-
Performance Portable Monte Carlo Particle Transport on Intel, NVIDIA, and AMD GPUs
Authors:
John Tramm,
Paul Romano,
Patrick Shriwise,
Amanda Lund,
Johannes Doerfert,
Patrick Steinbrecher,
Andrew Siegel,
Gavin Ridley
Abstract:
OpenMC is an open source Monte Carlo neutral particle transport application that has recently been ported to GPU using the OpenMP target offloading model. We examine the performance of OpenMC at scale on the Frontier, Polaris, and Aurora supercomputers, demonstrating that performance portability has been achieved by OpenMC across all three major GPU vendors (AMD, NVIDIA, and Intel). OpenMC's GPU p…
▽ More
OpenMC is an open source Monte Carlo neutral particle transport application that has recently been ported to GPU using the OpenMP target offloading model. We examine the performance of OpenMC at scale on the Frontier, Polaris, and Aurora supercomputers, demonstrating that performance portability has been achieved by OpenMC across all three major GPU vendors (AMD, NVIDIA, and Intel). OpenMC's GPU performance is compared to both the traditional CPU-based version of OpenMC as well as several other state-of-the-art CPU-based Monte Carlo particle transport applications. We also provide historical context by analyzing OpenMC's performance on several legacy GPU and CPU architectures. This work includes some of the first published results for a scientific simulation application at scale on a supercomputer featuring Intel's Max series "Ponte Vecchio" GPUs. It is also one of the first demonstrations of a large scientific production application using the OpenMP target offloading model to achieve high performance on all three major GPU platforms.
△ Less
Submitted 18 March, 2024;
originally announced March 2024.
-
Skewness, kurtosis and the 5th and 6th order cumulants of net baryon-number distributions from lattice QCD confront high-statistics STAR data
Authors:
A. Bazavov,
D. Bollweg,
H. -T. Ding,
P. Enns,
J. Goswami,
P. Hegde,
O. Kaczmarek,
F. Karsch,
R. Larsen,
Swagato Mukherjee,
H. Ohno,
P. Petreczky,
C. Schmidt,
S. Sharma,
P. Steinbrecher
Abstract:
We present new results on up to $6^{th}$ order cumulants of net baryon-number fluctuations at small values of the baryon chemical potential, $μ_B$, obtained in lattice QCD calculations with physical values of light and strange quark masses. Representation of the Taylor expansions of higher order cumulants in terms of the ratio of the two lowest order cumulants,…
▽ More
We present new results on up to $6^{th}$ order cumulants of net baryon-number fluctuations at small values of the baryon chemical potential, $μ_B$, obtained in lattice QCD calculations with physical values of light and strange quark masses. Representation of the Taylor expansions of higher order cumulants in terms of the ratio of the two lowest order cumulants, $M_B/σ_B^2=χ_1^B(T,μ_B)/χ_2^B(T,μ_B)$, allows for a parameter free comparison with data on net proton-number cumulants obtained by the STAR Collaboration in the Beam Energy Scan at RHIC. We show that recent high statistics data on skewness and kurtosis ratios of net proton-number distributions, obtained at beam energy $\sqrt{s_{_{NN}}}=54.4$ GeV, agree well with lattice QCD results on cumulants of net baryon-number fluctuations close to the pseudo-critical temperature, $T_{pc}(μ_B)$, for the chiral transition in QCD. We also present first results from a next-to-leading order expansion of $5^{th}$ and $6^{th}$ order cumulants on the line of pseudo-critical temperatures.
△ Less
Submitted 3 April, 2020; v1 submitted 23 January, 2020;
originally announced January 2020.
