AI存储架构设计工程师 上海 全职 芯片板块 职位描述 岗位职责：1. AI多级存储分层架构设计 - 统筹寄存器、Shared Memory、L0/L1/L2缓存、片上全局内存、主机系统内存分层存储架构； - 定义与设计GPGPU MMU、Cache架构； - 定义与设计异步拷贝引擎cp.async、张量访存、批量DMA、MMIO寄存器访问通路，区分volatile/mmio特殊访存语义硬件实现； - 针对大模型训练/推理场景优化带宽、延迟、冲突化解、读写合并、预取策略，平衡算力与存储吞吐。2. GPU内存一致性模型（MCM）架构定义与RTL约束 - 基于Memory Consistency Model，完成GPGPU片上存储、访存流水线、多级缓存、DMA、IOMMU/SMMU整体一致性架构设计； - 定义Memory Proxy（generic/async/tensormap/fabric）硬件通路隔离、跨Proxy同步、proxy fence屏障实现规范； - 梳理Scope（CTA/Cluster/GPGPU/System）硬件可见域、VMID/SID/PASID地址翻译隔离逻辑； - 负责Release/Acquire、Fence-SC、Barrier、异步访存cp.async等同步原语硬件架构设计。2. AI加速卡虚拟化I/O存储架构设计 - 设计SR-IOV/MIG硬件虚拟化存储隔离方案，基于Stream ID、VMID实现虚拟机间DMA内存硬件隔离； - 完成SVA共享虚拟地址、PASID子流ID、IOMMU页表翻译、Stage1/Stage2两级地址映射架构； - 解决多虚拟机、多MIG切片并发访存冲突、数据竞争、混合尺寸访存原子性问题，输出虚拟化存储隔离规范。3. 数据竞争、原子性与同步硬件方案落地
ASIC-PD team is hiring both junior and senior engineers, whose work scope is physical design from RTL to GSDII: design quality check, synthesis, formal check, partitioning, constraint (for both design and process), async check, timing analysis/fixing/signoff, also
NVIDIA has been transforming computer graphics, PC gaming, and accelerated computing for more than 25 years. It’s a unique legacy of innovation that’s fueled by great technology—and amazing people. Today, we’re tapping into the unlimited potential
Company: Qualcomm China Job Area:Engineering Group, Engineering Group Software Engineering General Summary: Role Summary We are seeking a AI solution Software Engineer focused on end-to-end latency optimization across heterogeneous compute IPs (GPU, NPU, ISP, CPU). You
NVIDIA has been transforming computer graphics, PC gaming, and accelerated computing for more than 25 years. It’s a unique legacy of innovation that’s fueled by great technology—and amazing people. Today, we’re tapping into the unlimited potential
About New Energy Nexus New Energy Nexus (NEX) is an international NGO that supports climate solutions and builds a more equitable clean energy economy. Our laser focus is on clean energy entrepreneurs: providing them with funds,