From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Wed Dec 12 2007 - 12:20:48 PST
王磊 wrote: > Dear Sir, > > Is there a mechanism that I can set checkpoints at my will，and > restart my program from one checkpoint which is > > pointed out by me ? > > And examples may be few ，so I still have many questions of the source > in libcr.h . Would you please > > offer us more examples ? > > Thank you ！ The purpose of BLCR is to allow checkpoints to be taken at any time a user chooses and to restart from any of the checkpoints taken. The BLCR User's Guide (http://upc-bugs.lbl.gov/blcr/doc/html/BLCR_Users_Guide.html) includes a brief description of the command line tools in BLCR. The following transcript shows me checkpointing and restarting the "counting" example program in the BLCR source tree. This program is linked to libcr.so, but has no other changes to make it checkpoint. In this example I ask cr_checkpoint to kill (with SIGTERM) the program, but that is not required. $ cd examples/counting/ $ make [...make output removed...] $ ./counting & [1] 21040 Counting demo starting with pid 21040 Count = 0 Count = 1 Count = 2 [...some output removed...] Count = 22 Count = 23 Count = 24 $ cr_checkpoint 21040 --term [1]+ Terminated ./counting $ ls context.21040 counting counting.o Makefile $ cr_restart context.21040 Count = 25 Count = 26 Count = 27 [...some output removed...] Count = 118 Count = 119 $ For more programming examples, I am afraid there is not much I can offer you. There is a lot of code in the tests directory that uses BLCR's libcr in various ways, but that is not a good place to try to learn. If you have specific questions, you can ask this list for help. It can be very hard for both the person asking and those who answer when English is not the first language for both. If I have not answered what you wanted to ask, please try to ask again and I'll try again to answer. -Paul -- Paul H. Hargrove PHHargrove_at_lbl_dot_gov Future Technologies Group HPC Research Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900