Andi,Cliff
Thanks for your in puts
Excuse me for the late response, I was on leave didn't see the reply
As you suggested I was thinking we can suggest distros to take the
latest version of numactl ie version 2.
But there is limitation, distro kernel is based on 2.6.18 and latest
numactl version 2.0 uses system call move_pages which is implemented on
2.6.18-rc1 kernel and above. Numactl build fails with distro kernel.
I tried to back port the patch, but looks like lot of functions
definitions have changed
So is it possible to keep the patch that I suggested, it serves our
purposed for the time being and later as the distros move to later
kernels we can suggest to take Numactl version 2 and above
Some observation about the patch
Andi Kleen wrote:
>> Index: numactl-0.9.8/Makefile
>> ===================================================================
>> --- numactl-0.9.8.orig/Makefile 2008-04-22 14:46:54.000000000 +0530
>> +++ numactl-0.9.8/Makefile 2008-04-22 14:47:22.000000000 +0530
>> @@ -1,6 +1,6 @@
>> # these can (and should) be overridden on the make command line for
production
>> # use
>> -CFLAGS := -g -Wall -O0
>> +CFLAGS := -g -Wall -O0 -D_GNU_SOURCE
>
> Please put the _GNU_SOURCE define into the source file. Users should
be able to easily
I tried that option, #define _GNU_SOURCE, but it was not working, hence
I provided this hack.
+#define BITS_PER_BYTE 8
+int get_max_cpu(void)
+{
+ int max = 0;
+ int index;
+ cpu_set_t mask;
+ if( numa_sched_getaffinity(0, sizeof(cpu_set_t), &mask) < 0 ){
+ perror("numa_sched_getaffinity:");
Ah noticed another problem. Please don't rely on the glibc cpuset_t,
there's no guarantee it's big enough.
Instead should I use a buffer with 2048 bits ?
Please let me know your thoughts, accordingly I can modify the patch
Thanks
Sharyathi
--
To unsubscribe from this list: send the line "unsubscribe linux-numa" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
