flat assembler
Message board for the users of flat assembler.
Index
> Windows > Using memory on video card as swap |
Author |
|
LocoDelAssembly 27 Jul 2007, 19:20
I wonder if something around this exists under Windows.
I had the idea an hour ago and when I searched on Google I couldn't find nothing relevant on Windows If it doesn't exists, do you think it would be possible by using a driver which create a DirectDraw surface to act as a cache for the swap file? I mean at driver level (not patching the graphic driver) because at the API level I think the surface could be lost in some situations. |
|||
27 Jul 2007, 19:20 |
|
LocoDelAssembly 27 Jul 2007, 20:19
bogdanontanu, I think that VRAM is faster than an IDE HDD, don't you think the same?
About USB stick, it really sucks and is clearly slower than a HDD. And no, the computer is too old for memory upgrade but enjoys of a GeForce FX 5200 with 128 MB VRAM which a big part is unused most of the time Also remember that I mean use it as cache for the swap file, not making it part of the physical RAM memory so these slow reads would be noticed in the swap-in operations only. Additionally the driver could reduce the cache size when legit DirectX use is detected, but in my particular case the most hard graphic task the computer does is web browsing. |
|||
27 Jul 2007, 20:19 |
|
f0dder 28 Jul 2007, 12:27
Loco: USB sticks have plenty faster random access time though.
Using video memory for non-graphics stuff is a fun idea, but... *shrug*. Obviously even (relatively) very slow read from video memory is going to be faster than a harddrive, but 128 megabytes (and of course even less because parts of it will be used at runtime) is just too little to be usable in a paging situation. Interesting idea, but dunno if it's useful or doable. |
|||
28 Jul 2007, 12:27 |
|
Maverick 28 Jul 2007, 12:54
Another cool idea would be to use the speakers, with a vocal synthesizer seeking collaboration with the user. I can imagine the PC saying "hey, it's you that benefit from me, so move your ass, grab a piece of paper, a pen, and start writing down: page 43435 = 00101010111101010100011010001001011...[thousands years later]....01011111000 keep it in a safe place, I'll need to ask you the data when the page has to be loaded again in physical memory. Thank you for your collaboration."
The only problem I can see is that it needs a microphone to work. |
|||
28 Jul 2007, 12:54 |
|
LocoDelAssembly 28 Jul 2007, 17:05
f0dder, yes is very few memory, actually I expect LOWER performance on situations where more that 64 MB of paging is in constant use. However I think that in a "calm" situation (which is more frecuently) it could have its benefits.
About the USB memory sticks, are them faster to write than a HDD? Apart, I don't want to write much on such media, you can't write on it forever, with thousands of writes you end up destroying the memory while HDDs lasts more. Maverick, hey, I don't need thousands of years to write a 4 K page on paper and I have a microphone. When can we start to develop you idea? PS: The computer already has 320 MB of RAM, this extra memory would be only for a little push. |
|||
28 Jul 2007, 17:05 |
|
f0dder 29 Jul 2007, 17:43
LocoDelAssembly wrote:
Slower to read and write than harddrives, although late-model usb sticks are getting decent performance (do be careful that some manufacturers list mbit and not mbyte transfer rates, without clear indication). It's the access time of the flash drives that make them interesting, much much faster than harddrives because there's no moving mechanical parts. And the amount of write cycles before the memory is ruined has also been improved vastly... |
|||
29 Jul 2007, 17:43 |
|
rugxulo 31 Jul 2007, 06:21
Doesn't the PS3 share its VRAM with its main memory?
|
|||
31 Jul 2007, 06:21 |
|
< Last Thread | Next Thread > |
Forum Rules:
|
Copyright © 1999-2024, Tomasz Grysztar. Also on GitHub, YouTube.
Website powered by rwasa.