Accelerates the five-phase method using Accelerad

[sida du]

Hello everyone,

I’ve used the five-phase method to generate an annual HDR image of a room and compared the results using Radiance with Accelerad. I found that in my results, the solar coefficients of the image calculated using Accelerad_rcontrib in the calculation of solar direct radiance simulation were all black. However, the Radiance was different.

The Radiance code used is " oconv -f materials_B.rad room_B.rad skies/suns_3.rad> octrees/sunCoefficients_3.oct; time vwrays -vf views/south.vf -x 800 -y 800 -pj 0.7 -ff | rcontrib -w- -i -ab 1 -ad 256 -lw 1.0e-3 -dc 1 -dt 0 -dj 0 -ffc -n 16 `vwrays -vf views/south.vf -x 800 -y 800 -d` -o matrices/cds/hdrIllSpace/southM6%04d.hdr -e MF:3 -f reinhart.cal -b  rbin -bn Nrbins -m solar octrees/sunCoefficients_3.oct “. (there was no BSDF file was entered in the octree file)

And Accelerad code is "vwrays-vf views/ south.vf-x 800-y 800-pj 0.7-ff > south.rays;

accelerad_rcontrib-x 800-y 800-w - - i-ab 1-ad 256-lw 1.0e-3-dc 1-dt 0-dj 0-ffc-n 16-o matrices/cds/hdrIllSpace/southM6%04d.hdr -e MF:3 -f reinhart.cal -b rbin -bn Nrbins -m solar octrees/sunCoefficients_3.oct < south.rays“

One of the results of the current calculation is as follows: the left is the radiance calculation and the right is the Accelerad calculation

I don't know why this has happened, and what will be done to solve this problem?

Any help would be appreciated,

Cheers,

Sida

Radiance

Accelerad

[Nathaniel Jones]

Hi Sida,

The spacing in your Accelerad command is not correct. I don't know if that's a result of the way you pasted it here or not, but if you literally put that command into Accelerad, it would not be understood.

I also noticed that you have -w- in your command to turn off warning messages, so if Accelerad did encounter an error, it would not tell you about it. You should try again without -w-.

Other than that, there is no obvious problem in the commands you provided.

Nathaniel

[sida du]

Hey Nathaniel,

As you said the command is correct, the missing Spacing are due to replication issues. This is the result of running the command after removing -w-.

As you can see, there are no errors reported, but there is still the problem of not calculating the correct value.

Cheers,

Sida

[Nathaniel Jones]

Hi Sida,

Okay, so there are no warnings being printed, so this suggests there may be some other issue with the input being provided. Have you tried running a basic rendering with Accelerad rpict or AcceleradRT to verify that Accelerad understands the geometry correctly? There are some other items listed in the documentation that could also lead to the issue you are seeing.

Nathaniel

[sida du]

Hey Nathaniel,

Thank you very much for your advice. I will carefully check the accuracy of the geometry again to see if there is any problem.

Cheers,

Sida

[sida du]

Hey Nathaniel,
When I used 5-PM to calculate HDR images, I found that in the third part, the Direct Sun Coefficients used to calculate HDR images did not appear light leakage between walls, but this phenomenon appeared in the Accelerad acceleration calculation. 

Then I tried to solve this problem. No success, so I would like to ask what causes this phenomenon, and then what can be solved.

CPU

Accelerad

Cheers,

Sida

[Nathaniel Jones]

Hi Sida,

This is likely because Accelerad uses single-precision arithmetic for ray calculations. A simple solution is to give thickness to your walls by modeling the exterior surface as well.

Nathaniel

[sida du]

Hey Nathaniel,

Thank you very much for your answer.

I'll give it a try.

Cheers,

Sida

[sida du]

Hey Nathaniel,

After taking your suggestions, the problem has been solved. I'm very interested and curious about what causes this phenomenon, so I want to ask, why does the single-precision algorithm used by Accelerad cause this phenomenon?
Thanks again for your reply!

Cheers,

Sida

[Nathaniel Jones]

Hi Sida,

It's because single-precision math solves the ray intersection position with less accuracy than double-precision math, so the probability of a miss, though still small, becomes several orders of magnitude higher at polygon edges.

Nathaniel

[sida du]

Hey Nathaniel,

Thank you so much for your detailed response to my question！I have many things need to study, i hope i can continue in this field.

Best regards,

Sida

[sida du]

Hey Nathaniel,

I also have a question, that is, why is the single precision method used instead of the double precision method, is to consider the problem of computational efficiency?

Cheers,

Sida

[Nathaniel Jones]

Hi Sida,

Yes, single-precision math is more computationally efficient, and this is how the GPU ray tracing library used by Accelerad works.

Nathaniel

[sida du]

Hey Nathaniel,

Ok, Thank you, It's a question of precision versus time.

Cheer,

Sida

[Nathaniel Jones]

Hi Sida,

No, it is a question of the computational resources that are available on the GPU.

Nathaniel