m********e
发帖数: 5088 | 1 我想用fortran来找1个矩阵的特征值，打算使用netlib里面的lanczos，但是我看不懂里
面的解释，也不知道该怎么使用。
有没有大牛用过那个package，比如我现在有个矩阵M，该怎么用lanczos呢 |
|
O****C
发帖数: 368 | 2 完全不能打开。
我已经重装几次都不行。麻烦你们帮我看看error report。 谢谢。
Microsoft Error Reporting log version: 2.0
Error Signature:
Exception: EXC_BAD_ACCESS
Date/Time: 2014-09-10 22:55:06 +0000
Application Name: Microsoft OneNote
Application Bundle ID: com.microsoft.onenote.mac
Application Signature: ONMC
Application Version: 15.2.140713
Crashed Module Name: MicrosoftOffice
Crashed Module Version: 15.2.140713
Crashed Module Offset: 0x00071f94
Blame Module Name: MicrosoftOffice
Blame Module Version: 15.2.140713
Blame Modu... 阅读全帖 |
|
|
D***n
发帖数: 6804 | 4 MLlib uses the linear algebra package Breeze, which depends on netlib-java,
and jblas. netlib-java and jblas depend on native Fortran routines. You need
to install the gfortran runtime library if it is not already present on
your nodes. MLlib will throw a linking error if it cannot detect these
libraries automatically.
LOL
Spark用的是你仇恨的Fortran做后台计算，没libgfortain都运行不了。
你还扯什么蛋？你现在都没安装过SPARK么。 |
|
D***n
发帖数: 6804 | 5 netlib-java is a wrapper for low-level BLAS, LAPACK and ARPACK that performs
as fast as the C / Fortran interfaces.
只是个Java Warpper而已，真正干活的还是Fortran. 你好意思拿Warpper来比speed？
通过这个事情，鉴定你连大名鼎鼎的netlib都不知道，谈个毛的分布式计算。 |
|
|
|
h***o
发帖数: 539 | 8 你的是要算什么？
最近我刚解决了算associated legendre function的问题，算这个函数
的时候有时候会要算到很大或者很小的数，以前用real*16，但巨慢
后来到netlib上找了一个，它就是用分离指数部分的方法，快多了。
netlib上找吧，如果有的话。 |
|
s********k
发帖数: 212 | 9
netlib bah :-)
www.netlib.org
plenty of deposit of codes for scientific computing in fortran, c/c++
or pascal :-)
Good luck :-) |
|
|
x****y
发帖数: 1853 | 11 matlab 这运算速度跟fortran比算过家家么?
oak ridge那个netlib十有九个都是fortran写的. 你估计是在国内用盗版
VC++出身的吧？ |
|
r**********g
发帖数: 22734 | 12 不做码工学什么java?老实用matlab就是了。科学计算哪有用java的，ampl, ipopt哪个
用的不是c++？去netlib下个算法，不是.f77就是.c |
|
w**z
发帖数: 8232 | 13 http://stackoverflow.com/questions/825221/where-can-i-find-the-
It's a math problem more than CS. Well, don't know how far you want to go..
That is the implementation from math lib
public static double sqrt(double a) {
return StrictMath.sqrt(a); // default impl. delegates to StrictMath
// Note that hardware sqrt instructions
// frequently can be directly used by JITs
// and should be much faster than doing
// Math.... 阅读全帖 |
|
|
|
m*****r
发帖数: 130 | 16 这个是标称，cpu 的规格上有。你也可以自己测试一把，譬如netlib上有个api，或者
你不用它自己
简单测试一下也差不多。 |
|
a*******t
发帖数: 891 | 17 Ingrain has a software + hardware solution
nCipher, Protegrity, NetLib, Application Security are the other the products
I looked at briefly, software based.
Resource was not a concern when i was looking into this so I don't know
about that.
I
email? |
|
z*****k
发帖数: 600 | 18 Image procesisng also may need NetLib, GSL, ffmpeg, ImageMagick, livemedia.
Qt will overtake GTK down the road since it has become LGPL in January. But
GTK is still based on C native for C people which I like. Besides, SDL is
simple and fast.
gcc/g++ is the predominant compiler. But IPP for Linux is free (you need
check licensing details for commercial use)
Linux is generally much better than windows for image processing for
research. |
|
|
s*****k
发帖数: 604 | 20 矩阵运算和 fft 用 MKL 或者 ACML
一直没搞清楚netlib上的东西是什么用license |
|
|
L***n
发帖数: 6727 | 22 gotoblas2内核全是汇编，外面加个fortran和C的wrapper，再包括一个netlib的
reference，那个倒是fortran写的古董，Goto写了篇paper讲他的思路，主要是避免
TLB miss，还有cache miss但是相对来说他对TLB更重视一点，结合他的文章读读他
的汇编代码挺有帮助的，每个函数都不是很长，不过优化还是在算法上搞是正路 |
|
D***n
发帖数: 6804 | 23 愚蠢之至的思路，除了显示你对计算一无所知还能说明啥？
你知道netlib，Lapack, Atlas是什么江湖地位么？你知道这些库花了多少人多少精力
和银子么？还短期无法完成代码，哈哈哈！
Scipy里面也用同样的库，所以你拿SPARK去比，只能比出同样的速度来。
Before building, you will also need to install packages that NumPy and SciPy
depend on
BLAS and LAPACK libraries (optional but strongly recommended for NumPy,
required for SciPy): typically ATLAS + LAPACK, or MKL, or ACML
C and Fortran compilers (typically gcc and gfortran). |
|
|
|
t********o
发帖数: 48 | 26 【 以下文字转载自 Programming 讨论区 】
【 原文由 tsingditto 所发表 】
I download the source code of BLAS
from www.netlib.org/blas
But there is no Makefile, who can
tell me how to compile them
and get my own library file?
or give me a Makefile.
Thanks a lot. |
|
P*****P
发帖数: 57 | 27 i also tried to install the lapack from netlib.org, though failed |
|
b*****y
发帖数: 163 | 28 There is a win32 version of clapack.
You can check them out from netlib. |
|
g****d
发帖数: 64 | 29 www.netlib.org has some f90 interface. That might be useful.
Most of fortran compiler will add an underscore after each file name, such as
filename_ . So when you might need to call filename_(...) instead of
filename(..) |
|
f**l
发帖数: 2041 | 30 LAPACK是矩阵计算的最好的库. 其前身为LINPACK.
Matlab这个软件用的就是这些库.
查查netlib, 应该可以查到. LAPACK应该包含并行功能. |
|
|
f**l
发帖数: 2041 | 32 if stiff, dassl is the best.
if nonstiff, there is a classic code using adams method.
You may check it on netlib also.
ODE solver不需要自己搞的. |
|
|
w**d
发帖数: 2334 | 34 Thanks. I am using TNT from netlib. It seems good. It does not fail even
for singular A.
BTW, what is IT++? |
|
h***o
发帖数: 539 | 35 netlib.org里肯定有一大堆吧，自己慢慢刨 |
|
h***o
发帖数: 539 | 36 netlib上眼睛都找花了，但还是没找到。ft |
|
a******s
发帖数: 232 | 37 你用netlib上的code，写的人怎么牛了我还是不知道 :) |
|
|
b********e
发帖数: 109 | 39 I searched netlib. But didn't find one. |
|
f******n
发帖数: 176 | 40 找一本linear algebra的书，把Lanczos弄明白。
里 |
|
m********e
发帖数: 5088 | 41 呵呵，只是自己比较懒，我情愿把lanczos当作一个黑盒，如果可以像matlab里面那样
[V,E]=eigs(M)就好了，一个命令多简单。可惜还要定义很多东西，麻烦 |
|
w*******e
发帖数: 6 | 42 我觉得如果对于sparse matrix 不是很熟悉的话, 有个 叫作sparse 的C程序库很好用,
它也提供了Fortran 的界面. 如果想在自己的程序中用的话, 这个库还是比较容易
上手的. 比较容易找到的是1.3 (netlib上就是这个版本), 大概因为是比较老的C程
, 所以编译常常会有些问题. 幸运的是, 最近作者推出了新版本, 所以编译也不是问
题了. 如果感兴的话, google一下就可以找到了. |
|
|
|
|
j**u
发帖数: 6059 | 46 google吧。学MPI的话，MPI complete reference不错，在netlib。 |
|
|
|
|
|
