 | Diffusion model inference benchmarking, profiling and optimizing |
| DLRMBenchmark | 0 | 活跃 | Repository for Showcasing DLRM v2 Functionality on AMD |
| dynamo | 0 | fork | A Datacenter Scale Distributed Inference Serving Framework |
| faster-action-models | 0 | 活跃 | Faster inference for World Action Models (FastWAM, DiT4DiT) on AMD GPUs: GEAK optimized kernels, fewer denoising steps, and Compact Action Experts. |
| hipBLASLt | 0 | fork | hipBLASLt is a library that provides general matrix-matrix operations with a flexible API and extends functionalities beyond a traditional BLAS library |
| LoKRA | 0 | 活跃 | Low Kruskal Rank Adaptation (ICML 2026) |
| mlcommons-inference | 0 | fork | Reference implementations of MLPerf® inference benchmarks |
| mlperf-common | 0 | 活跃 | mlperf-common |
| pr_pundit | 0 | 活跃 | An agent for OSS contributions |
| prime_amd | 0 | 活跃 | Prime repository with functionality enabled on AMD |
| Repo2Reproduce | 0 | 活跃 | Agentic system for reproducing research paper results from official paper and code on AMD ROCm GPUs. |
| ssd | 0 | fork | A lightweight inference engine supporting speculative speculative decoding (SSD). |
| torchforge | 0 | fork | PyTorch-native post-training at scale |
| vllm-2026 | 0 | fork | A high-throughput and memory-efficient inference and serving engine for LLMs |
AMD-Ecosystem:213 个仓库(AMD 开源生态移植与适配)
活跃非 fork 23 · fork 190 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| gsplat-amd | 39 | 活跃 | — |
| llama.cpp | 27 | fork | LLM inference in C/C++ |
| hipDF | 19 | 活跃 | hipDF - GPU DataFrame Library |
| flashinfer | 12 | fork | FlashInfer+ROCm: ROCm port of FlashInfer |
| hipMM | 8 | 活跃 | HIP Memory Manager (ROCm-DS) |
| hipVS | 7 | 活跃 | — |
| warp | 7 | fork | A Python framework for accelerated simulation, data generation and spatial computing. |
| hipGRAPH | 6 | 活跃 | Portable, accelerated graph algorithms for GPUs |
| ray | 6 | fork | Ray is a unified framework for scaling AI and Python applications. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads. |
| rocGRAPH | 6 | 活跃 | Accelerated graph algorithms for AMD GPUs |
| xgboost | 6 | fork | Scalable, Portable and Distributed Gradient Boosting (GBDT, GBRT or GBM) Library, for Python, R, Java, Scala, C++ and more. Runs on single machine, Hadoop, Spark, Dask, Flink and DataFlow |
| taichi | 5 | fork | Productive, portable, and performant GPU programming in Python. |
| dgl | 4 | fork | Python package built to ease deep learning on graph, on top of existing DL frameworks. |
| hipRaft | 4 | 活跃 | — |
| moat | 4 | 活跃 | Porting popular CUDA projects to ROCm/HIP, one repository at a time. Driven by Claude. |
| triton-inference-server-server | 3 | fork | The Triton Inference Server provides an optimized cloud and edge inferencing solution. |
| ComfyUI | 2 | fork | The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. |
| Genesis | 2 | fork | A generative world for general-purpose robotics & embodied AI learning on rocm |
| hipCIM | 2 | fork | cuCIM - RAPIDS GPU-accelerated image processing library |
| LichtFeld-Studio | 2 | fork | Train, inspect, edit, automate, and export 3D Gaussian Splatting scenes from a single native application. |
| quadrants | 2 | fork | quadrants on rocm |
| rocm-rag | 2 | 活跃 | — |
| Stanford-Megatron-LM | 2 | fork | Ongoing research training transformer models at scale |
| verl | 2 | fork | verl: Volcano Engine Reinforcement Learning for LLMs |
| ai-ecosystem | 1 | 活跃 | This repository contains documentation for the ROCm AI ecosystem: framework installation and setup, large-scale model training, LLM and diffusion serving, and performance optimization on AMD GPUs. The ecosystem builds… |
| datascience | 1 | 活跃 | AMD ROCm-DS - GitHub Home |
| ElasticFusion | 1 | fork | Real-time dense visual SLAM system |
| lifescience-examples | 1 | 活跃 | — |
| MASt3R-SLAM | 1 | fork | [CVPR 2025] MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors |
| megablocks | 1 | fork | — |
| MONAI | 1 | fork | AI Toolkit for Healthcare Imaging |
| monai-model-zoo | 1 | fork | MONAI Model Zoo that hosts models in the MONAI Bundle format. |
| MONAILabel | 1 | fork | MONAI Label is an intelligent open source image labeling and learning tool. |
| ROCmDS-cmake | 1 | 活跃 | This is a collection of CMake modules that are useful for all ROCm-DS projects. By sharing the code in a single place it makes rolling out CMake fixes easier. |
| simulation | 1 | 活跃 | — |
| triton-inference-server-core | 1 | fork | The core library and APIs implementing the Triton Inference Server. |
| .github | 0 | 活跃 | AMD Ecosystem Github Repo |
| 3DGS-LM | 0 | fork | 3DGS-LM accelerates Gaussian-Splatting optimization by replacing the ADAM optimizer with Levenberg-Marquardt. (ICCV 2025) |
| 3DUNDERWORLD-SLS-GPU_CPU | 0 | fork | A structured light scanner |
| 3P-ADMM-PC2 | 0 | fork | — |
| accelerated-scan | 0 | fork | Accelerated First Order Parallel Associative Scan |
| aihwkit | 0 | fork | IBM Analog Hardware Acceleration Kit |
| alien | 0 | fork | ALIEN is a CUDA-powered artificial life simulation program. |
| amgcl | 0 | fork | C++ library for solving large sparse linear systems with algebraic multigrid method |
| anari-visionaray | 0 | fork | — |
| arbor | 0 | fork | The Arbor multi-compartment neural network simulation library. |
| arrayfire | 0 | fork | ArrayFire: a general purpose GPU library. |
| AutoDock-GPU | 0 | fork | AutoDock for GPUs and other accelerators |
| bam | 0 | fork | — |
| barney | 0 | fork | A Scalable (and Optionally, Data-Parallel) ANARI Multi-GPU Path Tracer |
| baspacho | 0 | fork | Direct solver for sparse SPD matrices for nonlinear optimization. Implements supernodal Cholesky decomposition algorithm, and supports GPU (CUDA). |
| bellhopcuda | 0 | fork | CUDA and C++ port of BELLHOP / BELLHOP3D underwater acoustics simulator |
| brax | 0 | fork | Massively parallel rigidbody physics simulation on accelerator hardware. |
| brian2cuda | 0 | fork | A brian2 extension to simulate spiking neural networks on GPUs |
| catboost | 0 | fork | A fast, scalable, high performance Gradient Boosting on Decision Trees library, used for ranking, classification, regression and other machine learning tasks for Python, R, Java, C++. Supports computation on CPU and GPU. |
| Clair3 | 0 | fork | Clair3 - Symphonizing pileup and full-alignment for deep learning-based long-read variant calling |
| CogVideo | 0 | fork | text and image to video generation: CogVideoX (2024) and CogVideo (ICLR 2023) |
| colmap | 0 | fork | COLMAP - Structure-from-Motion and Multi-View Stereo |
| CTranslate2 | 0 | fork | Fast inference engine for Transformer models |
| CubbyFlow | 0 | fork | Voxel-based fluid simulation engine for computer games |
| cuBQL | 0 | fork | A CUDA BVH Build And Query Library |
| cubvh | 0 | fork | Mesh tools. |
| cuda-efficient-features | 0 | fork | A CUDA implementation of keypoint detection and descriptor extraction |
