GPU equivalent of "whos" or "feature memstat" ?

Is there any equivalent of "whos" or "feature memstat" for the gpu either documented or un-documented?

Also, how the heck does this happen?

   feature memstats

    Physical Memory (RAM):
        In Use:                             2265 MB (   8d98c000)
        Free:                               1713 MB (   6b1b0000)
        Total:                              3979 MB (   f8b3c000)
    Page File (Swap space):
        In Use:                             3121 MB (   c3107000)
        Free:                               4835 MB (  12e3a3000)
        Total:                              7956 MB (  1f14aa000)
    Virtual Memory (Address Space):
        In Use:                             2887 MB (   b4728000)
        Free:                            8385720 MB (7ff4b8b8000)
        Total:                           8388607 MB (7fffffe0000)
    Largest Contiguous Free Blocks:
         1. [at 7ffffff0000]      17592177655808 MB (fffff80000010000)
         2. [at   18000b000]             8376285 MB (7fcfdd65000)
         3. [at 7feff841000]                4100 MB (  10044f000)
         4. [at    c6fa0000]                1933 MB (   78d90000)
         5. [at 7fe7dd92000]                1782 MB (   6f64e000)
         6. [at   13fd59000]                1026 MB (   402a7000)
         7. [at 7fef056c000]                 117 MB (    7544000)
         8. [at    962b0000]                  99 MB (    6380000)
         9. [at    b5c68000]                  65 MB (    4138000)
        10. [at 7feed63c000]                  32 MB (    2034000)
                                            ======= ==========
                                        17592186041250 MB (ffffffff3a279000)

ans =  1.8447e+019

(Note: 1. [at 7ffffff0000] 17592177655808 MB (fffff80000010000) )

How in the world do I have 16 EXABYTES of memory?

Answer 1

[Edric Ellis]

You can use the gpuDevice command to show you how much memory your GPU has, and how much is being used. For example, on my machine right now I see this:

>> gpuDevice

ans = 

    parallel.gpu.CUDADevice handle
    Package: parallel.gpu

    Properties:
                        Name: 'Tesla C1060'
                       Index: 1
           ComputeCapability: '1.3'
              SupportsDouble: 1
               DriverVersion: 4
          MaxThreadsPerBlock: 512
            MaxShmemPerBlock: 16384
          MaxThreadBlockSize: [512 512 64]
                 MaxGridSize: [65535 65535]
                   SIMDWidth: 32
                 TotalMemory: 4.2948e+09
                  FreeMemory: 4.2122e+09
         MultiprocessorCount: 30
                ClockRateKHz: 1296000
                 ComputeMode: 'Default'
        GPUOverlapsTransfers: 1
      KernelExecutionTimeout: 0
            CanMapHostMemory: 1
             DeviceSupported: 1
              DeviceSelected: 1

[Gregory Vernon]

Ahh yes! I should have thought of that.

Answer 2

[Matt J]

See this File Exchange submission,

https://www.mathworks.com/matlabcentral/fileexchange/88024-whosgpu