Package Details: siftgpu 0.5.400-8

Git Clone URL: https://aur.archlinux.org/siftgpu.git (read-only)
Package Base: siftgpu
Description: Sift Features over GPU using GLSL or CUDA
Upstream URL: http://ccwu.me/
Keywords: cuda feature-extraction glsl
Licenses: custom
Submitter: dbrgn
Maintainer: bartus
Last Packager: bartus
Votes: 4
Popularity: 0.005617
First Submitted: 2015-10-17 19:22
Last Updated: 2018-06-05 12:37

Required by (1)

Sources (2)

Latest Comments

1 2 Next › Last »

bartus commented on 2018-06-05 13:21

@ILMostro7: fixed.

ILMostro7 commented on 2018-06-01 19:32

The optdepends doesn't take a version for packages listed. Otherwise, you get the following: ==> ERROR: optdepends contains invalid characters: '>='

bartus commented on 2018-04-03 21:08

@ILMostro7: I think, I find the issue, the problem arose when an extra copy of libcudart.so on your system triggers cuda kernel build. Now it checks only for /opt/cuda/lib64/libcudart.so.

bartus commented on 2018-03-29 19:36

@ILMostro7: siftgpu works also with GLSL and opencl besides cuda. All three are run time detected by makefile script.

  9 # enable CUDA-based SiftGPU?
 10 simple_find_cuda = $(shell locate libcudart.so)
 11 ifneq ($(simple_find_cuda), )
 12         siftgpu_enable_cuda = 1
 13 else
 14         siftgpu_enable_cuda = 0
 15 endif
 16 

It's strange that it fails on a system without cuda ... I can build siftgpu in chroot without cuda with no problem.

ILMostro7 commented on 2018-03-29 19:11

Should cuda be a dependency, rather than optdepends? I get a failure if cuda is not installed, as the installation fails to find "cuda.h".

bartus commented on 2018-03-28 23:08

@Harvie: Yes, looks like it works, updating.

Harvie commented on 2018-03-28 22:03

I've removed the gcc5 hack and it seems to work

Harvie commented on 2018-03-28 21:57

Can we build this with gcc7?

dbrgn commented on 2017-04-19 07:45

I'm disowning this package since I don't use it myself anymore.

op3 commented on 2017-04-19 07:24

http://wwwx.cs.unc.edu/~ccwu/cgi-bin/siftgpu.cgi does not exist anymore. I found a (modified) copy of it on GitHub: https://github.com/pitzer/SiftGPU

It can be downloaded as a tar archive from https://github.com/pitzer/SiftGPU/archive/master.tar.gz
You also have to replace cd "$srcdir/$folder" with cd "$srcdir/$folder-master" and recalculate the sha256sum.