PerfXLab, with its core team from the Chinese Academy of Sciences, is dedicated to the research and development of computing software stack technology, gathering three major fields: HPC, AI, and RISC-V. PerfXLab is an international leading accelerated computer technology solution provider, enabling computing chips, scientific computing, and high-performance computing application software companies. PerfXLab has a decade of technical precipitation in the field of high-performance computing and is committed to the development of arithmetic basic software. Even in the era when the computing technology system is dominated by international giants, we still insist on developing software for the underlying and fundamental technologies because we firmly believe that China must build an independent and controllable computing technology system and that computing hardware technology and computing infrastructure software are the infrastructures of the new generation of information technology. Today, we have built up our strengths and are actively working with major domestic computing hardware partners to meet the challenges of the new generation of computing technology.
PerfXAPI integrates all the necessary industry-standard technologies for building computing hardware platforms and takes full advantage of the performance of the latest processors and gas pedals. Supported technologies include PerfMPL, oneAPI, OpenCL, SYCL, ONNX runtime, OpenCV, and more.
PerfMPL mainly contains key mathematical libraries such as BLAS, FFT, IPP, MATH, and deep learning libraries, and is widely adapted to ARM, RISC-V, x86 CPU processors, and builds and optimizes core computing library components for NPU, GPU, DPU, and other accelerator card hardware.
PerfXPy is a high performance heterogeneous Python computing platform for data scientists.
HPC SDK industry solutions integrate the company's core technology, PerfXAPI, with an extensive ecosystem of domain-specific optimization libraries, and a fully tested, supported, and maintained set of software products. The result is to help the HPC application ecosystem easily use endless amounts of computing power while achieving efficient and green computing power costs.