CUDA：利用显卡进行超级科学运算
NVIDIA的CUDA官方主页:nvidia/object/cuda_home_cn.html51CTO的开勇博客:openhero.blog.51cto/
CUDA?是一种由NVIDIA推出的通用并行计算架构,该架构使GPU可以解决杂乱的计算问题。它包括了CUDA命令集架构(ISA)以及GPU内部的并行计算引擎。开发人员现在可以使用C言语来为CUDA?架构编写程序,C言语是使用最广泛的一种高档编程言语。所编写出的程序于是就可以在支持CUDA?的管理器上以超高功能运行。将来还会支持其它言语,包括FORTRAN以及C++。
CUDA技术的效果即是用来进行科/学运算的。尽管现在CPU的办公才能很强悍,但是因为CPU的通用性而集成了逻辑控制、逻辑运算等单元,形成其在数值运算方面的效率低下。而GPU自身即是为了运算而生,因为了图形管理要用到很多的运算才能。GPU的设计和通用CPU的设计不一样,GPU是流式管理器,所以在运算才能方面比CPU强几十倍或几百倍。夸大点说,某个科/学运算问题在一般PC上可能要用一个月的时间,而在GPU运算只需求三个小时(引自某教授的讲课)。
lecture6floating-point2008.ppt
lecture7casestudyvmd2008.ppt
lecture8casestudy2008.ppt
lecture9cudaconclusion2008.ppt
(NVIDIA's CUDA official home page: nvidia/object/cuda_home_cn.html
51CTO's Kaiyong Blog: openhero.blog.51cto/
CUDA? is a general-purpose parallel computing architecture introduced by NVIDIA that enables GPUs to solve complex computing problems. It includes the CUDA Command Set Architecture (ISA) and the parallel computing engine inside the GPU. Developers can now write programs for the CUDA? architecture using C, the most widely used high-level programming language. Programs written can then run with super high functionality on a CUDA?-capable manager. Other languages ??will be supported in the future, including FORTRAN and C.
The effect of CUDA technology is used to perform scientific/scientific operations. Although the office ability of CPU is very powerful, but because of the versatility of CPU, it integrates logic control, logic operation and other units, resulting in its low efficiency in numerical operation. The GPU itself is born for computing, because graphics management requires a lot of computing power. The design of the GPU is different from that of the general-purpose CPU. The GPU is a stream manager, so it is dozens or hundreds of times stronger than the CPU in terms of computing power. To exaggerate, a certain scientific/scientific computing problem may take a month on a general PC, but only three hours on a GPU (quoted from a lecture by a professor).
lecture6floating-point2008.ppt
lecture7casestudyvmd2008.ppt
lecture8casestudy2008.ppt
lecture9cudaconclusion2008.ppt)
页:
[1]