Memory hierarchy

The use of concurrence or parallelism in chemistry applications is not new. In the 1980s chemistry applications evolved to take advantage of multiple vector registers on vector supercomputers and attached processors by restructuring the software to utilize matrix—vector and matrix—matrix operations. In fact, once the software had been adapted to vector supercomputers, many applications ran faster on serial machines because of improved use of these machines memory hierarchies. The use of asynchronous disk operations (overlapped computation and disk reads or writes) and a few loosely coupled computers and workstations are other concurrency optimizations that were used before the development of current MPP technology. The challenge of the... 

Usually the start-up cost is large compared to the transfer cost for a single data unit, and both costs increase sharply (order of magnitude) for each level in the memory hierarchy. Because parallel computers have more levels than sequential ones, these transfer costs are significantly more intrusive on parallel computers and are evident in the performance of parallel applications. 

Memory organization Memory hierarchy, number of banks per block, cache associativity, memory/cache bus width/line size, memory interface protocol... 

S. A. Kuhn, M. B. Kleiner, P. Ramm, W. Weber. Performance improvement of the memory hierarchy of RISC-systems by application of 3-D technology. IEEE Trans. On Components, Packaging, and Manufacturing Technology, Part B Advanced Packaging, Vol. 19 No. 4,... 

Considering the whole memory hierarchy with all misses will finally hit the... 

R.E. Matick, T. J. Heller, M. Ignatowski. Analytical analysis of finite cache penalty and cycles per instruction of a multiprocessor memory hierarchy using miss rates and queuing theory. IBM J. Research Development, Vol. 45, No. 6, Nov. 2001, pp. 819 - 842. 

The memory hierarchy in a hypothetical MIMD machine constructed from the nodes in Figure 2.15 and the network in Figure 2.12. Logically equivalent data location types are subdivided by how many network hops are needed to reach them, Whop (intra-node hops for local memory, inter-node hops for remote memory). The estimate of the time required to access memory at the given location is access (this time is hypothetical and not based on any particular hardware). The level of treatment typically used for each memory level is given for the several layers of software... 

There are several aspects to memory organization. We will take a top down approach in discussing them. Memory Hierarchy... 

The strategy used to remedy this problem is called memory hierarchy. Memory hierarchy works because of the locahty property of memory references due to the sequentially fetched program instructions and the conjugation of related data. In a hierarchical memory system there are many levels of memory hierarchies. A small amount of very fast memory is usually allocated and brought right next to the central processing unit to help match up the speed of the CPU and memory. As the distance becomes greater between the... 

CPU and memory, the performance requirement for the memory is relaxed. At the same time, the size of the memory grows larger to accommodate the overall memory size requirement. Some of the memory hierarchies are registers, cache, main memory, and disk. Figure 8.45 illustrates the general memory hierarchy employed in a traditional system. When a memory reference is made, the processor accesses the memory at the top of the hierarchy. If the desired data is in the higher hierarchy, a bit is encountered and information is obtained quickly. Otherwise a miss is encountered. The requested in-... 

In modern computing systems, there maybe several sublevels of cache within the hierarchy of cache. The general principle of memory hierarchy is that the farther away from the CPU it is, the larger its size, slower its speed, and the cheaper its price per memory unit becomes. Because the memory space addressable by... 

Memory hierarchy Organize memoryin levels to make the speed ofmemory comparable to the processor. Memory read The process of retrieving information from memory. 

Cache Small amount of fast memory, physically close to CPU, used as storage of a block of data needed immediately by the processor. Caches exist in a memory hierarchy. There is a small but very fast LI (level one) cache if that misses, then the access is passed on to the bigger but slower L2 (level two) cache, and if that misses, the access goes to the main memory (or L3 cache, if it exists). 

Introduction Defining a Computer Architecture Single Processor Systems Multiple Processor Systems Memory Hierarchy Implementation Considerations... 

One reason that RISC architectures work better than traditional CISC machines is due to the use of large on-chip caches and register sets. Since locality of reference effects (described in the section on memory hierarchy) dominate most instruction and data reference behavior, the use of an on-chip cache and large register sets can reduce the number of instructions and data fetched per instruction execution. Most RISC machines use pipelining to overlap instruction execution, further reducing the clock period. Compiler techniques are used to exploit the natural parallelism inherent in sequentially executed programs. 