| CUDA-ScanMatcher-ICP | 0 | fork | A high performance CUDA implementation of Scan Matching via the Iterative Closest Point Algorithm |
| cudaKDTree | 0 | fork | — |
| CudaSift | 0 | fork | A CUDA implementation of SIFT for NVidia GPUs (1.2 ms on a GTX 1060) |
| CuMesh | 0 | fork | Cuda mesh utils. |
| cuPDLP-C | 0 | fork | Code for solving LP on GPU using first-order methods |
| cuPDLPx | 0 | fork | A GPU-Accelerated First-Order LP Solver |
| cupoch | 0 | fork | Robotics with GPU computing |
| CuRast | 0 | fork | Cuda-Based Software Rasterization for Billions of Triangles |
| cuSZ | 0 | fork | A GPU accelerated error-bounded lossy compression for scientific data. |
| CV-CUDA | 0 | fork | CV-CUDA™ is an open-source, GPU accelerated library for cloud-scale image processing and computer vision. |
| dart | 0 | fork | Research-focused C++23 physics engine for robotics, animation, and machine learning, with Python bindings |
| dask-hip | 0 | 活跃 | First release of rocm based rocm-ds dask-hip |
| DEM-Engine | 0 | fork | A dual-GPU DEM solver with complex grain geometry support |
| dgSPARSE-Lib | 0 | fork | PyTorch-Based Fast and Efficient Processing for Various Machine Learning Applications with Diverse Sparsity |
| dietgpu | 0 | fork | GPU implementation of a fast generalized ANS (asymmetric numeral system) entropy encoder and decoder, with extensions for lossless compression of numerical and other data types in HPC/ML applications. |
| diff-surfel-rasterizations | 0 | fork | Collection of differentiable surfel rasterizers with different feature channels. |
| diff-surfel-tracing | 0 | fork | A fully differentiable 2D Gaussian ray tracer built on 2DGS and OptiX. |
| DiffPhysDrone | 0 | fork | Published on Nature Machine Intelligence! The first real robot(quadrotor) based on differentiable physics training. |
| egg.c | 0 | fork | EGGROLL in C, integer-first training |
| EnvGS | 0 | fork | [CVPR 2025] EnvGS: Modeling View-Dependent Appearance with Environment Gaussian |
| espresso | 0 | fork | The ESPResSo package |
| evo2 | 0 | fork | Genome modeling and design across all domains of life |
| evogp | 0 | fork | A GPU-accelerated library for Tree-based Genetic Programming, leveraging PyTorch and custom CUDA kernels for high-performance evolutionary computation. It supports symbolic regression, classification, and policy optim… |
| faiss | 0 | fork | A library for efficient similarity search and clustering of dense vectors. |
| FaithC | 0 | fork | [CVPR 2026 (Oral)] Official Torch/CUDA Implementation of Faithful Contouring |
| Fast-Poisson-Image-Editing | 0 | fork | A fast poisson image editing implementation that can utilize multi-core CPU or GPU to handle a high-resolution image input. |
| faster-gaussian-splatting | 0 | fork | An efficient and research-friendly Gaussian Splatting framework described in the CVPR'26 paper "Faster-GS: Analyzing and Improving Gaussian Splatting Optimization" |
| FastGeodis | 0 | fork | Fast Implementation of Generalised Geodesic Distance Transform for CPU (OpenMP) and GPU (CUDA) |
| fdtd3d | 0 | fork | fdtd3d is an open source 1D, 2D, 3D FDTD electromagnetics solver with MPI, OpenMP and CUDA support for x64, ARM, ARM64, RISC-V, PowerPC, Wasm architectures |
| ffpa-attn | 0 | fork | 🤖FFPA: Extends FlashAttention-2 via Split-D for large headdims, 1.5x~3×↑🎉 vs SDPA, up to 430T🎉 on H200. |
| finance | 0 | 活跃 | — |
| FLAMEGPU2 | 0 | fork | FLAME GPU 2 is a GPU accelerated agent based modelling framework for CUDA C++ and Python |
| flashinfer-bench | 0 | fork | Building the Virtuous Cycle for AI-driven LLM Systems |
| flower | 0 | fork | Flower: A Friendly Federated AI Framework |
| foldmason | 0 | fork | Multiple Protein Structure Alignment at Scale with FoldMason |
| foldseek | 0 | fork | Foldseek enables fast and sensitive comparisons of large structure sets. |
| fused-ssim | 0 | fork | Lightning fast differentiable SSIM. |
| gaussian_splatting | 0 | fork | Unofficial implementation of 3D Gaussian Splatting in PyTorch + CUDA with MIT license |
| gdtk | 0 | fork | The Gas Dynamics Toolkit (GDTk) is a set of software tools for simulating high speed fluid flow, maintained at The University of Queensland and the University of Southern Queensland, Australia. |
| GOMC | 0 | fork | GOMC - GPU Optimized Monte Carlo is a parallel molecular simulation code designed for high-performance simulation of large systems |
| GooFit | 0 | fork | Code repository for the massively-parallel framework for maximum-likelihood fits, implemented in CUDA/OpenMP |
| google-deepvariant | 0 | fork | DeepVariant is an analysis pipeline that uses a deep neural network to call genetic variants from next-generation DNA sequencing data. |
| gpu4pyscf | 0 | fork | A plugin to use Nvidia GPU in PySCF package |
| GPU_IPC | 0 | fork | This is the first fully GPU Optimized IPC framework |
| Gpufit | 0 | fork | GPU-accelerated Levenberg-Marquardt curve fitting in CUDA |
| GPUMD | 0 | fork | Graphics Processing Units Molecular Dynamics |
| gpuRIR | 0 | fork | Python library for Room Impulse Response (RIR) simulation with GPU acceleration |
| gRASPA | 0 | fork | GPU Monte Carlo Simulation Code with a taste of RASPA |
| gsplat | 0 | fork | CUDA accelerated rasterization of gaussian splatting |
| gtsam_points | 0 | fork | A collection of GTSAM factors and optimizers for point cloud SLAM |
| HEonGPU | 0 | fork | HEonGPU is a high-performance library that optimizes Fully Homomorphic Encryption (FHE) on GPUs. Leveraging GPU parallelism, it reduces computational load through concurrent execution. Its multi-stream architecture mi… |
| hip-ucxx | 0 | 活跃 | First release of hip-ucxx for the ROCmDS 26.03 release supporting ROCm 7.2 |
| icicle | 0 | fork | A hardware acceleration library for compute intensive cryptography :ice_cube: |
| IsaacLab | 0 | fork | Unified framework for robot learning built on NVIDIA Isaac Sim |
| k2 | 0 | fork | FSA/FST algorithms, differentiable, with PyTorch compatibility. |
| kaldi | 0 | fork | kaldi-asr/kaldi is the official location of the Kaldi project. |
| kaldifeat | 0 | fork | Kaldi-compatible online & offline feature extraction with PyTorch, supporting CUDA, batch processing, chunk processing, and autograd - Provide C++ & Python API |
| kaolin | 0 | fork | A PyTorch Library for Accelerating 3D Deep Learning Research |
| KittenGpuLBVH | 0 | fork | A high performance and friendly GPU LBVH implementation. |
| LC-framework | 0 | fork | — |
| lc0 | 0 | fork | Open source neural network chess engine with GPU acceleration and broad hardware support. |
| LEAP | 0 | fork | comprehensive library of 3D transmission Computed Tomography (CT) algorithms with Python and C++ APIs, a PyQt GUI, and fully integrated with PyTorch |
| libSGM | 0 | fork | Stereo Semi Global Matching by cuda |
| lifescience | 0 | 活跃 | AMD Life Science |
