Skip to content

Tegrajpeg and perf code - #576

Merged
xlz merged 8 commits into
OpenKinect:masterfrom
xlz:tegrajpeg
Feb 18, 2016
Merged

xlz merged 8 commits into
OpenKinect:masterfrom
xlz:tegrajpeg

Conversation

@xlz

@xlz xlz commented Feb 18, 2016

Copy link
Copy Markdown
Member

There are some changes here that may be hard to test. I have tested these quite some time, but I'll give it some days of review.

  • Add Tegra JPEG decoder - this performance is better than previously obtained, probably because of general improvement of underlying code
  • Make CUDA 6.5 work with C++11 (compile .cu with c++98, and link it with c++11 objects - bad)
  • ./Protonect -gpu=1 cl (when the first gpu is nvidia and it doesn't work)
  • Set thread names so it's easier to look at top - cross platform code, harder to test
  • Print detailed stats for perf

I have put together some benchmark results, check out https://github.com/OpenKinect/libfreenect2/wiki/Performance

@xlz

xlz commented Feb 18, 2016

Copy link
Copy Markdown
Member Author

Hi Jetson users! The support for TK1 finally arrives in the libfreenect2 upstream, and NEW performace record: 90HZ depth and 80 Hz JPEG!

Call for testing @anshulvj @YoniSmolin @cretaceous-creature @ibaranov-cp @RyanGordon @l0g1x @andreykurenkov @wjchen84 @vishu2287 @bingyo

@cretaceous-creature

Copy link
Copy Markdown

GJ!

2016-02-18 15:20 GMT+09:00 Lingzhu Xiang notifications@github.com:

Hi Jetson users! The support for TK1 finally arrives in the libfreenect2
upstream, and NEW performace record: 90HZ depth and 80 Hz JPEG!

Call for testing @anshulvj https://github.com/anshulvj @YoniSmolin
https://github.com/YoniSmolin @cretaceous-creature
https://github.com/cretaceous-creature @ibaranov-cp
https://github.com/ibaranov-cp @RyanGordon
https://github.com/RyanGordon @l0g1x https://github.com/l0g1x
@andreykurenkov https://github.com/andreykurenkov @wjchen84
https://github.com/wjchen84 @vishu2287 https://github.com/vishu2287
@bingyo https://github.com/bingyo

—
Reply to this email directly or view it on GitHub
#576 (comment)
.

xlz added 4 commits February 18, 2016 10:36
The internal logging API is used by the CUDA processor.

For CUDA 6.5 and -DENABLE_CXX11=ON, the cuda object is compiled
with C++98 and other objects with C++11. Thus remove std::string
for being incompatible ABI across C++98 and C++11.
xlz added 4 commits February 18, 2016 15:06
CUDA 7.0 is the first version that supports C++11.

Though linking C++11 objects with non-C++11 ones is problematic.
Use cmake -DENABLE_PROFILING=ON (OFF by default).
@xlz
xlz merged commit 8efb59b into OpenKinect:master Feb 18, 2016
@xlz
xlz deleted the tegrajpeg branch February 18, 2016 20:51
@anshulvj

Copy link
Copy Markdown

Hi Lingzhu,

Do you have a bibtex for your libfreenect2 library?

On Wed, Feb 17, 2016 at 11:20 PM, Lingzhu Xiang notifications@github.com
wrote:

Hi Jetson users! The support for TK1 finally arrives in the libfreenect2
upstream, and NEW performace record: 90HZ depth and 80 Hz JPEG!

Call for testing @anshulvj https://github.com/anshulvj @YoniSmolin
https://github.com/YoniSmolin @cretaceous-creature
https://github.com/cretaceous-creature @ibaranov-cp
https://github.com/ibaranov-cp @RyanGordon
https://github.com/RyanGordon @l0g1x https://github.com/l0g1x
@andreykurenkov https://github.com/andreykurenkov @wjchen84
https://github.com/wjchen84 @vishu2287 https://github.com/vishu2287
@bingyo https://github.com/bingyo

—
Reply to this email directly or view it on GitHub
#576 (comment)
.

@xlz

xlz commented Feb 21, 2016

Copy link
Copy Markdown
Member Author

@anshulvj

Copy link
Copy Markdown

Thanks!

On Sun, Feb 21, 2016 at 4:17 PM, Lingzhu Xiang notifications@github.com
wrote:

@anshulvj https://github.com/anshulvj Yes,
https://zenodo.org/record/45314/export/hx

—
Reply to this email directly or view it on GitHub
#576 (comment)
.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants