<?xml version="1.0" encoding="utf-8" ?><rss version="2.0"><channel><title><![CDATA[bit262426738的专栏]]></title><description><![CDATA[]]></description><link>https://blog.youkuaiyun.com/bit262426738</link><language>zh-cn</language><generator>https://blog.youkuaiyun.com/</generator><copyright><![CDATA[Copyright &copy; bit262426738]]></copyright><item><title><![CDATA[各家处理器资源网站]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/46508811</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/46508811</guid><author>bit262426738</author><pubDate>Mon, 15 Jun 2015 22:13:10 +0800</pubDate><description><![CDATA[1、Intel官方文档：
http://www.intel.com/content/www/us/en/processors/architectures-software-developer-manuals.html

2、AMD官方文档：
http://developer.amd.com/resources/documentation-articles/developer-guides-]]></description><category></category></item><item><title><![CDATA[Integrated gem5 + GPGPU-Sim Simulator]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/46472919</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/46472919</guid><author>bit262426738</author><pubDate>Fri, 12 Jun 2015 16:15:51 +0800</pubDate><description><![CDATA[原文地址：http://cpu-gpu-sim.ece.wisc.edu/










Integrated gem5 + GPGPU-Sim Simulator



Last modified on: 04/16/2015 02:52:36 CST






Overview
Simulation Flow
Package Layout]]></description><category></category></item><item><title><![CDATA[GPGPU-SIM Code Study]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/46472825</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/46472825</guid><author>bit262426738</author><pubDate>Fri, 12 Jun 2015 16:08:20 +0800</pubDate><description><![CDATA[原文地址http://people.cs.pitt.edu/~yongli/notes/gpgpu/GPGPUSIMNotes.html

GPGPU-SIM Code Study (version: 3.1.2)

                                                                                       ]]></description><category></category></item><item><title><![CDATA[nVidia GPGPU vs AMD Radeon HD Graphics执行模式对比]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/42882573</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/42882573</guid><author>bit262426738</author><pubDate>Mon, 19 Jan 2015 21:29:25 +0800</pubDate><description><![CDATA[大家做高性能计算的朋友，想必对CPU的执行模式已经非常熟悉了吧。当代高级些的CPU一般采用超标量流水线，使得毗邻几条相互独立的指令能够并行执行——这称为指令集并行（ILP，Instruction-Level Parallelism）；而像x86引入的SSE（Streaming SIMD Extension）、AVX（Advanced Vector Extension），以及ARM的NEON技术都属]]></description><category></category></item><item><title><![CDATA[安卓手机GPU OpenCL总结]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/42738127</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/42738127</guid><author>bit262426738</author><pubDate>Thu, 15 Jan 2015 10:47:54 +0800</pubDate><description><![CDATA[安卓手机GPU OpenCL总结




2014-07-24 07:30:26优快云-wcj0626-点击数:276








 前段时间，把市面上手机GPU OpenCL支持情况做了一个总结。总结如下：目前,手机 GPU 市面有四个公司产品:Qualcomm, Imagination Technologies,ARM, Vivante,分别对应的产品如下: ]]></description><category></category></item><item><title><![CDATA[linux下vim命令详解]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/36699869</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/36699869</guid><author>bit262426738</author><pubDate>Thu, 03 Jul 2014 21:16:44 +0800</pubDate><description><![CDATA[linux下vim命令详解




高级一些的编辑器，都会包含宏功能，vim当然不能缺少了，在vim中使用宏是非常方便的：

:qx     开始记录宏，并将结果存入寄存器x
q     退出记录模式
@x     播放记录在x寄存器中的宏命令
稍微解释一下，当在normal模式下输入:qx后，你对文本的所有编辑动作将会被记录下来，再次输入q即退出了记录模
式，然后输入@x对]]></description><category></category></item><item><title><![CDATA[Questions about OPENCL global and local work size]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/25888495</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/25888495</guid><author>bit262426738</author><pubDate>Thu, 15 May 2014 14:15:26 +0800</pubDate><description><![CDATA[favorite
11




searching the nvidia forums I found these questions, which are also of interest to me, but nobody had answered them in the
 last four days or so. Can you help?
Original forum p]]></description><category></category></item><item><title><![CDATA[OPENCL如何获取获取最小线程并行粒度]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/25695781</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/25695781</guid><author>bit262426738</author><pubDate>Tue, 13 May 2014 11:12:15 +0800</pubDate><description><![CDATA[由于OpenCL是为各类处理器设备而打造的开发标准的计算语言。因此跟CUDA不太一样的是，其对设备特征查询的项更上层，而没有提供一些更为底层的特征查询。比如，你用OpenCL的设备查询API只能获取最大work group size，但无法获取到最小线程并行粒度。

但是，由于最小线程并行粒度对于OpenCL应用领域最广的GPU而言确实是一个比较重要的参数。如果你的work group的wor]]></description><category></category></item><item><title><![CDATA[clGetDeviceInfo]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24718157</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24718157</guid><author>bit262426738</author><pubDate>Tue, 29 Apr 2014 15:22:15 +0800</pubDate><description><![CDATA[clGetDeviceInfo

Get information about an OpenCL device.









cl_int clGetDeviceInfo(

cl_device_id device,



 

cl_device_info param_name,



 

size_t param_value_size]]></description><category></category></item><item><title><![CDATA[矩阵乘优化的思考]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24449683</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24449683</guid><author>bit262426738</author><pubDate>Fri, 25 Apr 2014 08:08:33 +0800</pubDate><description><![CDATA[所谓矩阵的乘法，就是现有两个矩阵，arr和brr。进行相乘运算，即：arr的第一行乘brr的第一列对应位置之后作和，作为结果的第一行第一列。arr的第一行乘brr的第二列对应位置之后作和，作为结果的第一行第二列。。。。一次类推~

没有做任何优化的代码：
__kernel void MatrixMul(__global const float *arr, __global const flo]]></description><category></category></item><item><title><![CDATA[内存的物理结构和工作原理]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24317145</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24317145</guid><author>bit262426738</author><pubDate>Tue, 22 Apr 2014 17:06:42 +0800</pubDate><description><![CDATA[内存的物理结构和工作原理

作者：xiaofeng　来源：PConline　发布时间：2012-11-01　点击数：1163


　　内存也叫主存，是PC系统存放数据与指令的半导体存储器单元，也叫主存储器（Main Memory），通常分为只读存储器（ROM-Read Only Memory）、随机存储器（RAM-Red Access Memory）和高速缓存存储器（Cache）。我们平]]></description><category></category></item><item><title><![CDATA[N卡Kepler 架构]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24270519</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24270519</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 20:32:22 +0800</pubDate><description><![CDATA[Kepler (microarchitecture)

From Wikipedia, the free encyclopedia










This article needs more links
 to other articles to help integrate it into the encyclopedia. Please
 help imp]]></description><category></category></item><item><title><![CDATA[N卡Fermi架构]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24269373</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24269373</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 19:45:33 +0800</pubDate><description><![CDATA[Fermi (microarchitecture)

From Wikipedia, the free encyclopedia




Fermi is the codename for a GPU microarchitecture developed
 by Nvidia as the successor to the Tesla microarchitecture.
 I]]></description><category></category></item><item><title><![CDATA[cuda学习笔记之异步并行执行]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24256041</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24256041</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 15:27:23 +0800</pubDate><description><![CDATA[分类： CUDA2010-01-03 21:13 2526人阅读 评论(1) 收藏 举报

cuda多线程direct3d集群api编译器

异步函数使得主机端与设备端并行执行：控制在设备还没有完成前就被返回给主机线程；
    包括：
      kernel启动； 以Async为后缀的内存拷贝函数； device到device内存拷贝函数；存储器初始化函数，比如cudaMemset]]></description><category></category></item><item><title><![CDATA[CUDA下在Host端分配的几种内存模式]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24255971</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24255971</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 15:26:20 +0800</pubDate><description><![CDATA[平时我们使用的内存都是Pageable(交换页)的，而另一个模式就是Pinned(Page-locked)，实质是强制让系统在物理内存中完成内存申请和释放的工作，不参与页交换，从而提高系统效率，需要使用cudaHostAlloc和cudaFreeHost来分配和释放。

　　优点

　　1、带宽更高

　　2、内核处理和内存拷贝可同时进行

　　3、可以内存映射(mapped)
]]></description><category></category></item><item><title><![CDATA[CUDA学习笔记之Tesla图形与计算架构和通用计算模型]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24255805</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24255805</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 15:24:30 +0800</pubDate><description><![CDATA[Tesla图形与计算架构：流处理器阵列（scalable streaming processor array，SPA）+存储器系统，由片上互联网络连接；
存储器系统：存储器控制器（MMC）、固定功能的光栅操作单元（raster operation processors，ROP），二级纹理缓存；
            MMC：负责控制片外的DRAM显存，每个存储器可以提供64bit位宽；
 ]]></description><category></category></item><item><title><![CDATA[CUDA学习笔记之程序优化]]></title><link>https://blog.youkuaiyun.com/bit262426738/article/details/24254011</link><guid>https://blog.youkuaiyun.com/bit262426738/article/details/24254011</guid><author>bit262426738</author><pubDate>Mon, 21 Apr 2014 15:02:37 +0800</pubDate><description><![CDATA[CUDA学习笔记之程序优化

分类： CUDA2010-01-05 17:18 3833人阅读 评论(3) 收藏 举报
cuda优化conflict存储算法数学计算

CUDA程序优化
CUDA程序优化应该考虑的点：
精度：只在关键步骤使用双精度，其他部分仍然使用单精度浮点以获得指令吞吐量和精度的平衡；
           延迟：需要首先缓冲一部分数据，缓冲的大小应该可以保证每个]]></description><category></category></item></channel></rss>