Open
Conversation
Also some simplifications.
1. Make sure we use predefined values for M_LN2f if available
2. Corrected CLARG() for colors
3. The "precise" algorithm (instead of quite good the approximation) has slightly less
banding artifacts, so
- The OpenCL implementation uses precise algorithm and possibly looses some precision
in optimizing code as using the native variant.
- The CPU code chooses the algorithm depending on -ffast-math switch tested via _FAST_MATH__
4. Avoid a few double->float compiler conversions.
Testing 0012-gdnd-45 via --dump-diff-pipe shows some remaining diffs in this module, diffs are ~1e-6,
overall pixel diffs seem to be unrelated to current code.
jenshannoschwalm
added
bugfix
Collaborator
|
Looks fine to me. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fix highlights clip mode if no temperature data is provided
Got aware of this while testing some integration tests, also some simplifications.
Graduated ND improvements & maintenance
Testing 0012-gdnd-45 via --dump-diff-pipe shows some remaining diffs in this module, diffs are ~1e-6, overall pixel diffs seem to be unrelated to current code.
@ralfbrown what do you think of such switches?
Tested complete integration suite without regressions.
@TurboGit while analysing big-cpu/gpu diff counts i checked:
0028-highpass-overlay
CPU & GPU version differ by 587539 pixels
commit 8bd644d introduced a fast box-blur-like OpenCL algorithm so diff results are expected as we don't do a box blur but a gaussian weighted approx in a single run. We have a box blur in bloom - also failing and needing analysis - so i will check that and possibly provide a fix for both.
0043-dithering-fs
CPU & GPU version differ by 820153 pixels
difference is due to tea states so expected and not a problem at all, every start is with different random data. Not sure if this is a worthwhile test at all.
0045-vignetting
CPU & GPU version differ by 719921 pixels
difference is due to tea states initialized differently so GPU/CPU diffs are expected. CPU always starts with same tea state so CPU vs ref test is safe.