From: "Bert Wesarg" <bert.wesarg@xxxxxxxxxxxxxx> Subject: Re: [RESEND][RFC] lscpu - CPU architecture information helper Date: Tue, 8 Jul 2008 18:04:21 +0200 > On Tue, Jul 8, 2008 at 14:54, Cai Qian <qcai@xxxxxxxxxx> wrote: > > Hi, > > > > Thanks for all your comments. I have rewritten the tool in C, and made > > several changes. The new output is like the following, > > > > $ /usr/bin/lscpu > > CPU(s): 8 > > Thread(s) per core: 2 > > Core(s) per socket: 2 > > CPU socket(s): 2 > > NUMA node(s): 1 > > Vendor ID: GenuineIntel > > CPU family: Itanium 2 > > Model: 0 > > CPU MHz: 1598.000005 > > L1d cache: 16K > > L1i cache: 16K > > L2d cache: 256K > > L2i cache: 1024K > > L3 cache: 12288K > > > > $ /usr/bin/lscpu -p > > #The following is the parsable format, which can be fed to other > > #programs. Each different item in every column has a unique ID > > #starting from zero. > > # > > #CPU,Core,Socket,Node,L1d,L1i,L2d,L2i,L3 > > 0,0,0,0,0,0,0,0,0 > > 1,0,0,0,0,0,0,0,0 > > 2,1,0,0,1,1,1,0,0 > > 3,1,0,0,1,1,1,0,0 > > 4,2,1,0,2,2,2,1,1 > > 5,2,1,0,2,2,2,1,1 > > 6,3,1,0,3,3,3,1,1 > > 7,3,1,0,3,3,3,1,1 > > > > If you are happy about the output. I'll tidy up the code a little bit, > > do more testing, and create a manpage. > > > > Thanks, > > CaiQian > > > > /* > > lscpu - CPU architecture information helper > > Copyright (C) 2008 Cai Qian <qcai@xxxxxxxxxx> > > > > This program is free software: you can redistribute it and/or modify > > it under the terms of the GNU General Public License as published by > > the Free Software Foundation, either version 3 of the License, or > > (at your option) any later version. > > > > This program is distributed in the hope that it will be useful, > > but WITHOUT ANY WARRANTY; without even the implied warranty of > > MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > > GNU General Public License for more details. > > > > You should have received a copy of the GNU General Public License > > along with this program. If not, see <http://www.gnu.org/licenses/>. > > */ > > > > #include <ctype.h> > > #include <dirent.h> > > #include <err.h> > > #include <errno.h> > > #include <fcntl.h> > > #include <getopt.h> > > #include <math.h> > > #include <stdio.h> > > #include <stdlib.h> > > #include <string.h> > > #include <sys/stat.h> > > #include <sys/types.h> > > #include <sys/utsname.h> > > #include <unistd.h> > > > > #define BASE 2 > > #define CACHE_MAX 100 > > #define MARK_LABEL ":" > > > > /* Calculate on number of silbings from a decimal mapping. */ > > int > > sibling (double mapping) > > { > > int i = 0, j = 0; > > > > while (mapping != 0) > > { > > i++; > > j = (int) (log (mapping) / log (BASE)); > > mapping -= pow (BASE, j); > > } > > > > return i; > > } > > > > /* Convert hexdecimal number from a mapping file to decimal.*/ > > double > > decimal (char *file) > So you extend your supported number of cpus from 32 to 52, great. You > should either look inside the kernel code, kernel documentationm or > into the libbitmask(3) from Paul Jackson/SGI [1]. > Hmm, it has been tested successfully on a 64 CPUs SGI machine though, $ lscpu CPU(s): 64 Thread(s) per core: 1 Core(s) per socket: 1 CPU socket(s): 64 NUMA node(s): 32 Vendor ID: GenuineIntel CPU family: Itanium 2 Model: 2 CPU MHz: 1500.000000 L1d cache: 16K L1i cache: 16K L2 cache: 256K L3 cache: 4096K $ lscpu -p #The following is the parsable format, which can be fed to other #programs. Each different item in every column has a unique ID #starting from zero. # #CPU,Core,Socket,Node,L1d,L1i,L2,L3 