- Research Article
56
- 10.1016/j.jpdc.2019.09.012
SIMD programming using Intel vector extensions
- Sep 28, 2019
- Journal of Parallel and Distributed Computing
- Hossein Amiri + 1 more +1
SIMD programming using Intel vector extensions
Satnav SDRs present many benefits in terms of flexibility and configurability. However, due to the high bandwidth signals involved in satnav SDR processing, the software must be highly optimized for the host platform in order to achieve acceptable runtimes. Modules such as sample decoding, carrier replica generation, carrier wipeoff, and correlation are computationally intensive components that benefit from accelerations. An architecture has been developed which allows for the components that must be high performance to be swapped out with versions that make use of accelerators such as FPGAs, GPUs, and vectorized instructions in the CPU (Gunawardena, 2021). Other modules are left in higher level language for ease of use. This differs from some previous projects which do not provide the same level of configurability and granularity that allows for modules to be swapped out without changing code. For the purposes of this paper, we will target general CPUs using Single Instruction Multiple Data (SIMD), multithreading, and bitwise methods. Both SIMD and multithreading accelerations have been previously discussed as applied to sample decoding (Reich et al., 2022). The target system for this paper is x86 64, extended with the Streaming SIMD Extensions (SSE) and Advanced Vector eXtensions (AVX) instruction sets. These 128-bit and 256-bit operations are used in conjunction with bitwise methodologies to perform many operations per instruction in parallel. In addition, these tasks are divided among several threads running concurrently. None of these parallelization techniques are invoked by default and must be specifically written into the program. This paper discusses the process of including these accelerations as well as extensive benchmarking results as compared to a naive C++ implementation. The results of this paper are intended to provide guidance on future optimizations to satnav SDR receivers.
SIMD programming using Intel vector extensions
SIMD programming using Intel vector extensions
Performance evaluation of implicit and explicit SIMDization
Performance evaluation of implicit and explicit SIMDization
A Review of SIMD Multimedia Extensions and their Usage in Scientific and Engineering Applications
The volume and complexity of data processed by today's personal computers are increasing exponentially, placing incredible demands on the microprocessors. In the meantime, computing performance that can be achieved by increasing the clock speed of a microprocessor is reaching to physical limits thus making the architectural solutions more prominent. Due to this an important architectural feature is added to recent microprocessors, single instruction multiple data (SIMD), which is a set of instructions that can speed up an application performance by allowing basic operation to be performed on multiple data elements in parallel with fewer instructions. The SIMD computational technique was introduced in the IA-32 Intel® architecture with MMX technology and then further enhanced with Intel's introduction of streaming SIMD extensions (SSE), SSE 2 (SSE2) and SSE 3 (SSE3). Although programming using these SIMD extensions enables software to achieve higher performance, several exiting scientific applications are not affected. This paper gives an overview of SIMD multimedia extensions. The features of these extensions are introduced. Available methods for programming with multimedia instruction sets are discussed. It also reviews recent trends to use multimedia extensions to accelerate many applications such as multimedia, scientific and engineering applications, and argues for further use in other significant computationally intensive applications.
Read moreV-PFORDelta: Data Compression for Energy Efficient Computation of Time Series
Chip multiprocessors (CMPs) and heterogeneous architectures have become predominant in all market segments, from embedded to high performance computing. These architectures exacerbate on-chip data requirements, creating additional pressure on the memory subsystem. Consequently, efficient utilization of on-chip memory space becomes critical for data intensive applications. A promising means of addressing this challenge is to use an effective compression method to reduce the data transmitted along the memory hierarchy. In this paper we present V-PFORDelta, a real-time vectorized integer differential compression method for memory bound applications. We evaluate the effectiveness of our SIMD (Single Instruction Multiple Data stream) based compression method on an industrial hydrological time series data processing kernel. We analyzed both Streaming SIMD Extensions (SSE) and Advanced Vector Extensions 2 (AVX2) versions of the compression method. Results show that the performance and energy efficiency can be improved up to a factor of 3.1 and 8.2, respectively. The proposed method not only outperforms the uncompressed SIMD implementations of the hydrological kernel, but also reduces the data storage requirements by a factor of 1.56x to 3.38x, depending on the analyzed dataset.
