'radix_sort: failed to get memory buffer' when executing accumarray on gpuArrays of certain size

Hello,

I'm trying to use accumarray on large gpuArrays, but get the error 'radix_sort: failed to get memory buffer'.

This is a minimal example that gives me the error:

a = randi(intmax, 2^28-2048, 1, 'gpuArray');
b = gpuArray(randi(3, 2^28-2048, 3, 'uint16'));
c = accumarray(b,a);

When I do the same with arrays of size [2^28-2047 1] and [2^28-2047 3] it works.

This is my gpuDevice after creating a and b:

 CUDADevice with properties:  
 
                      Name: 'GeForce GTX 1080 Ti'
                     Index: 1
         ComputeCapability: '6.1'
            SupportsDouble: 1
             DriverVersion: 10.1000
            ToolkitVersion: 9.1000
        MaxThreadsPerBlock: 1024
          MaxShmemPerBlock: 49152
        MaxThreadBlockSize: [1024 1024 64]
               MaxGridSize: [2.1475e+09 65535 65535]
                 SIMDWidth: 32
               TotalMemory: 1.1718e+10
           AvailableMemory: 7.6692e+09
       MultiprocessorCount: 28
              ClockRateKHz: 1683000
               ComputeMode: 'Default'
      GPUOverlapsTransfers: 1
    KernelExecutionTimeout: 1
          CanMapHostMemory: 1
           DeviceSupported: 1
            DeviceSelected: 1

Shouldn't this be enough memory for this kind of operation?

I'm running version 9.5.0.944444 (R2018b) on Linux.

I can work around this problem but I'd like to understand it so I can adapt my code accordingly.

Best wishes,

Daniel

Answer 1

[Ganesh Regoti]

I have run the code on TitanV and it works fine. The array of larger size is working fine. So, I think there is no memory issue in it.

Try to clear the memory of GPU device through reset command. Here is the link

https://www.mathworks.com/help/parallel-computing/parallel.gpu.gpudevice.reset.html

Now, try to re-run the code.

[Daniel Hähnke]

Thanks Ganesh.

If I reset the gpuDevice as you suggested it works (even if I use 2^28 and 3*2^28 number of elements for a and b). However, if after I run it once I reassign a and b to arrays of different sizes (e.g. subtracting 1 row), I again get the radix_sort error.

>> reset(gpuDevice)
>> a = randi(intmax, 2^28, 1, 'gpuArray');
>> b = gpuArray(randi(3, 2^28, 3, 'uint16'));
>> c = accumarray(b,a);
>> whos
  Name              Size             Bytes  Class       Attributes

  a         268435456x1                  4  gpuArray
  b         268435456x3                  4  gpuArray
  c                 3x3x3                4  gpuArray

>> a = randi(intmax, 2^28-1, 1, 'gpuArray');
>> b = gpuArray(randi(3, 2^28-1, 3, 'uint16'));
>> c = accumarray(b,a);
radix_sort: failed to get memory buffer

Does this mean that MATLAB doesn't properly free the RAM?

[Ganesh Regoti]

Hi Daniel,

Run the following command after excuting each line after the first assignment of a b c.

nvidia-smi 

Can you share the output, so I can go through it.

[Daniel Hähnke]

Thank you. This is what I get:

>> reset(gpuDevice)
>> unix('nvidia-smi');
Mon Aug  5 11:29:00 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  0%   47C    P2    57W / 280W |    245MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB   217MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+
>> a = randi(intmax, 2^28, 1, 'gpuArray');
>> unix('nvidia-smi');
Mon Aug  5 11:29:18 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  0%   48C    P2    57W / 280W |   2297MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  2269MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+
>> b = gpuArray(randi(3, 2^28, 3, 'uint16'));
>> unix('nvidia-smi');
Mon Aug  5 11:29:41 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  0%   50C    P2    62W / 280W |   3833MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  3805MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+
>> c = accumarray(b,a);
>> unix('nvidia-smi');
Mon Aug  5 11:29:57 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  0%   51C    P2    70W / 280W |   4857MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  4829MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+

>> a = randi(intmax, 2^28-1, 1, 'gpuArray');
>> unix('nvidia-smi');
Mon Aug  5 11:30:26 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  0%   54C    P2    71W / 280W |   5881MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  5853MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+
>> b = gpuArray(randi(3, 2^28-1, 3, 'uint16'));
>> unix('nvidia-smi');
Mon Aug  5 11:30:52 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  6%   55C    P2    66W / 280W |   5369MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  5341MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+
>> c = accumarray(b,a);
radix_sort: failed to get memory buffer
   
>> unix('nvidia-smi');
Mon Aug  5 11:31:02 2019
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.43       Driver Version: 418.43       CUDA Version: 10.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:03:00.0 Off |                  N/A |
|  7%   56C    P2    76W / 280W |   4857MiB / 11175MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
   
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      3275      C   ...alysis/MATLAB/R2018b/bin/glnxa64/MATLAB  4829MiB |
|    0      3520      G   /usr/lib/xorg/Xorg                             9MiB |
|    0      3558      G   /usr/bin/gnome-shell                           6MiB |
+-----------------------------------------------------------------------------+

[Ganesh Regoti]

Hi Daniel,

There is no concept of cleaning the memory implicitly in GPU. Only ways to clean memory are

1. Restart the session
2. Reset the Device

So once the memory is created, it stays in the GPU memory until you clean it explicitly.

Coming to your case of accumArray function, the cause for the issue is at some of executing the internal code, a memory buffer is created which is exceeding the RAM limit.

[Daniel Hähnke]

Thanks. So that means it's not possible to free up the memory of a variable when it's reassigned?

Is this an issue with GPUs in general (i.e. do they only support a full wipe?) or can MATLAB just not do it?

[Joss Knight]

This isn't strictly true. MATLAB holds onto a quarter of GPU memory, once assigned, as an optimisation to prevent unnecessary device synchronization. Memory is then re-used. MATLAB will never return an out-of-memory error because it is holding onto the memory of a variable that has gone out of scope. However, it appears that in this case the NVIDIA thrust library is allocating its own memory buffer and MATLAB doesn't know about this, so it doesn't know to free up its memory pool to make space. This should be fixed for you in MATLAB R2019a.

In the meantime, try

feature('GpuAllocPoolSizeKb', 0);

as a temporary measure to turn off the pooling of memory that's causing this issue.

Answer 2

[Joss Knight]

There is an issue in an NVIDIA library that is not functioning correctly when memory is limited. This is fixed in CUDA 10 / MATLAB R2019a.

In the meantime, try

poolSize = feature('GpuAllocPoolSizeKb', 0);

as a temporary measure to turn off the pooling of memory that's underlying this issue. When you are ready to enable pooling again use

feature('GpuAllocPoolSizeKb', poolSize);

This is advisable since turning off pooling will significantly reduce performance.