MATLAB PARALLEL COMPUTING TOOLBOX - S Guide de l'utilisateur Page 285
- Page / 656
- Table des matières
- DEPANNAGE
- MARQUE LIVRES
Noté. / 5. Basé sur avis des utilisateurs
Run CUDA or PTX Code on GPU
9-27
Use gpuArray Variables
It might be more efficient to use gpuArray objects as input when running a kernel:
k = parallel.gpu.CUDAKernel('conv.ptx','conv.cu');
i1 = gpuArray(rand(100,1,'single'));
i2 = gpuArray(rand(100,1,'single'));
result1 = feval(k,i1,i2);
Because the output is a gpuArray, you can now perform other operations using this
input or output data without further transfers between the MATLAB workspace and the
GPU. When all your GPU computations are complete, gather your final result data into
the MATLAB workspace:
result2 = feval(k,i1,i2);
r1 = gather(result1);
r2 = gather(result2);
Determine Input and Output Correspondence
When calling [out1, out2] = feval(kernel, in1, in2, in3), the inputs in1,
in2, and in3 correspond to each of the input arguments to the C function within your
CU file. The outputs out1 and out2 store the values of the first and second non-const
pointer input arguments to the C function after the C kernel has been executed.
For example, if the C kernel within a CU file has the following signature:
void reallySimple( float * pInOut, float c )
the corresponding kernel object (k) in MATLAB has the following properties:
MaxNumLHSArguments: 1
NumRHSArguments: 2
ArgumentTypes: {'inout single vector' 'in single scalar'}
Therefore, to use the kernel object from this code with feval, you need to provide feval
two input arguments (in addition to the kernel object), and you can use one output
argument:
y = feval(k,x1,x2)
- Parallel Computing Toolbox™ 1
- User's Guide 1
- How to Contact MathWorks 2
- Revision History 3
- Contents 5
- Program Independent Jobs 11
- Program Communicating Jobs 12
- GPU Computing 12
- Objects — Alphabetical List 14
- Functions — Alphabetical List 14
- Glossary 14
- Getting Started 15
- 1 Getting Started 16
- Partition Large Data Sets 19
- Run a Batch Job 22
- Run a Batch Parallel Loop 23
- Parallel for-Loops (parfor) 29
- Introduction to parfor 30
- Create a parfor-Loop 32
- Create a parfor-Loop 33
- 2 Parallel for-Loops (parfor) 34
- parfor Limitations 39
- Using Objects in parfor-Loops 41
- Handle Classes 41
- Nested Functions 43
- Nested Loops 43
- Nested spmd Statements 45
- Break and Return Statements 45
- P-Code Scripts 45
- Unambiguous Variable Names 47
- Transparency 47
- Loop Variable 53
- Sliced Variables 55
- Broadcast Variables 59
- Reduction Variables 60
- Reduction Variables 65
- Temporary Variables 67
- Temporary Variables 69
- Improving parfor Performance 70
- Improving parfor Performance 71
- Parallel Pools 72
- Parallel Pools 73
- Start a Parallel Pool 73
- When to Use spmd 80
- Define an spmd Statement 81
- Display Output 83
- Introduction to Composites 84
- Distribute Arrays 88
- Create Codistributed Arrays 89
- P>> help magic 97
- P>> x = pi 97
- P>> x = labindex 98
- P>> all = numlabs 98
- P>> getLocalPart(I) 102
- P>> pmode exit 103
- Parallel Command Window 104
- Click tiling icon 105
- Select layout 105
- Select vertical 106
- Drag to adjust 106
- Click tabbed output 107
- Select only two tabs 107
- Multiple labs 108
- Using Graphics in pmode 112
- Connectivity Testing 113
- Hostname Resolution 113
- Socket Connections 113
- Introduction 116
- Nondistributed Arrays 116
- Codistributed Arrays 118
- A = zeros(80, 1000); 119
- D = codistributed(A) 119
- Local Arrays 124
- Restoring the Full Array 127
- 2-Dimensional Distribution 130
- Parallelizing a for-Loop 134
- plot(1:numDataSets, res); 135
- Programming Overview 141
- 6 Programming Overview 142
- Toolbox and Server Components 143
- Life Cycle of a Job 147
- Parallel Preferences 152
- Parallel Preferences 153
- Clusters and Cluster Profiles 154
- Job Monitor 169
- Job Monitor 171
- Programming Tips 172
- Programming Tips 173
- Using the pause Function 175
- Interrupting a Job 175
- Speeding Up a Job 175
- Control Random Number Streams 177
- Client and GPU 179
- Worker CPU and Worker GPU 181
- Profiling Parallel Code 182
- Viewing Parallel Profile Data 183
- Profiling Parallel Code 185
- Benchmarking Performance 191
- Troubleshooting and Debugging 192
- Caused by: 195
- Start Parallel Pool 197
- Compare Parallel mapreduce 197
- See Also 199
- Cluster Preparation 201
- Output Format and Order 201
- Calculate Mean Delay 201
- Related Examples 203
- More About 203
- 7 Program Independent Jobs 212
- MJS Cluster 217
- Properties 217
- Share Code with the Workers 224
- Share Code with the Workers 227
- Overview 229
- Client node 231
- MATLAB client 231
- Scheduler 231
- MATLAB Worker Decode Function 235
- Program Communicating Jobs 249
- 8 Program Communicating Jobs 250
- Code in the Client 251
- 9 GPU Computing 260
- Establish Arrays on a GPU 261
- Establish Arrays on a GPU 265
- 100 100 265
- Supported MATLAB Code 273
- Run CUDA or PTX Code on GPU 278
- Create a CUDAKernel Object 279
- Run CUDA or PTX Code on GPU 281
- Compile a GPU MEX-File 291
- Comparison to a CUDA Kernel 291
- Constructor 302
- Description 302
- codistributed 303
- Properties 304
- Composite 306
- CUDAKernel 307
- distributed 311
- GPUDevice 314
- Constructors 319
- Container Hierarchy 319
- parallel.Cluster 321
- HPC Server 322
- PBS Pro and TORQUE 322
- MJS Jobs 332
- CJS Jobs 332
- RemoteClusterAccess 341
- Examples 344
- Input Arguments 346
- arrayfun 349
- Arguments 352
- Create gpuArray False Matrix 430
- Create gpuArray Inf Matrix 484
- Output Arguments 513
- mpiprofile 519
- C Syntax 524
- Create gpuArray NaN Matrix 549
- Create gpuArray Rand Matrix 605
- Create gpuArray True Matrix 641
- Glossary 653
- Glossary-5 655
Produits connexes et manuels pour Logiciel MATLAB PARALLEL COMPUTING TOOLBOX - S
(112 pages)© 2020, manymanuals.fr. Tous droits réservés | 0.029 s |
Manymanuals.com
Manymanuals.de
Manymanuals.fr
Manymanuals.it
Manymanuals.pl
Manymanuals.cz
Manymanuals.es
Manymanuals-pt.com
Commentaires sur ces manuels