Read moreSIMD code generation in data-parallel programming
Today';s desktop PCs feature a variety of parallel processing units. Developing applications that exploit this parallelism is a demanding task, and a programmer has to obtain detailed knowledge about the hardware for efficient implementation. CGiS is a data-parallel programming language providing a unified abstraction for two parallel processing units: graphics processing units (GPUs) and the vector processing units of CPUs. The CGiS compiler framework fully virtualizes the differences in capability and accessibility by mapping an abstract data-parallel programming model on those targets. The applicability of CGiS for GPUs has been shown in previous work; this work focuses on applying the abstract programming model of CGiS to CPUs with SIMD (Single Instruction Multiple Data) instruction sets. We have identified, adapted and implemented a set of program analyses to expose and access the available parallelism. The code generation phase is based on selected optimization algorithms tailored to SIMD code generation. Via code generation profiles, it is possible to adapt the code generation strategy to different target architectures. To assess the effectiveness of our approach, we have implemented backends for the two most widespread SIMD instruction sets, namely Intel';s Streaming SIMD Extensions and Freescale';s AltiVec. Additionally, we integrated a prototypical backend for the Cell Broadband Engine as an example for a multi-core architecture. Our experimental results show excellent average performance gains by a factor of 3 compared to standard scalar C++ implementations and underline the viability of this approach: real-world applications can be implemented easily with CGiS and result in efficient code.
Read moreA high performance FFT library with single instruction multiple data (SIMD) architecture
Fast Fourier Transform (FFT) is the basis of Digital Signal Processing (DSP). In this paper, a high performance FFT library using radix-2 decimation in frequency (DIF) algorithm is presented which is well suited for SIMD architecture. SIMD architecture microprocessors, such as Intel and AMD, allow parallel floating point operations on contiguous data in memory. A 128-point FFT based radix-2 DIF algorithm is implemented on the Intel architecture. All arithmetic operations in FFT are optimized by SSE assembly. Twiddle factors and binary reverse array are also optimized for SIMD architecture. The library is implemented using C and Intel Streaming SIMD Extensions (SSE) assembly instructions. The performance comparison with Fastest Fourier Transform in the West (FFTW) library shows that the proposed FFT library is faster.
Read moreNeat SIMD: Elegant vectorization in C++ by using specialized templates
Most of today's processors provide strong support for single instruction multiple data (SIMD, or vector) operations. But because modern compilers often fail to generate optimal SIMD code, explicitly written SIMD code is still the most performant. However, native SIMD code has some major disadvantages, mainly that it is quite cumbersome, that it is not easily ported to different instruction sets, and that typically multiple versions of similar code have to be written; e.g. a scalar, a streaming SIMD extensions (SSE) and an advanced vector extensions (AVX) version. Neat SIMD resolves all these disadvantages by using specialized templates in C++ that enable one neat version of easily portable code to be used for both scalar and SIMD instructions without sacrificing any performance in comparison to native SIMD code. Neat SIMD has no dependencies to any external library nor is it restricted to any specific instruction set or to a newer C++ version. The examples shown here use SSE and AVX. This publication is written such that all crucial code elements and implementation hints are given here. An interested software engineer can use those code elements, reconstruct the remaining code with ease, and start to use effective Neat SIMD code for own purposes.
