On Sun, Oct 31, 2010 at 11:26 PM, loody <miloody@xxxxxxxxx> wrote:
> Dear all:
> I have some questions about virtual memory (VM) subsystem of the Linux
> kernel.
> 1. From linux doc, there is drop_caches in /proc/ which can let us to
> free the cache that kernel used, such as dentries and inodes, etc.
> I know i can echo 1, 2 or 3 to free different level of caches in the
> linux kernel, but is there any more elegant or programming rule that I
> can follow to let it used more efficiently?
I think you fail to understand why drop_caches exists. It is not to
enhance performance. It is to enhance performance testing.
As such the main userspace api seems sufficient.
If you really want to control the cache, I think you need to implement
it in userspace for your app and do your disk i/o with o_direct.
> 2. what kind of system call will increase the size of cache?
> I write a test program as below but from top, the size of cache seems
> no change
>
> #include<stdio.h>
> #include<stdlib.h>
> int main(void)
> {
> int index;
> char input;
> char * point[1024];
>
> for(index=0;index<1024;index++)
> {
> point[index]=malloc(1024*1024);
> printf("please input any charter for going down:\n");
> scanf("%c",&input);
> }
>
> }
>
> the memory usage of test.mips increased but the cached size is always
> 4916k as below:
>
> Mem: 8352K used, 132800K free, 0K shrd, 60K buff, 4916K cached
> Load average: 0.00 0.00 0.00
> PID USER STATUS VSZ PPID %CPU %MEM COMMAND
> 400 0 SW 668 1 0.3 0.4 busybox.telnetd
> 498 0 RW 584 408 0.3 0.4 top
> 500 0 SW 10156 317 0.0 7.1 test.mips
> 317 0 SW 596 1 0.0 0.4 sh
> 408 0 SW 592 400 0.0 0.4 sh
> 1 0 SW 584 0 0.0 0.4 init
> 5 0 SW< 0 2 0.0 0.0 khelper
> 86 0 SW< 0 2 0.0 0.0 khubd
> 3 0 SW< 0 2 0.0 0.0 ksoftirqd/0
> 76 0 SW< 0 2 0.0 0.0 kblockd/0
> 4 0 SW< 0 2 0.0 0.0 events/0
> 2 0 SW< 0 0 0.0 0.0 kthreadd
> 112 0 SW 0 2 0.0 0.0 pdflush
> 113 0 SW< 0 2 0.0 0.0 kswapd0
> 8 0 SW< 0 2 0.0 0.0 async/mgr
> 116 0 SW< 0 2 0.0 0.0 crypto/0
> 252 0 SW< 0 2 0.0 0.0 mtdblockd
> 254 0 SW< 0 2 0.0 0.0 nftld
> 255 0 SW< 0 2 0.0 0.0 inftld
> 286 0 SW< 0 2 0.0 0.0 rpciod/0
> 253 0 SW< 0 2 0.0 0.0 ftld
> 89 0 SW< 0 2 0.0 0.0 kseriod
> 111 0 SW 0 2 0.0 0.0 pdflush
> 448 0 SW< 0 2 0.0 0.0 scsi_eh_1
> 449 0 SW< 0 2 0.0 0.0 usb-storage
> 114 0 SW< 0 2 0.0 0.0 nfsiod
> 115 0 SW< 0 2 0.0 0.0 cifsoplockd
>
> Mem: 8352K used, 132800K free, 0K shrd, 60K buff, 4916K cached
> Load average: 0.04 0.01 0.00
> PID USER STATUS VSZ PPID %CPU %MEM COMMAND
> 501 0 RW 584 408 0.5 0.4 top
> 400 0 SW 668 1 0.3 0.4 busybox.telnetd
> 500 0 SW 37912 317 0.0 26.8 test.mips
> 317 0 SW 596 1 0.0 0.4 sh
> 408 0 SW 592 400 0.0 0.4 sh
> 1 0 SW 584 0 0.0 0.4 init
> 5 0 SW< 0 2 0.0 0.0 khelper
> 86 0 SW< 0 2 0.0 0.0 khubd
> 3 0 SW< 0 2 0.0 0.0 ksoftirqd/0
> 76 0 SW< 0 2 0.0 0.0 kblockd/0
> 4 0 SW< 0 2 0.0 0.0 events/0
> 2 0 SW< 0 0 0.0 0.0 kthreadd
> 112 0 SW 0 2 0.0 0.0 pdflush
> 113 0 SW< 0 2 0.0 0.0 kswapd0
> 8 0 SW< 0 2 0.0 0.0 async/mgr
> 116 0 SW< 0 2 0.0 0.0 crypto/0
> 252 0 SW< 0 2 0.0 0.0 mtdblockd
> 254 0 SW< 0 2 0.0 0.0 nftld
> 255 0 SW< 0 2 0.0 0.0 inftld
> 286 0 SW< 0 2 0.0 0.0 rpciod/0
> 253 0 SW< 0 2 0.0 0.0 ftld
> 89 0 SW< 0 2 0.0 0.0 kseriod
> 111 0 SW 0 2 0.0 0.0 pdflush
> 448 0 SW< 0 2 0.0 0.0 scsi_eh_1
> 449 0 SW< 0 2 0.0 0.0 usb-storage
> 114 0 SW< 0 2 0.0 0.0 nfsiod
> 115 0 SW< 0 2 0.0 0.0 cifsoplockd
>
> appreciate your kind help,
> miloody
Greg
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
