#rb none
#jira
#ROBOMERGE-SOURCE: CL 4984075 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4984147 by ben marsh in Main branch]
#rb none
#jira UE-69957
#ROBOMERGE-SOURCE: CL 4983814 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4983818 by ben marsh in Main branch]
FLocalVertexFactory goes from 1072 bytes -> 480
#rb rolando.caloca
[FYI] daniel.wright,krzysztof.narkowicz
#ROBOMERGE-SOURCE: CL 4977718 via CL 4979513 via CL 4980026
[CL 4983397 by marcus wassmer in Main branch]
#rb none
[FYI] rolando.caloca,peter.sauerbrei
#ROBOMERGE-OWNER: marcus.wassmer
#ROBOMERGE-AUTHOR: marcus.wassmer
#ROBOMERGE-SOURCE: CL 4977685 via CL 4979506 via CL 4980019
[CL 4983374 by marcus wassmer in Main branch]
#rb ryan.gerleve
#ROBOMERGE-OWNER: brian.bekich
#ROBOMERGE-AUTHOR: brian.bekich
#ROBOMERGE-SOURCE: CL 4977682 via CL 4979505 via CL 4980018
[CL 4983195 by brian bekich in Main branch]
There can sometimes be a race condition where one of the global lists (such as the package names or paths lists) are updated due to content being mounted, while one of these background threads is using that list.
#jira
#rb Chris.Gagnon
#rnx
#ROBOMERGE-SOURCE: CL 4982820 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4982824 by jamie dale in Main branch]
Support separate MSAA and resolve targets on Metal MRT for iOS devices - this means the Forward renderer can support MSAA the same way as on macOS. It isn't optimal for iOS GPUs but it at least gets it to run without validation errors - though it doesn't yet render correctly. However now that Metal is using the RenderPassInfo descriptor we can start making changes to fix and improve it.
#rb none
#jira UERNDR-320
On iOS Metal is a bit more specific about what LoadAction::NoAction really means - it will reallocate the backing store for the target to uninitialised memory, abandoning whatever was already there. This fixes corruption in the Forward renderer when MSAA is enabled.
#rb none
#jira N/A
#ROBOMERGE-SOURCE: CL 4982542 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4982548 by mark satterthwaite in Main branch]
#rb rolando.caloca
[FYI] mark.satterthwaite, luke.thatcher,dmitriy.dyomin
#ROBOMERGE-OWNER: marcus.wassmer
#ROBOMERGE-AUTHOR: marcus.wassmer
#ROBOMERGE-SOURCE: CL 4977376 via CL 4977378 via CL 4979036
[CL 4982426 by marcus wassmer in Main branch]
#jira UE-69802
#rb simon.tovey
#ROBOMERGE-SOURCE: CL 4981865 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981866 by patrick boutot in Main branch]
#rb Marcus.Wassmer
#rnx
#jira
#ROBOMERGE-OWNER: rolando.caloca
#ROBOMERGE-AUTHOR: rolando.caloca
#ROBOMERGE-SOURCE: CL 4977338 via CL 4977347 via CL 4979031
[CL 4981812 by rolando caloca in Main branch]
#jira UE-68976
#rb none
#ROBOMERGE-SOURCE: CL 4981769 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981770 by allan bentham in Main branch]
The bug was being exibited when there was already a navmesh instance on the map and another navmesh bounds volume had been added. The whole detour navmesh got dropped, but only the area covered by the new bounds had been rebuild. I've fixed it so that as part of nav bounds update a new detour navmesh instance gets created and areas covered by _all_ bounds get dirtied.
#jira UE-66340
#rb Stephen.Holmes
#ROBOMERGE-SOURCE: CL 4981450 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981451 by mieszko zielinski in Main branch]
Added formats requested + additional missing formats. Both d3d11 and d3d12
#rb rolando.caloca
#JIRA UE-63199
#ROBOMERGE-SOURCE: CL 4981375 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981376 by jonas meyer in Main branch]
Added cvar (r.Mobile.MeshSortingMethod) to optionally enable strict front to back mesh sorting, similar to sorting in 4.19 (enabled for SunTemple).
Default is sort by state.
On GPU bound devices when scene has < 300 drawcalls it's faster to use strict front to back mesh sorting.
#jira UE-60801
#rb none
#ROBOMERGE-SOURCE: CL 4981186 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981187 by dmitriy dyomin in Main branch]
- engine changes to use 1.38.20 toolchain
#jira UE-69753 Multithreaded HTML5 loads into a black screen
#rnx
#rb none
#ROBOMERGE-SOURCE: CL 4981071 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4981072 by nick shin in Main branch]
- thirdparty libs HTML5 build scripts updates
#jira UE-69753 Multithreaded HTML5 loads into a black screen
#rnx
#rb none
#ROBOMERGE-SOURCE: CL 4980842 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4980844 by nick shin in Main branch]
This hack is motivated because the render target pool and render graph requires to much work for 4.22 to supporting arbitrary SRV creation for a FRDGTexture properly.
#rb none
[FYI] daniel.wright
#jira UE-69715
#ROBOMERGE-SOURCE: CL 4979860 in //UE4/Release-4.22/...
#ROBOMERGE-BOT: RELEASE (Release-4.22 -> Main)
[CL 4979862 by guillaume abadie in Main branch]
