Problem in running openACC in Fortan with RTX2080ti

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

I am trying to run a Fortran code compiled with
CFT1=“pgf95 -g -O4 -i8 -r8 -mp=allcores -Mcuda rdc=false -ta:tesla -acc=autopar -ta=multicore -tp=haswell -Mlarge_arrays -mcmodel=medium”

Everything is ok with compilation process , but when Im trying to run the code , to use a RTX 2080ti GPU , I had this problem.
In fact, I would like to use parallelism in parts of the fortan code to use openAcc , and it works in an old version of the model , and now dont works anymore. (error attached)
Any idea of what I’m doing wrong ?
Thanks in advance

My error
zeta lives at 0x1357a40 size 1560648 partially present
Present table dump for device[1]: NVIDIA Tesla GPU 0, compute capability 7.5, threadid=4
host:0x1357a40 device:0x7f036ec00000 size:1449272 presentcount:1+0 line:651 name:zeta
allocated block device:0x7f036ec00000 size:1449472 thread:7
FATAL ERROR: variable in data clause is partially present on the device: name=zeta
file:/run/media/marciocintra/DADOS/MODELOS/Run_NE_INTER2/Compile/step2d_.f step2d_fb_tile line:651

2

Hi cintra,

It looks like there’s difference between the size difference between how zeta was created and the size it is when being used. Looks like it was created with a size in bytes of “1449272” but at line 651 of step3d_fb_tile, the size being used in 1560648.

First, check that the array shape syntax in the data clauses (i.e. the “zeta[:size]”) match between where you created the array and where it’s used.

Otherwise, has zeta been reallocated to a different size after it was created on the device?

If you can, please provide a reproducing example and I can better help diagnose.

Best Regards,
Mat

3

Ok Mat,

Thanks a lot …