From 7575bbf5f73df3989c6fbc27e2d343826873de65 Mon Sep 17 00:00:00 2001 From: csmartdalton Date: Fri, 11 Oct 2024 17:04:54 +0000 Subject: [PATCH] Optimize atomic rendering for input attachments Pack clip data into an RGBA8 attachment and turn on src-over blending for both color and clip. Handle advanced blend modes by rearranging the math such that the correct color isn't reached until *AFTER* the hardware blend state is applied. This allows us to preserve clip and color contents by just emitting a=0 instead of loading the current value. It also saves flops by offloading the blending work onto the ROP blending unit, and serves as a hint to the hardware that it doesn't need to read or write anything when a == 0. Diffs= 1b5e50fce Optimize atomic rendering for input attachments (#8310) --- .rive_head | 2 +- submodules/rive-runtime | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/.rive_head b/.rive_head index d985c4e..5b664a8 100644 --- a/.rive_head +++ b/.rive_head @@ -1 +1 @@ -1cc5f2b6f6cf02d15bcb3654343e4f3322a95402 +1b5e50fcec4a38fc5056c0dff630d115f547cfb9 diff --git a/submodules/rive-runtime b/submodules/rive-runtime index fc31e5e..ee9325b 160000 --- a/submodules/rive-runtime +++ b/submodules/rive-runtime @@ -1 +1 @@ -Subproject commit fc31e5e7572d17817463f12e13e1cf39f19aff84 +Subproject commit ee9325b0bdfb967ea2b8fc174da11e59ba825c2b