GINFO

From Jacket Wiki

Jump to: navigation, search
Back to Jacket Function List, Jacket Basics

Information on Jacket runtime and detected graphics cards

Supported Syntax

ginfo            % print general information about all GPUs
r = ginfo        % get information about the current GPU
r = ginfo(N)     %             ... about the Nth GPU
r = ginfo('all') %             ... about all GPUs

Description

The optional output struct from GINFO contains various information about the system, Jacket, your NVIDIA driver, and your CUDA-capable devices.

>> ginfo
Jacket v2.0 (build ddb354c) by AccelerEyes (64-bit Linux)
License Type: Designated Computer (/home/pavan/jacket/engine/jlicense.dat)
Licenses: JMC, SDK, DLA, SLA, 4 GPUs
CUDA toolkit 4.0, driver 280.13
GPU1 Quadro 6000, 6144 MB, Compute 2.0 (single,double) (in use)
GPU2 Quadro 4000, 2048 MB, Compute 2.0 (single,double)
Display Device: GPU0 Quadro 6000
Memory Usage: 5677 MB free (6144 MB total)
>> info = ginfo
info = 
       system: '64-bit Linux'
       driver: 280.1300
      toolkit: 4
      version: '2.0 (build ddb354c)'
      default: 'single'
    gpu_count: 2
      current: 0
         name: {'Quadro 6000'}
    gpu_total: 6.4416e+09
     gpu_free: 5.9526e+09
      compute: 2
        flops: 513856000
>> info = ginfo(2)
info = 
       system: '64-bit Linux'
       driver: 280.1300
      toolkit: 4
      version: '2.0 (build ddb354c)'
      default: 'single'
    gpu_count: 2
      current: 0
         name: {'Quadro 4000'}
    gpu_total: 2.1472e+09
     gpu_free: 2.1016e+09
      compute: 2
        flops: 243200000
>> info = ginfo('all')
info = 
       system: '64-bit Linux'
       driver: 280.1300
      toolkit: 4
      version: '2.0 (build ddb354c)'
      default: 'single'
    gpu_count: 2
      current: 0
         name: {'Quadro 6000'  'Quadro 4000'}
    gpu_total: [6.4416e+09 2.1472e+09]
     gpu_free: [5.9526e+09 2.1016e+09]
      compute: [2 2]
        flops: [513856000 243200000]


Note that not all GPU memory can be used for data and computation, since some is allocated by your operating system for graphics purposes as well as CUDA itself. Also note that GEVAL must be used to include some GPU variables in the GPU memory usage count, due to the lazy optimizations included in the Jacket runtime.

You can get diagnostic debug information on the state of memory allocation, kernels loaded, etc., with this (unsupported) synax:

ginfo(1,1)

See Also

GSELECT, GACTIVATE, GHELP

Personal tools