| LightGBM | 0 | fork | A fast, distributed, high performance gradient boosting (GBT, GBDT, GBRT, GBM or MART) framework based on decision tree algorithms, used for ranking, classification and many other machine learning tasks. |
| LiteGS | 0 | fork | A refactored codebase for Gaussian Splatting. Training 3DGS in 50 seconds! |
| llm-awq | 0 | fork | [MLSys 2024 Best Paper Award] AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration |
| llm.c | 0 | fork | LLM training in simple, raw C/CUDA |
| llmextension | 0 | 活跃 | — |
| LMCache | 0 | fork | LMCache: Supercharge Your LLM with the Fastest KV Cache Layer |
| mahout | 0 | fork | Apache Mahout - an environment for quickly creating scalable, performant machine learning applications. |
| ManiSkill | 0 | fork | Manipulation Skill Framework, an open source GPU parallelized robotics simulator and benchmark |
| marian-dev | 0 | fork | Fast Neural Machine Translation in C++ - development repository |
| mcx | 0 | fork | Monte Carlo eXtreme (MCX) - Physically accurate and validated GPU ray-tracer |
| metaeuk | 0 | fork | MetaEuk - sensitive, high-throughput gene discovery and annotation for large-scale eukaryotic metagenomics |
| mHC.cu | 0 | fork | mHC kernels implemented in CUDA |
| minibwa | 0 | fork | Successor of bwa-mem for short-read alignment |
| minimap2 | 0 | fork | A versatile pairwise aligner for genomic and spliced nucleotide sequences |
| mjlab | 0 | fork | Isaac Lab API, powered by MuJoCo-Warp, for RL and robotics research |
| MMseqs2 | 0 | fork | MMseqs2: ultra fast and sensitive search and clustering suite |
| MPPI-Generic | 0 | fork | Templated C++/CUDA implementation of Model Predictive Path Integral Control (MPPI) |
| mujoco | 0 | fork | Multi-Joint dynamics with Contact. A general purpose physics simulator. |
| mujoco_warp | 0 | fork | GPU-optimized version of the MuJoCo physics simulator. |
| mumax3 | 0 | fork | GPU-accelerated micromagnetic simulator |
| nerfacc | 0 | fork | A General NeRF Acceleration Toolbox in PyTorch. |
| nerfacc-amd | 0 | 活跃 | — |
| newton | 0 | fork | An open-source, GPU-accelerated physics simulation engine built upon NVIDIA Warp, specifically targeting roboticists and simulation researchers. |
| nextflow | 0 | fork | A DSL for data-driven computational pipelines |
| ntransformer | 0 | fork | High-efficiency LLM inference engine in C++/CUDA. Run Llama 70B on RTX 3090. |
| NVFlare | 0 | fork | NVIDIA Federated Learning Application Runtime Environment |
| OCTproZ | 0 | fork | OCTproZ is an open source software for optical coherence tomography processing and visualization. |
| ohm | 0 | fork | An efficient, extensible occupancy map supporting probabilistic occupancy, normal distribution transforms in CPU and GPU. |
| omp-bowtie2-prime | 0 | fork | A simplified version of bowtie2 that uses OpenMP based parallelization |
| oneflow | 0 | fork | OneFlow is a deep learning framework designed to be user-friendly, scalable and efficient. |
| op43dgs | 0 | fork | [ECCV'24] On the Error Analysis of 3D Gaussian Splatting and an Optimal Projection Strategy |
| Open3D | 0 | fork | Open3D: A Modern Library for 3D Data Processing |
| opencv | 0 | fork | Open Source Computer Vision Library |
| opencv_contrib | 0 | fork | Repository for OpenCV's extra modules |
| openfold | 0 | fork | Trainable, memory-efficient, and GPU-friendly PyTorch reproduction of AlphaFold 2 |
| openfold-3 | 0 | fork | A fully open source biomolecular structure prediction model based on AlphaFold3 |
| OpenUSD | 0 | fork | Universal Scene Description |
| openvdb | 0 | fork | OpenVDB - Sparse volume data structure and tools |
| Paddle | 0 | fork | PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice (『飞桨』核心框架,深度学习&机器学习高性能单机、分布式训练和跨平台部署) |
| PaddleNLP | 0 | fork | Easy-to-use and powerful LLM and SLM library with awesome model zoo. |
| PaddleOCR | 0 | fork | Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages. |
| PaddleRec | 0 | fork | Recommendation Algorithm |
| PaddleSpeech | 0 | fork | Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotti… |
| PhysX | 0 | fork | NVIDIA PhysX SDK |
| plumed2 | 0 | fork | Development version of plumed 2 |
| plvs | 0 | fork | PLVS is a real-time SLAM system with points, lines, volumetric mapping and 3D unsupervised incremental segmentation. |
| Pointcept | 0 | fork | Pointcept: Perceive the world with sparse points, a codebase for point cloud perception research. Latest works: Utonia (ICML'26), Concerto (NeurIPS'25), Sonata (CVPR'25 Highlight), PTv3 (CVPR'24 Oral) |
| popsift | 0 | fork | PopSift is an implementation of the SIFT algorithm in CUDA. |
| prismatic | 0 | fork | C++/CUDA package for parallelized simulation of image formation in Scanning Transmission Electron Microscopy (STEM) using the PRISM and multislice algorithms |
| qrack | 0 | fork | Comprehensive, GPU accelerated framework for developing universal virtual quantum processors |
| Quest | 0 | fork | [ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference |
| QUICK | 0 | fork | QUICK: A GPU-enabled ab intio quantum chemistry software package |
| rmagine | 0 | fork | A robot wants to simulate too |
| rmcl | 0 | fork | Mobile Robot Localization in 3D Triangle Meshes & Geometric Scene Graphs |
| RoboticsDiffusionTransformer | 0 | fork | RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation |
| rocmds-logger | 0 | 活跃 | — |
| rocOpt | 0 | fork | GPU accelerated decision optimization |
| RWKV-CUDA | 0 | fork | The CUDA version of the RWKV language model ( https://github.com/BlinkDL/RWKV-LM ) |
| RXMesh | 0 | fork | GPU-accelerated triangle mesh processing |
| SCAMP | 0 | fork | The fastest way to compute matrix profiles on CPU and GPU! |
| sglang | 0 | fork | SGLang is a high-performance serving framework for large language models and multimodal models. |
| splatad | 0 | fork | SplatAD: Real-Time Lidar and Camera Rendering with 3D Gaussian Splatting for Autonomous Driving |
| sppark | 0 | fork | Zero-knowledge template library |
| stdgpu | 0 | fork | stdgpu: Efficient STL-like Data Structures on the GPU |
| STRUMPACK | 0 | fork | Structured Matrix Package (LBNL) |
| symforce | 0 | fork | Fast symbolic computation, code generation, and nonlinear optimization for robotics |
| thundergbm | 0 | fork | ThunderGBM: Fast GBDTs and Random Forests on GPUs |
| ThunderGBM-amd | 0 | 活跃 | — |
| TIGRE | 0 | fork | TIGRE: Tomographic Iterative GPU-based Reconstruction Toolbox |
| tiny-vllm | 0 | fork | Build your own high performance LLM inference engine in C++ and CUDA - a smaller version of vLLM |
| torch-linear-assignment | 0 | fork | Batch computation of the linear assignment problem on GPU. |
| triton-inference-server-backend | 0 | fork | Common source, scripts and utilities for creating Triton backends. |
| triton-inference-server-onnxruntime_backend | 0 | fork | The Triton backend for the ONNX Runtime. |
| triton-inference-server-python_backend | 0 | fork | Triton backend that enables pre-process, post-processing and other logic to be implemented in Python. |
