Indic threads pune12-accelerating computation in html 5
Download as PPTX, PDF1 like1,214 views
The document discusses the integration of OpenCL and WebCL for accelerating computations in HTML5 applications, highlighting the importance of multicore computing and parallel algorithms. It introduces OpenCL as a framework for writing parallel algorithms across heterogeneous platforms and provides examples of its execution model and applications. The conclusion emphasizes the significant performance gains achievable through OpenCL in client-side environments, advocating for parallelizable algorithms and further optimizations.
![Example of adding two vectors
Serial version
For(i=1 to n)
c[i]= a[i]+b[i];
Using OpenCL
_kernel add(a,b,c)
{
int i =get_global_id(); //get thread id
c[i]=a[i]+b[i];
}](https://image.slidesharecdn.com/indicthreads-pune12-acceleratingcomputationinhtml5-121217123118-phpapp02/85/Indic-threads-pune12-accelerating-computation-in-html-5-8-320.jpg)
![OpenCL-Execution Model
_kernel add(a,b,c)
1. Kernel {
2. Work-items int i =get_global_id();//get thread/workitem id
c[i]=a[i]+b[i];
3. Work group }
4. ND-range
5. Program
6. Memory
objects
7. Command
queues](https://image.slidesharecdn.com/indicthreads-pune12-acceleratingcomputationinhtml5-121217123118-phpapp02/85/Indic-threads-pune12-accelerating-computation-in-html-5-11-320.jpg)
![Coding with WebCL
platforms = WebCL.getPlatformIDs();
context = WebCL.createContextFromType([WebCL.CL_CONTEXT_PLATFORM,
platforms[0]], WebCL.CL_DEVICE_TYPE_CPU);
devices = context .getContextInfo(WebCL.CL_CONTEXT_DEVICES);
program = context .createProgramWithSource(kernelSrc);
kernelfunction1 = program.createKernel(“function1");
buffparam = context.createBuffer(WebCL.CL_MEM_READ_WRITE, bufSize);
cmdQueue = context.createCommandQueue(devices[0], 0);
cmdQueue.enqueueWriteBuffer(buffparam , true, 0, bufSize, parameter, []);
kernelfunction1.setKernelArg(0, buffparam , WebCL.types.float2);
cmdQueue.enqueueNDRangeKernel(kernelfunction1 , 1, [], totalWorkitems,
totalWorkgroups, []);
cmdQueue.finish ();
cmdQueue.enqueueReadBuffer(‘xyz’, true, 0, bufSize, ‘xyzParam’, []);](https://image.slidesharecdn.com/indicthreads-pune12-acceleratingcomputationinhtml5-121217123118-phpapp02/85/Indic-threads-pune12-accelerating-computation-in-html-5-22-320.jpg)
Indic threads pune12-accelerating computation in html 5
- 1. Accelerating computation in html 5 Ashish Shah SAS R&D INDIA
- 2. Outline • Multicore Computing • Problem statement • Demo • Introduction to OpenCL and WebCL • Conclusion • References
- 4. Problem statement Layout algorithm for node-linked graphs Layout Algorithm
- 5. DEMO Demo 1 – Serial version Demo 2 - Parallel version with multi-core CPU Demo 3 - Parallel version with many-core GPU
- 6. Performance analysis Time in ms Number of particles
- 7. Introduction to OpenCL • Open Compute Language, C- like language. • Framework for writing parallel algorithms • Heterogeneous platforms • Developed by Apple • Is an open standard and controlled by Khronos group
- 8. Example of adding two vectors Serial version For(i=1 to n) c[i]= a[i]+b[i]; Using OpenCL _kernel add(a,b,c) { int i =get_global_id(); //get thread id c[i]=a[i]+b[i]; }
- 9. OpenCL Architecture 1. Platform model 2. Execution model 3. Memory model 4. Programming model
- 10. OpenCL -Platform • Device • Host Host Intel CPU GPU 2 Compute Device 1 (GPU 1) Compute unite (Cores)
- 11. OpenCL-Execution Model _kernel add(a,b,c) 1. Kernel { 2. Work-items int i =get_global_id();//get thread/workitem id c[i]=a[i]+b[i]; 3. Work group } 4. ND-range 5. Program 6. Memory objects 7. Command queues
- 12. Memory Model in OpenCL Compute Device Private register Private register Private register Compute unit 0 Compute unit 1 Compute unit 2 Local memory/cache Local memory/cache Local memory/cache Global constant memory-DRAM Global Memory -DRAM
- 13. Programming model 1. Data parallel-single function on multiple data 2. Task parallel-Multiple functions on single data
- 14. OpenCL Stack Java,c,.net, HTML,.java, WebCL Applications .NET,c,c++ kernals OpenCL-Api Compiler String data context Memory Api’s OpenCL Framework OpenCL Runtime Command queues, buffer objects, kernel Device driver execution OpenCL Device (GPU/CPU hardware)
- 15. Essential Development Tasks C-code with restrictions Initialize Initiate Execute Read back Parallelize Code Kernel OpenCL kernels and kernel data to host environment data
- 16. Essential Development Tasks • Query compute device • Create context • Compile kernels Initialize Initiate Execute Read back Parallelize Code Kernel OpenCL kernels and kernel data to host environment data
- 17. Essential Development Tasks • Create memory objects • Map data structures to OpenCL supported data structures. • Initialize kernel parameters Initialize Initiate Execute Read back Parallelize Code Kernel OpenCL kernels and kernel data to host environment data
- 18. Essential Development Tasks • Specify number of threads to execute task • Trigger the execution of kernel- sync or async Initialize Initiate Execute Read back Parallelize Code Kernel OpenCL kernels and kernel data to host environment data
- 19. Essential Development Tasks • Map to application datastructure Initialize Initiate Execute Read back Parallelize Code Kernel OpenCL kernels and kernel data to host environment data
- 20. Introduction to WebCL • Java Script bindings for OpenCL • First announced in March 2011 by Khronos • API definition underway • Prototype plugin is available only for Firefox browser
- 21. Binding OpenCL to WebCL CPU Host application JavaScript WebCL OpenCL OpenCL Framework compliant device
- 22. Coding with WebCL platforms = WebCL.getPlatformIDs(); context = WebCL.createContextFromType([WebCL.CL_CONTEXT_PLATFORM, platforms[0]], WebCL.CL_DEVICE_TYPE_CPU); devices = context .getContextInfo(WebCL.CL_CONTEXT_DEVICES); program = context .createProgramWithSource(kernelSrc); kernelfunction1 = program.createKernel(“function1"); buffparam = context.createBuffer(WebCL.CL_MEM_READ_WRITE, bufSize); cmdQueue = context.createCommandQueue(devices[0], 0); cmdQueue.enqueueWriteBuffer(buffparam , true, 0, bufSize, parameter, []); kernelfunction1.setKernelArg(0, buffparam , WebCL.types.float2); cmdQueue.enqueueNDRangeKernel(kernelfunction1 , 1, [], totalWorkitems, totalWorkgroups, []); cmdQueue.finish (); cmdQueue.enqueueReadBuffer(‘xyz’, true, 0, bufSize, ‘xyzParam’, []);
- 23. Applications of OpenCL • Database mining • Neural networks • Physics based simulation,mechanics • Image processing • Speech processing • Weather forecasting and climate research • Bioinformatics
- 24. Conclusion • Significant performance gains in using OpenCL for computations in client-side environments like HTML5 • Algorithms need to be ‘parallelizable’ • Further optimizations can be achieved by exploiting memory model
- 25. Software/Hardware used in demo application Hardware Intel(R) Core(TM)2 Quad core CPU Q8400 @ 2.66GHz Nvidia 160m Quadro 8 cores @ 580 MHz Software OpenCL runtime for CPU http://software.intel.com/en-us/articles/vcsource- tools-opencl-sdk/ OpenCL runtime for GPU http://www.nvidia.com/object/quadro_nvs_notebook. html WebCL plugin for Firefox http://webcl.nokiaresearch.com/
- 26. References http://www.macresearch.org/opencl http://en.wikipedia.org/wiki/GPGPU http://www.khronos.org/webcl/
Editor's Notes
- #3: Html applications use multicores
- #4: What are multicores?Compare Cpu and GpuAdvantages of using GPUHow is it possible to do so.Moors lawSimple questions
- #5: -Explain network diagram-how to render-performance issue-solve it using parallel computation opencl technologyUse different color force directed.Real time use ,giving imp to diagram than textReplace Force directed algorithm with simple lay out algo.
- #6: Fontext for demoPause and explain details,with variation in speech
- #8: Brief OpenCl intro.Dreaft and open specification.Parallel only!!!Thread amnagement,synch is simple
- #10: Why this models?How architecture helps in improving performance
- #11: HierarcyHosta calls
- #12: What to do with execution model.Acces threads,Create memory,execute program,CommandsWorkgrou-cores corrospondence
- #18: Explain mapping data structures
- #22: More details on WebCL
- #27: variation in speechConclusion… and question