-
Meson Screening Masses in (2+1)-Flavor QCD
Authors:
Alexei Bazavov,
Simon Dentinger,
Heng-Tong Ding,
Prasad Hegde,
Olaf Kaczmarek,
Frithjof Karsch,
Edwin Laermann,
Anirban Lahiri,
Swagato Mukherjee,
Hiroshi Ohno,
Peter Petreczky,
Rishabh Thakkar,
Hauke Sandmeyer,
Christian Schmidt,
Sayantan Sharma,
Patrick Steinbrecher
Abstract:
We present lattice QCD results for mesonic screening masses in the temperature range 140 MeV $\lesssim T \lesssim$ 2500 MeV. Our calculations were carried out using (2+1)-flavors of the Highly Improved Staggered Quark (HISQ) action, with a physical value for the strange quark mass and two values of the light quark mass corresponding to pion masses of 160 MeV and 140 MeV. Continuum-extrapolated res…
▽ More
We present lattice QCD results for mesonic screening masses in the temperature range 140 MeV $\lesssim T \lesssim$ 2500 MeV. Our calculations were carried out using (2+1)-flavors of the Highly Improved Staggered Quark (HISQ) action, with a physical value for the strange quark mass and two values of the light quark mass corresponding to pion masses of 160 MeV and 140 MeV. Continuum-extrapolated results were obtained using calculations with a variety of lattice spacings corresponding to temporal lattice extents $N_τ= 6 - 16$. We discuss the implications of these results for the effective restoration of various symmetries in the high temperature phase of QCD, as well as the approach toward the perturbative limit.
△ Less
Submitted 4 September, 2019; v1 submitted 26 August, 2019;
originally announced August 2019.
-
Chiral phase transition temperature in (2+1)-Flavor QCD
Authors:
H. -T. Ding,
P. Hegde,
O. Kaczmarek,
F. Karsch,
Anirban Lahiri,
S. -T. Li,
Swagato Mukherjee,
H. Ohno,
P. Petreczky,
C. Schmidt,
P. Steinbrecher
Abstract:
We present a lattice QCD based determination of the chiral phase transition temperature in QCD with two degenerate, massless quarks and a physical strange quark mass. We propose and calculate two novel estimators for the chiral transition temperature for several values of the light quark masses, corresponding to Goldstone pion masses in the range of…
▽ More
We present a lattice QCD based determination of the chiral phase transition temperature in QCD with two degenerate, massless quarks and a physical strange quark mass. We propose and calculate two novel estimators for the chiral transition temperature for several values of the light quark masses, corresponding to Goldstone pion masses in the range of $58~{\rm MeV}\lesssim m_π\lesssim 163~{\rm MeV}$. The chiral phase transition temperature is determined by extrapolating to vanishing pion mass using universal scaling analysis. Finite volume effects are controlled by extrapolating to the thermodynamic limit using spatial lattice extents in the range of $2.8$-$4.5$ times the inverse of the pion mass. Continuum extrapolations are carried out by using three different values of the lattice cut-off, corresponding to lattices with temporal extent $N_τ=6,\ 8$ and $12$. After thermodynamic, continuum and chiral extrapolations we find the chiral phase transition temperature $T_c^0=132^{+3}_{-6}$ MeV.
△ Less
Submitted 10 August, 2019; v1 submitted 12 March, 2019;
originally announced March 2019.
-
Chiral crossover in QCD at zero and non-zero chemical potentials
Authors:
A. Bazavov,
H. -T. Ding,
P. Hegde,
O. Kaczmarek,
F. Karsch,
N. Karthik,
E. Laermann,
Anirban Lahiri,
R. Larsen,
S. -T. Li,
Swagato Mukherjee,
H. Ohno,
P. Petreczky,
H. Sandmeyer,
C. Schmidt,
S. Sharma,
P. Steinbrecher
Abstract:
We present results for pseudo-critical temperatures of QCD chiral crossovers at zero and non-zero values of baryon ($B$), strangeness ($S$), electric charge ($Q$), and isospin ($I$) chemical potentials $μ_{X=B,Q,S,I}$. The results were obtained using lattice QCD calculations carried out with two degenerate up and down dynamical quarks and a dynamical strange quark, with quark masses corresponding…
▽ More
We present results for pseudo-critical temperatures of QCD chiral crossovers at zero and non-zero values of baryon ($B$), strangeness ($S$), electric charge ($Q$), and isospin ($I$) chemical potentials $μ_{X=B,Q,S,I}$. The results were obtained using lattice QCD calculations carried out with two degenerate up and down dynamical quarks and a dynamical strange quark, with quark masses corresponding to physical values of pion and kaon masses in the continuum limit. By parameterizing pseudo-critical temperatures as $ T_c(μ_X) = T_c(0) \left[ 1 -κ_2^{X}(μ_{X}/T_c(0))^2 -κ_4^{X}(μ_{X}/T_c(0))^4 \right] $, we determined $κ_2^X$ and $κ_4^X$ from Taylor expansions of chiral observables in $μ_X$. We obtained a precise result for $T_c(0)=(156.5\pm1.5)\;\mathrm{MeV}$. For analogous thermal conditions at the chemical freeze-out of relativistic heavy-ion collisions, i.e., $μ_{S}(T,μ_{B})$ and $μ_{Q}(T,μ_{B})$ fixed from strangeness-neutrality and isospin-imbalance, we found $κ_2^B=0.012(4)$ and $κ_4^B=0.000(4)$. For $μ_{B}\lesssim300\;\mathrm{MeV}$, the chemical freeze-out takes place in the vicinity of the QCD phase boundary, which coincides with the lines of constant energy density of $0.42(6)\;\mathrm{GeV/fm}^3$ and constant entropy density of $3.7(5)\;\mathrm{fm}^{-3}$.
△ Less
Submitted 12 June, 2019; v1 submitted 19 December, 2018;
originally announced December 2018.
-
The QCD crossover at zero and non-zero baryon densities from Lattice QCD
Authors:
Patrick Steinbrecher
Abstract:
We map out the QCD crossover line $\frac{T_c(μ_B)}{T_c(0)} = 1 - κ_2 \left( \frac{μ_B}{T_c(0)} \right)^2 - κ_4 \left( \frac{μ_B}{T_c(0)} \right)^4 + \mathcal{O}(μ_B^6)$ for the first time up to $\mathcal{O}(μ_B^4)$ for a strangeness neutral system by performing a Taylor expansion of chiral observables in temperature $T$ and chemical potentials $μ$. At vanishing chemical potential, we report a cros…
▽ More
We map out the QCD crossover line $\frac{T_c(μ_B)}{T_c(0)} = 1 - κ_2 \left( \frac{μ_B}{T_c(0)} \right)^2 - κ_4 \left( \frac{μ_B}{T_c(0)} \right)^4 + \mathcal{O}(μ_B^6)$ for the first time up to $\mathcal{O}(μ_B^4)$ for a strangeness neutral system by performing a Taylor expansion of chiral observables in temperature $T$ and chemical potentials $μ$. At vanishing chemical potential, we report a crossover temperature $T_c(0) = (156.5 \pm 1.5)\;\mathrm{MeV}$ defined by the average of several second-order chiral susceptibilities. For a system with thermal conditions appropriate for a heavy-ion collision, we determined a curvature from the subtracted condensate as $κ_2 = 0.0120(20)$ and from the disconnected susceptibility as $κ_2 = 0.0123(30)$. The next order $κ_4$ is significantly smaller. We also report the crossover temperature as a function of the chemical potentials for: baryon-number, electric charge, strangeness and isospin. Additionally, we find that $T_c(μ_B)$ is in agreement with lines of constant energy density and constant entropy density. Along this crossover line, we study net baryon-number fluctuations and show that their increase is substantially smaller compared to that obtained in HRG model calculations. Similarly, we analyze chiral susceptibility fluctuations along the crossover line and show that these are constant. We conclude that no signs for a narrowing of the crossover region can be found for baryon chemical potential $μ_B < 250\;\mathrm{MeV}$.
△ Less
Submitted 15 July, 2018;
originally announced July 2018.
-
Skewness and kurtosis of net baryon-number distributions at small values of the baryon chemical potential
Authors:
HotQCD Collaboration,
A. Bazavov,
H. -T. Ding,
P. Hegde,
O. Kaczmarek,
F. Karsch,
E. Laermann,
Swagato Mukherjee,
H. Ohno,
P. Petreczky,
E. Rinaldi,
H. Sandmeyer,
C. Schmidt,
Chris Schroeder,
S. Sharma,
W. Soeldner,
R. A. Soltz,
P. Steinbrecher,
P. M. Vranas
Abstract:
We present results for the ratios of mean ($M_B$), variance ($σ_B^2$), skewness ($S_B)$ and kurtosis ($κ_B$) of net baryon-number fluctuations obtained in lattice QCD calculations with a physical light to strange quark mass ratio. Using next-to-leading order Taylor expansions in baryon chemical potential we find that qualitative features of these ratios closely resemble the corresponding experimen…
▽ More
We present results for the ratios of mean ($M_B$), variance ($σ_B^2$), skewness ($S_B)$ and kurtosis ($κ_B$) of net baryon-number fluctuations obtained in lattice QCD calculations with a physical light to strange quark mass ratio. Using next-to-leading order Taylor expansions in baryon chemical potential we find that qualitative features of these ratios closely resemble the corresponding experimentally measured cumulants ratios of net proton-number fluctuations for beam energies down to $\sqrt{s_{_{NN}}} \ge 19.6$ GeV. We show that the difference in cumulant ratios for the mean net baryon-number, $M_B/σ_B^2=χ_1^B(T,μ_B)/χ_2^B(T,μ_B)$ and the normalized skewness, $S_Bσ_B=χ_3^B(T,μ_B)/χ_2^B(T,μ_B)$, naturally arises in QCD thermodynamics. Moreover, we establish a close relation between skewness and kurtosis ratios, $S_Bσ_B^3/M_B=χ_3^B(T,μ_B)/χ_1^B(T,μ_B)$ and $κ_Bσ_B^2=χ_4^B(T,μ_B)/χ_2^B(T,μ_B)$, valid at small values of the baryon chemical potential.
△ Less
Submitted 1 November, 2017; v1 submitted 16 August, 2017;
originally announced August 2017.
-
The QCD Equation of State to $\mathcal{O}(μ_B^6)$ from Lattice QCD
Authors:
A. Bazavov,
H. -T. Ding,
P. Hegde,
O. Kaczmarek,
F. Karsch,
E. Laermann,
Y. Maezawa,
S. Mukherjee,
H. Ohno,
P. Petreczky,
H. Sandmeyer,
P. Steinbrecher,
C. Schmidt,
S. Sharma,
W. Soeldner,
M. Wagner
Abstract:
We calculated the QCD equation of state using Taylor expansions that include contributions from up to sixth order in the baryon, strangeness and electric charge chemical potentials. Calculations have been performed with the Highly Improved Staggered Quark action in the temperature range $T\in [135~{\rm MeV}, 330~{\rm MeV}]$ using up to four different sets of lattice cut-offs corresponding to latti…
▽ More
We calculated the QCD equation of state using Taylor expansions that include contributions from up to sixth order in the baryon, strangeness and electric charge chemical potentials. Calculations have been performed with the Highly Improved Staggered Quark action in the temperature range $T\in [135~{\rm MeV}, 330~{\rm MeV}]$ using up to four different sets of lattice cut-offs corresponding to lattices of size $N_σ^3\times N_τ$ with aspect ratio $N_σ/N_τ=4$ and $N_τ=6-16$. The strange quark mass is tuned to its physical value and we use two strange to light quark mass ratios $m_s/m_l=20$ and $27$, which in the continuum limit correspond to a pion mass of about $160$ MeV and $140$ MeV espectively. Sixth-order results for Taylor expansion coefficients are used to estimate truncation errors of the fourth-order expansion. We show that truncation errors are small for baryon chemical potentials less then twice the temperature ($μ_B\le 2T$). The fourth-order equation of state thus is suitable for the modeling of dense matter created in heavy ion collisions with center-of-mass energies down to $\sqrt{s_{NN}}\sim 12$ GeV. We provide a parametrization of basic thermodynamic quantities that can be readily used in hydrodynamic simulation codes. The results on up to sixth order expansion coefficients of bulk thermodynamics are used for the calculation of lines of constant pressure, energy and entropy densities in the $T$-$μ_B$ plane and are compared with the crossover line for the QCD chiral transition as well as with experimental results on freeze-out parameters in heavy ion collisions. These coefficients also provide estimates for the location of a possible critical point. We argue that results on sixth order expansion coefficients disfavor the existence of a critical point in the QCD phase diagram for $μ_B/T\le 2$ and $T/T_c(μ_B=0) > 0.9$.
△ Less
Submitted 9 March, 2017; v1 submitted 16 January, 2017;
originally announced January 2017.