| triton-inference-server-pytorch_backend | 0 | fork | The Triton backend for the PyTorch TorchScript models. |
| triton-inference-server-tensorflow_backend | 0 | fork | The Triton backend for TensorFlow. |
| triton-inference-server-third_party | 0 | fork | Third-party source packages that are modified for use in Triton. |
| triton-inference-server-vllm_backend | 0 | fork | — |
| TTT3R | 0 | fork | [ICLR 2026] A simple state update rule to enhance length generalization for CUT3R |
| TurboFNO | 0 | fork | The first fully fused FFT–GEMM–iFFT GPU kernel. |
| unified-cache-management | 0 | fork | Persist and reuse KV Cache to speedup your LLM. |
| Velvet | 0 | fork | A CUDA-accelerated cloth simulation engine based on Extended Position Based Dynamics (XPBD). |
| visionaray | 0 | fork | A C++-based, cross platform ray tracing library |
| yalm | 0 | fork | Yet Another Language Model: LLM inference in C++/CUDA, no libraries except for I/O |
| YarnBall | 0 | fork | A massively parallel high performance GPU Cosserat Rods Simulator |
| ZhiLight | 0 | fork | A highly optimized LLM inference acceleration engine for Llama and its variants. |
AMDResearch:23 个仓库(AMD 研究院与教学原型)
活跃非 fork 21 · fork 2 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| Ryzers | 89 | 活跃 | Ryzen AI Dockerfiles for ML, Robotics and Development Environments |
| Riallto | 87 | 活跃 | The Riallto Open Source Project from AMD |
| aup-learning-cloud | 69 | 活跃 | AUP Learning Cloud is a customized JupyterHub platform that delivers an intuitive, hands‑on AI learning experience with AMD‑accelerated toolkits. |
| DAGEE | 50 | 活跃 | Directed Acyclic Graph Execution Engine (DAGEE) is a C++ library that enables programmers to express computation and data movement, as task graphs that are scheduled concurrently and asynchronously on both CPUs and GPUs. |
| intellikit | 37 | 活跃 | IntelliKit is a collection of intelligent tools designed to make GPU kernel development, profiling, and validation accessible to LLMs and human developers alike. Built for AMD ROCm, these tools provide clean abstracti… |
| NPUEval | 34 | 活跃 | NPUEval is an LLM evaluation dataset written specifically to target AIE kernel code generation on RyzenAI hardware. |
| intelliperf | 23 | 活跃 | Automated bottleneck detection and solution orchestration |
| hpcfund | 21 | 活跃 | AMD HPC Research Fund Cloud |
| omniprobe | 18 | 活跃 | Simply log all kernel durations |
| aup-ai-tutorials | 15 | 活跃 | AUP AI Tutorials |
| introduction_to_hip | 14 | 活跃 | — |
| AMDGeniePIM | 8 | 活跃 | This repository contains AMD GeniePIM, an analytical performance model for modeling GEMV execution on emerging PIM architectures. |
| atmi | 6 | fork | Asynchronous Task and Memory Interface, or ATMI, is a runtime framework and programming model for heterogeneous CPU-GPU systems. It provides a consistent, declarative API to create task graphs on CPUs and GPUs (integr… |
| instrument-amdgpu-kernels | 6 | fork | LLVM/MLIR based compiler instrumentation of AMD GPU kernels |
| aup-teaching-labs | 5 | 活跃 | Provide a series of AMD learning materials to support teaching and instruction. |
| ai4science-studio | 4 | 活跃 | How-to guides and AMD/ROCm optimization recipes for open AI-for-science models. |
| dh_comms | 3 | 活跃 | Facilities to allow device code to push arbitrary data onto a FIFO queue that can be processed at runtime by host code. |
| kerneldb | 3 | 活跃 | A library for interacting with CDNA code objects to support performance analysis |
| OmniHub | 2 | 活跃 | Tools for AI/ML Workload Analysis and Characterization |
| AMD-DMA-Latte | 1 | 活跃 | A collection of DMA-based ML collective implementations and microbenchmarks which harness advanced DMA engine features on AMD Instinct GPUs. |
| AdaHOP | 0 | 活跃 | — |
| geometric-nn-ops | 0 | 活跃 | — |
| intelliperf-action | 0 | 活跃 | — |
GPUOpen-LibrariesAndSDKs:85 个仓库(GPUOpen SDK 与库)
活跃非 fork 83 · fork 1 · 归档 1 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| VulkanMemoryAllocator | 3,533 | 活跃 | Easy to integrate Vulkan memory allocation library |
| FidelityFX-SDK | 1,829 | 活跃 | The main repository for the FidelityFX SDK. |
| Cauldron | 1,114 | 活跃 | A simple framework for rapid prototyping on Vulkan or Direct3D 12 |
| RadeonRays_SDK | 1,101 | 活跃 | Radeon Rays is ray intersection acceleration library for hardware and software multiplatforms using CPU and GPU |
| D3D12MemoryAllocator | 1,049 | 活跃 | Easy to integrate memory allocation library for Direct3D 12 |
| V-EZ | 892 | 活跃 | — |
| AMF | 754 | 活跃 | The Advanced Media Framework (AMF) SDK provides developers with optimal access to AMD devices for multimedia processing |
| RadeonProRenderBlenderAddon | 554 | 活跃 | This hardware-agnostic rendering plug-in for Blender uses accurate ray-tracing technology to produce images and animations of your scenes, and provides real-time interactive rendering and continuous adjustment of effe… |
| Capsaicin | 471 | 活跃 | AMD ARR team rendering framework |
| BlenderUSDHydraAddon | 399 | 活跃 | This add-on allows you to assemble and compose USD data with Blender data and render it all using various renderers via Hydra. |
| RenderPipelineShaders | 360 | 活跃 | Render Pipeline Shaders SDK |
| RadeonProRender-Baikal | 333 | 活跃 | — |
| AGS_SDK | 308 | 活跃 | AMD GPU Services (AGS) library and samples |
| RadeonProRenderSDK | 287 | 活跃 | AMD Radeon™ ProRender is a powerful physically-based path traced rendering engine that enables creative professionals to produce stunningly photorealistic images. |
| Orochi | 260 | 活跃 | — |
| RadeonProRenderUSD | 242 | 活跃 | This plug-in allows GPU or CPU accelerated viewport rendering on all OpenCL 1.2 hardware for the open source USD and Hydra system. You can build this plug-in as a USDView plug-in or a Houdini plug-in. |
| MxGPU-Virtualization | 208 | 归档 | — |
| brotli_g_sdk | 183 | 活跃 | Brotli-G SDK provides an improved lossless compression ratio with GPU decompression support than the standard Brotli compression algorithm maintained by the IETF (also known as RFC7932) |
| Radeon-ReLive-VR | 164 | 活跃 | — |
| HelloD3D12 | 161 | 活跃 | Introductory DirectX 12 sample |
| OCL-SDK | 154 | 活跃 | — |
| TAN | 149 | 活跃 | AMD TrueAudio Next is a software development kit for GPU accelerated audio signal processing |
| ForwardPlus11 | 148 | 活跃 | AMD Forward+ sample based on DirectX 11 |
| display-library | 146 | 活跃 | AMD Display Library SDK |
| HIPRT | 146 | 活跃 | — |
| WorkGraphPlayground | 140 | 活跃 | A DirectX12-based C++-application that allows graphics programmers to learn and experiment with the new Work Graphs feature using HLSL shaders. It runs on Windows 10 and Windows 11 systems. |
| glTFSample | 131 | 活跃 | A simple demo to show off the capabilities of the Cauldron framework |
| GPUParticles11 | 128 | 活跃 | AMD GPU particles sample based on DirectX 11 |
| HelloVulkan | 123 | 活跃 | Introductory Vulkan sample |
