转自新浪博客： http://blog.sina.com.cn/s/blog_48b9e1f90100fm56.html
一、CUDA 编程模型
2009-10-21
CUDA的代码分成两部分，一部分在host(CPU)上运行，是普通的C代码；另一部分在device(GPU)上运行，是并行代码，称为kernel，由nvcc进行编译。
Kernel产生的所有线程成为Grid。在并行部分结束后，程序回到串行部分即到host上运行。
在CUDA中，host和device有不同的内存空间。所以在device上执行kernel时，程序员需要把host memory上的数据传送到分配的device memory上。在device执行完以后，需要把结果从device传送回host，并释放device memory。CUDA runtime system提供了API给程序员做这些事情。
int pthread_create(
pthread_t *tid,
const pthread_attr_t *attr,
void*(*start_routine)(void*),
void *arg
); 继续阅读
多谢大家最近的关注，我们的大学生关注资讯网站于今天正式改版了。
杯具洗具：关注大学生文化，关注杯具洗具的人生
给大家带来的不便还请谅解，我们会继续为大家带来更多更好的大学生资讯。谢谢！
Basic knowledge
LISP: LISt Processor. It is a function language for the list.
Data-driven machine: different from the old machines. It execute with the data flow.
Pseudo-result: not a actual-result but can be used in the next function as a semi-result.
Processing element: the basic unit of processing. We often call it PE.
Lazy uation: uation of a computation is delayed until the following computation requires the actual argument values.
A new control mechanism: use a data-driven architecture (one of non von-Neumann computers) to exhibits full potential for parallelism both in hardware and software.
Parallel: Divide a program into different piece and execurate in serval processing element at the same time. To do so, we can accelerate the speech of processing time and use the time and space wisely.
Semi-result: A cons operation include a pseudo-result or a semi-result
Actual-result: result with the real data after execurating.
Packet oriented architecture: data transference between each section in a PE is done by a packet in a pipeline manner as well as between each PE
Pseudo-result lifetime: a time interval from the time when a new pseudo-result is created, to the time when the value of the result becomes actual-result.
First section, the organization of the data-driven machine is described.
Function uation scheme –
to achieve eager uation with pseudo-result, allows some degree of overlapping of computation.
Machine Organization –
multiprocessing system with a number of identical PEs in wich each PE is connected via a packet communication network. 继续阅读
Now I have a desktop, a laptop, a sony play station 3 with 3 mouse and 3 keyboards.
Programming in Fedora 9 and Yellow Dog Linux with Cell BE SDK and IBM Cell simulator.
The main theme is develop a waiting mechanism for data driven machine and uate the mechanism.
I want to write a formal essay and publish it before I go back home.
Need to work hard now.
One year ago, I registed a host for personal blog. But the speed is not very fast, in another word, it is very slow sometimes. So I have to cancel it. Later on, I opened a new blog on sina.com.cn and moved all my msn and qzone articles into it. It seems run very well.
So today, I’m back to this blog again and try to figure this out. This is just another world of nobita gu who is at tsukuba university right now.
Hope things go well this time.
Good night.