|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| I didn't suppose enterprise software to work this way. |
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| Windows hasn't been capable of searching file names since NT4, they ruined it in Windows 2000 by assuming that you want to search the contents of files |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| The sum of all "commit size"s in processes tab is magnitude smaller than "Commit" in "perfomance" tab in task manager.
What am I missing?
I now also know about how I should formalize my question about memory usage which I tried to discuss above: how can I calculate the amount of memory which a program may allocate without causing swapping using the measurements from resource monitor? |
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| The processes tab shows just processes, the performance tab shows the entire OS, kernel and all caches.
Total physical RAM - total commit + caches + system code (at a very late hour guess). |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| Excuse me please, I have underestimated the number of processes which Opera created. The sum of all commit sizes in processes sums quite well to commit in perfomance tab.
Quote C:Amie - 2015-07-03 2:39 AM
Total physical RAM - total commit + caches + system code (at a very late hour guess).
I thought out quite simple way of testing it. I will write a program which incrementally allocates memory and does a pass over all allocated memory after each allocation. The first increase of cycle length will be the sign of the swapping.
I will report the results. Edited by Alt Bass 2015-07-03 10:41 AM
|
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| Isn't relying on the timer length for the generation of a hard fault a little risky? Hard faults aren't exclusively caused by page operations, apps can actually request them! |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| Quote A hard fault happens when the address in memory of part of a program is no longer in main memory,
If an application is a several kilobytes thing without any imports other than something basic which does not do anything except allocating and calculating using the allocated memory, what can be the source of hard fault for it? Edited by Alt Bass 2015-07-03 7:17 PM
|
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| I mean, other than page fault. |
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| Oh if you're only interest on whether or not YOUR application is hard faulting, that's less of an issue, unless the resource scheduler is busy.
You can simply prevent your program from paging at all by performing a locked allocation:
https://msdn.microsoft.com/en-us/library/aa366895(VS.85 ).aspx |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| Let me specify it then.
Imagine that application is allocating memory quickly. I want to know which is the moment after which the Indows won't be able to serve malloc (or however Indows calls it, it does not matter) without swapping out something else including the memory used by application.
I cannot see any formalizeable way of detecting whether
the application is causing other applications to swap (other than monitoring all working sets of suspended processes and noting the moment when they start decresing)
or
the application is causing the file cache flushing thus degrading the FS performance.
I do not necessary need to prevent the testing program from partially swapping. If OS swaps the testing program patially, that would be an interesting result.
Edited by Alt Bass 2015-07-04 11:26 PM
|
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| Why not monitor the Total Virtual Memory used value for expansion?
MEMORYSTATUSEX memInfo;
DWORDLONG virtualMemUsed = memInfo.ullTotalPageFile - memInfo.ullAvailPageFile; |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| Please elaborate. |
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| I have discussed too much questions not related to KB installation here already. So...
I have found another nasty property of Windows. Minimized programs will be swapped out if PC has intensive HDD usage for some time
The program requiring much enough memory to work will take inexcusably long time to maximize/restore. Like, a Photoshop will require 20-30 seconds to restore even if the physical memory usage will be under 70% all the time (8 GB of RAM).
Why is that so?
Edited by Alt Bass 2015-08-21 9:54 PM
|
|
|
|
H/PC Sensei Posts: | 1,169 |
Location: | Russia | Status: | |
| BSOD
System memory dump ..... 100%.
Contact your system administrator blablabla.
After reboot, there is no kernel memory dump file.
wat |
|
|
|
Administrator H/PC Oracle Posts: | 17,987 |
Location: | United Kingdom | Status: | |
| is it doing a dump or a mini dump? |
|
|