Iterative Compilation Optimization Based on Metric Learning and Collaborative Filtering

Pass selection and phase ordering are two critical compiler auto-tuning problems. Traditional heuristic methods cannot effectively address these NP-hard problems especially given the increasing number of compiler passes and diverse hardware architectures. Recent research efforts have attempted to address these problems through machine learning. However, the large search space of candidate pass sequences, the large numbers of redundant and irrelevant features, and the lack of training program instances make it difficult to learn models well. Several methods have tried to use expert knowledge to simplify the problems, such as using only the compiler passes or subsequences in the standard levels (e.g., -O1, -O2, and -O3) provided by compiler designers. However, these methods ignore other useful compiler passes that are not contained in the standard levels. Principal component analysis (PCA) and exploratory factor analysis (EFA) have been utilized to reduce the redundancy of feature data. However, these unsupervised methods retain all the information irrelevant to the performance of compilation optimization, which may mislead the subsequent model learning. To solve these problems, we propose a compiler pass selection and phase ordering approach, called Iterative Compilation based on Metric learning and Collaborative filtering (ICMC) . First, we propose a data-driven method to construct pass subsequences according to the observed collaborative interactions and dependency among passes on a given program set. Therefore, we can make use of all available compiler passes and prune the search space. Then, a supervised metric learning method is utilized to retain useful feature information for compilation optimization while removing both the irrelevant and the redundant information. Based on the learned similarity metric, a neighborhood-based collaborative filtering method is employed to iteratively recommend a few superior compiler passes for each target program. Last, an iterative data enhancement method is designed to alleviate the problem of lacking training program instances and to enhance the performance of iterative pass recommendations. The experimental results using the LLVM compiler on all 32 cBench programs show the following: (1) ICMC significantly outperforms several state-of-the-art compiler phase ordering methods, (2) it performs the same or better than the standard level -O3 on all the test programs, and (3) it can reach an average performance speedup of 1.20 (up to 1.46) compared with the standard level -O3.

Active Segregation Dynamics in the Living Cell

In this paper, we bring together our efforts in identifying and understanding nonequilibrium phase segregation driven by active processes in the living cell, with special focus on the segregation of cell membrane components driven by active contractile stresses arising from cortical actomyosin. This also has implications for active segregation dynamics in membraneless regions within the cytoplasm and nucleus (3d). We formulate an active version of the Flory-Huggins theory that incorporates a contribution from fluctuating active stresses. Apart from knitting together some of our past theoretical work in a comprehensive narrative, we highlight some new results, and establish a cor- respondence with recent studies on Active Model B/B+. We point to the many unusual aspects of the dynamics of active phase segregation, such as (i) anomalous growth dynamics, (ii) coarsening accompanied by propulsion and coalescence of domains that exhibit nonreciprocal effects, (iii) seg- regation into mesoscale domains, (iv) emergence of a nonequilibrium phase segregated steady state characterised by strong macroscopic fluctuations (fluctuation dominated phase ordering (FDPO)), and (v) mesoscale segregation even above the equilibrium Tc. Apart from its implications for actively driven segregation of binary fluids, these ideas are at the heart of an Active Emulsion description of the lateral organisation of molecules on the plasma membrane of living cells, whose full molecular elaboration appears elsewhere.

Multifaceted phase ordering kinetics of an antiferromagnetic spin-1 condensate

We study phase domain coarsening in the long time limit after a quench of magnetic field in a quasi one-dimensional spin-1 antiferromagnetic condensate. We observe that the growth of correlation length obeys scaling laws predicted by the two different models of phase ordering kinetics, namely the binary mixture and vector field. We derive regimes of clear realization for both of them. We demonstrate appearance of atypical scaling laws, which emerge in intermediate regions.

Spontaneously coherent orbital coupling of counterrotating exciton polaritons in annular perovskite microcavities

Exciton-polariton condensation is regarded as a spontaneous macroscopic quantum phenomenon with phase ordering and collective coherence. By engineering artificial annular potential landscapes in halide perovskite semiconductor microcavities, we experimentally and theoretically demonstrate the room-temperature spontaneous formation of a coherent superposition of exciton-polariton orbital states with symmetric petal-shaped patterns in real space, resulting from symmetry breaking due to the anisotropic effective potential of the birefringent perovskite crystals. The lobe numbers of such petal-shaped polariton condensates can be precisely controlled by tuning the annular potential geometry. These petal-shaped condensates form in multiple orbital states, carrying locked alternating π phase shifts and vortex–antivortex superposition cores, arising from the coupling of counterrotating exciton-polaritons in the confined circular waveguide. Our geometrically patterned microcavity exhibits promise for realizing room-temperature topological polaritonic devices and optical polaritonic switches based on periodic annular potentials.

Stability of Ordered B2-βO and Disordered BCC-β Phases in Tial - A First Principles Study

Either at higher temperatures or when a certain alloying element content is exceeded, γ-TiAl alloys contain the β phase (bcc) or its ordered derivate βo (B2). The relatively soft β phase can facilitate hot deformation, but βo is detrimental for creep strength and ductility. Thus, knowledge about βo→β phase transformation is desirable. Surprisingly, even for the binary Ti-Al system it is under discussion whether the ordered βo phase exists. Also, the effect of alloying elements on the β phase ordering is still unclear. In the present work the ordering of the β phase in binary Ti-(39,42,45)Al and ternary Ti-42Al-2X alloys (X=Fe, Cr, Nb, Ta, Mo) which was experimentally investigated by neutron and high energy X-ray diffraction is compared with the results of first principles calculations using density functional theory. Except for Cr the experimentally determined and the predicted behavior correspond.