/[opencvs]/eyes/gpu/bg_opencl_work.c
ViewVC logotype

Log of /eyes/gpu/bg_opencl_work.c

Parent Directory Parent Directory | Revision Log Revision Log


Links to HEAD: (view) (annotate)
Sticky Tag:

Revision 1.20 - (view) (annotate) - [select for diffs]
Sun Nov 29 00:35:20 2015 UTC (2 years, 11 months ago) by hib
Branch: MAIN
CVS Tags: HEAD
Changes since 1.19: +27 -14 lines
Diff to previous 1.19
work on japan

Revision 1.19 - (view) (annotate) - [select for diffs]
Wed Oct 21 18:57:30 2015 UTC (3 years ago) by hib
Branch: MAIN
Changes since 1.18: +262 -106 lines
Diff to previous 1.18
get stuff from cookie

Revision 1.18 - (view) (annotate) - [select for diffs]
Mon Oct 6 06:39:04 2014 UTC (4 years, 1 month ago) by hib
Branch: MAIN
Changes since 1.17: +106 -262 lines
Diff to previous 1.17
trying to incorporate changes that was in burner. in the case of gpu - cookie might be more accurate

Revision 1.17 - (view) (annotate) - [select for diffs]
Wed Apr 9 04:31:36 2014 UTC (4 years, 7 months ago) by hib
Branch: MAIN
Changes since 1.16: +5 -5 lines
Diff to previous 1.16
cleaned this up alot. Now we can handl millions of points and the machine is more stable and bug free.

Revision 1.16 - (view) (annotate) - [select for diffs]
Fri Apr 4 17:22:25 2014 UTC (4 years, 7 months ago) by hib
Branch: MAIN
Changes since 1.15: +4 -6 lines
Diff to previous 1.15
cool

Revision 1.15 - (view) (annotate) - [select for diffs]
Thu Apr 3 15:27:05 2014 UTC (4 years, 7 months ago) by hib
Branch: MAIN
Changes since 1.14: +1 -1 lines
Diff to previous 1.14
cool. WIP

Revision 1.14 - (view) (annotate) - [select for diffs]
Tue Apr 1 23:04:01 2014 UTC (4 years, 7 months ago) by hib
Branch: MAIN
Changes since 1.13: +63 -45 lines
Diff to previous 1.13
got antique created.

Revision 1.13 - (view) (annotate) - [select for diffs]
Sun Feb 2 11:15:46 2014 UTC (4 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.12: +14 -6 lines
Diff to previous 1.12
adjusted this so it can handle size 0
Also made it do sincos by defualt but you can turn it off with a def

Revision 1.12 - (view) (annotate) - [select for diffs]
Wed Aug 14 00:44:33 2013 UTC (5 years, 3 months ago) by hib
Branch: MAIN
Changes since 1.11: +198 -66 lines
Diff to previous 1.11
Added nyquest for the gpu based builds

Revision 1.11 - (view) (annotate) - [select for diffs]
Thu Feb 28 21:26:47 2013 UTC (5 years, 8 months ago) by hib
Branch: MAIN
Changes since 1.10: +1 -0 lines
Diff to previous 1.10
fp_to_txt - allows for level transitions, and also spartain mode
gpu code - now works well, and 60 times faster! yeah for holograms!
half.c - new program to cut a double .fp file in half/half (1 quarter area)

Revision 1.10 - (view) (annotate) - [select for diffs]
Wed Feb 20 00:50:58 2013 UTC (5 years, 8 months ago) by hib
Branch: MAIN
Changes since 1.9: +1 -36 lines
Diff to previous 1.9
got rid of many debugging messages

Revision 1.9 - (view) (annotate) - [select for diffs]
Fri Feb 15 07:36:39 2013 UTC (5 years, 8 months ago) by hib
Branch: MAIN
Changes since 1.8: +20 -18 lines
Diff to previous 1.8
Well,  build_double works now.  I added some .pl scrpts here to help with things
But the sincod takes longer, and adding colors doesn't really cut it much.
I did 2 540 dpi versions and you know what?  It looked better without attenuation.
The center was still interesting,  but the outside less so.  It might grab more which makes me conflicted,  but it would capture less.

So I am going to stick with no attenuation ,   and modify the double program just to do the cosine - or make a single version or something
I am also going to push it to 2880x1440,  and maybee further.

It would be cool to do a 2 foot or 3 foot wide strip that is 8 feet long, as the center parts are the coolest,  but then the heart would be lost.

Conflicted,  want to do both - one with and without attenuation.

Revision 1.8 - (view) (annotate) - [select for diffs]
Tue Feb 12 07:36:56 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.7: +29 -31 lines
Diff to previous 1.7
optomizing...

Revision 1.7 - (view) (annotate) - [select for diffs]
Tue Feb 12 06:34:23 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.6: +666 -158 lines
Diff to previous 1.6
OK - now the multi-gpu works.
Time to speed it up.

Revision 1.6 - (view) (annotate) - [select for diffs]
Tue Feb 12 06:05:26 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.5: +30 -18 lines
Diff to previous 1.5
Got it to work with a single gpu fast.
the trick was to call CLFlush RIGHT after the Nqueue otherwise
the card gets frozen and won't queue up the second job in time.
Also moved from ATI SDK 2.3 to 2.8 which helps with the cpu hogging a bit.
Now I am going to go back to multi gpu because thats better!

Revision 1.5 - (view) (annotate) - [select for diffs]
Mon Feb 11 17:37:02 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.4: +185 -285 lines
Diff to previous 1.4
fuck.  the multi gpu code just didn't work.  It would
wait till both are done.  fuck that.
This is going back to v1.2 the code that worked as a single gpu
Maybe I can get two jobs running at once?

Revision 1.4 - (view) (annotate) - [select for diffs]
Mon Feb 11 16:57:41 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.3: +13 -14 lines
Diff to previous 1.3
broken.
I cant get both gpus to work together

Revision 1.3 - (view) (annotate) - [select for diffs]
Sun Feb 10 15:18:24 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.2: +276 -175 lines
Diff to previous 1.2
Got the opencl to use both GPUS - which wastes both cpus but what the heck.

Revision 1.2 - (view) (annotate) - [select for diffs]
Sun Feb 10 01:01:35 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
Changes since 1.1: +97 -30 lines
Diff to previous 1.1
It works, and it is hella fast.
for 1024x1024 it takes 2 1/2 minutes without attenuation, 2 with attenuation

Revision 1.1 - (view) (annotate) - [select for diffs]
Thu Feb 7 06:55:39 2013 UTC (5 years, 9 months ago) by hib
Branch: MAIN
converting this to opencl - this version rolls out a large program based on the points
But it fails with only 1000 points.  So I will have to have a loop in it.

This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, select a symbolic revision name using the selection box, or choose 'Use Text Field' and enter a numeric revision.

  Diffs between and
  Type of Diff should be a

Sort log by:

  ViewVC Help
Powered by ViewVC 1.1.5