Loop Fusion Research Articles

Cd loop fusion enhances the immunogenicity and the potential transmission blocking activity of Plasmodium falciparum generative cell specific 1 (GCS1) antigen

TBVs are suggested to inhibit parasite transmission from humans to Anopheles mosquitoes. For the transmission of Plasmodium parasite, a variety of factors are included in gametes fusion phase. In this step, conserved male-specific generative cell specific 1 antigen is necessary for fusion of cytoplasmic membranes of micro- and macro-gametocytes and zygot formation. The partial blocking activities of elicited antibodies against either the HAP2-GCS1 domain or the cd loop of this antigen have been recorded to hinder the transmission of Plasmodium species in Anopheles mid-gut. Thus, the objective of the present study was to investigate if the cd loop-fusion can enhance the quantity and quality of humoral and cellular immune responses against Plasmodium falciparum GCS1 in comparison to non-fusion antigen (without cd loop), in the adjuvanted and non-adjuvanted mouse groups. The immunogenicity of two constructs of P. falciparum generative cell specific 1 antigen, a fusion protein composed of cd loop and HAP2-GCS1 domain (cd-HAP) and another recombinant PfGCS1 containing solo HAP2-GCS1 domain (HAP2) were assessed to impede Plasmodium gametocytes integration before zygote formation. The antibodies profiling, titer, and avidity of induced antibodies were measured by the immunized mice sera, and the released cytokines (IL-5, TNF, and INF-γ) were analyzed in the supernatants of stimulated splenocytes. Furthermore, the inhibitory potency of the elicited antibodies against HAP2 and cd-HAP was measured during oocyst development by Standard Membrane Feeding Assay (SMFA). The comparative results in the present study showed the higher titer of IgG antibodies and IgG2a subclass, avidity, and transmission-reducing activity (TRA = 72.5 %) when mice were immunized by cd-HAP rather than HAP2. Moreover, our findings confirmed intensified Th1-directed immune responses in group 4 received cd-HAP/Poly(I:C). These findings declared the potential ability of cd loop fusion (cd-HAP) to upsurge humoral and cellular immune responses. However, the immune responses may switch to stronger Th1-type using alternative formulations. Explicitly, the cd-HAP-based vaccine may enhance the overall efficiency of immune responses and present a promising implementation in aiming malaria transmission.

Connes fusion of spinors on loop space

The loop space of a string manifold supports an infinite-dimensional Fock space bundle, which is an analog of the spinor bundle on a spin manifold. This spinor bundle on loop space appears in the description of two-dimensional sigma models as the bundle of states over the configuration space of the superstring. We construct a product on this bundle that covers the fusion of loops, i.e. the merging of two loops along a common segment. For this purpose, we exhibit it as a bundle of bimodules over a certain von Neumann algebra bundle, and realize our product fibrewise using the Connes fusion of von Neumann bimodules. Our main technique is to establish novel relations between string structures, loop fusion, and the Connes fusion of Fock spaces. The fusion product on the spinor bundle on loop space was proposed by Stolz and Teichner as part of a programme to explore the relation between generalized cohomology theories, functorial field theories, and index theory. It is related to the pair of pants worldsheet of the superstring, to the extension of the corresponding smooth functorial field theory down to the point, and to a higher-categorical bundle on the underlying string manifold, the stringor bundle.

Acceleration of a Production-Level Unstructured Grid Finite Volume CFD Code on GPU

Due to the complex topological relationship, poor data locality, and data racing problems in unstructured CFD computing, how to parallelize the finite volume method algorithms in shared memory to efficiently explore the hardware capabilities of many-core GPUs has become a significant challenge. Based on a production-level unstructured CFD software, three shared memory parallel programming strategies, atomic operation, colouring, and reduction were designed and implemented by deeply analysing its computing behaviour and memory access mode. Several data locality optimization methods—grid reordering, loop fusion, and multi-level memory access—were proposed. Aimed at the sequential attribute of LU-SGS solution, two methods based on cell colouring and hyperplane were implemented. All the parallel methods and optimization techniques implemented were comprehensively analysed and evaluated by the three-dimensional grid of the M6 wing and CHN-T1 aeroplane. The results show that using the Cuthill–McKee grid renumbering and loop fusion optimization techniques can improve memory access performance by 10%. The proposed reduction strategy, combined with multi-level memory access optimization, has a significant acceleration effect, speeding up the hot spot subroutine with data races three times. Compared with the serial CPU version, the overall speed-up of the GPU codes can reach 127. Compared with the parallel CPU version, the overall speed-up of the GPU codes can achieve more than thirty times the result in the same Message Passing Interface (MPI) ranks.

