Results 1 - 10
of
18,214
Table 3: Overhead of software-managed address translation
1996
Cited by 1
TABLE II Address translation behavior
Table 3. Sumary of diferences in the virtual memory protection schemes During the memory address translation process, the TIMPS temporal access check is invoked just after the access mode permision check (see Figure 4). In this section, three virtual memory protection schemes for TIMPS are presented. Details of the specific hardware extensions and software components required to suport temporal access control are discused elsewhere [1].
"... In PAGE 14: ... Details of the specific hardware extensions and software components required to suport temporal access control are discused elsewhere [1]. Table3 summarizes the major differences of the three schemes. Hardware Reference Clock TIMPS requires a hardware reference clock that can be used to determine the time of aces during an aces check.... ..."
Table 3: Baseline application performance. \Seq quot; and \Sp1 quot; report sequential running time and speedup, respectively, on an Alewife machine without SVM. \SVM-Seq quot; reports sequential running time with SVM. \Ovhd quot; is the amount of SVM overhead. \SVM-Par quot; reports running time with SVM on a 32-node Alewife machine, \Sp2 quot; reports speedup with SVM on 32 nodes.
"... In PAGE 17: ...4. Table3 provides baseline performance numbers for our applications on Alewife without the overheads of software shared memory that would be incurred by a DSMP. The rst two columns report performance numbers on Alewife without any software address translation overhead, i.... In PAGE 18: ...3. Finally, the last column in Table3 , labeled \Sp2, quot; is the speedup attained on 32 nodes with software address translation (the ratio of the \SVM-Seq quot; and \SVM-Par quot; columns). Except for the Jacobi application, an application known for its excellent speedup, all our applications exhibit only modest to good speedups.... In PAGE 18: ... The synchronization components include both the overhead of executing synchronization code and waiting on synchronization conditions. Finally, the 32-processor SMP node size data points (the rightmost bars in Figures 6 through 13) are exactly the runtimes reported in the \SVM-Par quot; column of Table3 . As described earlier, these bars represent performance on an all-hardware DSM, so there is no MGS component.... ..."
Table 3: Baseline application performance. \Seq quot; and \Sp1 quot; report sequential running time and
2000
"... In PAGE 17: ...4. Table3 provides baseline performance numbers for our applications on Alewife without the overheads of software shared memory that would be incurred by a DSMP. The rst two columns report performance numbers on Alewife without any software address translation overhead, i.... In PAGE 18: ...3. Finally, the last column in Table3 , labeled \Sp2, quot; is the speedup attained on 32 nodes with software address translation (the ratio of the \SVM-Seq quot; and \SVM-Par quot; columns). Except for the Jacobi application, an application known for its excellent speedup, all our applications exhibit only modest to good speedups.... In PAGE 18: ... The synchronization components include both the overhead of executing synchronization code and waiting on synchronization conditions. Finally, the 32-processor SMP node size data points (the rightmost bars in Figures 6 through 13) are exactly the runtimes reported in the \SVM-Par quot; column of Table3 . As described earlier, these bars represent performance on an all-hardware DSM, so there is no MGS component.... ..."
Table 3: Baseline application performance. #5CSeq quot; and #5CSp1 quot; report sequential running time and
2000
"... In PAGE 17: ...4. Table3 provides baseline performance numbers for our applications on Alewife without the overheads of software shared memory that would be incurred by a DSMP. The #0Crst two columns report performance numbers on Alewife without any software address translation overhead, i.... In PAGE 18: ...3. Finally, the last column in Table3 , labeled #5CSp2, quot; is the speedup attained on 32 nodes with software address translation #28the ratio of the #5CSVM-Seq quot; and #5CSVM-Par quot; columns#29. Except for the Jacobi application, an application known for its excellent speedup, all our applications exhibit only modest to good speedups.... In PAGE 18: ... The synchronization components include both the overhead of executing synchronization code and waiting on synchronization conditions. Finally, the 32-processor SMP node size data points #28the rightmost bars in Figures 6 through 13#29 are exactly the runtimes reported in the #5CSVM-Par quot; column of Table3 . As described earlier, these bars represent performance on an all-hardware DSM, so there is no MGS component.... ..."
Table 64 - Translate address page - SEND DIAGNOSTIC
"... In PAGE 8: ............. 82 Table64 - Translate address page - SEND DIAGNOSTIC .... ..."
Table 1: VMMC-2 address translation cost
1997
Cited by 75
Table 1: VMMC-2 address translation cost
1997
Cited by 75
Results 1 - 10
of
18,214