NTW-Mt

Abstract

This paper describes a parallel discrete event simulator, Neuron Time Warp-Multi Thread (NTW-MT), developed for the simulation of reaction diffusion models of neurons. The simulator was developed as part of the NEURON project and is intended to be included in NEURON. It relies upon a stochastic discrete event model developed for chemical reactions. NTW-MT is optimistic and thread-based, in which communication latency among threads within the same process is minimized by pointers. We investigate the performance of NTW-MT on a reaction-diffusion model for the transmission of calcium waves in a neuron. Calcium plays a fundamental role in the second messenger system of a neuron. However, the mechanism by which calcium waves are transmitted is not entirely understood. Stochastic models are more realistic than deterministic models for small populations of ions such as those found in apical dendrites. To be more precise, we simulate a stochastic discrete event model for calcium wave propagation on an unbranched apical dendrite of a hippocampal pyramidal neuron. We examine the performance of NTW-MT on this calcium wave model and compare it to the performance of (1) a process based op- * This author is affiliated with State timistic simulator and (2) a threaded simulator which uses a single priority (SQ) queue for each thread. Our multithreaded simulator is shown to achieve superior performance to these simulators.

Parallel Discrete Event Simulation of Stochastic Reaction and Diffusion Using Reverse Computation

“…In this paper, we employ reverse computation to address the memory usage issue in NTW-MT [10], a multi-threaded parallel discrete event simulator for stochastic reaction-diffusion simulation in neurons. NTW-MT is optimistic and employs state-saving for restoring state during roll-back, which leads to high memory consumption and cache misses.…”

“…One notable intracellular calcium dynamics is the so-called Ca 2+ -induced-Ca 2+ -release (CICR) [21] which controls a diverse array of cellular processes including fertilization, gene transcription, muscle contraction and even cell death [22]. A brief introduction to CICR dynamics can be found in [10].…”

“…The first one is the simplified calcium wave model in [10] which consists of three species: (1) Ca 2+ er , calcium in Endoplasmic Reticulum (ER) of a cell (2) Ca 2+ cyt , calcium in cytoplasm (3) Inositol 1,4,5-triphosphate (IP 3 ). As the real wave model is complex we simplified it by assuming (1) a IP 3 receptor (IP 3 R ) opens when the concentration of IP 3 and Ca 2+ are both higher than some respective threshold (2) an opening IP 3 receptor channel will close for a period of time determined by an exponential distribution.…”

“…We developed a multi-threaded PDES simulator, NTW-MT [10], used to simulate stochastic reaction and diffusion in neurons. It makes use of the NSM algorithm [8] to do stochastic calculations, and diffusion of ions between neighbouring subvolumes is represented as events.…”

“…However, as analysed above, in our simulation, the processed events can be reversely processed in arbitrary order. Let us take the order e3 → e4 → e2 as an example, then the state trace during roll-back becomes s5 → ( Since the connection among LPs is static, a LP uses a list of input channels to receive events from its adjacent LPs, and processed events from the same sender are also stored in the corresponding channel to maximize cache locality, and more details can be found in [10]. Owing to the orderindependent property of our reverse computation algorithm, the incorrectly-processed events can be processed by channels without collecting and sorting all incorrectly-processed events into strictly backward order.…”