| LiquidVR | 108 | 活跃 | The LiquidVR™ SDK is a platform based on DirectX 11 designed to simplify and optimize VR development |
| Schola | 97 | 活跃 | Schola is a plugin for enabling Reinforcement Learning (RL) in Unreal Engine. It provides tools to help developers create environments, define agents, and connect to python-based RL frameworks such as OpenAI Gym, RLli… |
| ADLX | 91 | 活跃 | AMD Device Library eXtra |
| RadeonML | 87 | 活跃 | — |
| TiledLighting11 | 78 | 活跃 | AMD compute-based tiled lighting sample based on DirectX 11 |
| HIPRTSDK | 70 | 活跃 | — |
| DGF-SDK | 68 | 活跃 | The main repository for the AMD Dense Geometry Format SDK |
| Tessellation | 66 | 活跃 | OpenGL sample that demonstrates terrain tessellation on the GPU |
| nBodyD3D12 | 60 | fork | This repo contains the DirectX Graphics samples that demonstrate how to build graphics intensive applications on Windows. |
| GameEngineIntegrations | 55 | 活跃 | Information about example integrations of GPUOpen technology into game engines |
| cpu-core-counts | 54 | 活跃 | A sample demonstrating how to correctly detect physical core and logical processor counts on AMD processors. |
| RadeonProRenderMayaPlugin | 51 | 活跃 | This hardware-agnostic rendering plug-in for Maya uses accurate ray-tracing technology to produce images and animations of your scenes, and provides real-time interactive rendering and continuous adjustment of effects. |
| RadeonImageFilter | 48 | 活跃 | — |
| Vulkan-Samples | 47 | 活跃 | based on Khronos Vulkan-Samples |
| SSAA11 | 45 | 活跃 | AMD supersample anti-aliasing (SSAA) sample based on DirectX 11 |
| AntiLag2-SDK | 42 | 活跃 | The main repository for the AMD Radeon Anti-Lag 2 SDK. |
| RenderStudioKit | 40 | 活跃 | — |
| MLAA11 | 39 | 活跃 | AMD morphological anti-aliasing (MLAA) sample based on DirectX 11 |
| RapidFire_SDK | 39 | 活跃 | RapidFire SDK - interface for cloud gaming and virtualization (VDI) |
| WorkGraphComputeRasterizer | 38 | 活跃 | A compute/workgraph workload running inside the Cauldron framework |
| SilhouetteTessellation11 | 33 | 活跃 | AMD tessellation sample based on DirectX 11 |
| WorkGraphsMeshNodeSample | 32 | 活跃 | — |
| DirectStorageSample | 31 | 活跃 | Demonstrating the advantages of using DirectStorage over standard file I/O asset loading. Includes the API and changes required to make such a pipeline work. |
| SPARSEtextures | 31 | 活跃 | OpenGL sample demonstrating the GL_AMD_sparse_texture extension introduced by the AMD FirePro W and Radeon HD 7000 series |
| MiniDXNN | 30 | 活跃 | An implementation of MLP using DX12 cooperative vector. |
| SeparableFilter11 | 27 | 活跃 | AMD separable filter sample based on DirectX 11 |
| CrossfireAPI11 | 26 | 活跃 | AMD Crossfire API sample for DirectX 11 |
| RadeonProRenderUE | 26 | 活跃 | — |
| Streaming-SDK | 26 | 活跃 | Advanced Interactive Streaming SDK |
| DepthBoundsTest11 | 25 | 活跃 | AMD depth bounds test DirectX 11 driver extension sample |
| Cauldron-Media | 24 | 活跃 | A selection of media files needed by Cauldron samples |
| WorkGraphsIvyGenerationSample | 24 | 活跃 | — |
| RadeonProRenderANARI | 22 | 活跃 | — |
| DirectGMA_P2P | 21 | 活跃 | Samples showing FIrePro DirectGMA features in OpenGL and OpenCL |
| FreesyncPremiumProSample | 21 | 活跃 | A simple demo to show Freesync Premium Pro integration into DX12 and Vulkan API. |
| WorkGraphsHelloWorkGraphs | 21 | 活跃 | — |
| UnityIntegrations | 20 | 活跃 | Examples that demonstrate game integration with some of the libraries and SDKs available on GPUOpen. |
| DOPPEngine | 18 | 活跃 | — |
| Framelock | 18 | 活跃 | OpenGL sample that demonstrates how to enable Framelock in an application |
| ScholaExamples | 18 | 活跃 | Schola Examples is an Unreal Engine project containing sample environments developed with the Schola plugin for Unreal Engine. Schola provides tools to help developers create environments, define agents, and connect t… |
| OutOfOrderRasterization | 16 | 活跃 | Vulkan sample built on the Anvil framework |
| VkMBCNT | 16 | 活跃 | Vulkan mbcnt sample that shows how to use the AMD_shader_ballot extension and mbcnt to perform a fast reduction within a wavefront |
| Barycentrics11 | 13 | 活跃 | Barycentric coordinates GCN shader extension sample for DirectX 11 |
| DOPP | 13 | 活跃 | — |
| WorkGraphsDirectX-Graphics-Samples | 13 | 活跃 | A fork of https://github.com/microsoft/DirectX-Graphics-Samples modified to include a sample for Work Graphs |
| Barycentrics12 | 12 | 活跃 | Barycentric coordinates GCN shader extension sample for DirectX 12 |
| VkD3DDeviceMapping | 11 | 活跃 | Simple example to show how to map devices between Vulkan and Direct3D |
| WorkGraphsHelloMeshNodes | 11 | 活跃 | — |
| RadeonProRenderMaxPlugin | 10 | 活跃 | — |
| RadeonProRender-Tests | 7 | 活跃 | — |
| RadeonProRenderSDKKernels | 4 | 活跃 | — |
| RadeonProRenderSharedComponents | 3 | 活跃 | — |
| CapsaicinTestMedia | 2 | 活跃 | Capsaicin Framework test media assets |
| RadeonProRenderMayaUSD | 2 | 活跃 | — |
| SPIRV-Tools | 2 | 活跃 | Mirror of https://github.com/KhronosGroup/SPIRV-Tools |
| DirectXShaderCompiler | 0 | 活跃 | Mirror of https://github.com/microsoft/DirectXShaderCompiler |
GPUOpen-Effects:22 个仓库(GPUOpen 图形特效)
活跃非 fork 21 · fork 1 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| FidelityFX-FSR | 2,187 | 活跃 | FidelityFX Super Resolution |
| FidelityFX-FSR2 | 2,081 | 活跃 | FidelityFX Super Resolution 2 |
| TressFX | 847 | 活跃 | DirectX 12 and Vulkan libraries that provide convenient access to realistically rendered and simulated hair and fur |
| FEMFX | 511 | 活跃 | — |
| FidelityFX | 372 | 活跃 | A set of AMD provided GPU effects with provided source and sample implementations |
| FidelityFX-SSSR | 312 | 活跃 | Stochastic Screen Space Reflections (SSSR) |
| FidelityFX-CAS | 232 | 活跃 | Contrast Adaptive Sharpening (CAS) |
| GeometryFX | 227 | 活跃 | DirectX 11 library that provides convenient access to compute-based triangle filtering (CTF) |
| FidelityFX-SPD | 213 | 活跃 | Single Pass Downsampler (SPD) |
| FidelityFX-CLI | 203 | 活跃 | FidelityFX CLI |
| ShadowFX | 166 | 活跃 | DirectX 11 and 12 library that provides a scalable and GCN-optimized solution for deferred shadow filtering |
| FidelityFX-CACAO | 159 | 活跃 | Combined Adaptive Compute Ambient Occlusion (CACAO) |
| FidelityFX-Denoiser | 155 | 活跃 | FidelityFX Denoiser |
| FidelityFX-FSR2-Unity-URP | 139 | 活跃 | FidelityFX FSR 2 for the Unity URP |
| FidelityFX-ParallelSort | 125 | 活跃 | FidelityFX Parallel Sort |
| FidelityFX-FSR-Unity-URP | 116 | 活跃 | FidelityFX FSR 1.0 for the Unity URP |
| FidelityFX-LPM | 103 | 活跃 | Luma Preserving Mapper (LPM) |
| DepthOfFieldFX | 92 | 活跃 | DirectX 11 library that provides a GCN-optimized Compute Shader implementation of Depth of Field |
| Hybrid-Shadows | 66 | 活跃 | A sample demonstrating hybrid ray tracing and rasterisation for shadow rendering and use of the FidelityFX Denoiser. |
| FidelityFX-VariableShading | 56 | 活跃 | FidelityFX Variable Shading |
