PPT-Warp-Level Divergence in GPUs:

Characterization Impact and Mitigation Ping Xiang Yi Yang Huiyang Zhou 1 The 20th IEEE International Symposium On High Performance Computer Architecture Orlando Florida . Dept. of Electrical and Computer Engineering North Carolina State UniversityRaleigh, NC, USA {pxiang, hzhou}@ncsu.edu *Dept. of Computing Systems Architecture NEC Laboratories AmericaPrinceton, NJ, Instructor Notes. This lecture deals with how work groups are scheduled for execution on the compute units of devices. Also explain the effects of divergence of work items within a group and its negative effect on performance. to Improve GPGPU Performance. Rachata. . Ausavarungnirun. Saugata. . Ghose, . Onur. . Kayiran, Gabriel H. . Loh. . Chita . Das, . Mahmut. . Kandemir. , . Onur. . Mutlu. Overview of This Talk. Problem: . T. Rogers, M O’Conner, and T. . Aamodt. MICRO 2012. Goal. Understand the relationship between schedulers (warp/wavefront) and locality behaviors . Distinguish between inter-wavefront and intra-wavefront locality. Farzad Khorasani. , . Rajiv Gupta. , . Laxmi. N. . Bhuyan. University of California Riverside. Scalable SIMD-Efficient Graph Processing on GPUs. Graph Processing. Building blocks of data analytics.. T. Rogers, M O’Conner, and T. . Aamodt. MICRO 2012. Goal. Understand the relationship between schedulers (warp/wavefront) and locality behaviors . Distinguish between inter-wavefront and intra-wavefront locality. Farzad Khorasani. , . Rajiv Gupta. , . Laxmi. N. . Bhuyan. University of California Riverside. Scalable SIMD-Efficient Graph Processing on GPUs. Graph Processing. Building blocks of data analytics.. Collaborative Context Collection. Farzad Khorasani, Rajiv Gupta, . Laxmi. N. . Bhuyan. UC Riverside. The 48th Annual IEEE/ACM International Symposium on . Microarchitecture (MICRO), 2015. One PC for the SIMD group (warp):. Reducing Branch Divergence in GPU Programs. . T. D. Han and T. Abdelrahman. GPGPU 2011. Reading. T. D. Han and T. Abdelrahman, “Reducing Branch Divergence in GPGPU Programs,” GPGPU 2011. Goal. Improve the utilization of the SIMD core. Niladrish Chatterjee. Mike O’Connor. Gabriel H. . Loh. Nuwan. . Jayasena. Rajeev . Balasubramonian. Irregular GPGPU Applications. Conventional GPGPU workloads access vector or matrix-based data structures. K. ainz. Overview. About myself. Motivation. GPU hardware and system architecture. GPU programming languages. GPU programming paradigms. Pitfalls and best practice. Reduction and tiling examples. State-of-the-art . panel discussions. HSF and its role in performance? . V: This is the question that we should keep in mind throughout the panel discussion . What is Computational Efficiency for you?. Holistic Performance Assessment? . Profiling, AWS Cluster. Synchronization. Ideal case for parallelism: . no resources shared between threads. no communication between threads. . Many algorithms that require just a little bit of resource sharing can still be accelerated by massive parallelism of GPU. Goes around to every warp . and issue if ready (R). If warp is not ready (W), . skip and issue next ready warp. Issue: Warps all run at the same speed,. potentially all reaching memory access. phase together and stalling..