Read morePerformance Evaluation of Matrix-Matrix Multiplications Using Intel's Advanced Vector Extensions (AVX)
Performance Evaluation of Matrix-Matrix Multiplications Using Intel's Advanced Vector Extensions (AVX)
SIMD Implementation of the Aho-Corasick Algorithm using Intel AVX2
The Aho-Corasick (AC) algorithm is a multiple pattern exact string-matching algorithm proposed by Alfred V. Aho and Margaret J. Corasick. It is used to locate all occurrences of a finite set of patterns within an input text simultaneously. The AC algorithm is in the heart of many applications including digital forensics such as digital signatures demonstrating the authenticity of a digital message or document, full text search (utility programs such as grep, awk and sed of Unix systems), information retrieval (biological sequence analysis and gene identification), intrusion detection systems (IDS) in computer networks like SNORT, web filtering, spam filters, and antimalware solutions (virus scanner). In this paper we present a vectorized version of the AC algorithm designed with the use of packed instructions based on the Intel streaming SIMD (Single Instruction Multiple Data) extensions AVX2 (Advanced Vector Extensions 2.0) technology. This paper shows that the vectorized AC algorithm reduces significantly the time matching process comparing to the implementation of the original AC algorithm.
Read moreHadamard Transform Improvement for HEVC using Intel AVX-512
High Efficiency Video Coding (HEVC) doubles the data compression ratio compared to previous generation compression technology, Moving Picture Expert Group-Advanced Video Codec (MPEG-AVC/H.264) without sacrificing the image quality. However, this superior compression comes at the cost of more computation payload resulting in longer time for encoding and decoding. This work proposes the vectorization on HEVC data heavy computation algorithm, Hadamard Transform or Sum of Absolute Transform Difference (SATD) and Sum of Absolute Difference (SAD) to achieve optimized compression performance. Single Instruction Multiple Data (SIMD) acceleration will be based on the Intel AVX-512 (Advanced Vector Extension) Instruction Set Architecture (ISA). Since HEVC supports more coding tree block (CTB) sizes, SATD and SAD algorithms eventually become more complex compared to AVC. As a result, SATD and SAD algorithms with various block sizes will be subjected to SIMD acceleration. We provide performance evaluation based on different SIMD ISA and without SIMD implementation on HEVC SATD and SAD and found that AVX-512 optimized implementation performed faster when compared to non- optimized SATD and SAD but showed signs of reduced performance when compared to SSE optimized SATD and SAD.
Read morePerformance optimization of block matching in 3D echocardiography
The introduction of real time 3D echocardiography allows for deformation tracking in three dimensions, without the limitations of 2D methods. However, the processing needs of 3D methods are much higher. The aim of the study was to optimize the performance of 3D block matching by using a Single Instruction Multiple Data (SIMD) model, a technique employed to achieve data level parallelism. Two implementations of SIMD have been tested. The first is based on Streaming SIMD Extensions (SSE), the second uses CUDA, a SIMD architecture proposed by NVIDIA, which is available on several graphics cards. The proposed methods have been validated on synthetic and patient data. With the use of SIMD architecture, the overall processing time is significantly reduced, thus making 3D speckle tracking feasible in a clinical setting. Apart from the implemented sum of square differences (SAD), other matching criteria (e.g. sum of squared differences, normalized cross correlation) can be implemented efficiently (especially on the GPU) thus further improving the accuracy of the block matching process.
Read moreParallel Seismic Modeling Based on OpenMP+AVX and Optimization Strategy
This paper describes parallel simulation of the memory/computation-intensive acoustic wave equation with CPU template buffer optimization. Considering the 8-core CPU shared storage platform as an example, we obtain a one-time speed-up ratio of 6.7× compared with the serial program by using a coarse-grained OpenMP parallel scheme. Then, data is vectorized on the template buffer using the single instruction-multiple data (SIMD) technique to further exploit the computing potential of the CPUs. We apply an 8-channel parallel vector to simulate seismic wavefields with the 256-bit advanced vector extensions (AVX) instruction set. This increases the computing bandwidth, thus eliminating a significant volume of the computing instructions and obtaining a secondary speed-up ratio of 3–7×. In addition, we use 32-byte data alignment, shortest data direction vectorization, and loop tiling optimization algorithm to achieve faster program execution. Finally, we analyze the factors affecting the secondary speed-up of AVX through three-dimensional modeling experiments with the salt model. The results indicate that the memory, cache, and register can better cooperate with each other and the speed-up is increased by optimizing the AVX algorithm.
