Why am I getting inconsistent reports between FEATURE MEMSTATS and MEMORY in MATLAB 7.6 (R2008a)?

1 view (last 30 days)
The following Tech Note seems to indicate that a matrix can be as large as the "Maximum Free Contiguous Block Size" reported by executing the 'feature memstats'command in MATLAB:
https://www.mathworks.com/help/matlab/performance-and-memory.html
However, this command resulted in the following output on my 64-bit Windows Server 2003 machine:
Physical Memory (RAM):
In Use: 1596 MB ( 63ce2000)
Free: 31002 MB ( 791a63000)
Total: 32599 MB ( 7f5745000)
Page File (Swap space):
In Use: 1870 MB ( 74eaa000)
Free: 33887 MB ( 845fc5000)
Total: 35758 MB ( 8bae6f000)
Virtual Memory (Address Space):
In Use: 928 MB ( 3a041000)
Free: 8387679 MB (7ffc5f9f000)
Total: 8388607 MB (7fffffe0000)
Largest Contiguous Free Blocks:
1. [at 180401000] 8379758 MB (7fdd6e4f000)
2. [at 7fff0000] 3072 MB ( c0010000)
...
ans =
8.7868e+012
However, executing the following command:
A = zeros(2^40,1,'uint8');
results in an "Out of Memory" error.
Further, executing MEMORY results in the following output:
Maximum possible array: 33885 MB (3.553e+010 bytes) *
Memory available for all arrays: 33885 MB (3.553e+010 bytes) *
Memory used by MATLAB: 331 MB (3.474e+008 bytes)
Physical Memory (RAM): 32599 MB (3.418e+010 bytes)

Accepted Answer

MathWorks Support Team
MathWorks Support Team on 18 Apr 2023
Edited: MathWorks Support Team on 19 Apr 2023
The use of FEATURE MEMSTATS to estimate the largest block of contiguous memory is limited to 32-bit Windows platforms as described in the tech note,
On 64-bit platforms, the commands FEATURE MEMSTATS and MEMORY return differing interpretations of available memory.
Executing the following command:
feature memstats
reports the largest memory address block available on the system. This number is in terms of the available memory addresses based on the theoretical limit/process Limit of the architecture (i.e. 32-bit Windows or 64-bit Windows) and not the physical resources that are available.

More Answers (0)

Categories

Find more on Performance and Memory in Help Center and File Exchange

Products


Release

R2008a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!