HILL,M. 1987. Aspects of cache memory and instruction buffer performance. Ph.D. dissertation. University of California, Berkeley, Berkeley, Calif.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....order and data layout transformation techniques which change the layout of the program data in memory. Which optimization is appropriate for a certain program usually depends on the characteristics of the cache misses. The misses for a cache C can be partitioned into the following three types [Hil87] compulsory misses: The compulsory misses of a cache C are the misses which arise when a data is requested for the first time in a program. Since the data was never referenced before the data cannot be in the cache. Compulsory misses are often also called cold or startup misses. capacity ....

....and a low level interface which provides the basis for source level performance analysis software. Furthermore, utility routines layered on top of the PAPI library are provided to allow a dynamic runtime controlling of the library via socket connections. 7.2. 2 Simulation Cache Simulation Tycho [Hil87, HS89] is a trace driven cache simulator that can simulate many alternative configurations of one cache level like direct mapped, set associative, and fully associative caches with one pass through an address trace to produce a table of miss ratios for all caches. The cache parameters which ....

M.D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Computer Science Division (EECS), University of California, Berkeley, California, USA, November 1987.

....of direct mapped cache as the cache hierarchy. The address stream seen by main memory is the stream of addresses that miss in the cache. Using the model of uniprocessor cache misses by Hill, the addresses in this miss stream can be classified as compulsory, capacity, or conflict cache misses [Hill 87] For this discussion, assume that cache blocks are fetched and allocated on read misses. Compulsory cache misses are for blocks whose addresses are seen for the first time. Capacity misses are generated as a result of not having enough space to cache all the blocks being accessed. Conflict ....

M. D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. Ph.D. dissertation, University of California at Berkeley, Computer Sciences Division, Technical Report UCB/CSD 87/381, November 1987.

....the memory references generated by the code) but we do not take into account the particular configuration of the cache which is determinant in the number of cache misses. Therefore, D j does not correspond to a number of cache misses. We use the classification of cache misses provided by Hill [14] which is based on their causes. Compulsory misses occur the first time a data element is referenced, capacity misses occur when the size of the cache is insufficient to hold the entire working set of the program, and conflict misses occur when several cache lines of the working set map to the ....

M. D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Berkeley Computer Science Division, Univ. California, Berkeley, 1987.

....global cycle counter provided on most microprocessors. We discuss mechanisms and their implementations in the sections that follow. 4 Timekeeping Metrics to Identify and Avoid Conflict Misses Canonically, cache misses are classified into 3 categories: cold miss, conflict miss and capacity miss [4]. Cold misses occur when a cache line is loaded into the cache the first time. Conflict misses are those misses which can be eliminated by a fully associative cache. Capacity misses are those which will miss even with a fully associative cache. Interpreting Hill s definitions with generational ....

M. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, University of California at Berkeley, Nov. 1987.

....The result is an automated design technique to find power efficient memory hierarchies and generate the corresponding optimized code. 1. Introduction Memory hierarchy design has been introduced long ago to improve the data access (bandwidth) to match the increasing performance of the CPU [8][24] 23] The wellknown idea of using memory hierarchy to minimise the power consumption, is based on the fact that memory power consumption depends primarily on the access frequency and the size of the memory [30] Power savings can be obtained by accessing heavily used data from smaller memories ....

M. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Univ. of California at Berkeley, Nov 1987.

....view of the attacker. In our example attack we assume the existence of a processor from which we can glean the result of cache accesses. In order to simulate the information an attacker might gain from a real device, we linked our algorithm implementations to the Dinero cache simulation library [15] which was configured to model a 1 kilobyte, four byte per line, direct mapped variant of that found in a real processor core [19, 3, 4] The cache simulator produces a list of accesses for each run of the algorithm where each access in the list relates to the S box structures being read. 4 ....

M.D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. Technical Report CSD-87-381, University of California, Berkeley, Department. of Computer Sciences, November 1987.

....to attack them. In our example attack we assume the existence of a processor from which we can glean the result of cache accesses. In order to simulate the information an attacker might gain from a real device, we linked our algorithm implementations to the Dinero cache simulation library [9] which was configured to model a 1 kilobyte, four byte per line, direct mapped variant of that found in a real smart card device [2] The cache simulator produces a list of accesses for each run of the algorithm where each access in the list relates to the S box structures being read. 3 Example ....

M.D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. Technical Report CSD-87-381, University of California, Berkeley, Department. of Computer Sciences, November 1987.

.... be able to vary parameters along several dimensions and also be able to produce any of several different metrics for 40 Reference Name Range of Parameters Metrics Overhead Sets Line Assoc Write Policy Sector [Mattson70] Stack Simulation Fixed Fixed Vary None No Misses [Hill87] Forest Simulation Vary Fixed 1 way None No Misses 5 [Hill87] All Associativity Vary Fixed Vary None No Misses 30 [Thompson89] Fixed Fixed Vary W back Yes Misses,Write Backs 100 [Wang90] Vary Fixed Vary W back No Misses,Write Backs 65 [Sugumar93] ....

