Get in Touch
 Duration 21 hours

Course Outline

Overview of Biren GPU Architecture

  • Introduction to Biren and its key use cases.
  • Hardware configuration: cores, memory structure, and compute clusters.
  • Benchmarking against NVIDIA and AMD GPU standards.

Configuring the Biren Development Environment

  • Installing the Biren SDK and runtime components.
  • Explaining the toolchain and compiler workflow.
  • Establishing basic project structures and build procedures.

Developing GPU Code with the Biren Stack

  • Detailed look at thread and block organization models.
  • Managing memory allocation and data transfer operations.
  • Creating kernels and defining launch patterns.

Migrating from CUDA to Biren

  • Techniques for translating CUDA code to Biren syntax.
  • Mapping common APIs and necessary adaptations.
  • Practical labs for code conversion and practice.

Debugging and Performance Profiling

  • Utilizing Biren’s integrated debugger and profiler tools.
  • Pinpointing performance bottlenecks.
  • Analyzing memory access patterns for optimization.

Advanced Optimization Strategies

  • Managing thread scheduling and instruction pipelining.
  • Applying loop unrolling and leveraging shared memory.
  • Fine-tuning kernels for maximum throughput.

Real-World Case Studies and Examples

  • Training models using Biren accelerators.
  • Porting and profiling vision or NLP-based models.
  • Performance comparison against CUDA/NVIDIA solutions.

Summary and Future Directions

Requirements

  • A solid understanding of GPU architecture and parallel processing concepts.
  • Prior experience working with CUDA, OpenCL, or comparable GPU programming frameworks.
  • Familiarity with deep learning libraries such as PyTorch or TensorFlow.

Target Audience

  • HPC developers.
  • AI infrastructure engineers.
  • Specialists in performance optimization.

Testimonials (2)

Related Categories