GPU - Revision history

Daniel Castaño at 06:44, 2 September 2016

2016-09-02T06:44:17Z

Daniel Castaño: /* Using the GPU in an alignment project */

2016-07-12T15:24:00Z

Using the GPU in an alignment project

Daniel Castaño at 16:01, 3 June 2016

2016-06-03T16:01:11Z

Daniel Castaño at 16:00, 3 June 2016

2016-06-03T16:00:41Z

Daniel Castaño at 15:59, 3 June 2016

2016-06-03T15:59:10Z

Daniel Castaño at 15:55, 3 June 2016

2016-06-03T15:55:23Z

Daniel Castaño at 15:19, 29 April 2016

2016-04-29T15:19:22Z

Daniel Castaño at 10:49, 22 April 2016

2016-04-22T10:49:15Z

Daniel Castaño at 10:48, 22 April 2016

2016-04-22T10:48:40Z

Daniel Castaño at 10:35, 22 April 2016

2016-04-22T10:35:45Z

← Older revision		Revision as of 06:44, 2 September 2016
Line 77:		Line 77:

	[[File:Nvidia-smi example.png\|thumb\|center\|400px\| A screenshot on <tt>nvidia-smi</tt> showing three devices]]		[[File:Nvidia-smi example.png\|thumb\|center\|400px\| A screenshot on <tt>nvidia-smi</tt> showing three devices]]
		+
		+	== GPUs in classification ==
		+
		+	GPUs in classification are only available in the context of MRA alignment. Multireference alignment inherits the "GPU-friendliness" of single reference alignment: the cc of many rotations of the template against the particle can be computed inside the GPU without any transfer of data into the CPU or the hard disk.
		+
		+	In PCA, the situation is totally different: computation of the cross correlation matrix is not a compute-intensive process, but rather a data-intensive process: we don't rotate several times the same data particle. Particles need to be read from disk, rotated and compared against other particles (which have undergone the same process). For this reason, a GPU version of cc-matrix computation for PCA analysis is not available, as it would not provide any speed-up in comparison with a CPU application. The only effective way to speed up PCA is the use of parallel CPU computing.

@@ Line 56: / Line 56: @@
 Your execution environment should include a {{t|LD_LIBRARY_PATH}} environment variable that includes the location of the CUDA libraries.
-You probably need to inform your UNIX shell with the command:
+You probably need to inform your UNIX shell with a command similar to this:
 <tt>export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda75/lib</tt>
-(replace with the location of the CUDA libraries in your system).
+Here, replace with the location of the CUDA libraries in your system. If you have several CUDA installations, choose the one that you used to compile ''Dynamo'' in your system.
 Also if you are going to work under Matlab you should update your {{t|LD_LIBRARY_PATH}} variable in the Linux shell before starting Matlab.

@@ Line 33: / Line 33: @@
 {{t|make all}}
 to recompile the executables in your system.
 === CUDA libraries ===

@@ Line 23: / Line 23: @@
 <tt>CUDA_ROOT=/usr/local/cuda-7.5</tt>
-have been edited to the correct CUDA libraries on your system. The <tt>CUDA_ROOT</tt> variable exepected here should  correspond to <tt><CUDA_ROOT>/bin/nvcc</tt>
+have been edited to the correct CUDA libraries on your system. The <tt>CUDA_ROOT</tt> variable expected here can be infered from the path to the detected <tt>nvcc</tt> compiler by <tt><CUDA_ROOT>/bin/nvcc</tt>
-Then you can just type
+Once your {{t|makefile}} is correctly formated, you can just type:
 {{t|make clean}}

@@ Line 19: / Line 19: @@
 {{t|source config.sh}}
-This will automatically edit the makefile file in the folder, informing it on the location of CUDA in your system. Then you can just type
+This will automatically edit the <tt>makefile</tt> file in the folder, informing it on the location of CUDA in your system. In order to be sure, you can also edit the {{t|makefile}} text file and make sure that the line
 {{t|make clean}}

@@ Line 6: / Line 6: @@
 A regular distribution of ''Dynamo'' will include GPU precompiled GPU executables. As they need to be linked to libraries that might not be present in your system, it is advised that you recompile them:
 After untarring the tar package in location {{t|DYNAMO_ROOT}}, go to
-{{t|cd DYNAMO_ROOT}}
-and make certain that you have CUDA active in your shell, for instance look for the NVIDIA {{t|nvcc}} compiler.
+<tt>cd DYNAMO_ROOT/cuda</tt>
-{{t|which nvcc}}
 If  this is succcessful run
 {{t|source config.sh}}
 This will automatically edit the makefile file in the folder, informing it on the location of CUDA in your system. Then you can just type
 {{t|make clean}}
 to delete the executables in the distribution, and then
 {{t|make all}}
 to recompile the executables in your system.

@@ Line 2: / Line 2: @@
 GPU stands for "Graphic Processing Unit".
 If you have access to a dedicated GPU device, the iteration procedure will run much faster.  While ''Dynamo'' can run its  GPU functionalities in most Nvidia cards, a real advantage will be measured for those cards that are conceived for scientific computation. The Tesla series and the Titan series are examples of this.
 == Installation==
@@ Line 22: / Line 20: @@
 === CUDA libraries ===
 You need to have CUDA installed on your system. This might require coordinating with your system administrator.
 ==== Versions====
@@ Line 59: / Line 56: @@
 == Checking your system ==
+{{main|GPU identifiers|GPU identifiers}}
 In a linux machine {{t|nvidia-smi}} will give you a status summary of the installed, reachable GPUs. In the example image, you would use an {{t|identifier_set}} of 1,2 to. Alternatively, you could run different projects on device 1 and 2. In any case, device 0 is not intended for GPU computing.
 [[File:Nvidia-smi example.png|thumb|center|400px| A screenshot on <tt>nvidia-smi</tt> showing three devices]]

← Older revision		Revision as of 10:49, 22 April 2016
Line 62:		Line 62:
	In a linux machine {{t\|nvidia-smi}} will give you a status summary of the installed, reachable GPUs. In the example image, you would use an {{t\|identifier_set}} of 1,2 to. Alternatively, you could run different projects on device 1 and 2. In any case, device 0 is not intended for GPU computing.		In a linux machine {{t\|nvidia-smi}} will give you a status summary of the installed, reachable GPUs. In the example image, you would use an {{t\|identifier_set}} of 1,2 to. Alternatively, you could run different projects on device 1 and 2. In any case, device 0 is not intended for GPU computing.

−	[[File:Nvidia-smi example.png\|thumb\|~~right~~\|~~300px~~\| A screenshot on <tt>nvidia-smi</tt> showing three devices]]	+	[[File:Nvidia-smi example.png\|thumb\|center\|400px\| A screenshot on <tt>nvidia-smi</tt> showing three devices]]

@@ Line 1: / Line 1: @@
 [[Category:installation]]
-GPU stands for "Graphic Processing Unit"
+GPU stands for "Graphic Processing Unit".
@@ Line 51: / Line 52: @@
 Go to the ''Computing environment'' GUI.
 Select 'matlab gpu' or 'system gpu'
-Set the {{t|gpu identifier set}} field to the device numbers given to your GPUs by the operative system. You can select a subset of the available GPUs.
+Set the {{t|gpu identifier set}} field to the device numbers given to your GPUs by the operative system. You can select a subset of the available GPUs. If you have a GPU that controls the screen display, you can technically use it for GPU computations, but it is not advisable, as it will be typically much slower than GPUs intended for computation.
 === With the command line ===
 There are two project parameters related the the GPU: the {{t|destination}} parameter and the {{t|gpu_identifier_set}} parameter.
-The {{t|destination}} must be set to {{t|matlab_gpu}} (to run projects inside Matlab) or to {{t|system_gpu}}
+[[File:Nvidia-smi example.png|thumb|right|300px| A screenshot on <tt>nvidia-smi</tt> showing three devices]]

← Older revision		Revision as of 10:35, 22 April 2016
Line 49:		Line 49:

	=== With the {{t\|dcp}} GUI ===		=== With the {{t\|dcp}} GUI ===
		+	Go to the ''Computing environment'' GUI.
		+	Select 'matlab gpu' or 'system gpu'
		+	Set the {{t\|gpu identifier set}} field to the device numbers given to your GPUs by the operative system. You can select a subset of the available GPUs.

	=== With the command line ===		=== With the command line ===