A significant part of the material of this glossary was adapted from material originally written by Gregory V. Wilson which appeared as "A Glossary of Parallel Computing Terminology" (IEEE Parallel & Distributed Technology, February 1993), and is being re-printed in the same author's "Practical Parallel Programming" (MIT Press, 1995). Several people have contributed additions to this glossary, especially Jack Dongarra, Geoffrey Fox and many of my colleagues at Edinburgh and Syracuse.
Original version is from NPAC at <URL:http://nhse.npac.syr.edu/hpccgloss/>>Original author: Ken Hawick, firstname.lastname@example.org
See also the index of all letters and the full list of entries (very large)
Sections: A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
T1 (n.) Transmission facility at digital service (DS1) level 1 with 1.544Mbps in North America and 2.048Mbps in Europe. See also T3.
T3 (n.) Transmission facility at digital service (DS3) level 3 with 44.736Mbps. STS1 or OC1 at 51.84Mbps is the Sonet equivalent for broadband services. Sometimes called a 45meg circuit. See also T1.
task farming (n.) A technique for implementing self-scheduling calculations. In a task farm, a source process generates a pool of jobs, while a sink process consumes results. In between, one or more worker processes repeatedly claim jobs from the source, turn them into results, dispatch those results to the sink, and claim their next jobs. If the number of jobs is much greater than the number of workers, task farming can be an effective way to load balance a computation.
TCP (n.) transmission control protocol is the main transport protocol in the internet protocol suite. It provides reliable, stateful and connection/stream oriented end to end connectivity.
TCP/IP (n.) is a compound acronym used synonymously with transmission control protocol TCP which is an internet protocol.
telnet (n.) An application that provides virtual terminal services for a wide variety of remote systems. It allows a user at one site to interact with applications at other sites as if the user's terminal is local.
TeraFLOPS (n.) 10^12 FLOPS.
thrashing (n.) a phenomenon of virtual memory systems that occurs when the program, by the manner in which it is referencing its data and instructions, regularly causes the next memory locations referenced to be overwritten by recent or current instructions. The result is that referenced items are rarely in the machines physical memory and almost always must be fetched from secondary storage, usually a disk. Cache thrashing involves a similar situation between cache and physical memory.
thread (n.) a lightweight or small granularity process.
throughput(n.) number of results produced per unit time.
tiling (n.) A regular division of a mesh into patches, or tiles. Tiling is the most common way to do geometric decomposition.
time sharing (adj.) Dividing the effort of a processor among many programs so they can run concurrently. Time sharing is usually managed by an operating system. See also space sharing.
time-processor product (n.) The product of the time taken to execute a program and the number of processors used to achieve that time, often used as a measure of goodness for parallel algorithms. See also Amdahl's Law, efficiency, Gustafson's Law, speedup.
TLB (n.) translation look-aside buffer; the memory cache of the most recently used page table entries within the memory management unit.
token ring (n.) Token ring is an IBM based LAN protocol that uses a ring shaped network topology. Token Ring has speeds at 4Mbps and 16Mbps. A distinguishing packet is transferred from machine to machine and only the machine that is in control of the token is able to transmit.
topology (n.) A family of graphs created using the same general rule or that share certain properties. The processors in a multicomputer, and the circuits in a switch, are usually laid out using one of several topologies, including the mesh, the hypercube, the butterfly, the torus and the shuffle exchange network. See also bisection bandwidth, diameter.
torus (n.) A topology in which nodes form a regular cyclic d-dimensional grid, and each edge is parallel to a grid axis and joins two nodes that are adjacent along that axis. The architecture of some multicomputers is a two or three dimensional torus. See also hypercube, mesh.
TP4 (n.) OSI's transport protocol class 4. This includes error detection and recovery and is the most powerful OSI transport layer protocol. TP4 is the OSI equivalent of the internet's TCP.
trace scheduling(n.) A compiler optimization technique that vectorizes the most likely path through a program as if it were a single basic block, includes extra instructions at each branch to undo any ill effects of having made a wrong guess, vectorizes the next most likely branches and so on.
transposition table (n.) a hash table that stores previously evaluated game positions.
transputer (n.) A single integrated circuit which contains a CPU, communications links, memory and some cache memory. The name transputer refers to a proprietary series of chips manufactured by Inmos, although other node chips have had similar characteristics.
tree (n.) a connected, undirected, acyclic graph. The most commonly encountered tree in computer science is the regular binary tree, in which a root node has two children but no parent, each interior node has a single parent and two children, and leaf nodes have a single parent but no children.
true dependence (n.) See dependence.
true ratio (n.) the frequency with which the "true" branch of a Fortran IF-test occurs. If the true ratio is know at compile time, some compilers can take advantage of this knowledge. However the true ratio is often data dependent and cannot effectively be dealt with automatically. See also interactive vectorizer.
tuple (n.) An ordered sequence of fixed length of values of arbitrary types. Tuples are used for both data storage and interprocess communication in the generative communication paradigm. See also tuple space.
tuple space (n.) A repository for tuples in a generative communication system. Tuple space is an associative memory.