0,0,0,0,0,0,0,0 1,1,1,0,1,1,1,1 2,2,2,1,2,2,2,2 3,3,3,1,3,3,3,3 4,4,4,2,4,4,4,4 5,5,5,2,5,5,5,5 6,6,6,3,6,6,6,6 7,7,7,3,7,7,7,7 8,8,8,4,8,8,8,8 9,9,9,4,9,9,9,9 10,10,10,5,10,10,10,10 11,11,11,5,11,11,11,11 12,12,12,6,12,12,12,12 13,13,13,6,13,13,13,13 14,14,14,7,14,14,14,14 15,15,15,7,15,15,15,15 16,16,16,8,16,16,16,16 17,17,17,8,17,17,17,17 18,18,18,9,18,18,18,18 19,19,19,9,19,19,19,19 20,20,20,10,20,20,20,20 21,21,21,10,21,21,21,21 22,22,22,11,22,22,22,22 23,23,23,11,23,23,23,23 24,24,24,12,24,24,24,24 25,25,25,12,25,25,25,25 26,26,26,13,26,26,26,26 27,27,27,13,27,27,27,27 28,28,28,14,28,28,28,28 29,29,29,14,29,29,29,29 30,30,30,15,30,30,30,30 31,31,31,15,31,31,31,31 32,32,32,16,32,32,32,32 33,33,33,16,33,33,33,33 34,34,34,17,34,34,34,34 35,35,35,17,35,35,35,35 36,36,36,18,36,36,36,36 37,37,37,18,37,37,37,37 38,38,38,19,38,38,38,38 39,39,39,19,39,39,39,39 40,40,40,20,40,40,40,40 41,41,41,20,41,41,41,41 42,42,42,21,42,42,42,42 43,43,43,21,43,43,43,43 44,44,44,22,44,44,44,44 45,45,45,22,45,45,45,45 46,46,46,23,46,46,46,46 47,47,47,23,47,47,47,47 48,48,48,24,48,48,48,48 49,49,49,24,49,49,49,49 50,50,50,25,50,50,50,50 51,51,51,25,51,51,51,51 52,52,52,26,52,52,52,52 53,53,53,26,53,53,53,53 54,54,54,27,54,54,54,54 55,55,55,27,55,55,55,55 56,56,56,28,56,56,56,56 57,57,57,28,57,57,57,57 58,58,58,29,58,58,58,58 59,59,59,29,59,59,59,59 60,60,60,30,60,60,60,60 61,61,61,30,61,61,61,61 62,62,62,31,62,62,62,62 63,63,63,31,63,63,63,63 > > /* number of CPUs */ > > for (;;) > > { > > sprintf (buf, "%s/cpu/cpu%d", syspath, cpu); > > if (stat (buf, &info) == 0) > > cpu++; > > else > > break; > > } > Again, no support for holes in the cpu range. But I'm currently unsure > if holes are actually possible. What about using readdir and check for > a cpu%d pattern of the dir name with sscanf? > There should not be a hole. Otherwise, a Kernel bug? I have tried to offline a CPU, but it does still has a entry there. > > if (have_topology) > > { > > /* number of threads */ > > sprintf (buf, "%s/topology/thread_siblings", cpu0path); > > thread = sibling (decimal (buf)); > > > > /* number of cores */ > > sprintf (buf, "%s/topology/core_siblings", cpu0path); > > core = sibling (decimal (buf)) / thread; > > > > /* number of sockets */ > > socket = cpu / core / thread; > Sockets can also be counted by the physical_package_id topology > attribute. But some older architectures have a bug with this [2]. > > > if (have_node) > > { > > /* number of NUMA node */ > > for (;;) > > { > > sprintf (buf, "%s/node/node%d", syspath, node); > > if (stat (buf, &info) == 0) > > node++; > > else > > break; > > } > > > > /* information about how nodes share different CPUs */ > > sprintf (buf, "%s/cpumap", node0path); > > nodecpu = sibling (decimal (buf)); > What about nodes >0, in node0/cpumap are only the cpus from node0, you > should also check the cpumaps from the other nodes. > So there is a possibility that different nodes have different number of CPUs? If so, that looks like too complicated to handle in this program, and I'll probably remove NUMA information at all. > By the way, you check only cpu0 for topology and cache infos, I don't > know if it is possible to have cpus of different topology/cache types > in one system. Anyone? > I don't know either, but the program at the moment does not handle the system installed with different physical processors. CaiQian > Regards > Bert > > [1] http://oss.sgi.com/projects/cpusets/ > [2] http://lkml.org/lkml/2008/5/13/429 -- To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
