estimate memory size in matlab
364 views (last 30 days)
Show older comments
Hi,
Is there anyway of estimating memory before execution?So that i can cut down my problem size.(out of memory error)
if i try to give x = ones(10^6);
size of x is out of memory.
Please help in this regards.
Thanks, Sita
2 Comments
James VanderVeen
on 7 Jan 2022
The memory size in Bytes required by an rectangular array is (# of rows)*(# of cols)*(size of one element).
E.g. Double precision variables are by default 64-bits or 8-bytes
For a double precision array with 25850 rows and colums the memory size is:
format longG
arraysize = 25850*25850*8;
fprintf("The array will need %.0f bytes in memory", arraysize)
1 GB is 1024 MB
1 MB is 1024 KB
1 KB is 1024 B
Therefore;
inKB = arraysize/1024;
inMB = arraysize/1024^2;
inGB = arraysize/1024^3;
fprintf("The array will need %.5f KB or %.5f MB or %.5f GB of memory", inKB, inMB, inGB)
The MATLAB online version I'm using has a maximum array size of 5.0 GB, but if I increase the array to 25851 by 25851, it says I exceed the 5.0 GB limit even though the new array is only 4.9797 GB in size. This appears to mean that there is about 22 MB of data to store information about the array.
Answers (4)
Chetan Aswathanarayana
on 18 Jun 2013
I' am not sure if we can do this currently without creating a variable. However if the variable is already in the workspace, then the below command would give the bytes allocated.
>>whos x
However you could type the below command, you would then know the maximum memory avaliable in your system
>>memory
This is what I got on my PC:
Maximum possible array: 18848 MB (1.976e+10 bytes) *
Memory available for all arrays: 18848 MB (1.976e+10 bytes) *
Memory used by MATLAB: 1680 MB (1.761e+09 bytes)
Physical Memory (RAM): 12279 MB (1.288e+10 bytes)
0 Comments
Iain
on 18 Jun 2013
The generic size of any variable is:
sum of (each element of the array * bytes in that element) + some overhead.
Normal arrays: Double, uint64 & int64 formats have 8 bytes per element. single, uint32 & int32 formats have 4 bytes per element. some character types, uint16 & int16 formats have 2 bytes per element. logical, most character types, uint8 & int8 formats have 1 byte per element. Overheads are "small" - I think this would be a few 10s of bytes for normal arrays.
Cell arrays and structures are much more detailed and have higher overheads.
3 Comments
Haiyin Amania
on 24 Dec 2018
where do you find this info from ? because I'm still trying to figure out how I can get the data overheads recorded in matlab. thanks.
Bruno Luong
on 7 Jan 2022
Using C Mex you can find the size by sizeof(MxArray)
It changes from version to version.
See Also
Categories
Find more on Data Type Identification in Help Center and File Exchange
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!