Simulation study shows cold TLB misses in reverse address translation dominate latency for small collectives in multi-GPU pods, causing up to 1.4x degradation, while larger ones see diminishing returns.
Rogers, Evan Schneider, Jean-Luc Vay, and P
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 6roles
background 1polarities
background 1representative citing papers
EMP maps importance scores to effective sample size N_eff and prunes the lowest N - N_eff components, with a derived lower bound on retained effective mass and upper bound on loss increase.
Lagrangian tracers show mixing with low-entropy seeds drives most condensation in cluster cores; magnetic fields cause earlier divergence, higher vorticity, lower Mach numbers, and slower cold-cloud motion via tension.
PICO is a benchmarking framework for collective operations that decouples portable setup from platform execution, supplies reference MPI implementations, and shows default choices can be up to 5x slower with up to 44% end-to-end training time reductions in simulator replays.
HieraSparse delivers a hierarchical semi-structured sparse KV attention system that achieves 1.2x KV compression and 4.57x decode attention speedup versus prior unstructured sparsity methods at equivalent sparsity, plus up to 1.85x prefill speedup and 1.37x/1.77x speedups with magnitude pruning and
Aurora reached 1.01 EF/s FP64 HPL and 11.64 EF/s HPL-MxP through locality-aware mapping, CPU-GPU pipelining, mixed-precision orchestration, and hybrid resilience on a large Intel GPU-based system.
citing papers explorer
-
Analyzing Reverse Address Translation Overheads in Multi-GPU Scale-Up Pods
Simulation study shows cold TLB misses in reverse address translation dominate latency for small collectives in multi-GPU pods, causing up to 1.4x degradation, while larger ones see diminishing returns.
-
Effective Model Pruning: Measure The Redundancy of Model Components
EMP maps importance scores to effective sample size N_eff and prunes the lowest N - N_eff components, with a derived lower bound on retained effective mass and upper bound on loss increase.
-
XMAGNET -- Stir before serving: a Lagrangian perspective on mixing-driven condensation in the intracluster medium
Lagrangian tracers show mixing with low-entropy seeds drives most condensation in cluster cores; magnetic fields cause earlier divergence, higher vorticity, lower Mach numbers, and slower cold-cloud motion via tension.
-
PICO: Performance Insights for Collective Operations
PICO is a benchmarking framework for collective operations that decouples portable setup from platform execution, supplies reference MPI implementations, and shows default choices can be up to 5x slower with up to 44% end-to-end training time reductions in simulator replays.
-
HieraSparse: Hierarchical Semi-Structured Sparse KV Attention
HieraSparse delivers a hierarchical semi-structured sparse KV attention system that achieves 1.2x KV compression and 4.57x decode attention speedup versus prior unstructured sparsity methods at equivalent sparsity, plus up to 1.85x prefill speedup and 1.37x/1.77x speedups with magnitude pruning and
-
Sustaining Exascale Performance: Lessons from HPL and HPL-MxP on Aurora
Aurora reached 1.01 EF/s FP64 HPL and 11.64 EF/s HPL-MxP through locality-aware mapping, CPU-GPU pipelining, mixed-precision orchestration, and hybrid resilience on a large Intel GPU-based system.