High-performance computer systems use a multiple level memory hierarchy ranging from small, fast cache memory to larger, slower main memory to improve performance. Parallelism can be introduced into a system through the memory hierarchy as depicted in Fig. 19.7. A cache is a small, high-speed buffer used to temporarily hold those portions of memory that are currently in use in the processor. 

Memory hierarchy management A collection of transformations that rewrite the program to change the order in which it accesses memory locations. On machines with cache memories, reordering the references can increase the extent to which values already in the cache are reused, and thus decrease the aggregate amount of tie spent waiting on values to be fetched from memory. 

Memory hierarchy management Every modem computer has a cache hierarchy, designed to reduce the average amount of time required to retrieve a value from memory. To improve the performance of programs, many compilers transform loop nests to improve data locality—i.e., they reorder memory references in a way that increases the likelihood that data elements will found in the cache when they are needed. To accomplish this, the compiler must transform the loops so that they repeatedly iterate over blocks of data that are small enough to fit in the cache. 

The study of computer architecture amounts to analyzing two basic principles, performance and cost for a set of computer architecture alternatives that meet functionality requirements. This appears to be straightforward when considered at a general level, but when these topics are looked at in detail the water can become muddy. If a systematic approach is used the concepts can be straightforward. Fortunately, central processing units, memory hierarchy, I/O systems, etc., are layered, and these layers may be considered as levels of abstractions. Each level of abstraction has its own abstractions and objects. By studying aspects of computer science in this fashion, it is possible to censor out details that are irrelevant to the level of abstraction being considered and to concentrate on the task at hand. 

This sequential circuit is the fastest and most expensive part of the memory hierarchy. The registers are the part of the memory hierarchy that are directly a part of the processor datapath. Because these are the fastest memory it is desirable to have all of the active data present in them. 

The memory in the memory hierarchy of a computer system is used to store information, instructions, and data that will be used by the computer system. Memory is often classified as registers, cache memory, main memory, hard disk, floppy disk, and tapes. These are pictured in a hierarchal form in Fig. 10 with locations within each type of memory randomly accessible except for tapes. Tapes are sequentially accessible, and in the long run each disk data unit is accessible in equal time, but at a given time the access time for a particular unit is dependent on the location of the disk components. The term access designates the memory activities that are associated with either a read or a write. Randomly accessible means that a memory location may be read or written in the same amount of time irregardless of the order of accesses of memory locations, and sequentially accessible means that the time required to access a memory location is dependent on location of the immediate prior memory access. 

There are other forms of information storage such as CD-ROM and cassettes. The control store is a memory unit but is not considered in the memory hierarchy because it is only used to store microprogram instructions. Memory hierarchy locations are used to store abstraction... 

In the memory hierarchy, the memory level in the higher location of the pyramid of Fig. 10 is usually physically closer to the processor, faster, more frequently accessed, smaller in size, and each bit is more expensive. Some of the characteristics of memory types are access time, bandwidth, size limit, management responsibility, and location within the computer system (within processor or on the external bus). 

Size, cost, and speed are the major design parameters in the memory hierarchy. The principle behind memory hierarchy design is to keep the cost per unit of memory as close as possible to that of the least expensive memory and keep the average access time as close as possible to that of the fastest memory. To accomplish this, the design must use a minimal amount of the memory type at the top of the pyramid of Fig. 10 and attempt to keep information that will be accessed in the near future in memory at the top of the pyramid. Removable disks and tapes are of unlimited size, and that means that the user may continue to buy more of the media until his needs are met. 

The similarity exhibited here between the multi-variable and the memory hierarchy formulations of the orientational problem is obviously not a general result in fact, it derives from the choice of a variable and its time derivative in the multi-variable theory. Other variations on these themes can readily be devised for example, one could couple Q, and 4 (orthogonal ized to Qj, however) (8) or one could combine higher-order memory functions with a multi-variable theory or one could couple translational and rotational variables (9), The enormous flexibility of the G,L,E, means that the intuition of the user will play a particularly significant role in determining the success of the outcome. To illustrate this point, we now briefly recapitulate how the 6,L,E, applies to a quite different orientational problem namely, anisotropic rotational diffusion (10). 

---