-
Notifications
You must be signed in to change notification settings - Fork 115
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
T>hybrid color 2 does not work in openCL #685
Comments
same with T>DIFS auxColor
|
What I see that it works properly when checkForBailoutCondition is disabled for t>hybridColoring |
It is weird, for T>DIFS auxColor disabling bailout check produces color in
full mode.
T>HybridColor2 disabling bailout does not fix it
Maybe two different issues
HOWEVER I need to revisit my c++ code (it looks wrong), so wait until i
advise that I am happy with that code, before you look at the c+++ versus
openCL issues :)
…On Wed, Nov 13, 2019 at 8:08 AM Krzysztof Marczak ***@***.***> wrote:
What I see that it works properly when checkForBailoutCondition is
disabled for t>hybridColoring
But I don't know why yet.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#685?email_source=notifications&email_token=ADBNSHG23MYUZCBY6MIOQD3QTL5LZA5CNFSM4IDMMPF2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOED3SH7I#issuecomment-553067517>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADBNSHA7DSF6ZOWWZ254QMTQTL5LZANCNFSM4IDMMPFQ>
.
|
OK , I have checked the c++ code, so the c++ versus openCL issues can be
re-looked at.
The effect of color iter = 4 x max iter can be noticed with difs color
under certain settings. So i suggest it becomes a user defined parameter,
My preference is default 1.0, but it could be 4 or the historic 10
…On Wed, Nov 13, 2019 at 5:47 PM carbuncle grim ***@***.***> wrote:
It is weird, for T>DIFS auxColor disabling bailout check produces color
in full mode.
T>HybridColor2 disabling bailout does not fix it
Maybe two different issues
HOWEVER I need to revisit my c++ code (it looks wrong), so wait until
i advise that I am happy with that code, before you look at the c+++ versus
openCL issues :)
On Wed, Nov 13, 2019 at 8:08 AM Krzysztof Marczak <
***@***.***> wrote:
> What I see that it works properly when checkForBailoutCondition is
> disabled for t>hybridColoring
> But I don't know why yet.
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <#685?email_source=notifications&email_token=ADBNSHG23MYUZCBY6MIOQD3QTL5LZA5CNFSM4IDMMPF2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOED3SH7I#issuecomment-553067517>,
> or unsubscribe
> <https://github.com/notifications/unsubscribe-auth/ADBNSHA7DSF6ZOWWZ254QMTQTL5LZANCNFSM4IDMMPFQ>
> .
>
|
There may be more than one issue. Here i am just using color by numbers. But there seems to be a difference between iteration based color. I will recheck a few things tomorrow.
|
and here is openCL c++ diference with aux.color
|
I will start again an investigation with first settings which are in this issue. I need to find why it doesn't work in OpenCL mode when check for bailout condition is enabled on second formula tab. Last example is too complicated to analyze this effect. |
Hybryd Color formula fails because it is considered as a death computation (result before and after iteration is the same). Now I'm wondering why it works in CPU mode. I'm close to fix this problem. |
Check if it works now |
a) all settings files in #685 now work, including those that did not use any T>Hybrid color2 or T>DIFS Hybrid Color. Thanks, i would never have found that :) b) I will think about possible backwards compatibility . Materials Editor - Pre - 2.15 orbit trap color checkbox can be moved into a Pre -2.20 Compatibility checkgroupbox c) I will remove code : d) I still have some problems with some options in T>Hybrid color2 with openCL. I will relook at this, but it does not effect the release of v2.20 |
Everything appears to be working as expected. TODO is general backwards compatibility issue related to bailout fix. I will look at that tomorrow |
a) Great! |
f) this is a new test for aux.color, hybrid mode - extra hybrid mode controls - only aux.color active. Latest build shows some minor differenec with openCL
|
g) further testing produced major difference in pseudo Kleinian openCL. The good thing is the bailout fix has removed the ugly color cuts that we use to get with pKlien. Left image is openCL before the fix. If I increase color speed on the latest build image, color stays perfect, no cuts.
|
So we are not taking any further actions in this topic. Can we close this issue? |
Original issue is fixed, T>hybrid color2 working as expected, (and T>DIFS hybrid color) However my latest test for general backwards compatibility issues related to the fix in bailout code, appears to show a major difference in color with pseudoKleinian in hybrid mode. Please check the latest settings above in openCL mode for reproducibility. we can : |
The issue title is now confusing. Made new issue #742 |
The text was updated successfully, but these errors were encountered: