Smart Logic Tech
Posted 2y ago

数学库开发工程师

Smart Logic Tech
Beijing or Shanghai or Xi'an or Chengdu or Hangzhou
OnsiteFull Time
Responsibilities
  • developing library
  • designing algorithms
  • optimizing performance
Requirements
  • Design and implement low-level math libraries for custom chips
  • Strong C/C++ and assembly skills
  • Deep linear algebra and numerical analysis, and experience with heterogeneous parallelism
Technical tools mentioned
CC++assemblyCUDAHIPOpenCLcuBLASMKL

Job description

岗位描述:
1、定制化数学库开发: 针对项目指定的自研芯片,从零到一设计并开发底层的核心数学运算库(如稠密/稀疏线性代数库 BLAS/Sparse、FFT 库等)。
2、大规模并行算法设计: 深入剖析目标异构硬件的并发执行模型,突破传统数值算法的物理瓶颈,设计高效的多层级并行计算方案。
3、软硬件协同与系统级调优: 站在系统全局视角,设计高效的计算与数据搬运流水线。通过精细的任务切分、存储层次复用与异步并发调度,最大化隐藏访存延迟,充分利用硬件的极限吞吐量(FLOPS)与内存带宽。
4、数值稳定性与架构演进: 结合数值分析理论,保障极高并发、混合精度计算环境下的数值准确性与稳定性。与芯片架构(IC)、编译器及上层应用团队深度协作,为下一代芯片的软硬件协同设计(Co-design)提供关键数据支撑与架构建议

岗位要求:
1、计算机、电子信息、应用数学等相关专业,本科及以上学历
2、精通 C/C++,具备优秀的汇编语言阅读编写能力者优先,精通底层与异构编程,有丰富的大规模数据并行计算设计及开发经验,能够独立完成复杂算法的并行化重构。
3、具备优秀的线性代数与数值分析基础,深刻理解各类矩阵运算、求解器的数学推导及底层逻辑。
4、熟悉 SIMD/SIMT 或 VLIW 执行模型,透彻理解多级存储层次(Memory Hierarchy)、缓存一致性协议、异步数据搬运以及指令流水线隐藏技术。
5、跟踪行业最新数学库优化策略,并形成技术方案落地。

加分项:
掌握主流异构编程模型(如cuda、hip、opencl)者优先。
参与过国产芯片/异构加速卡的底层算子库(如类似 cuBLAS, MKL 的替代品)从零搭建者优先
具有极强的极客精神,曾在没有完善 Profiling 工具的情况下,通过打点、仿真器(Simulator)或底层汇编成功进行过极致性能 Debug 者优先

About Smart Logic Tech

Designs and develops high-performance reconfigurable processor chip architecture.

Year founded
2016
Employees
500
Organization type
Private
Latest investment
Raised $100.00M Private Equity (2022) — led by China International Capital Corporation (CICC), Jinshi Investment, Zhongxin Juyuan
Headquarters
CN

Similar jobs

Math Library Developer roles near Beijing, Beijing
1y
Save
Mark Applied
Hide
2026校招-芯片数学库开发工程师-北京
Beijing, Beijing, China
OnsiteFull Time
Smart Logic Technology
Smart Logic Technology: Designs and develops system-on-chip semiconductor solutions.
Familiarity with math library algorithms (linear algebra, matrix ops, numerical computing); proficiency in Python/C/C++; knowledge of computer architectures (x86/ARM); experience with parallel/distributed computing preferred; familiarity with LAPACK, MKL, CUDA preferred.
LAPACK, MKL, CUDA, Python, C, C++
1y
Save
Mark Applied
Hide
实习生-芯片数学库开发工程师
Beijing, Beijing, China
OnsiteInternship
SmartLogic Tech
SmartLogic Tech: Developing AI-driven solutions for numerical calculation and life sciences.
Familiarity with basic math functions, linear algebra, matrix operations and numerical computation; experience with LAPACK, MKL, CUDA preferred; knowledge of x86/ARM architecture; proficiency in Python/C/C++; parallel or distributed storage experience preferred.
LAPACK, MKL, CUDA, Python, C, C++