Message boards : Number crunching : Ralph support OpenCL ?
Previous · 1 · 2 · 3 · 4 · 5 . . . 8 · Next
Author | Message |
---|---|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
sgaboinc Send message Joined: 8 Jul 14 Posts: 20 Credit: 4,159 RAC: 0 |
a muse on vectorized computing SSE/AVX/AVX2/OpenCL e.g. GPU there has been quite a bit of talk about vectorized computing i.e. the use os GPU and AVX2 for highly vectorized computing etc i actually did a little bit of experiment, i'm running on haswell i7 4771 (non-k) asus h87-pro motherboard & 16G 1600ghz ram i tried openblas http://www.openblas.net/ https://github.com/xianyi/OpenBLAS ./dlinpack.goto these are the benchmarks SIZE Residual Decompose Solve Total 100 : 4.529710e-14 503.14 MFlops 2000.00 MFlops 514.36 MFlops 500 : 1.103118e-12 8171.54 MFlops 3676.47 MFlops 8112.38 MFlops 1000 : 5.629275e-12 45060.27 MFlops 2747.25 MFlops 43075.87 MFlops 5000 : 1.195055e-11 104392.81 MFlops 3275.04 MFlops 102495.20 MFlops 10000 : 1.529443e-11 129210.71 MFlops 3465.54 MFlops 127819.77 MFlops ok quite impressive ~128 Gflops on a haswell i7 desktop PC running at only 3.7ghz! that almost compare to an 'old' supercomputer Numerical wind tunnel in Tokyo https://en.wikipedia.org/wiki/Numerical_Wind_Tunnel_%28Japan%29 but what become immediately very apparent is also that only very large matrices 10,000 x 10,000 benefits from the vectorized codes (i.e. AVX2) in the *decompose* part. if you have tiny matrices say 100x100 in size that gives a paltry 514.36 Mflops, less than 100 times (or could say 1/200) of that speed of 10,000 x 10,000. The other apparent thing is the *solve* part of the computation, you could see that while the decompose part which involves a matrix multiplication (e.g. DGEMM) can reach speeds of 128 Ghz, *but* the *solve* part *did not benefit* from all that AVX2 vectorized codes showing little improvements for different matrices sizes! this has major implications, it means that whether you have a good cpu with AVX2 etc or that you have a large GPU that can process say vectorized / parallel 1000s floating point calcs per clock cycle. But if your problems are small (e.g. 100x100) or that it cannot benefit from such vectorized codes much of that GPU capacity and even for this instance AVX2 may simply be *unused* and will *not benefit* from all that expensive vectorized hardware (e.g. AVX2 and expensive GPU cards capable of processing possibly thousands of vector computation per cycle, e.g. thousands of gpu simd cores) i'd guess this reflect in a way Amdahl's law https://en.wikipedia.org/wiki/Amdahl%27s_law Gene Amdahl passed away recently & perhaps this could be a little tribute to him for having 'seen so far ahead' from back then. http://www.latimes.com/local/obituaries/la-me-gene-amdahl-20151116-story.html |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
This week, during SC15, Kronos Group released OpenCl 2.1 specifications, tools and compilers (with SPIR 1.0). Soon, the release of OpenCl C++ kernel. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
CodeXL 1.8 has a lot of new features: Support for Microsoft Windows® 10 HSA Profiler Power Profiling discrete GPUs Power Profiling API Offline build and analysis of OpenGL shaders New Static Analyzer enhanced ISA view GPU Debugger support up to OpenGL 4.5 64-bit offline build for OpenCL™ kernels CPU Profiling for the “Carrizo†APU CPU Profiling on VMWare |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
New CodeXl is out. CodeXL 1.9 includes: Build and Static Analysis of these types of OpenGL® shaders on Microsoft Windows® and Linux®: Vertex Tessellation Evaluation Geometric Fragment Compute Faster processing of CPU profile data; Support for longer CPU profiling sessions; API for controlling CPU Profiling data collection from app code; HSAIL Debugger (beta) Support for Ubuntu 15.04 Many user-experience enhancements plus more. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Code XL 2.0 is here (and now is open source!!). |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Now CodeXl is 2.1 and is on GitHub |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
|
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Visual Studio 2017 is here. And Codeplay releases the first open beta of ComputeCpp, their implementation of SYCL to bring C++ to OpenCl |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
New KDevelop 5.1 supports OpenCl (and c/c++/c#,java,bash,fortran,pascal,python,etc...) and runs also on Windows and not only on Linux. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
Arm opens up Compute Library with OpenCl Arm Compute Library |
robertmiles Send message Joined: 13 Jan 09 Posts: 103 Credit: 331,865 RAC: 0 |
I'll consider starting to make some BOINC applications for OpenCL AFTER I find and take an online class in OpenCL that is intended for programming GPUs, not FPGAs. I cannot travel enough to use in-person classes instead. BOINC now supports use of CUDA and OpenCL; I've seen no information on whether it can also handle the various other ways of programming GPUs that have already been mentioned in this thread. I'd expect library compatibility problems to block use of some of them at least until future versions of BOINC build in replacements for the incompatible sections of the libraries. |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
OpenCl 2.2 is official. - Support to OpencCl C++ - Spir 1.2 - SyCl 2.2 and other intersting things |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
During IWOCL Kronos Group has annunced that they wants to unify OpenCl and Vulkan Api. That's very interesting! |
[VENETO] boboviz Send message Joined: 9 Apr 08 Posts: 905 Credit: 1,892,541 RAC: 294 |
ROCm is now at 1.6 version |
Message boards :
Number crunching :
Ralph support OpenCL ?
©2024 University of Washington
http://www.bakerlab.org