| Hybrid-Stochastic-Reflections | 56 | 活跃 | A sample demonstrating hybrid ray tracing and rasterisation for reflections. |
| pbrt-v4 | 33 | fork | Source code to pbrt, the ray tracer described in the forthcoming 4th edition of the "Physically Based Rendering: From Theory to Implementation" book. |
GPUOpen-Tools:66 个仓库(GPUOpen 分析与调试工具)
活跃非 fork 58 · fork 8 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| compressonator | 1,463 | 活跃 | Tool suite for Texture and 3D Model Compression, Optimization and Analysis using CPUs, GPUs and APUs |
| GPU-Reshape | 494 | 活跃 | GPU Reshape (GRS) is an API & vendor agnostic instrumentation framework, with instruction level validation. |
| radeon_gpu_analyzer | 483 | 活跃 | The Radeon GPU Analyzer (RGA) is an offline compiler and code analysis tool for Vulkan, DirectX, OpenGL, and OpenCL. |
| radeon_gpu_profiler | 454 | 活跃 | Radeon GPU Profiler (RGP) is a tool from AMD that allows for deep inspection of GPU workloads. |
| ocat | 365 | 活跃 | The Open Capture and Analytics Tool (OCAT) provides an FPS overlay and performance measurement for D3D11, D3D12, and Vulkan |
| frame_latency_meter | 260 | 活跃 | — |
| gpu_performance_api | 260 | 活跃 | GPU Performance API for AMD GPUs |
| radeon_raytracing_analyzer | 256 | 活跃 | The Radeon Raytracing Analyzer (RRA) is a tool to visualize and inspect Bounding Volume Hierarchies (BVH) for ray tracing applications. |
| radeon_memory_visualizer | 141 | 活跃 | The Radeon Memory Visualizer (RMV) is a software tool that will allow users to analyze video memory usage on AMD Radeon GPUs. |
| radeon_compute_profiler | 85 | 活跃 | The Radeon Compute Profiler (RCP) is a performance analysis tool that gathers data from the API run-time and GPU for OpenCL™ and ROCm/HSA applications. This information can be used by developers to discover bottleneck… |
| radeon_gpu_detective | 67 | 活跃 | Tool for post-mortem analysis of GPU crashes. |
| isa_spec_manager | 62 | 活跃 | Utilities for accessing AMD's Machine-Readable GPU ISA Specifications. |
| vscode_extensions | 29 | 活跃 | Handy extensions for Visual Studio Code to work on AMD GPU technologies |
| radeon_developer_panel | 22 | 活跃 | The Radeon Developer Panel (RDP) is a software tool that allows users to capture RGP profiles, RMV traces, RRA scenes, and RGD crash analysis dumps on Radeon GPUs. |
| smoldr | 22 | 活跃 | A simple scripting tool to run DX12 shaders on HW using a text input file. |
| dev_driver_tools | 19 | 活跃 | Developer Driver Tools components (Radeon Developer Service and Radeon Developer Panel) |
| device_info | 19 | 活跃 | — |
| adl | 10 | 活跃 | — |
| appsdk | 10 | 活跃 | — |
| qt_common | 8 | 活跃 | — |
| common_lib_ext_zlib_1.2.8 | 7 | 活跃 | — |
| common_src_amddxext | 6 | 活跃 | — |
| dynamic_library_module | 5 | 活跃 | — |
| Fossilize | 5 | fork | A serialization format for various persistent Vulkan object types. |
| llm_tools | 5 | 活跃 | Hub for LLM tooling features, such as MCP servers, skills, agents and more. |
| renderdoc | 5 | fork | RenderDoc is a stand-alone graphics debugging tool. |
| common_src_amdtbasetools | 4 | 活跃 | Common base code shared by multiple tool projects |
| common_src_amdtoswrappers | 4 | 活跃 | Common OS Abstraction layer shared by multiple tool projects |
| common_src_amdvkext | 4 | 活跃 | — |
| Lyra | 4 | fork | A simple to use, composable, command line parser for C++ 11 and beyond |
| opengl | 4 | 活跃 | — |
| tsingleton | 4 | 活跃 | — |
| adl_util | 3 | 活跃 | — |
| Catch2 | 3 | fork | A modern, C++-native, header-only, test framework for unit-tests, TDD and BDD - using C++11, C++14, C++17 and later (or C++03 on the Catch1.x branch) |
| cmake-modules | 3 | fork | My collection of CMake modules |
| common_lib_amd_acl | 3 | 活跃 | — |
| common_lib_amd_ags_4.0.0 | 3 | 活跃 | — |
| common_lib_ext_opengles | 3 | 活跃 | — |
| common_lib_ext_utf8cpp | 3 | 活跃 | — |
| common_src_misc | 3 | 活跃 | — |
| googletest | 3 | fork | Google Test |
| tinyxml_2 | 3 | 活跃 | — |
| update_check_api | 3 | 活跃 | The UpdateCheckAPI repository provides utility source code for AMD Tools to check for new releases available through GitHub's Release API. |
| windows_kits | 3 | 活跃 | — |
| comgr_utils | 2 | 活跃 | — |
| common_lib_ext_boost_1.59 | 2 | 活跃 | — |
| common_lib_ext_glew_1.9 | 2 | 活跃 | — |
| common_lib_ext_opencv_2.49 | 2 | 活跃 | — |
| common_lib_ext_tinyxml2_5.0.1 | 2 | 活跃 | — |
| common_lib_ext_zlib_1.2.10 | 2 | 活跃 | — |
| common_src_acl_module_manager | 2 | 活跃 | — |
| common_src_amdtmutex | 2 | 活跃 | — |
| common_src_celf | 2 | 活跃 | — |
| common_src_cmake_modules | 2 | 活跃 | — |
| common_src_miniz | 2 | 活跃 | — |
| common_src_version_info | 2 | 活跃 | — |
| common_src_vsprops | 2 | 活跃 | Common Visual Studio build settings shared across multiple projects |
| system_info_utils | 2 | 活跃 | — |
| common_lib_ext_openexr_2.2 | 1 | 活跃 | — |
| common_lib_ext_yaml_cpp | 1 | 活跃 | — |
| cxxopts | 1 | fork | Lightweight C++ command line option parser |
| volk | 1 | fork | Meta loader for Vulkan API |
| devdriver | 0 | 活跃 | — |
| qt_isa_gui | 0 | 活跃 | — |
| sdl_utils | 0 | 活跃 | — |
| spm_db | 0 | 活跃 | — |
GPUOpen-Drivers:14 个仓库(GPUOpen 开源图形驱动)
活跃非 fork 6 · fork 4 · 归档 5 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| AMDVLK | 1,977 | 活跃 | AMD Open Source Driver For Vulkan |
| pal | 413 | 归档 | Platform Abstraction Library |
| xgl | 238 | 归档 | Vulkan API Layer |
| llpc | 185 | 归档 | LLVM-Based Pipeline Compiler |
| gpurt | 85 | 归档 | GPU Ray Tracing Library |
| spvgen | 36 | 活跃 | Library to Generate SPIR-V Binary |
| llvm | 26 | fork | — |
| llvm-dialects | 24 | 活跃 | LLVM Dialects Library |
| AMD-Gfx-Drivers | 18 | 活跃 | Forum for AMD OpenGL and Vulkan graphics drivers |
| amd-vulkan-versions | 15 | 活跃 | Machine-readable mapping from public driver version to AMD Vulkan component version |
| libamdrdf | 12 | 活跃 | AMD Radeon Data File library |
| llvm-project | 10 | 归档 / fork | This is the canonical git mirror of the LLVM subversion repository. The repository does not accept github pull requests at this moment. Please submit your patches at http://reviews.llvm.org. |
| MetroHash | 10 | fork | MetroHash v1: Exceptionally fast and statistically robust hash functions |
| CWPack | 2 | fork | An ANSI C encoder/decoder for the MessagePack serialization format / msgpack.org[C] |
11.2 九个外围官方组织
AMD-HPC:14 个仓库(AMD HPC 应用性能团队)
活跃非 fork 10 · fork 4 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| CoralGemm | 16 | 活跃 | — |
| rocBudAI | 8 | 活跃 | This is rocBudAI -- your ROCm AI Buddy! |
| ShibuyaStream | 5 | 活跃 | — |
| rooflineExtractor | 4 | 活跃 | A tool for automated roofline analysis on HPC and AI workloads |
| amd-sbd | 3 | fork | — |
| fp64monitor | 2 | 活跃 | — |
| abinit | 1 | fork | The official github mirror of the Abinit repository. We welcome bug fixes and improvements. Note that most of the active developments are hosted on our https://gitlab.abinit.org/ server. Before embarking on making si… |
| HMMBench | 1 | 活跃 | Set of benchmarks to test HMM functionality and performance |
