The best LLMs (o1, o3-mini-high) classify GPU kernels as compute- or bandwidth-bound from source with 64% accuracy, versus 100% when given profiled arithmetic intensity.
Critical exponents and fluctuations at BEC in a 2D harmonically trapped ideal gas
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
The critical properties displayed by an ideal 2D Bose gas trapped in a harmonic potential are determined and characterized in an exact numerical fashion. Beyond thermodynamics, addressed in terms of the global pressure and volume which are the appropriate variables of a fluid confined in a non-uniform harmonic potential, the density-density correlation function is also calculated and the corresponding correlation length is found. Evaluation of all these quantities as Bose-Einstein condensation (BEC) is approached manifest its critical continuous phase transition character. The divergence of the correlation length as the critical temperature is reached, unveils the expected spatial scale invariance proper of a critical transition. The logarithmic singularities of this transition are traced back to the non-analytic behavior of the thermodynamic variables at vanishing chemical potential, which is the onset of BEC. The critical exponents associated with the ideal BEC transition in the 2D inhomogeneous fluid reveals its own universality class.
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Can Large Language Models Predict Parallel Code Performance?
The best LLMs (o1, o3-mini-high) classify GPU kernels as compute- or bandwidth-bound from source with 64% accuracy, versus 100% when given profiled arithmetic intensity.