Analyzing CUDA workloads using a detailed GPU simulator

By Ali Bakhoda, George L. Yuan, Wilson W. L. Fung, Henry Wong, Tor M. Aamodt. IEEE International Symposium on Performance Analysis of Systems and Software, 2009

1,701 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture

View Paper: https://www.semanticscholar.org/paper/2d6f002477015469075954c6748a1a85af352c94

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Hardware and Architecture

Citations in the corpus

OWL: cooperative thread array aware scheduling techniques for improving GPGPU performance (2013, methodology): “We execute these applications on GPGPU-Sim, which simulates the baseline architecture described in Table 1.”
Processing Data Where It Makes Sense: Enabling In-Memory Computation (2019, background): “, gem5 [216], zsim [217], gem5-gpu [218], GPGPUSim [219]) to provide a robust environment that can evaluate how various PIM architectures affect the entire compute stack, and can allow designers to identify memory, workload, and system characteristics that affect the efficiency of PIM execution.”
A characterization of the Rodinia benchmark suite with comparison to contemporary CMP workloads (2010, methodology): “developed GPGPU-Sim [2] and use it to analyze various CUDA programs.”
Kernel Fusion: An Effective Method for Better Power Efficiency on Multithreaded GPU (2010, methodology): “As one of the earliest GPU simulator, GPGPU-Sim provides a cycle-accurate simulation model of contemporary GPUs running CUDA and OpenCL workloads [9].”
Accel-Sim: An Extensible Simulation Framework for Validated GPU Modeling (2018, methodology): “We compare Accel-Sim against the state-of-the-art NVIDIA simulation model from GPGPU-Sim 3.x [1], [5].”
MASK: Redesigning the GPU Memory Hierarchy to Support Multi-Application Concurrency (2018, methodology): “To evaluate MASK , we model the NVIDIA Maxwell architecture [98], and the TLB-fill bypassing, cache bypassing, and memory scheduling mechanisms in MASK , using the Mosaic simulator [17], which is based on GPGPU-Sim 3.2.2 [20].”
An analytical model for a GPU architecture with memory-level and thread-level parallelism awareness (2009, methodology): “[6] recently implemented a GPU simulator and analyzed the performance of CUDA applications using the simulation output.”
Pegasus: Coordinated Scheduling for Virtualized Accelerator-based Systems (2011, background): “However, even for today’s platforms, the majority of applications do not occupy the entire accelerator [2, 18].”
Ramulator: A Fast and Extensible DRAM Simulator (2016, neutral): “GPGPU-Sim (2009) [3] GDDR3, GDDR5 McSimA+ (2013) [2] DDR3 Integrated gem5 (2014) [9] DDR3, ∗ LPDDR3, ∗ WIO∗”
Cache-Conscious Wavefront Scheduling (2012, methodology): “This small difference is a result of variability in the GPGPU-Sim memory system that SAGCS does not take into account.”
Accelerating CUDA graph algorithms at maximum warp (2011, background): “NN [8], another neural network application, processes separate data streams in each thread.”
A detailed and flexible cycle-accurate Network-on-Chip simulator (2013, methodology): “An earlier version of BookSim was used to model the on-chip network for the GPGPU-sim simulator [11].”
GPUWattch: enabling energy optimizations in GPGPUs (2013, methodology): “Although GPU performance models, such as GPGPU-Sim [4], Multi2Sim [34] and MacSim [1], have enabled performance oriented research on branch divergence [11,12,27], memory bandwidth pressure [5, 32], and so forth, similar eﬀorts to investigate and optimize GPU energy-eﬃciency problems have been…”
A quantitative study of irregular programs on GPUs (2012, background): “study non-graphics kernels on a GPU simulator [4].”
Multi2Sim: A simulation framework for CPU-GPU computing (2012, methodology): “Simulation performance has been also evaluated for an architectural simulation on GPGPUSim, an NVIDIA-based GPU simulator [10].”
Analyzing Machine Learning Workloads Using a Detailed GPU Simulator (2018, methodology): “The developers of GPGPU-Sim gave us some help by explaining the process they followed to getting GPGPU-Sim working [4]: ﬁrst, they validated individual instructions one-by-one by comparing execution on real GPU hardware with execution on GPGPU-Sim.”
Demystifying GPU microarchitecture through microbenchmarking (2010, neutral): “No caching effects were observed.”
Buddy Compression: Enabling Larger Memory for Deep Learning and HPC Workloads on GPUs (2019, methodology): “It is two orders of magnitude faster than GPGPUSim [81], enabling the simulation of longer programs (right, linear regression lines drawn).”
Zorua: A holistic approach to resource virtualization in GPUs (2016, methodology): “N-Queens Solver (NQU) [11] [5] S:10496-47232 (T:64-288)”
Convolution engine: balancing efficiency & flexibility in specialized computing (2013, methodology): “264 SAD algorithm [28] using GPUGPUSim simulator [6] with GPUWattch energy model [21].”