| AI-GPU-Pauli-tensor-network | 0 | 活跃 | A C++/HIP implementation of real-time quantum dynamics for the 2D transverse-field Ising model, using Tensor Network (TN) methods in the Heisenberg picture on AMD GPUs. |
| benchWarmer | 0 | 活跃 | — |
| CvxCompress | 0 | fork | Chevron Developed Compression library. Part of the COFII framework. |
| hip-api-wrappers | 0 | 活跃 | — |
| hpcperf | 0 | 活跃 | — |
| specfem3d | 0 | fork | SPECFEM3D Cartesian simulates acoustic (fluid), elastic (solid), coupled acoustic/elastic, or poroelastic seismic wave propagation in any type of conforming mesh of hexahedra (structured & unstructured). |
AMDESE:51 个仓库(AMD 嵌入式/服务器软件工程(SEV、OpenBMC、EPYC Linux))
活跃非 fork 27 · fork 22 · 归档 3 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| AMDSEV | 381 | 活跃 | AMD Secure Encrypted Virtualization |
| linux-svsm | 147 | 活跃 | Linux SVSM (Secure VM Service Module) for secure x86 virtualization in Rust |
| sev-tool | 86 | 归档 | AMD SEV Tool |
| sev-guest | 59 | 归档 | Tools, scripts, and configuration files necessary to demonstrate an end-to-end remote attestation example with SEV-SNP. |
| OpenBMC | 43 | 活跃 | OpenBMC for Genoa SP5 socket platforms |
| linux | 35 | fork | Linux kernel source tree |
| amd-perf-tools | 34 | 活跃 | Documentation and diagnostic utilities for running Linux perf on AMD hardware |
| amd-bmc-ras | 30 | 活跃 | — |
| amd_ucode_info | 27 | 活跃 | Parse and display information about an amd-ucode (CPU microcode) container file |
| mem-encryption-tests | 26 | 活跃 | Tests related to Memory Encryption |
| qemu | 26 | 活跃 | QEMU fork |
| sched-scoreboard | 16 | 活跃 | Scheduler scoreboard is a single toolkit to capture and report all the data related to the Linux Kernel Scheduler which can help analyze performance issues potentially caused due to modelling or the heuristics adopted… |
| linux-aspeed | 11 | fork | OpenBMC Linux kernel source tree |
| ovmf | 10 | 活跃 | EDK2 / OVMF related projects |
| amd-power-control | 9 | 活跃 | OpenBMC power control for AMD CRBs. |
| linux-sdxi | 9 | fork | Linux kernel source tree |
| bmc-ppr | 6 | 活跃 | — |
| libhsmp | 6 | 活跃 | The libhsmp repository is in the process of being deprecated. The repo will remain available for the near future though. All new work on HSMP test suites and commands are a part of the AMDESE/hsmputils repository. |
| linux-kvm | 6 | fork | Linux kernel source tree |
| elves | 5 | fork | AMD EPYC Linux Validation Exerciser Suite (ELVES) |
| amd-power-cap | 4 | 活跃 | — |
| bmc-cpuinfo | 4 | 活跃 | — |
| vr-firmware-update | 4 | 活跃 | — |
| epyc-feature-doc | 3 | 活跃 | AMD EPYC Linux feature Documentation project |
| workload-insight-tool | 3 | 活跃 | Workload Insight Tool (syswit) is a tool to analyze workloads on Linux. It collects global and process related system stats exposed by the kernel (via procfs and sysfs files) in a time-series manner using collector. C… |
| amd_hsmp | 2 | 活跃 | — |
| qemu-iommu | 2 | 活跃 | — |
| rasdaemon | 2 | fork | Rasdaemon is a RAS (Reliability, Availability and Serviceability) logging tool. It records memory errors, using the EDAC tracing events. EDAC is a Linux kernel subsystem with handles detection of ECC errors from memor… |
| u-boot | 2 | fork | OpenBMC "Das U-Boot" Source Tree |
| amd-bmc-ubm | 1 | 活跃 | — |
| amd-ipmi-oem | 1 | 活跃 | — |
| amd-lcd-lib | 1 | 活跃 | — |
| avocado-misc-tests | 1 | fork | Community maintained Avocado tests repository |
| bmcweb | 1 | fork | A do everything Redfish, KVM, GUI, and DBus webserver for OpenBMC |
| elves_legacy | 1 | 活跃 | AMD EPYC Linux Validation Exerciser Suite (ELVES) |
| ledmon | 1 | fork | Enclosure LED Utilities |
| linux-iommu | 1 | fork | Linux kernel source tree |
| linux-sdci | 1 | fork | Linux kernel source tree |
| openbmc-build-scripts | 1 | 归档 / fork | — |
| pmic-error-injection-bmc | 1 | 活跃 | — |
| qemu-openbmc | 1 | fork | Official QEMU mirror. Please see https://www.qemu.org/contribute/ for how to submit changes to QEMU. Pull Requests are ignored. Please only use release tarballs from the QEMU website. |
| sev-apitest-kerneldriver | 1 | 活跃 | Example Kernel driver for SEV apitest |
| avocado | 0 | fork | Avocado is a set of tools and libraries to help with automated testing. One can call it a test framework with benefits. Native tests are written in Python and they follow the unittest pattern, but any executable can s… |
| avocado-vt | 0 | fork | Avocado VT Plugin |
| freebsd-src | 0 | fork | The FreeBSD src tree publish-only repository. Experimenting with 'simple' pull requests.... |
| libpldm | 0 | fork | — |
| linux-mm | 0 | fork | Linux kernel source tree |
| pldm | 0 | fork | — |
| sev-apitest | 0 | 活跃 | AMD SEV Apitest |
| tp-libvirt | 0 | fork | Test Provider for Libvirt and related virtualization backends |
| tp-qemu | 0 | fork | Virt Test Provider for qemu and other related virtualization backends |
pensando:24 个仓库(AMD Pensando 网络与 DPU)
活跃非 fork 8 · fork 16 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| dsc-drivers | 8 | 活跃 | — |
| dsc-tools | 4 | 活跃 | — |
| pensando-elk | 4 | 活跃 | ELK based analytics for Pensando Systems |
| dsc-linux-patches | 3 | 活跃 | Patches to pensando/dsc-linux repo |
| goipmi | 3 | fork | Go IPMI library |
| pypi | 3 | 活跃 | — |
| trex-core | 2 | fork | trex-core site |
| beats | 0 | fork | :tropical_fish: Beats - Lightweight shippers for Elasticsearch & Logstash |
| box | 0 | fork | A mruby-based Builder for Docker Images |
| diskimage-builder | 0 | fork | Image building tools for OpenStack |
| docker-registry-client | 0 | fork | A Go API client for the v2 Docker Registry API |
| dsc-artifacts | 0 | 活跃 | — |
| external-qemu | 0 | 活跃 | — |
| go-cisco-nx-api | 0 | fork | Cisco NX-OS API client library written in Go. |
| gogstash | 0 | fork | Logstash like, written in golang |
| grpc | 0 | fork | C++ based gRPC (C++, Python, Ruby, Objective-C, PHP, C#) |
| ipxe | 0 | 活跃 | — |
| overmount | 0 | fork | OverlayFS layer mounter used for creating OCI/Docker images - not ready for use |
| ovn | 0 | fork | Open Virtual Network |
| ovn-vif | 0 | fork | — |
| ovs | 0 | fork | Open vSwitch |
| progress | 0 | fork | Progress meter implemented as an io.Reader with channels |
| rebase-action | 0 | fork | GitHub Action to automatically rebase PRs |
| tftp | 0 | fork | TFTP server and client library for Golang |
AMDEPYC:15 个仓库(AMD EPYC 平台软件)
活跃非 fork 9 · fork 2 · 归档 4 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| kubernetes-power-manager | 5 | 活跃 | Kubernetes Operator to provide cluster users with a mechanism to dynamically request adjustment of worker node power management settings |
| CENTOS-MILAN-SUPPORT-PATCHES | 4 | 归档 | — |
| sev-certify | 3 | 活跃 | — |
| sev_component_test | 3 | 活跃 | — |
| dispatch | 2 | 活跃 | Tool for dispatching EFI jobs from GitHub over HTTP boot |
| Host_Processor_Microbenchmarks | 2 | 活跃 | — |
| kata-runtime-2.x-SEV | 2 | 归档 | — |
