Challenges in using accelerators via programming languages
summary
Covers the rise of hardware accelerators used to handle large-scale data processing and examines alternative programming models for those devices. Discusses challenges these accelerators and programming approaches pose specifically for the Eclipse ecosystem.
AI-written summary · about
chapters (25)
- 0:00E eclipse Summit India 2017
- 2:13Programmers job was easy
- 2:47But processors were changing
- 3:40Exploiting parallelism
- 5:20Processor performance growth
- 5:52How was the programmer affected?
- 6:23Changes that happened
- 7:37Also, power plays a role
- 9:12Optimizing for power
- 10:35The programmers challenge?
- 10:50Use a library like pthreads
- 11:48Sample code
- 12:57OpenMP to the rescue
- 14:36GPU Architecture evolution
- 16:25Why Massively Parallel Processors • A quiet revolution and potential build-up
- 16:56CPU and GPU are designed very differently
- 17:56CPUs: Latency Oriented Design
- 18:32GPUs: Throughput Oriented Design
- 18:51How does programming change?
- 19:05Model - Master Slave model
- 19:34Arrays of Parallel Threads
- 20:01Thread Blocks: Scalable Cooperation
- 20:40Programmers Challenges
- 21:18The promise of OpenAcc
- 21:47For future - OpenAcc
related talks
01h 21mPanel+Closing
01h 04mThe challenge of managing parallelism and data-movement for tensor computations on GPUs
01h 16mA New Golden Age for Computer Architecture
17mThe Real Challenge for RISC-V Vector Processors
10mA Programming Model for Efficient Inter-Kernel Control-Flow on Memory-Mapped Near-Data(…)
45mHackathon Presentations
50mGeometry on Steroids: Practical Experience Porting Heavy Computation to GPU
39mTechnology mix: Decoding the neural code -analyzing data form (...)