NCBI C Toolkit Cross Reference

C/make/readme.aix

  1 
  2 General Information for BLAST Developers on IBM-AIX Systems
  3 ===========================================================
  4 
  5  To compile the toolkit under AIX follow the steps in readme.unx
  6  SPECIAL NOTE:
  7  To build the 64-bit version type:
  8  $ env AIX_MODE=64 ./ncbi/make/makedis.csh or
  9  $ env AIX_MODE=64 ./ncbi/make/makedis.csh 2> log 1>&2    ! Korn shell users
 10  % env AIX_MODE=64 ./ncbi/make/makedis.csh >& log         ! C shell users
 11 
 12  AIX Specific information
 13  ------------------------
 14 
 15  The following files are the main platform dependent files
 16 
 17  1.  makedis.csh in ./ncbi/make
 18 
 19  2.  ibm64_auto.ncbi.mk and ibm_auto.ncbi.mk in ./ncbi/platform
 20 
 21  3.  ncbilcl.ibm in ./ncbi/corelib
 22 
 23 
 24 makedis.csh:
 25 
 26      This file identifies the machine model via uname -M and assigns
 27      the corresponding architecture.  Currently, AIX Power3 is the
 28      only architecture explicitly targeted.  All other architectures
 29      are identified as auto.  auto refers to the compiler flag that
 30      will check the architecture and build corresponding executables.
 31      The auto flag assumes that the execution environment is the same
 32      as the compilation environment.
 33 
 34 ibm_xxxx.ncbi.mk:
 35 
 36      Currently xxxx can be either auto64 or auto.  auto64 builds the
 37      64-bit version.
 38      This file contains compiler and loader flags specific for IBM.
 39 
 40 ncbilcl.ibm:
 41 
 42      Pre-processor defines / platform specific.
 43 
 44 
 45 
 46 Multi-threading
 47 ---------------
 48 
 49 
 50      When running multi-threaded it may be beneficial to set the
 51      following environment variables:
 52 
 53                  AIXTHREAD_SCOPE=S
 54                  MALLOCMULTIHEAP=true
 55 
 56      Addional benefit is possibly achieved by experimenting with 
 57      the value of the environment variables SPINLOOPTIME and 
 58      YIELDLOOPTIME. The user might initially try values of 500 
 59      for each of these and vary them to see the effect.  These 
 60      variables affect the behavior of a thread at the end of a 
 61      parallel region. Generally, best application (though not
 62      necessarily system throughput) performance is achieved by 
 63      putting an idle thread into a busy wait rather than letting 
 64      it sleep.  Setting SPINLOOPTIME and YIELDLOOPTIME to a suitable 
 65      "large" number encourages the former behavior.
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ]
This page was automatically generated by the LXR engine.
Visit the LXR main site for more information.