[Boost-bugs] [Boost C++ Libraries] #10155: thread::physical_concurrency() always returns 0 on PowerPC

Subject: [Boost-bugs] [Boost C++ Libraries] #10155: thread::physical_concurrency() always returns 0 on PowerPC
From: Boost C++ Libraries (noreply_at_[hidden])
Date: 2014-06-26 19:39:39

#10155: thread::physical_concurrency() always returns 0 on PowerPC
-------------------------------------------------+-------------------------
 Reporter:  Harris Munir <hamunir@…>             |      Owner:  anthonyw
     Type:  Bugs                                 |     Status:  new
Milestone:  To Be Determined                     |  Component:  thread
  Version:  Boost Release Branch                 |   Severity:  Problem
 Keywords:  thread, threads, pthread, powerpc,   |
  ppc, power                                     |
-------------------------------------------------+-------------------------
 The thread::physical_concurrency() function, located in
 src/pthread/thread.cpp, always returns 0 on the PowerPC platform. I
 believe this is because /proc/cpuinfo, which the function uses (on linux
 platforms) to determine the number of processors/cores on the machine, is
 formatted differently on PowerPC than it is on x86 and other platforms.
 Here is an example from the machine I am testing on (Ubuntu 14.04 LTS
 ppc64le):

 {{{
 $ cat /proc/cpuinfo
 processor       : 0
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 1
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 2
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 3
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 4
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 5
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 6
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 processor       : 7
 cpu             : POWER7 (raw), altivec supported
 clock           : 3220.000000MHz
 revision        : 2.3 (pvr 003f 0203)

 timebase        : 512000000
 platform        : pSeries
 model           : IBM pSeries (emulated by qemu)
 machine         : CHRP IBM pSeries (emulated by qemu)
 }}}

 Note that the "physical id" and "core id" lines that the function expects
 are not present. The function does not account for this, and returns 0 as
 the number of processors/cores on the system. This causes
 test_physical_concurrency_lib.test and test_physical_concurrency.test to
 fail.

 I have created a pull request with a tentative fix, which uses the
 hardware_concurrency() function in the same file to determine the number
 of processors/cores on Power systems. It seems to resolve the two failing
 test cases without any regressions. The request can be found here:
 https://github.com/boostorg/thread/pull/16

-- 
Ticket URL: <https://svn.boost.org/trac/boost/ticket/10155>
Boost C++ Libraries <http://www.boost.org/>
Boost provides free peer-reviewed portable C++ source libraries.