Compiler optimizations play a crucial role in achieving high processor performance by improving the efficiency of the machine code they generate through better algorithms and strategies.These optimizations ultimately ...Compiler optimizations play a crucial role in achieving high processor performance by improving the efficiency of the machine code they generate through better algorithms and strategies.These optimizations ultimately increase program execution efficiency.Over the past few decades,numerous compiler optimizations have been integrated into compilers,addressing research topics such as instruction scheduling and register allocation.However,selecting the optimal combination of compiler optimization options from a vast number of possibilities remains a challenging task.We proposed a Fast Compiler Autotuning Framework(FCAF),leveraging the systematic application of Design of Experiments(DoE)methods to rapidly narrow down the search space,aimed at efficiently enhancing the effectiveness of auto-tuning optimizations.FCAF achieves performance improvements comparable to state of the art tuning tools while significantly reducing tuning time on LoongArch and X86 platforms.Moreover,we used the orthogonal array design method to select a universal set of compilation options and applied these to the SPEC CPU 2006 benchmark suite.When compared to the GCC-O2 baseline,it increases by 3.89%for integer tasks and by 15.27%for floating-point tasks in benchmark scores.Further validation with real-world HPC applications confirms consistent performance gains.展开更多
A sliding mode and active disturbance rejection control(SM-ADRC)was employed to regulate the speed of a permanent magnet synchronous motor(PMSM).The major advantages of the proposed control scheme are that it can main...A sliding mode and active disturbance rejection control(SM-ADRC)was employed to regulate the speed of a permanent magnet synchronous motor(PMSM).The major advantages of the proposed control scheme are that it can maintain the original features of ADRC and make the parameters of ADRC transition smoothly.The proposed control scheme also ensures speed control accuracy and improves the robustness and anti-load disturbance ability of the system.Moreover,through the analysis of a d-axis current output equation,a novel current-loop SM-ADRC is presented to improve the system’s dynamic performance and inner ability of anti-load disturbance.Results of a simulation and experiments show that the improved sliding-mode ADRC system has the advantages of fast response,small overshoot,small steady-state error,wide speed range and high control accuracy.It shows that the system has strong anti-interference ability to reduce the influence of variations in rotational inertia,load and internal parameters.展开更多
An asymmetric MOSFET-C band-pass filter （BPF） with on chip charge pump auto-tuning is presented. It is implemented in UMC （United Manufacturing Corporation） 0.18 μm CMOS process technology. The filter system with...An asymmetric MOSFET-C band-pass filter （BPF） with on chip charge pump auto-tuning is presented. It is implemented in UMC （United Manufacturing Corporation） 0.18 μm CMOS process technology. The filter system with auto-tuning uses a master-slave technique for continuous tuning in which the charge pump outputs 2.663 V, much higher than the power supply voltage, to improve the linearity of the filter. The main filter with third order low-pass and second order high-pass properties is an asymmetric band-pass filter with bandwidth of 2.730-5.340 MHz. The in-band third order harmonic input intercept point （ⅡP3） is 16.621 dBm, with 50Ω as the source impedance. The input referred noise is about 47.455 μVrms. The main filter dissipates 3.528 mW while the auto-tuning system dissipates 2.412 mW from a 1.8 V power supply. The filter with the auto-tuning system occupies 0.592 mm2 and it can be utilized in GPS （global positioning system） and Bluetooth systems.展开更多
CC’s(Cloud Computing)networks are distributed and dynamic as signals appear/disappear or lose significance.MLTs(Machine learning Techniques)train datasets which sometime are inadequate in terms of sample for inferrin...CC’s(Cloud Computing)networks are distributed and dynamic as signals appear/disappear or lose significance.MLTs(Machine learning Techniques)train datasets which sometime are inadequate in terms of sample for inferring information.A dynamic strategy,DevMLOps(Development Machine Learning Operations)used in automatic selections and tunings of MLTs result in significant performance differences.But,the scheme has many disadvantages including continuity in training,more samples and training time in feature selections and increased classification execution times.RFEs(Recursive Feature Eliminations)are computationally very expensive in its operations as it traverses through each feature without considering correlations between them.This problem can be overcome by the use of Wrappers as they select better features by accounting for test and train datasets.The aim of this paper is to use DevQLMLOps for automated tuning and selections based on orchestrations and messaging between containers.The proposed AKFA(Adaptive Kernel Firefly Algorithm)is for selecting features for CNM(Cloud Network Monitoring)operations.AKFA methodology is demonstrated using CNSD(Cloud Network Security Dataset)with satisfactory results in the performance metrics like precision,recall,F-measure and accuracy used.展开更多
An active-RC low-pass filter of 5MHz cutoff frequency with a tuning architecture is proposed. It is implemented in 0. 18μm standard CMOS technology. The accuracy of the tuning system is improved to be within （ - 1.2...An active-RC low-pass filter of 5MHz cutoff frequency with a tuning architecture is proposed. It is implemented in 0. 18μm standard CMOS technology. The accuracy of the tuning system is improved to be within （ - 1.24%, ＋ 2.16%） in measurement. The chip area of the tuning system is only a quarter of that of the main-filter. After tuning is completed, the tuning system will be turned off automatically to save power and to avoid interference. The in-band 3rd order harmonic input intercept point （IIP3） is larger than 16. ldBm, with 50Ω as the source impedance. The input referred noise is about 36μVrms The measured group delay variation of the filter between 3 and 5MHz is only 24ns,and the filter power consump- tion is 3.6roW. This filter with the tuning system is realized easily and can be used in many wireless low-IF receiver applications, such as global position systems （GPS）, global system for mobile communications （GSM） and code division multiple access （CDMA） chips.展开更多
In the field of digital image processing,denoising is one of the basic problems.The challenges faced in image denoising are detecting impulse noise and designing a suitable filter.In this paper,we propose a methodolog...In the field of digital image processing,denoising is one of the basic problems.The challenges faced in image denoising are detecting impulse noise and designing a suitable filter.In this paper,we propose a methodology to remove the random impulse noise on the color image using a novel switching median filter.By using this novel technique,the occurrence of color artifacts has been avoided after noise removal which depends on auto-tuning threshold detection and a vector-type median filter noise remover.In the proposed technique,the random valued impulse noises with uniform distribution have been dealt with switching median filter.L_(2)Norm is employed to calculate the distribution distance rather than L1 Norm which is used to identify the optimal threshold value for auto-tuning filter.The switching auto-tuning detector automatically tunes the noisy pixels based on distance information of pixels distribution.The Normalized Mean Square Error(NMSE)is found to decrease for L_(2)Norm when compared with L1 Norm.The Peak Signal to Noise Ratio(PSNR)value and True Positive Rate(TPR)value improved with L_(2)Norm signifying effective noise removal.The efficiency of the present method is verified by conducting experiments on digital images.展开更多
文摘Compiler optimizations play a crucial role in achieving high processor performance by improving the efficiency of the machine code they generate through better algorithms and strategies.These optimizations ultimately increase program execution efficiency.Over the past few decades,numerous compiler optimizations have been integrated into compilers,addressing research topics such as instruction scheduling and register allocation.However,selecting the optimal combination of compiler optimization options from a vast number of possibilities remains a challenging task.We proposed a Fast Compiler Autotuning Framework(FCAF),leveraging the systematic application of Design of Experiments(DoE)methods to rapidly narrow down the search space,aimed at efficiently enhancing the effectiveness of auto-tuning optimizations.FCAF achieves performance improvements comparable to state of the art tuning tools while significantly reducing tuning time on LoongArch and X86 platforms.Moreover,we used the orthogonal array design method to select a universal set of compilation options and applied these to the SPEC CPU 2006 benchmark suite.When compared to the GCC-O2 baseline,it increases by 3.89%for integer tasks and by 15.27%for floating-point tasks in benchmark scores.Further validation with real-world HPC applications confirms consistent performance gains.
基金Project(2011AA11A10102) supported by the High-tech Research and Development Program of China
文摘A sliding mode and active disturbance rejection control(SM-ADRC)was employed to regulate the speed of a permanent magnet synchronous motor(PMSM).The major advantages of the proposed control scheme are that it can maintain the original features of ADRC and make the parameters of ADRC transition smoothly.The proposed control scheme also ensures speed control accuracy and improves the robustness and anti-load disturbance ability of the system.Moreover,through the analysis of a d-axis current output equation,a novel current-loop SM-ADRC is presented to improve the system’s dynamic performance and inner ability of anti-load disturbance.Results of a simulation and experiments show that the improved sliding-mode ADRC system has the advantages of fast response,small overshoot,small steady-state error,wide speed range and high control accuracy.It shows that the system has strong anti-interference ability to reduce the influence of variations in rotational inertia,load and internal parameters.
