Memory leaks during read about adios

[pnorbert]

Hi, thanks for the info. Do you think it matters? One is supposed to
finalize adios at the end of the application just before MPI_Finalize(). So
that adios_finalize() does not free every bit of memory, are there any
circumstances where this may cause trouble?

On Sat, Jan 16, 2016 at 4:28 PM, Saurabh Tendulkar <[email protected]

wrote:

Allocations by adios_read_hooks_init and adios_query_hooks_init leak
memory. Simple test case (had to append .txt for github) and valgrind
output attached. Adios 1.9.0, no MPI, valgrind 3.11, 64-bit Linux system.
adios.c.txt https://github.com/ornladios/ADIOS/files/93091/adios.c.txt
valgrind.txt https://github.com/ornladios/ADIOS/files/93092/valgrind.txt

—
Reply to this email directly or view it on GitHub
#58.

[sim-saurabh]

Not much trouble for me, only have to add suppressions for valgrind for our testing. Although it's not my usecase I could see one where adios is run as needed, while the application stays alive, so a separate module handles all adios stuff, so probably not ideal for it to leak memory every time.

[ax3l]

Huh, should check if the same leak occurs for

adios_init_noxml(MPI_Comm);
adios_finalize(rank);

... we call that several thousand times.

[pnorbert]

The query and read method hooks are not freed in adios_read_finalize. But
they are allocated only once at the first initialization so there is no
memory leak.

The write methods are actually allocated and freed in each
adios_init/adios_finalize steps.

Let me know if you find other types of memory leaks.

On Wed, Feb 17, 2016 at 12:04 PM, Axel Huebl [email protected]
wrote:

Huh, should check if the same leak occurs for

adios_init_noxml(MPI_Comm);adios_finalize(rank);

... we call that several thousand times.

—
Reply to this email directly or view it on GitHub
#58 (comment).

[ax3l]

all right, thank you!