METHODOLOGY FOR DEVELOPING A HIGH-SPEED COMPILER BASED ON THE MODIFIED LOOP FUSION OPTIMIZATION METHOD: MODELS AND TOOLS FOR ITS IMPLEMENTATION

In connection with the development of information technologies, the complexity of programming languages and, accordingly, applications developed using them, program optimization is of particular importance. In the process of optimization, the program is improved by reducing the code size, complexity, using less memory and provides a reduction in execution time without changing the internal function. In practice, optimization is implemented using compilers and their functions. Taking into account the above, the aim of the article was to develop a methodology for studying various aspects of building a high-speed compiler with a modified loop fusion optimization method, as well as models and tools for its implementation. In the course of the research, the features of its design using the modified loop fusion optimization method are outlined, descriptions of the flowchart of the modified loop fusion algorithm and the logical flowchart of the compiler development stages are given. As a result of the work, a compiler based on the modified loop fusion optimization method is proposed, using loop reversal to ensure their “legitimate” and profitable merging, which reduces the execution time of the program while maintaining its correctness. The efficiency of the proposed compiler is shown by comparing the compilation times of the test program obtained using it and using the well-known compiler x86-64 gcc 4.7.1.

FPGA-Based Design for Accelerating 3D Convolutional Neural Networks

Three-dimensional convolutional neural networks (3D CNNs) have been shown to be excellent for classification in human action recognition, hyperspectral image classification and many other aspects. However, 3D CNNs often contain convolution kernels of different sizes and a large number of operations, and the numerous operations make 3D CNNs consume a lot of time when executing inference on resource-constrained terminals. Aiming at these two points, this paper proposes a 3D CNN accelerator design based on loop optimization and weight reuse. The design adopts loop tiling, loop unrolling and loop fusion to optimize the direct convolution form of the 3D convolution operation; and guided by the optimized form of the 3D convolution operation, a module that can accelerate different 3D convolution operations is designed; and a memory access method based on weight reuse is also designed to reduce the memory access time of features and weights. Experimental results show that the proposed accelerator not only supports convolution kernels of different sizes, but also has a performance density of 1.59, which outperforms most existing accelerators.

Deep reinforcement learning in loop fusion problem

Loops’ execution time and resource consumption are one of the interest points and vital issues in the field of appraising complex scientific or computational algorithms. This issue caused the proposal of Loop performance optimization techniques such as fusion. In the literature, loop fusion merges the loops by taking into account a set of properties associated with the loops or the system on which the resulting code will be executed. The number of these factors and their interactions on the one hand, and the high runtime of available comprehensive approaches, on the other hand, reveals the need for a new method that could be concerned for further progress in solving this NP-hard problem. For the first time, Deep Reinforcement Learning Loop Fusion (DRLLF) advanced to be an ideal solution for the challenge in this article. For the proposed framework, a particular matrix is configured as the inputs of a deep neural network based on the information of the problem, namely data dependencies, data reuse, loops’ types, and computer system’s register size. These randomly generated matrixes are used in the training phase by reinforcement learning to get the imperative experience on predicting a profitable distribution over loops’ various fusion orders. In the evaluations performed, the presented algorithm was able to achieve the same or better performance in terms of speedup rate, comparing with the methods under study, approximately averaged in 7.36 percent better results. The considerable improvement observed in the results, besides the low run time, proves the comprehensiveness and superiority of this approach.

On the Transformation Optimization for Stencil Computation

Stencil computation optimizations have been investigated quite a lot, and various approaches have been proposed. Loop transformation is a vital kind of optimization in modern production compilers and has proved successful employment within compilers. In this paper, we combine the two aspects to study the potential benefits some common transformation recipes may have for stencils. The recipes consist of loop unrolling, loop fusion, address precalculation, redundancy elimination, instruction reordering, load balance, and a forward and backward update algorithm named semi-stencil. Experimental evaluations of diverse stencil kernels, including 1D, 2D, and 3D computation patterns, on two typical ARM and Intel platforms, demonstrate the respective effects of the transformation recipes. An average speedup of 1.65× is obtained, and the best is 1.88× for the single transformation recipes we analyze. The compound recipes demonstrate a maximum speedup of 1.92×.