-
Conserved Charge Fluctuations from Lattice QCD and the Beam Energy Scan
Authors:
F. Karsch,
A. Bazavov,
H. -T. Ding,
P. Hegde,
O. Kaczmarek,
E. Laermann,
Swagato Mukherjee,
H. Ohno,
P. Petreczky,
C. Schmidt,
S. Sharma,
W. Soeldner,
P. Steinbrecher,
M. Wagner
Abstract:
We discuss the next-to-leading order Taylor expansion of ratios of cumulants of net-baryon number fluctuations. We focus on the relation between the skewness ratio, $S_Bσ_B = χ_3^B/χ_1^B$, and the kurtosis ratio, $κ_Bσ_B^2 =χ_4^B/χ_2^B$. We show that differences in these two cumulant ratios are small for small values of the baryon chemical potential. The next-to-leading order correction to…
▽ More
We discuss the next-to-leading order Taylor expansion of ratios of cumulants of net-baryon number fluctuations. We focus on the relation between the skewness ratio, $S_Bσ_B = χ_3^B/χ_1^B$, and the kurtosis ratio, $κ_Bσ_B^2 =χ_4^B/χ_2^B$. We show that differences in these two cumulant ratios are small for small values of the baryon chemical potential. The next-to-leading order correction to $κ_Bσ_B^2$ however is approximately three times larger than that for $S_Bσ_B$. The former thus drops much more rapidly with increasing beam energy, $\sqrt{s_{NN}}$. We argue that these generic patterns are consistent with current data on cumulants of net-proton number fluctuations measured by the STAR Collaboration at $\sqrt{s_{NN}}\ge 19.6$~GeV.
△ Less
Submitted 22 December, 2015;
originally announced December 2015.
-
Conjugate gradient solvers on Intel Xeon Phi and NVIDIA GPUs
Authors:
O. Kaczmarek,
C. Schmidt,
P. Steinbrecher,
M. Wagner
Abstract:
Lattice Quantum Chromodynamics simulations typically spend most of the runtime in inversions of the Fermion Matrix. This part is therefore frequently optimized for various HPC architectures. Here we compare the performance of the Intel Xeon Phi to current Kepler-based NVIDIA Tesla GPUs running a conjugate gradient solver. By exposing more parallelism to the accelerator through inverting multiple v…
▽ More
Lattice Quantum Chromodynamics simulations typically spend most of the runtime in inversions of the Fermion Matrix. This part is therefore frequently optimized for various HPC architectures. Here we compare the performance of the Intel Xeon Phi to current Kepler-based NVIDIA Tesla GPUs running a conjugate gradient solver. By exposing more parallelism to the accelerator through inverting multiple vectors at the same time, we obtain a performance greater than 300 GFlop/s on both architectures. This more than doubles the performance of the inversions. We also give a short overview of the Knights Corner architecture, discuss some details of the implementation and the effort required to obtain the achieved performance.
△ Less
Submitted 17 November, 2014;
originally announced November 2014.
-
HISQ inverter on Intel Xeon Phi and NVIDIA GPUs
Authors:
O. Kaczmarek,
C. Schmidt,
P. Steinbrecher,
Swagato Mukherjee,
M. Wagner
Abstract:
The runtime of a Lattice QCD simulation is dominated by a small kernel, which calculates the product of a vector by a sparse matrix known as the "Dslash" operator. Therefore, this kernel is frequently optimized for various HPC architectures. In this contribution we compare the performance of the Intel Xeon Phi to current Kepler-based NVIDIA Tesla GPUs running a conjugate gradient solver. By exposi…
▽ More
The runtime of a Lattice QCD simulation is dominated by a small kernel, which calculates the product of a vector by a sparse matrix known as the "Dslash" operator. Therefore, this kernel is frequently optimized for various HPC architectures. In this contribution we compare the performance of the Intel Xeon Phi to current Kepler-based NVIDIA Tesla GPUs running a conjugate gradient solver. By exposing more parallelism to the accelerator through inverting multiple vectors at the same time we obtain a performance 250 GFlop/s on both architectures. This more than doubles the performance of the inversions. We give a short overview of both architectures, discuss some details of the implementation and the effort required to obtain the achieved performance.
△ Less
Submitted 4 September, 2014;
originally announced September 2014.
