dsview & ds input[DO NOT MERGE] #15322
Conversation
* fix paste component value bug * fix node.active undo bug
… false (cocos#15232) * feat: inspector/fbx/mountAllAnimationsOnPrefab default value false * feat: use getPropValue get value
* add reset function to call when start play * remove clear * reset when stop emitting * rename param & reuse position * call reset in stop
…) (cocos#15294) Co-authored-by: cocos-robot <cocos-robot@cocos.com>
* Marionette: adjust additive animation rule * Rename base -> ref * Importer UI
Interface Check ReportThis pull request does not change any public interfaces ! |
| auto resIter = std::find_if(attachments.begin(), attachments.end(), [resID](const AccessStatus& status) { | ||
| return status.vertID == resID; | ||
| }); | ||
|
|
||
| auto slotName = rasterView.slotName; | ||
| if (rasterView.accessType == AccessType::READ || rasterView.accessType == AccessType::READ_WRITE) { | ||
| slotName.insert(0, "__in"); |
There was a problem hiding this comment.
I think __in prefix is not needed?
There was a problem hiding this comment.
In shdc we process input attachment eg c0 in #pragma subpassColor c0 into __inc0.
Same time user coding like: pass.addRasterView(user_defined_c0_name, "c0", ...) according to effect, so it's still c0 in rasterView.slotName, which does not match what it is in layoutGraph.attributeIndex.
Not sure if name processing is neccesarry.
There was a problem hiding this comment.
I think in c++ and effect, __in is not used. __inc0 is an internal variable and will only be used by subpassLoad.
There was a problem hiding this comment.
In Vulkan backend subpass input attachment will be bind in descriptor as INPUT_ATTACHMENT, this is the only usage AFAIK.
There was a problem hiding this comment.
In that case, I think the name is not needed?
There was a problem hiding this comment.
just check the code, that's true, will remove later.
Re: #
Changelog
Continuous Integration
This pull request:
Compatibility Check
This pull request: