how many of you have killed your swapfile?

[Toadster]

given that I've upgraded to 8GB of ram, I dumped my swapfile... crazy?

hmm - I haven't done any official benchmarking - but the 'seat of the pants' feel is that I've gained some performance...

what is the 'cutoff' to when you should remove your swapfile? 4GB? 8GB? more?

 

[bsobel]

Unless you needed to page, you werent using it anyhow. I have 32gig, still leave it on, some API's need it.

 

[pallejr]

bsobel, what APIs are those?

 

[TaylorTech]

Page Files are useless IMO, at least with home desktops. Unless you're running CAD all day and night, don't bother.

If you have 4GB or more I can't think of many programs that use it.

 

[bsobel]

Originally posted by: pallejr
bsobel, what APIs are those?

Creating null memory mapped files use them. And system wise, creating dump file requires one as well.

 

[Nothinman]

Page Files are useless IMO, at least with home desktops.

Not at all, it gives the system more breathing room since it has a place to dump pages that aren't backed by some file on disk.

 

[TC91]

i have mine set to 2gb on the beginning of of seagate 500gb drive (the wd 640gb is the system drive)

 

[taltamir]

killing the swap forces programs that insist on using it to use the ram instead. that or BSOD.
The game "Spellforce the order of dawn" used to BSOD on me back in the day, I found out that by setting the swap to over 3GB it will no longer do it (I dont think it ever took more then 300MB of ram + swap total, it will just occsionally try to address a random RAM address at the 4GB range, the maximum ram address on a 32bit machine... that was back when 2GB was wasteful overkill).

Its an interesting prospect, properly written programs know what to put in ram and what to put in pagefile. You are basically forcing badly written programs to use the ram for a performance boost, and even WORSTLY (not a word, i know) written programs to BSOD your computer.

 

[pallejr]

Originally posted by: bsobel

Originally posted by: pallejr
bsobel, what APIs are those?

Creating null memory mapped files use them. And system wise, creating dump file requires one as well.

Dump files, yes. But all the memory allocation functions will work fine without a paging file.

 

[Nothinman]

Its an interesting prospect, properly written programs know what to put in ram and what to put in pagefile.

Properly written programs should just allocate memory and not care whether the OS has to touch the pagefile to satisfy the allocation.

 

[bsobel]

Originally posted by: pallejr

Originally posted by: bsobel

Originally posted by: pallejr
bsobel, what APIs are those?

Creating null memory mapped files use them. And system wise, creating dump file requires one as well.

Dump files, yes. But all the memory allocation functions will work fine without a paging file.

One again, null memory mapped files going to the paging file. If there is no paging file or its too small for the request, the request will fail. Your welcome to go look at MSDN.

 

[bsobel]

Its an interesting prospect, properly written programs know what to put in ram and what to put in pagefile. You are basically forcing badly written programs to use the ram for a performance boost, and even WORSTLY (not a word, i know) written programs to BSOD your computer.

This all basically just wrong. User mode programs should never be able to BSOD your system, if they can its indicative of a kernel or driver bug (period).

 

[pallejr]

Originally posted by: bsobel

One again, null memory mapped files going to the paging file. If there is no paging file or its too small for the request, the request will fail. Your welcome to go look at MSDN.

null memory mapped files? you mean pagefile backed memory objects? They can live just fine without a paging file. I take for granted that there is enough free memory (you can also run out of memory even with a paging file).

 

[bsobel]

Originally posted by: pallejr

Originally posted by: bsobel

One again, null memory mapped files going to the paging file. If there is no paging file or its too small for the request, the request will fail. Your welcome to go look at MSDN.

null memory mapped files? you mean pagefile backed memory objects? They can live just fine without a paging file. I take for granted that there is enough free memory (you can also run out of memory even with a paging file).

They require a pagefile to exist or the api fails.

 

[pallejr]

They actually don't. They are very common objects. Everybody uses them. So if they would fail without a paging file, it would be impossible to run Windows without a paging file, and it is not.

 

[bsobel]

Originally posted by: pallejr
They actually don't. They are very common objects. Everybody uses them. So if they would fail without a paging file, it would be impossible to run Windows without a paging file, and it is not.

I suggest you go throw some code together and give it a whirl, and no, they arent exactly common.

 

[pallejr]

I have. Sysinternals have a tool, process explorer, it can show you all kinds of things. Pagefile backed sections live fine on my system with no paging file.

 

[bsobel]

Originally posted by: pallejr
I have. Sysinternals have a tool, process explorer, it can show you all kinds of things. Pagefile backed sections live fine on my system with no paging file.

What OS are you running? These definately fail on XP, admittidly I havent retested on Vista (this used to be a big abobe photoshop issue, it would fail without a swapfile)

And I'm very familiar with process explorer, but thanks

 

[pallejr]

I've run 2000,xp,vista. It works like a charm. Because they're backed by the paging file, doesn't mean they cannot live without it.

 

[bsobel]

Originally posted by: pallejr
I've run 2000,xp,vista. It works like a charm. Because they're backed by the paging file, doesn't mean they cannot live without it.

Im afraid your wrong, that was definately the old Adobe issue. If its working for you, then the system created a temporary paging file at boot. Go right some code and see for yourself

 

[pallejr]

This issue has been discussed on this forum before. I provided some documentation about it. I am not wrong.

About be writing some code... As I said earlier, I have. And if you check with process explorer, you'll see these memory objects used a lot of places. And they work great with no paging file.

 

[bsobel]

Originally posted by: pallejr
This issue has been discussed on this forum before. I provided some documentation about it. I am not wrong.

About be writing some code... As I said earlier, I have. And if you check with process explorer, you'll see these memory objects used a lot of places. And they work great with no paging file.

Youll find the system created a temp paging file for you.

 

[pallejr]

No I will not. Windows *may* create a small temporary paging file, if the system runs out of memory, so you have the oppunity to properly configure the system. I have no such file.

Pagefile backed sections just means it can be backed by the paging file. It will not fail if there is just enough physical memory to hold it.

One purpose of creating such an object is to share data between multiple processes. It would be rather poor design, if that wouldn't be possible without a paging file.

 

[Deadtrees]

Based on the applications you use, you may not face any problems at all. That doesn't mean everybody won't face problems by not having a pagefile.