gputools:安装错误 [英] gputools: error in installation
问题描述
我正在使用NVidia Tesla 2050 GPU卡设置新的Dell Precision工作站。我想安装R的包gputools。我的操作系统是带有KDE 4.4的openSuse 11.3。
I am setting up a new Dell Precision workstation with an NVidia Tesla 2050 GPU card. I would like to install R's package gputools. My OS is openSuse 11.3 with KDE 4.4.
我下载了NVidia的CUDA Toolkit 3.2并安装在/ usr / local / cuda中,我也下载了最新版本的CULA工具集(版本R10)并安装在/ usr / local / cula中。
I downloaded NVidia's CUDA Toolkit 3.2 and installed it in /usr/local/cuda, I also downloaded the latest version of the CULA Tools set (version R10) and installed it in /usr/local/cula.
当尝试从R中安装gputools时使用:
install.packages gputools)
我得到以下错误消息:
When trying to install gputools from within R using: install.packages("gputools") I get the following error message:
classification.cu(735): error: argument of type "unsigned int *" is incompatible with parameter of type "size_t *"
classification.cu(735): error: argument of type "unsigned int *" is incompatible with parameter of type "size_t *"
classification.cu(1042): error: argument of type "unsigned int *" is incompatible with parameter of type "size_t *"
classification.cu(1042): error: argument of type "unsigned int *" is incompatible with parameter of type "size_t *"
4 errors detected in the compilation of "/tmp/tmpxft_00003d8d_00000000-12_classification.compute_12.cpp1.ii".
make: *** [classification.o] Error 2
ERROR: compilation failed for package ‘gputools’
* removing ‘/home/moswald/R/x86_64-unknown-linux-gnu-library/2.12/gputools’
The downloaded packages are in
‘/tmp/RtmphI30zE/downloaded_packages’
Warning message:
In install.packages("gputools", dependencies = TRUE) :
installation of package 'gputools' had non-zero exit status
使用较早版本的CULA和CUDA工具集也没有帮助。
Using older versions of the CULA and CUDA toolsets does not help either.
推荐答案
为了支持具有大量内存的设备(例如C2070具有6GB设备内存),CUDA-3.2使用size_t 来描述存储器的量,而不是如在CUDA <= 3.1中使用的无符号整数。这看起来像是使用CUDA-3.2的典型案例,其中需要CUDA-3.1。您可以通过添加
To support devices with large amounts of memory (such as the C2070 with 6GB device memory), CUDA-3.2 uses "size_t" to describe amounts of memory rather than "unsigned int" as was used in CUDA <= 3.1. This looks like a classic case of using CUDA-3.2 where CUDA-3.1 is expected. You may be able to hack things with the CUDA-3.2 compiler by adding
-DCUDA_FORCE_API_VERSION=3010
到您的nvcc命令行。
to your nvcc command-line.
这篇关于gputools:安装错误的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!