.... to produce any of several different metrics for 40 Reference Name Range of Parameters Metrics Overhead Sets Line Assoc Write Policy Sector [Mattson70] Stack Simulation Fixed Fixed Vary None No Misses [Hill87] Forest Simulation Vary Fixed 1 way None No Misses 5 [Hill87] All Associativity Vary Fixed Vary None No Misses 30 [Thompson89] Fixed Fixed Vary W back Yes Misses,Write Backs 100 [Wang90] Vary Fixed Vary W back No Misses,Write Backs 65 [Sugumar93] Cheetah Fixed Vary 1 way W thru No Misses, WB Stalls 120 ....

[Article contains additional citation context not shown here]

Hill, M. Aspects of cache memory and instruction buffer performance. Ph.D. dissertation, The University of California at Berkeley. 1987.

....cache design, trace driven simulation, multichip modules, pipelining, caches, cache access times, macromodels of delay. F 1I NTRODUCTION HE performance evaluation of cache based systems has received considerable attention [1] [2], 3] These studies have considered the impact of architectural level issues like cache size, associativity, line length, write policies, etc. However, different cache organizations, in particular, size, change a cache s access time, and, thus, also affect performance. Increasing cache size tends ....

M.D. Hill, "Aspects of Cache Memory and Instruction Buffer Performance," PhD thesis, Univ. of California, 1987.

....the DRAM is necessary. Thus, larger caches usually result in better performance. Beyond the cache size, a lot of parameters affect the cache performance, such as associativity, line size, or allocation policy. There are many publications which study the impact of those parameters on the miss ratio [1, 2, 3, 4, 5]. However, adding write back does not affect the miss ratio but still has a significant performance impact. Cache miss rate simulations therefore are no answer to the write back question, and different simulations are required. Their results are presented in this paper and these results will ....

M.D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Computer Science Devision (EECS), UC Berkeley, CA 94720, 1987.

No context found.

HILL,M. 1987. Aspects of cache memory and instruction buffer performance. Ph.D. dissertation. University of California, Berkeley, Berkeley, Calif.

No context found.

Hill, M. D. Aspects of Cache Memory and Instruction Buffer Performance. Computer Science Division, EECS, University of California, Berkeley, Technical Report UCB/CSD 87/381, PhD Dissertation, November 1987.

No context found.

Hill, M.D. Aspects of Cache Memory and Instruction Buffer Performance, Ph.D. dissertation, (published as Technical Report UCB/CSD 87/381), , November 1987.

No context found.

M. D. Hill. Aspects of cache memory and instruction buffer performance. PhD thesis, University of California at Berkeley, 1987.

No context found.

M. D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Computer Science Dept., University of California, Berkeley, 1987. Available as Technical Report UCB/CSD 87/381.

No context found.

Hill, M.D. "Aspects of cache memory and instruction buffer performance", Ph.D. Thesis, University of California, Berkeley, 1987.

No context found.

M. Hill, " Aspects of Cache Memory and Instruction Buffer Performance," Ph.D. Thesis, University of California at Berkeley, 1987.

No context found.

M. D. Hill. Aspects of Cache Memory and Instruction Buffer Performance. PhD thesis, Unversity of Berkeley, 1987.

No context found.

M. D. Hill. Aspects of cache memory and instruction buffer performance. PhD thesis, University of California, Berkeley, November 1987.

No context found.

M. D. Hill. Aspects of cache memory and instruction buffer performance. PhD thesis, University of California at Berkeley, 1987.

No context found.

Hill, M.D. Aspects of Cache Memory and Instruction Buffer Performance, Ph.D. dissertation, (published as Technical Report UCB/CSD 87/381), , November 1987.

No context found.

Hill, M.D., Aspects of Cache Memory and Instruction Buffer Performance, Ph.D. Thesis, U.C. Berkeley, Computer Science Division Tech. Rept. UCB/CSD 87/381, November 1987.

No context found.

M. D. Hill, `Aspects of cache memory and instruction buffer performance', Ph.D. Thesis, University of California at Berkeley, Berkeley, CA, November 1987. Also appears as Tech. Report UCB/CSD 87/381.

No context found.

HIL87 M. Hill. "Aspects of Cache Memory and Instruction Buffer Performance", Ph.D. Thesis, University of California at Berkeley, Computer Science Division, Technical Report UCB/CSD 87/381, November 1987.

No context found.

Hill, Mark D., "Aspects of Cache Memory and Instruction Buffer Performance," PhD dissertation, Computer Science Division, U.C. Berkeley, November 1987.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC