Delivering SDK library components for CPU, GPU, DSP, DL accelerator

MulticoreWare engineering consulting services include low level software like Kernel development, porting kernels, profiling & optimizing kernels, memory planning, instructions pipelining to middle level software like CNN framework parsers, CNN model format converters, model compression & optimization etc. With vast array of hardware micro-architecture (CPU, GPU, DSP, DL HW accelerator) awareness and its experience to deliver consulting services uniquely positions MulticoreWare to take up any general purpose as well as custom hardware accelerator that are targeted for AI/ML/DL applications.

Examples of areas research, development, and solutions delivered include:

• Kernel optimization with SIMD intrinsic
• Kernel designs to achieve CPP (Cycles per Pixels) spec
• Kernel optimization for various NN kernels (data-bound & compute-bound) on various DSP micro-architecture
• OpenVX, OpenCV, OpenMP and OpenCL libraries to various DL accelerator platforms
• End-to-end pipeline creation for embedded computer vision (CV) functions
• Parallel computing, pipeline stalls, Runtime engine optimization etc.
• End-to-end inference pipeline (Inception-V3, Yolo-v2/v3 etc.)
• High Performance workload benchmarking