Code Optimizations for Parallelization of Programs using Data Dependence Identifier

In a Parallelizing Compiler, code transformations help to reduce the data dependencies and identify parallelism in a code. In our earlier paper, we proposed a model Data Dependence Identifier (DDI), in which a program P is represented as graph GP . Using 􀀀􀀀 , we could identify data dependencies in a program and also perform transformations like dead code elimination and constant propagation. In this paper, we present algorithms for loop invariant code motion, live range analysis, node splitting and loop fusion code transformations using DDI in polynomial time.

Effective Loop Fusion in Polyhedral Compilation Using Fusion Conflict Graphs

Polyhedral auto-transformation frameworks are known to find efficient loop transformations that maximize locality and parallelism and minimize synchronization. While complex loop transformations are routinely modeled in these frameworks, they tend to rely on ad hoc heuristics for loop fusion. Although there exist multiple loop fusion models with cost functions to maximize locality and parallelism, these models involve separate optimization steps rather than seamlessly integrating with other loop transformations like loop permutation, scaling, and shifting. Incorporating parallelism-preserving loop fusion heuristics into existing affine transformation frameworks like Pluto, LLVM-Polly, PPCG, and PoCC requires solving a large number of Integer Linear Programming formulations, which increase auto-transformation times significantly. In this work, we incorporate polynomial time loop fusion heuristics into the Pluto-lp-dfp framework. We present a data structure called the fusion conflict graph (FCG), which enables us to efficiently model loop fusion in the presence of other affine loop transformations. We propose a clustering heuristic to group the vertices of the FCG, which further enables us to provide three different polynomial time greedy fusion heuristics, namely, maximal fusion , typed fusion , and hybrid fusion , while maintaining the compile time improvements of Pluto-lp-dfp over Pluto. Our experiments reveal that the hybrid fusion model, in conjunction with Pluto’s cost function, finds efficient transformations that outperform PoCC and Pluto by mean factors of 1.8× and 1.07×, respectively, with a maximum performance improvement of 14× over PoCC and 2.6× over Pluto.

Dynamic workload prediction and distribution in numerical modeling of solidification on multi‐/manycore architectures

SummaryThis work is a part of the global tendency to use modern computing systems for modeling the phase‐field phenomena. The main goal of this article is to improve the performance of a parallel application for the solidification modeling, assuming the dynamic intensity of computations in successive time steps when calculations are performed using a carefully selected group of nodes in the grid. A two‐step method is proposed to optimize the application for multi‐/manycore architectures. In the first step, the loop fusion is used to execute all kernels in a single nested loop and reduce the number of conditional operators. These modifications are vital to implementing the second step, which includes an algorithm for the dynamic workload prediction and load balancing across cores of a computing platform. Two versions of the algorithm are proposed—with the 1D and 2D maps used for predicting the computational domain within the grid. The proposed optimizations allow increasing the application performance significantly for all tested configurations of computing resources. The highest performance gain is achieved for two Intel Xeon Platinum 8180 CPUs, where the new code based on the 2D map yields the speedup of up to 2.74 times, while the usage of the proposed method with the 2D map for a single KNL accelerator permits reducing the execution time up to 1.91 times.

The Application of Robust Least Squares Method in Frequency Lock Loop Fusion for Global Navigation Satellite System Receivers

The tracking accuracy of a traditional Frequency Lock Loop (FLL) decreases significantly in a complex environment, thus reducing the overall performance of a satellite receiver. In order to ensure high tracking accuracy of a receiver in a complex environment, this paper proposes a new tracking loop combining the vector FLL (VFLL) with a robust least squares method, which accurately matches the weights of received signals of different qualities to ensure high positioning accuracy. The weights of received signals are selected at the signal level, not at the observation level. In this paper, the ranges of strong and weak signals of the loop are determined according to the different expressions of the distribution function at different signal strengths, and the concept of loop segmentation is introduced. The segmentation results of the FLL are taken as a basis of the weight selection, and then combined with the Institute of Geodesy and Geophysics (IGGIII) weight function to obtain the equivalent weight matrix; the experiments are conducted to prove the advantages of the proposed method over the traditional methods. The experimental results show that the proposed VFLL tracking method has strong denoising capability under both normal- signal and harsh application environment conditions. Accordingly, the proposed model has a promising application perspective.

Using an evolutionary approach based on shortest common supersequence problem for loop fusion

In the literature, loop fusion is an effective optimization technique which tries to enhance parallelizing compilers’ performance via memory hierarchy management, and all its competing criteria create an NP-hard problem. This paper proposes an evolutionary algorithm that aims to achieve a profitable loop order which maximizes fusion taking into account register size, parallelism and data reuse advancement. Besides, this method preserves prerequisite relations between the loops by encoding each distinct loop sequence as the shortest common supersequence (SCS) of the related dependence graph. Regarding the related optimization methods that only focus on fusion, this set of metrics, an evolutionary algorithm and also the shortest common supersequence problem have not been considered before in this area. Despite all the envisaged complexities, experimental results confirm the accuracy and advantage of the proposed approach. But due to evolutionary methods effect on raising the compilation time, the proposed algorithm is only applicable when this issue is not prominent, in comparison with the quality of the outcome.