Read moreInvestigating large integer arithmetic on Intel Xeon Phi SIMD extensions
In this paper, we investigate the Single Instruction Multiple Data (SIMD) extensions of the Intel Xeon Phi, and we analyze the benefits and limitations of its unique architecture in large integer arithmetic applications. The first generation Intel Xeon Phi, named Knights Corner, is a state-of-the-art many core coprocessor with up to 61 cores. Each Knights Corner core has 512-bit SIMD vectors, which are supported by a unique instruction set architecture, incompatible with existing SIMD schemes (SSE, AVX etc.). The second generation Intel Xeon Phi, Knights Landing, will include the standard 512-bit AVX512 SIMD extensions. We implement a multiplication scheme for large integers, which leverages the aforementioned SIMD extensions. Using this multiplication scheme, we discuss the SIMD extensions' performance for both Intel Xeon Phi generations. Preliminary results indicate that the Knights Corner SIMD speedup of large integer multiplication is limited by the absence of specific instructions that typically appear in common SIMD architectures. The emulation on Knights Landing (which includes backwards-compatible AVX512 extensions), on the contrary, shows that large integer multiplication can indeed benefit by the presence of 512-bit vectors, for commonly used 1024- and 2048-bit operands, compared to publicly available large arithmetic libraries.
Read moreSSE Vectorized and GPU Implementations of Arakawa's Formula for Numerical Integration of Equations of Fluid Motion
The numerical method presented by Arakawa in 1966[3] implements a finite difference scheme of the Jacobian for the solution of the equation of motion for two-dimensional incompressible flows, which diminishes nonlinear computational instability and permits long-term numerical integrations. This paper presents an efficient implementation of Arakawa's formula using vectorized Streaming SIMD Extension (SSE) and Advanced Vector Extension (AVX) instructions. Additionally, we have improved the performance of memory access in the code. Performance measurements show that the vectorized implementation is close to two times more efficient compared to an implementation without SSE. The AVX version will in the near future further improve the vectorized performance with an estimated factor of up to 1.8. Finally we compare our results to an implementation on a general purpose graphics processor (GPGPU) and to auto-vectorization by two compilers.
Read moreSIMD 명령어 기반 HEVC RExt 복호화기 고속화
In this paper, we introduce the fast decoding method with the SIMD (Single Instruction Multiple Data) instructions for HEVC RExt (High Efficiency Video Coding Range Extensions). Several tools of HEVC RExt such as intra prediction, interpolation, inverse-quantization, inverse-transform, and clipping modules can be classified as the proper modules for applying the SIMD instructions. In consideration of bit-depth increasement of RExt, intra prediction, interpolation, inverse-quantization, inversetransform, and clipping modules are accelerated by SSE (Streaming SIMD Extension) instructions. In addition, we propose effective implementations for interpolation filter, inverse-quantization, and clipping modules by utilizing a set of AVX2 (Advanced Vector eXtension 2) instructions that can use 256 bits register. The evaluation of the proposed methods were performed on the private HEVC RExt decoder developed based on HM 16.0. The experimental results show that the developed RExt decoder reduces 12% average decoding time, compared with the conventional sequential method. Keyword : HEVC, RExt, HEVC Range extensions, SIMD, Parallelization 특집논문 (Special Paper) 방송공학회논문지 제20권 제2호, 2015년 3월 (JBE Vol. 20, No. 2, March 2015) http://dx.doi.org/10.5909/JBE.2015.20.2.224 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) 목정수 외 3인 : SIMD 명령어 기반 HEVC RExt 복호화기 고속화 225 (Jung-Soo Mok et al.: SIMD Instruction-based Fast HEVC RExt Decoder)
Read more