Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 1 - 10
of
16,825
Table 1: Data lifetime statistics for heap allocated memory. Allocated is the total amount of heap memory allocated during each run. Written is the amount of allocated memory that was actually written, and Ideal Lifetime is the lifetime this written data would have if it were zeroed immediately after the last time it was read. Written amp; Freed is allocated bytes that were written and later freed, with Secure Deallocation Lifetime the lifetime of this data when it is zeroed at deallocation time. Finally, Written amp; Freed amp; Overwritten is allocated bytes that were written and freed, then later reallocated and overwritten by the new owner, with Natural Lifetime the lifetime of this data.
"... In PAGE 8: ... 5.3 Results Table1 summarizes the results of our experiments. We ran each application through our modified Valgrind, recorded timings for various memory events, and com- puted the resultant data lifetimes.... In PAGE 9: ...fter its last use, i.e. it very closely approximates ideal lifetime. Comparing the Written and Written amp; Freed columns in Table1 , we can see that most programs free most of the data that they use. Comparing Ideal Lifetime to Se- cure Deallocation Lifetime, we can also see that most do so promptly, within about a second of the end of the ideal lifetime.... In PAGE 9: ...1), it would take just over half a second, an insignificant penalty for a 46-minute experiment. GUI Programs Table1 reveals that GUI programs of- ten delay deallocation longer than other programs, result- ing in a much greater secure deallocation lifetime than others. One reason for this is that GUI programs generally use data for a short period of time while rendering a page of output, and then wait while the user digests the informa- tion.... In PAGE 12: ... When we combine these results with our observations about common application memory behavior, we see that zeroing speeds far outpace the rate at which memory is allocated and freed. Even the worst memory hogs we saw in Table1 only freed on the order of hundreds of MB of data throughout their entire lifetime, which incurs only a fraction of a second of penalty at the slowest, bus- bandwidth zeroing rate (2 GB/s). 7.... ..."
Table 1: Data lifetime statistics for heap allocated memory. Allocated is the total amount of heap memory allocated during each run. Written is the amount of allocated memory that was actually written, and Ideal Lifetime is the lifetime this written data would have if it were zeroed immediately after the last time it was read. Written amp; Freed is allocated bytes that were written and later freed, with Secure Deallocation Lifetime the lifetime of this data when it is zeroed at deallocation time. Finally, Written amp; Freed amp; Overwritten is allocated bytes that were written and freed, then later reallocated and overwritten by the new owner, with Natural Lifetime the lifetime of this data.
"... In PAGE 8: ... 5.3 Results Table1 summarizes the results of our experiments. We ran each application through our modi ed Valgrind, recorded timings for various memory events, and com- puted the resultant data lifetimes.... In PAGE 9: ...fter its last use, i.e. it very closely approximates ideal lifetime. Comparing the Written and Written amp; Freed columns in Table1 , we can see that most programs free most of the data that they use. Comparing Ideal Lifetime to Se- cure Deallocation Lifetime, we can also see that most do so promptly, within about a second of the end of the ideal lifetime.... In PAGE 9: ...1), it would take just over half a second, an insigni cant penalty for a 46-minute experiment. GUI Programs Table1 reveals that GUI programs of- ten delay deallocation longer than other programs, result- ing in a much greater secure deallocation lifetime than others. One reason for this is that GUI programs generally use data for a short period of time while rendering a page of output, and then wait while the user digests the informa- tion.... In PAGE 12: ... When we combine these results with our observations about common application memory behavior, we see that zeroing speeds far outpace the rate at which memory is allocated and freed. Even the worst memory hogs we saw in Table1 only freed on the order of hundreds of MB of data throughout their entire lifetime, which incurs only a fraction of a second of penalty at the slowest, bus- bandwidth zeroing rate (2 GB/s). 7.... ..."
Table 4.1: E ciency of the component operations The measurements were taken on a 40Mhz Sun IPX. Unaccessed means the page has neither been read nor written. Clean means the page has been read since the last call to mprotect. Dirty means the page has been written since the last call to mprotect. RW- RW means successive calls to mprotect always grant full access to the page. RO-RW means successive calls to mprotect alternate between restricting access and restoring access.