文摘An asymmetric MOSFET-C band-pass filter （BPF） with on chip charge pump auto-tuning is presented. It is implemented in UMC （United Manufacturing Corporation） 0.18 μm CMOS process technology. The filter system with auto-tuning uses a master-slave technique for continuous tuning in which the charge pump outputs 2.663 V, much higher than the power supply voltage, to improve the linearity of the filter. The main filter with third order low-pass and second order high-pass properties is an asymmetric band-pass filter with bandwidth of 2.730-5.340 MHz. The in-band third order harmonic input intercept point （ⅡP3） is 16.621 dBm, with 50Ω as the source impedance. The input referred noise is about 47.455 μVrms. The main filter dissipates 3.528 mW while the auto-tuning system dissipates 2.412 mW from a 1.8 V power supply. The filter with the auto-tuning system occupies 0.592 mm2 and it can be utilized in GPS （global positioning system） and Bluetooth systems.
文摘CC’s(Cloud Computing)networks are distributed and dynamic as signals appear/disappear or lose significance.MLTs(Machine learning Techniques)train datasets which sometime are inadequate in terms of sample for inferring information.A dynamic strategy,DevMLOps(Development Machine Learning Operations)used in automatic selections and tunings of MLTs result in significant performance differences.But,the scheme has many disadvantages including continuity in training,more samples and training time in feature selections and increased classification execution times.RFEs(Recursive Feature Eliminations)are computationally very expensive in its operations as it traverses through each feature without considering correlations between them.This problem can be overcome by the use of Wrappers as they select better features by accounting for test and train datasets.The aim of this paper is to use DevQLMLOps for automated tuning and selections based on orchestrations and messaging between containers.The proposed AKFA(Adaptive Kernel Firefly Algorithm)is for selecting features for CNM(Cloud Network Monitoring)operations.AKFA methodology is demonstrated using CNSD(Cloud Network Security Dataset)with satisfactory results in the performance metrics like precision,recall,F-measure and accuracy used.
文摘An active-RC low-pass filter of 5MHz cutoff frequency with a tuning architecture is proposed. It is implemented in 0. 18μm standard CMOS technology. The accuracy of the tuning system is improved to be within （ - 1.24%, ＋ 2.16%） in measurement. The chip area of the tuning system is only a quarter of that of the main-filter. After tuning is completed, the tuning system will be turned off automatically to save power and to avoid interference. The in-band 3rd order harmonic input intercept point （IIP3） is larger than 16. ldBm, with 50Ω as the source impedance. The input referred noise is about 36μVrms The measured group delay variation of the filter between 3 and 5MHz is only 24ns,and the filter power consump- tion is 3.6roW. This filter with the tuning system is realized easily and can be used in many wireless low-IF receiver applications, such as global position systems （GPS）, global system for mobile communications （GSM） and code division multiple access （CDMA） chips.
文摘In the field of digital image processing,denoising is one of the basic problems.The challenges faced in image denoising are detecting impulse noise and designing a suitable filter.In this paper,we propose a methodology to remove the random impulse noise on the color image using a novel switching median filter.By using this novel technique,the occurrence of color artifacts has been avoided after noise removal which depends on auto-tuning threshold detection and a vector-type median filter noise remover.In the proposed technique,the random valued impulse noises with uniform distribution have been dealt with switching median filter.L_(2)Norm is employed to calculate the distribution distance rather than L1 Norm which is used to identify the optimal threshold value for auto-tuning filter.The switching auto-tuning detector automatically tunes the noisy pixels based on distance information of pixels distribution.The Normalized Mean Square Error(NMSE)is found to decrease for L_(2)Norm when compared with L1 Norm.The Peak Signal to Noise Ratio(PSNR)value and True Positive Rate(TPR)value improved with L_(2)Norm signifying effective noise removal.The efficiency of the present method is verified by conducting experiments on digital images.