| 5GCore-UPF | 1 | 活跃 | 5G Core UPF Reference Architecture on AMD EPYC Platforms |
| linux | 1 | 归档 | — |
| nri-plugins | 1 | fork | A collection of community maintained NRI plugins |
| beacon | 0 | 活跃 | workload communication client for dispatch orchestration |
| devpath | 0 | 活跃 | UEFI Device Path parsing library |
| kata-runtime-1.x-SEV | 0 | 归档 | — |
| kubernetes | 0 | fork | Production-Grade Container Scheduling and Management |
| uefi-reset | 0 | 活跃 | UEFI module that resets the system when loaded |
AMDComputeLibraries:6 个仓库(AMD 历史计算库)
活跃非 fork 5 · fork 1 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| morton_filter | 90 | 活跃 | A compressed, sparse cuckoo filter (see https://www.vldb.org/pvldb/vol11/p1041-breslow.pdf) |
| ComputeApps | 25 | 活跃 | Compute applications. |
| OpenMPApps | 13 | 活跃 | OpenMP GPU Accelerated Applications |
| MultiGPUBandwidthTest | 5 | 活跃 | MultiGPUBandwidthTest |
| caffe | 3 | fork | Caffe: a fast open framework for deep learning. |
| Nekbone | 1 | 活跃 | — |
ROCm-LS:6 个仓库(ROCm Life Science(归档))
活跃非 fork 0 · fork 4 · 归档 6 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| hipCIM | 6 | 归档 | HIP port of the cuCIM library under the RAPIDS ecosystem. |
| monai | 3 | 归档 | First release of ROCm enabled Monai |
| GPU_biology | 0 | 归档 / fork | AMD MI250x ready container recipes and test cases for biology tools |
| hipCIMwip | 0 | 归档 / fork | hipCIM - AMD ROCm fork of RAPIDS cucim |
| model-zoo | 0 | 归档 / fork | MONAI Model Zoo that hosts models in the MONAI Bundle format. |
| MONAIwip | 0 | 归档 / fork | AI Toolkit for Healthcare Imaging |
ROCm-DS:1 个仓库(ROCm Data Science(归档))
活跃非 fork 0 · fork 1 · 归档 1 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| rocOpt | 0 | 归档 / fork | GPU accelerated decision optimization |
amd-enterprise-ai:6 个仓库(AMD 企业 AI 解决方案)
活跃非 fork 6 · fork 0 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| solution-blueprints | 22 | 活跃 | — |
| aim-build | 18 | 活跃 | — |
| amd-eai-apps | 14 | 活跃 | — |
| aim-deploy | 7 | 活跃 | — |
| aim-engine | 7 | 活跃 | — |
| aim-fine-tuning | 0 | 活跃 | AIM Fine-tuning offers tools and starting points to make fine-tuning easy to use, fast, and guaranteed to work in the context of AIMs. |
AMD-AIM:21 个仓库(AMD AI Group, AIM:Radeon / Cloud / GPU MODE)
活跃非 fork 14 · fork 7 · 归档 0 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| reference-kernels | 6 | fork | Official Problem Sets / Reference Kernels for the GPU MODE Leaderboard! |
| inference-skill | 5 | 活跃 | — |
| HIPGenerator | 3 | 活跃 | — |
| ai_sprint_shanghai | 1 | 活跃 | Radeon-adapted Triton workshop |
| InferenceX | 1 | fork | Open Source Continuous Inference Benchmarking Qwen3.5, DeepSeek, GPTOSS - GB200 NVL72 vs MI355X vs B200 vs GB300 NVL72 vs H100 & soon™ TPUv6e/v7/Trainium2/3 |
| PaddleX | 1 | fork | All-in-One Development Tool based on PaddlePaddle |
| Physical_AI_Challenge | 1 | 活跃 | — |
| radeon-cloud-docs | 1 | 活跃 | Radeon Cloud documentation — user guide and API reference (EN / 简体中文) |
| Relax | 1 | fork | An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale |
| sglang | 1 | fork | SGLang is a fast serving framework for large language models and vision language models. |
| TransferQueue | 1 | fork | This is the official **live** mirror of https://gitcode.com/Ascend/TransferQueue. Feel free to contribute! |
| AMD_Developers_Notebooks | 0 | 活跃 | Developer documentation and tutorials |
| Distill-run | 0 | 活跃 | — |
| FlagTree | 0 | fork | FlagTree is a unified compiler supporting multiple AI chip backends for custom Deep Learning operations, which is forked from triton-lang/triton. |
| hf-radeon-gpu-notebooks | 0 | 活跃 | Validated Hugging Face model notebooks adapted for AMD Radeon GPUs and Radeon Cloud local inference. |
| hunyuan-models | 0 | 活跃 | — |
| kfd-vgpu | 0 | 活跃 | Driver-enforced multi-tenant GPU partitioning for consumer AMD GPUs (RDNA3/RDNA4). One physical card → N kernel-confined virtual cards, native to Kubernetes. |
| llm-eval-hub | 0 | 活跃 | A self-hosted evaluation platform for OpenAI-compatible LLM APIs, providing automated accuracy, F1, exact match, latency, error analysis, and regression testing. |
| radeon-cloud-notebooks | 0 | 活跃 | Hands-on notebooks for AMD Radeon Cloud on a single 48GB W7900D, adapted from AI Academy (https://github.com/AMD-ROCm-Internal/ai-academy-internal). |
| sglang-radeon | 0 | 活跃 | Run SGLang on AMD Radeon RDNA3 (gfx110x) GPUs - an out-of-tree plugin |
| zhihu_rednote_articles | 0 | 活跃 | source file of the articles that will be launched on zhihu and rednote |
11.3 三个历史组织(归档)
ROCmSoftwarePlatform:9 个仓库(ROCm 旧组织名,仓库已迁移到 ROCm)
归档 9 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| cub-hip | 6 | 归档 | An implementation of CUB on the ROCM stack. This has been replaced by hipCUB. |
| MIOpenTensile | 3 | 归档 | — |
| Anakin | 2 | 归档 / fork | — |
| roctracer | 1 | 归档 / fork | ROCm Tracer Callback/Activity Library for Performance tracing AMD GPU's |
| .github | 0 | 归档 | ROCm Software Platform organization README |
| actions | 0 | 归档 / fork | — |
| rfcs | 0 | 归档 / fork | PyTorch RFCs (experimental) |
| ROCm | 0 | 归档 / fork | ROCm - Open Source Platform for HPC and Ultrascale GPU Computing |
| torchrec_private | 0 | 归档 | — |
RadeonOpenCompute:4 个仓库(ROCm 旧组织名,仓库已迁移到 ROCm)
归档 4 · 按 Star 数排序
| 仓库 | Star | 状态 | 官方简介 |
|---|---|---|---|
| ROCm_Documentation | 111 | 归档 | Legacy ROCm Software Platform Documentation |
| rdma-perftest | 11 | 归档 | — |
| rfcs | 2 | 归档 | RFCs for changes Requests for ROCm |
| .github | 0 | 归档 | ROCm GitHub Introduction |
12. 参考来源
官方文档
- AMD ROCm documentation
- What is ROCm?
- ROCm Core SDK components
- Transition guide from legacy ROCm release stream
- ROCm 10.1.0 compatibility matrix
- AMD GPU specifications
- AMD GPU architecture documentation
- ROCm programming guide
- HIP programming model
- HIP runtime API
- AMD CLR
- ROCm LLVM compiler infrastructure
- HIPIFY documentation
- MIOpen documentation
- rocBLAS documentation
- hipBLASLt documentation
- Composable Kernel documentation
- RCCL documentation
- rocSHMEM documentation
- hipFile documentation
- ROCprofiler-SDK documentation
- ROCgdb documentation
- ROCdbgapi documentation
- AMD ROCm AI ecosystem
- ATOM documentation
- Primus documentation
- AMD GPU Driver (amdgpu)
- Linux kernel drm/amdgpu documentation
- AMD CDNA architecture
- AMD Instinct MI355X specifications
- AMD Ryzen AI Software
- AMD Quark documentation
- Ryzen AI Software documentation (latest)
- Ryzen AI Linux installation
- Ryzen AI LLM overview
- AMD XDNA architecture
- AMD Ryzen AI 300 Series press release
- AMD Ryzen 8040 Series press release
- AMD Ryzen AI PRO 300 Series
- AMD Ryzen AI Embedded P121i
- AMD Ryzen AI 400 series (CES 2026)
- AMD Ryzen AI Embedded P100/X100
- Introducing FastFlow-SDK
- IRON AI Engine API tutorial (ISCA 2025)
- Linux kernel amdxdna documentation
- AMD Research open-source projects
- AMD AOCL
- AMD ZenDNN
- AMD open-source GenAI models
官方仓库