Hybrid Distributed/Shared Memory Model for the RI-MP2 Method in the Fragment Molecular Orbital Framework.

The general distributed data interface (GDDI) that was developed for the fragment molecular orbital (FMO) method is combined with the shared memory OpenMP parallel middleware to support a threading multilevel parallelism. First, GDDI partitions [logical] compute nodes into groups, which are statically or dynamically assigned to different fragments. A small number of processes are created on each compute node. Each process subsequently spawns multiple threads for the actual computation. The performance of the hybrid GDDI/OpenMP approach relative to the pure GDDI model was examined in terms of the FMO/RI-MP2 method; that is, the second-order Moller-Plesset (MP2) correlation energy was evaluated using the resolution-of-the-identity (RI) and the FMO approximations. The GDDI and OpenMP workload balances are handled by an arithmetic progression and a loop fusion, respectively. Other OpenMP properties, such as threadprivate or shared memory, are combined with the low memory demand of the RI two-electron integrals to enhance the performance. Benchmark calculations demonstrate that because the hybrid parallel model can make use of multiprocessor resources more efficiently than the regular distributed memory-based GDDI model, calculations for small to large water clusters containing 139-2165 molecules and an ionic liquid cluster exhibit node linear scaling and speedups of a factor of 10×.

Automated Tiling of Unstructured Mesh Computations with Application to Seismological Modeling

Sparse tiling is a technique to fuse loops that access common data, thus increasing data locality. Unlike traditional loop fusion or blocking, the loops may have different iteration spaces and access shared datasets through indirect memory accesses, such as A[map[i]]—hence the name “sparse.” One notable example of such loops arises in discontinuous-Galerkin finite element methods, because of the computation of numerical integrals over different domains (e.g., cells, facets). The major challenge with sparse tiling is implementation—not only is it cumbersome to understand and synthesize, but it is also onerous to maintain and generalize, as it requires a complete rewrite of the bulk of the numerical computation. In this article, we propose an approach to extend the applicability of sparse tiling based on raising the level of abstraction. Through a sequence of compiler passes, the mathematical specification of a problem is progressively lowered, and eventually sparse-tiled C for-loops are generated. Besides automation, we advance the state-of-the-art by introducing a revisited, more efficient sparse tiling algorithm; support for distributed-memory parallelism; a range of fine-grained optimizations for increased runtime performance; implementation in a publicly available library, SLOPE; and an in-depth study of the performance impact in Seigen, a real-world elastic wave equation solver for seismological problems, which shows speed-ups up to 1.28× on a platform consisting of 896 Intel Broadwell cores.

Using Arm’s scalable vector extension on stencil codes

Data-level parallelism is frequently ignored or underutilized. Achieved through vector/SIMD capabilities, it can provide substantial performance improvements on top of widely used techniques such as thread-level parallelism. However, manual vectorization is a tedious and costly process that needs to be repeated for each specific instruction set or register size. In addition, automatic compiler vectorization is susceptible to code complexity, and usually limited due to data and control dependencies. To address some of these issues, Arm recently released a new vector ISA, the scalable vector extension (SVE), which is vector-length agnostic (VLA). VLA enables the generation of binary files that run regardless of the physical vector register length. In this paper, we leverage the main characteristics of SVE to implement and optimize stencil computations, ubiquitous in scientific computing. We show that SVE enables easy deployment of textbook optimizations like loop unrolling, loop fusion, load trading or data reuse. Our detailed simulations using vector lengths ranging from 128 to 2048 bits show that these optimizations can lead to performance improvements over straightforward vectorized code of up to 1.57$$\times$$. In addition, we show that certain optimizations can hurt performance due to reduced arithmetic intensity and instruction overheads, and provide insight useful for compiler optimizers.

