Changeset 92443 in vbox
- Timestamp:
- Nov 15, 2021 9:38:37 PM (3 years ago)
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
trunk/src/VBox/VMM/VMMR3/NEMR3Native-win.cpp
r92442 r92443 2820 2820 * @endverbatim 2821 2821 * 2822 * NEM w/ simplified memory and but no MEM_LARGE_PAGES: 2823 * @verbatim 2824 bs3-memalloc-1: From 0x100000000 to 0xc20000000 2825 bs3-memalloc-1: Pages : 11 665 408 pages 2826 bs3-memalloc-1: MiBs : 45 568 MB 2827 bs3-memalloc-1: Alloc elapsed : 90 062 027 900 ns 2828 bs3-memalloc-1: Alloc elapsed in ticks : 269 754 826 466 ticks 2829 bs3-memalloc-1: Page alloc time : 7 720 ns/page 2830 bs3-memalloc-1: Page alloc time in ticks : 23 124 ticks/page 2831 bs3-memalloc-1: Alloc thruput : 129 526 pages/s 2832 bs3-memalloc-1: Alloc thruput in MiBs : 505 MB/s 2833 bs3-memalloc-1: Allocation speed : PASSED 2834 bs3-memalloc-1: Access elapsed : 3 596 017 220 ns 2835 bs3-memalloc-1: Access elapsed in ticks : 10 770 732 620 ticks 2836 bs3-memalloc-1: Page access time : 308 ns/page 2837 bs3-memalloc-1: Page access time in ticks : 923 ticks/page 2838 bs3-memalloc-1: Access thruput : 3 243 980 pages/s 2839 bs3-memalloc-1: Access thruput in MiBs : 12 671 MB/s 2840 bs3-memalloc-1: 2nd access : PASSED 2841 bs3-memalloc-1: Access elapsed : 133 060 160 ns 2842 bs3-memalloc-1: Access elapsed in ticks : 398 459 884 ticks 2843 bs3-memalloc-1: Page access time : 11 ns/page 2844 bs3-memalloc-1: Page access time in ticks : 34 ticks/page 2845 bs3-memalloc-1: Access thruput : 87 670 178 pages/s 2846 bs3-memalloc-1: Access thruput in MiBs : 342 461 MB/s 2847 bs3-memalloc-1: 3rd access : PASSED 2848 * @endverbatim 2849 * 2822 2850 * Same everything but native VT-x and VBox (stripped output a little): 2823 2851 * @verbatim … … 2875 2903 bs3-memalloc-1: 3rd access : PASSED 2876 2904 * @endverbatim 2905 * 2906 * Comparing large pages, therer is an allocation speed difference of two order 2907 * of magnitude. When disabling large pages in VBox the allocation numbers are 2908 * closer, and the is clear from the 2nd and 3rd access tests that VBox doesn't 2909 * spend enough memory on nested page tables as Hyper-V does. The similar 2nd 2910 * and 3rd access numbers the two large page testruns seems to hint strongly at 2911 * Hyper-V eventually getting the large pages in place too, only that it sucks 2912 * hundredfold in the setting up phase. 2913 * 2877 2914 * 2878 2915 *
Note:
See TracChangeset
for help on using the changeset viewer.