Improving performance portability for GPU-specific Open CL kernels on multi-core/many-core CPUs by analysis-based transformations

OpenCL is an open heterogeneous programming framework. Although OpenCL programs are func- tionally portable, they do not provide performance portability, so code transformation often plays an irreplaceable role. When adapting GPU-specific OpenCL kernels to run on multi-core/many-core CPUs, coarsening the thread granularity is necessary and thus has been extensively used. However, locality concerns exposed in GPU-specific OpenCL code are usually inherited without analysis, which may give side-effects on the CPU performance. Typi- cally, the use of OpenCL＇s local memory on multi-core/many-core CPUs may lead to an opposite performance effect, because local-memory arrays no longer match well with the hardware and the associated synchronizations are costly. To solve this dilemma, we actively analyze the memory access patterns using array-access descriptors derived from GPU-specific kernels, which can thus be adapted for CPUs by （1） removing all the unwanted local-memory arrays together with the obsolete barrier statements and （2） optimizing the coalesced kernel code with vectorization and locality re-exploitation. Moreover, we have developed an automated tool chain that makes this transformation of GPU-specific OpenCL kernels into a CPU-friendly form, which is accompanied with a scheduler that forms a new OpenCL runtime. Experiments show that the automated transformation can improve OpenCL kernel performance on a multi-core CPU by an average factor of 3.24. Satisfactory performance improvements axe also achieved on Intel＇s many-integrated-core coprocessor. The resultant performance on both architectures is better than or comparable with the corresponding OpenMP performance.

Optimal precursors of double-gyre regime transitions with an adjoint-free method

In this paper, we find the optimal precursors which can cause double-gyre regime transitions based on conditional nonlinear optimal perturbation (CNOP) method with Regional Ocean Modeling System (ROMS). Firstly, we simulate the multiple-equilibria regimes of double-gyre circulation under different viscosity coefficient and obtain the bifurcation diagram, then choose two equilibrium states (called jet-up state and jet-down state) as reference states respectively, propose Principal Component Analysis-based Simulated Annealing (PCASA) algorithm to solve CNOP-type initial perturbations which can induce double-gyre regime transitions between jet-up state and jet-down state. PCASA algorithm is an adjoint-free method which searches optimal solution randomly in the whole solution space. In addition, we investigate CNOP-type initial perturbations how to evolve with time. The results show:(1) the CNOP-type perturbations present a two-cell structure, and gradually evolves into a three-cell structure at predictive time;(2) by superimposing CNOP-type perturbations on the jet-up state and integrating ROMS, double-gyre circulation transfers from jet-up state to jet-down state, and vice versa, and random initial perturbations don't cause the transitions, which means CNOP-type perturbations are the optimal precursors of double-gyre regime transitions;(3) by analyzing the transition process of double-gyre regime transitions, we find that CNOP-type initial perturbations obtain energy from the background state through both barotropic and baroclinic instabilities, and barotropic instability contributes more significantly to the fast-growth of the perturbations. The optimal precursors and the dynamic mechanism of double-gyre regime transitions revealed in this paper have an important significance to enhance the predictability of double-gyre circulation.