Push versus pull-based loop fusion in query engines

AbstractDatabase query engines use pull-based or push-based approaches to avoid the materialization of data across query operators. In this paper, we study these two types of query engines in depth and present the limitations and advantages of each engine. Similarly, the programming languages community has developed loop fusion techniques to remove intermediate collections in the context of collection programming. We draw parallels between databases (DB) and programming language (PL) research by demonstrating the connection between pipelined query engines and loop fusion techniques. Based on this connection, we propose a new type of pull-based engine, inspired by a loop fusion technique, which combines the benefits of both approaches. Then, we experimentally evaluate the various engines, in the context of query compilation, for the first time in a fair environment, eliminating the biasing impact of ancillary optimizations that have traditionally only been used with one of the approaches. We show that for realistic analytical workloads, there is no considerable advantage for either form of pipelined query engine, as opposed to what recent research suggests. Also, by using micro-benchmarks, which demonstrate certain edge cases on which one approach or the other performs better, we show that our proposed engine dominates the existing engines by combining the benefits of both.

Mining non-redundant recurrent rules from a sequence database

Many methods have been studied for mining sequential patterns from a sequence database. In particular, Lo et al. have proposed the notion of recurrent rules and an algorithm called NR3 for mining them. Recurrent rules can express temporal constraints such as Whenever a series of precedent events eventually a series of consequent events occurs, and they are useful in various domains, especially in software specification and verification. Although the algorithm NR3 and its successor BOB for mining non-redundant recurrent rules have been given by Lo et al. mining recurrent rules still requires considerable computational costs. In this paper, we propose a new algorithm, called LF-NR3, to make NR3 more efficient, which is based on a familiar program transformation loop fusion. We apply the loop fusion technique to NR3, thereby simplifying the operations of the original mining algorithm. We also make use of a hash-based data structure to reduce loads of the manipulation of sequences repeatedly required in mining recurrent rules. We show the effectiveness of our proposed method based on experimental results of some datasets used in the literature.

Locality Transformations for Nested Recursive Iteration Spaces

There has been a significant amount of effort invested in designing scheduling transformations such as loop tiling and loop fusion that rearrange the execution of dynamic instances of loop nests to place operations that access the same data close together temporally. In recent years, there has been interest in designing similar transformations that operate on recursive programs, but until now these transformations have only considered simple scenarios: multiple recursions to be fused, or a recursion nested inside a simple loop. This paper develops the first set of scheduling transformations for nested recursions: recursive methods that call other recursive methods. These are the recursive analog to nested loops. We present a transformation called recursion twisting that automatically improves locality at all levels of the memory hierarchy, and show that this transformation can yield substantial performance improvements across several benchmarks that exhibit nested recursion.

Locality Transformations for Nested Recursive Iteration Spaces

There has been a significant amount of effort invested in designing scheduling transformations such as loop tiling and loop fusion that rearrange the execution of dynamic instances of loop nests to place operations that access the same data close together temporally. In recent years, there has been interest in designing similar transformations that operate on recursive programs, but until now these transformations have only considered simple scenarios: multiple recursions to be fused, or a recursion nested inside a simple loop. This paper develops the first set of scheduling transformations for nested recursions: recursive methods that call other recursive methods. These are the recursive analog to nested loops. We present a transformation called recursion twisting that automatically improves locality at all levels of the memory hierarchy, and show that this transformation can yield substantial performance improvements across several benchmarks that exhibit nested recursion.

A high performance data parallel tensor contraction framework: Application to coupled electro-mechanics

The paper presents aspects of implementation of a new high performance tensor contraction framework for the numerical analysis of coupled and multi-physics problems on streaming architectures. In addition to explicit SIMD instructions and smart expression templates, the framework introduces domain specific constructs for the tensor cross product and its associated algebra recently rediscovered by Bonet et al. (2015, 2016) in the context of solid mechanics. The two key ingredients of the presented expression template engine are as follows. First, the capability to mathematically transform complex chains of operations to simpler equivalent expressions, while potentially avoiding routes with higher levels of computational complexity and, second, to perform a compile time depth-first or breadth-first search to find the optimal contraction indices of a large tensor network in order to minimise the number of floating point operations. For optimisations of tensor contraction such as loop transformation, loop fusion and data locality optimisations, the framework relies heavily on compile time technologies rather than source-to-source translation or JIT techniques. Every aspect of the framework is examined through relevant performance benchmarks, including the impact of data parallelism on the performance of isomorphic and nonisomorphic tensor products, the FLOP and memory I/O optimality in the evaluation of tensor networks, the compilation cost and memory footprint of the framework and the performance of tensor cross product kernels. The framework is then applied to finite element analysis of coupled electro-mechanical problems to assess the speed-ups achieved in kernel-based numerical integration of complex electroelastic energy functionals. In this context, domain-aware expression templates combined with SIMD instructions are shown to provide a significant speed-up over the classical low-level style programming techniques.