Invited Talk
December 15, 2015
"Scaling of pseudospectral code Tarang for turbulence simulations"
Speaker: Prof. Mahendra Verma
Professor, Swarnajayanti Fellow, IITK, India
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here
December 15, 2015
"Scaling of pseudospectral code Tarang for turbulence simulations"
Speaker: Prof. Mahendra Verma
Professor, Swarnajayanti Fellow, IITK, India
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here
"Scaling of pseudospectral code Tarang for turbulence simulations"
Speaker: Prof. Mahendra Verma
Professor, Swarnajayanti Fellow, IITK, India
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here
Speaker: Prof. Mahendra Verma
Professor, Swarnajayanti Fellow, IITK, India
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here
Professor, Swarnajayanti Fellow, IITK, India
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here
Abstract:
Tarang is a general-purpose pseudospectral parallel code for simulating flows involving fluids, magnetohydrodynamics, and Rayleigh–Bénard convection in turbulence and instability regimes. The fast Fourier transform (FFT) is a major component of a pseudospectral code, and it takes around 75% of the total computer time. We performed scaling tests of FFT and Tarang on Shaheen, a IBM BlueGene/P machine, on grid up to 4096x4096x4096 grid with 65536 processors, and observed that the time per step for FFT as well as Tarang scales as p^{-2/3}, where p is the number of processors. This scaling is consistent with the bandwidth estimate for All-to-all communication for the torus network.
We also observe good ‘weak’ and ‘strong’ scaling. We are extending our scaling studies to Shaheen I, a CRAY XC40 supercomputer.
Reference: M. K. Verma, A. Chatterjee, K. S. Reddy, R. K. Yadav, S. Paul, M. Chandra, and R. Samtaney, Benchmarking and scaling studies of pseudospectral code Tarang for turbulence simulations, Pramana, 81, 617 (2013).
Bio:
Prof. Mahendra Verma is a Professor in the physics department of IIT Kanpur. His research interests are turbulence, computational fluid dynamics, and high-performance computing. He is a recipient of Swarnajayanti fellowship. He has written over 60 research papers and Textbook “Introduction to Mechanics”
For More Information about Prof. Mahendra Varma , Please Visit: Here