Hello,
I know very limited support is given for self-compiled applications, but can anyone provide some insight on the following error for BRP4?
Couldn't allocate 0 bytes of memory for sumspec at bottom level
Does anyone also know where I can find the latest source code for all the EAH applications?
http://albert.phys.uwm.edu/license.php looks out of date.
Thanks!
Copyright © 2024 Einstein@Home. All rights reserved.
Allocation Error
)
I'd try here instead which contains a link to this. On download and unzip, the timestamps are under six weeks old.
Now a ( directory recursive ) search of the files locates this :
// allocate memory for the harmonic summed spectra // in CUDA version, this includes the 1st harmonics // TODO : sumspec [0] has to be treated differently once the powerspectrum is left on device // initially and only copied async. later for( i = 1; i < 5; i++) { sumspec = (float *) calloc(fundamental_idx_hi, sizeof(float)); if(sumspec[i] == NULL) { logMessage(error, true, "Couldn't allocate %d bytes of memory for sumspec at bottom level.\n", fundamental_idx_hi*sizeof(float )); return(RADPUL_EMEM); } }
being lines 97 thru 109 of the file [i]demod_binary_cuda.cu within the /src/cuda/app/ directory.
I have no idea what that means at a glance, other than the obvious : the calloc() call failed and a NULL pointer was returned.
Cheers, Mike.
( edit ) Closer look :
- since sizeof(float) won't be zero then fundamental_idx_hi must be.
- fundamental_idx_hi is the 4th parameter ( an unsigned int ) to the set_up_harmonic_summing routine
- so why was that zero ?
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Thanks for your help
)
Thanks for your help Mike.
I figured this one out. I was trying to compile the BRP4 application on an big endian architecture (SPARC). The header files sent with the tasks is written for little endian systems.
After making a couple of changes to the source code, the application is no longer quitting immediately with the aforementioned error.
I should know in a day or so if I'm returning valid results.
RE: ... big endian
)
Yup, that would be an issue! :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Now I just need to find a
)
Now I just need to find a new(er) SPARC machine to run tasks on.
Apparently a 440 Mhz Ultrasparc IIi (circa late 90s) resulted in 62.50 granted credit after a run time of 97,933.95 seconds. Now where did I put that Raspberry Pi?
RE: Now where did I put
)
Excellent comparision, a heavily overclocked Raspberry Pi should come very close to it. And a modern Android smartphone will dance circles around the Sun...
But I like vintage computers! Congratulations for the port to Sparc!
Cheers
HB