Table 4.1: E ciency of the component operations The measurements were taken on a 40Mhz Sun IPX. Unaccessed means the page has neither been read nor written. Clean means the page has been read since the last call to mprotect. Dirty means the page has been written since the last call to mprotect. RW- RW means successive calls to mprotect always grant full access to the page. RO-RW means successive calls to mprotect alternate between restricting access and restoring access.
Table 1. The phases of the edit-compile-debug benchmark. The first two columns describe the phase of the benchmark. The third column gives the number of bytes read and written by each phase. The last column gives the size of the largest virtual memory image of the phase. The last row is not a phase in the benchmark but rather shows the total amount of memory required by the basic environment in which the benchmark is running.
1990
"... In PAGE 3: ... The results obtained on Sprite should be similar to results obtained with the same benchmarks on systems with mapped-files. The first benchmark that I used is an edit-compile-debug (ECD) benchmark that runs under the X11 window system on Sprite (see Table1 ). This benchmark represents work that is commonly done on Sprite, and is both VM and FS intensive.... ..."
Cited by 20
Table 4.9: Analysis of the write absorption provided by the MOB under the simulator workload. The last two sets of columns show the average number of dirty objects present in the MOB for each page that is written out to disk.
1995
Cited by 33
Table 4.9: Analysis of the write absorption provided by the MOB under the simulator workload. The last two sets of columns show the average number of dirty objects present in the MOB for each page that is written out to disk.
1995
Table 1. Results of an extrapolation of our nite size estimates (1500 N 104) based on (4) with di erent choices of N. a) N = 0; b) N BS=N; c) N CS=(N ln N). The numbers in parentheses represent statistical errors obtained by 2 analysis, in units of the last written digit.
Table 5. Basic I/O Commands Supported by LMS
2005
"... In PAGE 36: ... Storage Processing Operations In the same manner that the CPU of a computer executes instructions, a storage device controller such as Little Man Storage is capable of executing a pre-defined group of commands that create, delete, store, retrieve and process data. Although some storage devices support a wider range of operations, we limit LMS to five commands as shown in Table5 . LMS processes complete files and individual records must be identified in the application programs (since storage devices are unaware of logical records).... In PAGE 43: ... Program memory content for program 2 45 Load input register from memory location 5 (zero) 20 Add memory location 0 (column) to input register 62 Save result in memory location 2 (column step) 63 Save result in memory location 3 (row step) c6 lp1: Output from memory location 6 (symbol quot;* quot;) 44 Load input register from memory location 4 (neg one) 22 Add from memory location 2 (column step) 62 Save result in memory location 2 (colmn step) a9 Jump on zero to lp2: 83 Jump to lp1: 23 lp2: Add from memory location 3 (row step) 63 Save result in memory location 3 (row step) b6 Jump on zero to :hlt c7 Output from memory location 7 (new line) 45 Load input register from memory location 5 (zero) 20 Add memory location 0 (column) to input register 44 Load input register from memory location 4 (neg one) 20 Add memory location 0 (column) to input register 60 Save result in memory location 0 (column) 45 Load input register from memory location 4 (neg one) 20 Add memory location 0 (column) to input register 62 Save result in memory location 2 (column step) 83 Jump to lp1: eo hlt: Halt Table 6. Program memory content for program 3 06 column (size of triangle) 03 row (not used in program) 00 column step 00 row step ff negative one (allows decrementing ) 00 zero 2a symbol quot;* quot; 0d new line Table5 . Data memory content for program 3 20 Space 20 Space 48 H 45 E 4c L 4c L 4f O 20 Space 57 W 4f O 52 R 4c L 44 D 21 ! 0d New Line Table 3.... ..."
Table 1: Experimental determinations of the electron g-factor anomaly (a = (g ? 2)=2). The error is written in the parenthesis after the value and refers to the last digits. The square boxes represent atomic level experiments, the stars free electron spin precession ex- periments, and the diamonds free electron spin resonance experiments. A bullet indicates a reevaluation of earlier experiments with no new data taken.
"... In PAGE 11: ... In an experiment from 1963 by Farago, Gardiner, Muir and Rae [6] a transverse electric eld was explicitly introduced to control the number of cyclotron revolutions. The experimental precision (see Table1 ) was limited by unsurmountable technical... ..."
Results 1 - 10
of
16,825
