Cuda, OpenCL

  • Published on
    20-Jun-2015

  • View
    207

  • Download
    7

Embed Size (px)

Transcript

<ul><li> 1. ( top500.org) . CPU : GPU = 1 : n (n = 1, 2, 3,)</li></ul> <p> 2. GPGPU. . ., 5085/2, , 07.05.2013 3. GPUCPU vs GPU GPU , (.. Flow control) 4. GPGPU 2007 Cg, GLSL HLSL GLSL OpenGL HLSL Microsoft DirectX 1. . 2. . 3. 4. API. 5. NVIDIA Compute Unified Device Architecture C GPU GeForce 8, 9, 100, 200, 300, 400, 500, 600, 700 6. NVIDIA Compute Unified Device Architecture GPU (device) CPU (host) ( , device memory) (kernel) GPU CPU: GPU , 7. NVIDIA Compute Unified Device Architecture 8. NVIDIA Compute Unified Device Architecture SIMD (Single Instruction Multiple Data) SIMT (Single Instruction Multiple Thread) (, , , , , ) 9. NVIDIA Compute Unified Device Architecture 10. Open Compute Language Khronos Group: AMD, Intel, Motorola, Apple, NVIDIA, IBM, ARM 2008 . - OpenCL 1.0 1 2012 . - OpenCL 1.2 C99 WinZip 16.5 AMD APU - 45% 11. Open Compute Language CPU (host) OpenCL- (devices): CPU, GPU, DSP, 12. Open Compute Language WinZip, ArcSoft, Corel VideoStudio, GIMP, Mathematica 8, vReveal, Blender, BattleField 3, PyOpenCL, WebCL (JS- , Firefox WebKit), ScalaCL, Ruby-OpenCL PGI OpenCL Compiler, OpenCL Studio, ZiiLABD OpenCL SDK http://openclnews.com/apps 13. Open Compute Language AMD OpenCL University Kit AMD . 13 http://developer.amd.com/resources/heterogeneouscomputing/opencl-zone/ 14. HSA FoundationAMD, ARM, Samsung, Texas Instruments, Qualcomm, MediaTek, Imagination Technologies 15. Intel Larrabee -&gt; Xeon Phi x86 Pentium 16. Intel Xeon Phi 5110P: 60 Pentium 1GHz EM64T, 4 / 512 SIMD 512 L2 1 TFLOPS 6/8 Gb GDDR5 RHEL 6.x SuSE Linux 12 17. Intel Xeon Phi 18. NVIDIA 269 , , "", , , . , , " ", . , ... 2009 - - " ", 19. 1. HSA Foundation, NVIDIA, AMD, Khronos Group 2. thg.ru - Intel Xeon Phi 3. Wikipedia 4. habrahabr.ru 5. - " " http://www.parallel-compute.ru/ 6. opencl.org 7. gpgpu.org 8. 3-4 . - . . 20. !</p>