On 03/27/2016 03:08 AM, Tom Hughes wrote:
I haven't done that, but maybe the Open MPI library does. It does indeed seem that they are trying to using these features of valgrind. At least one of the first search hit with key you mention is a question a few days ago to valgrind list about how to mark a buffer as read only. :-) I have notice some other runtime checking that they enabled, and also a number valgrind leak reports in there code. although ompi_info claims the features are disabled. Maybe I follow up with them. Thanks!On 26/03/16 19:15, Burlen Loring wrote:I like to understand better the following reports. I have a buffer that I allocated and that gets filled by MPI recv. However, for every access to the buffer valgrind produce a report similar to the following first report shown here: Invalid read of size 4 Address 0x1788a6d0 is 0 bytes inside a block of size 30,121 alloc'd in subsequent reports "0 bytes inside" the 0 is replaced with the offset of the access in the buffer. Full output of the first few reports are shown below.That does look odd. The only obvious way I can think of to get a report like that is if you had explicitly told valgrind to consider that memory as inaccessible with the VALGRIND_MAKE_MEM_NOACCESS call.
Re: [Valgrind-users] can memchecker mark memory as read-only ? <https://sourceforge.net/p/valgrind/mailman/message/34933327/>
------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140
_______________________________________________ Valgrind-users mailing list Valgrind-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/valgrind-users
