[Radiance-general] Errors with Vertically Tall HDR Images

[Greg Ward]

Hi Alstan,

I didn't realize the Discourse server was down (again).  I'm having Daniel Fuller look into it....

Regarding your HDR file, the one you posted is indeed missing the last 500 or so scanlines, explaining your errors.  Either rtrace died prematurely, or your views.rays file didn't have the full compliment of views.  Running under Windows, we have had issues with RAW floating-point file termination, although I was hoping those had been worked out with the most recent compiler and libraries.

There are other, easier ways to accomplish what you are doing.  The simplest would be the rpict -S option, which combined with the -o option, does almost exactly what you want:  generates multiple views from the same octree with the same options, thus avoiding the start-up costs.  You can write out the many picture files using a "%d" somewhere in the -o output specification.  This avoids the need for pcompos to separate the outputs.

Cheers,
-Greg

> From: "J. Alstan Jakubiec" <alstan at jakubiec.net>
> Date: August 15, 2018 7:33:10 PM PDT
> 
> Dear Radiance friends, 
> (Sorry to post to the mailing list, but the Discourse server seems to be down.)
> 
> I have an issue when I am working with some rather tall HDR images (10,000 - 100,000 pixels tall). The application is that I am running annual discomfort calculations using Jan's gen_dgp_profile approach with many views on very large octrees. In this case, one call to rtrace with a big vwrays output file is better than 40+ to rpict where the octree needs to be loaded into memory each time. I tested this in a model with 38 views and a 500 x 500 pixel output per view, resulting in a 500 x 19,000 pixel image per timestep, which worked great! Now I am working with a model that has 56 views (a 500 x 28,000 pixel image), and I get systematic errors when passing the output to evalglare. 
> Here are the steps I am using:
> 
> I run vwrays 56 (or n) times: vwrays -ff -vta -vv 180 -vh 180 -vp ... -vd ... -x 500 -y 500 >> views.rays 
> rtrace -x 500 -y 28000 -ab 0 -lw 0.0000001 -st 0.15 -ffc !oconv ... < views.rays > image.hdr
> I cut the output using pcompos, and pass it to evalglare: pcompos -x 500 -y 500 image.hdr 0 pixel_offset | evalglare.exe -vta -vv 180 -vh 180 ...
> In this case pixel_offset is 0 for the last view, -500 for the second-to-last, etc.
> I can extract images for every offset -500 through -27500 except for the 0 offset. I get this error for the 0 offset (passing to ra_bmp in this case): 
> image.hdr: read error (y==499)
> ra_bmp: error reading Radiance picture
> 
> I am not sure quite what to make of this and hope that someone here might have an idea. I uploaded an example image here: https://www.dropbox.com/s/7gub8i4row9qmsz/image.hdr?dl=0 Is the error with the render itself or with the size of the HDR?
> Thank you,
> 
> Alstan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.radiance-online.org/pipermail/radiance-general/attachments/20180816/abb5e57a/attachment.html>