program killed running 3dTcorrMap

AFNI Message Board
1

Hi all,

I got a problem when I was running 3dTcorrMap for my Whole-Brain Voxel-Voxel

Correlation analysis, the program always kill itself without any reason like this:

3dTcorrMap -input y03.r01.bp+tlrc. -mask y03.r01.bp+tlrc. -polort -1 -CorrMap wb_map -CorrMask
++ 3dTcorrMap: AFNI version=AFNI_17.1.10 (Jun  6 2017) [64-bit]
++ Number of voxels in mask = 103704
++ Loading input dataset
++ CorrMap: this might take a long while and a lot of memory,
+          and will create big [436 billion [giga] bytes] output dataset ...
+ extracting mask-ed time series
+ normalizing extracted time series
++ Starting long long loop: OpenMP threads=8  10 trillion [tera] Flops
+ Single loop (1/50) duration: 25.4 secs = 0.42 mins
   Remaining (49/50) time est.: 1244 secs = 20.7 mins = 0.35 hrs = 0.014 days
++ Voxel loop: 0123Killed

Does anyone has idea about this problem?THANKS!!!

Best regards,
Sue

2

Hi Sue,

“Killed” is generally a message from the operating system when
it terminates a program due to lack of resources (usually when
it runs out of memory).

In this, the text output shows: “+ and will create big [436 billion
[giga] bytes] output dataset …”.

I am not sure whether that means it will use 1/2 terabyte of RAM,
probably not, but in any case, it seems useful to pare down the
analysis.

Did you supply a mask? What is the data resolution (hopefully
not 1 mm^3)?

  • rick
3

Hi Rick,

To save the space, I have applied the GMe mask on my data and all of my grids were resampled to 2 mm^3.
According to the given information, it seems like that it has began to compute my data but somehow crashed suddenly, I really have no idea about that.

Best regards,
Sue

4

Do you mean it is now showing something different from “Killed”?
Can you run “top” or anything, to see how much RAM is being used?

  • rick
5

It seems to be the problem of the memory space because my %CPU is over 90% when running 3dTcorrMap :(.
It’s really impossible for me to update my computer recently, so I guess I could import my data into the MATLAB to finish this analysis, but my code couldn’t work.
Here’s my core command in AFNI.MATLAB:
“[x_size,y_size,z_size,TR]=size(Vmask);
TcorrMap=zeros(x_size,y_size,z_size);
for x=1:x_size
for y=1:y_size
for z=1:z_size
voxel_time=squeeze(Vmask(x,y,z,:));
if unique(voxel_time)~=0
TcorrMap=corr(Vmask(x,y,z,:));
end
end
end
end”
Is there any problem in my way of using this “corr” function?

Best regards,
Sue

6

Hi Sue,

Were you able to figure out anything with this? Note that
the CPU usage of 90% does not imply how much memory
was being used.

Nothing leaps out about your code, but MATLAB is not a
language that I use. Another option would be to use
3dAutoTcorrelate, with which you could break the output
into pieces.

  • rick