From b6cf9b73482e23e0d528396dfa367a8fceb290f1 Mon Sep 17 00:00:00 2001 From: Emil Velikov Date: Wed, 13 Mar 2019 17:47:20 +0000 Subject: [PATCH] cherry-ignore: add explicit 19.0 performance optimisations nir: lower IO to vector intel: make use of the optimisation They address a 10% regression in a synthetic benchmark. No real world use-cases seems to be affected. Signed-off-by: Emil Velikov --- bin/.cherry-ignore | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/bin/.cherry-ignore b/bin/.cherry-ignore index 47e6f1453bc..1041997a21e 100644 --- a/bin/.cherry-ignore +++ b/bin/.cherry-ignore @@ -46,3 +46,9 @@ cbd1ad6165f0aea7fb7c6fd1b36ad5317dd65cb7 st/mesa: require RGBA2, RGB4, and RGBA4 # revert The commit was reverted shortly after it was applied on master 47fc359822494935852de1e70e4d840b2fe6a25c anv: release memory allocated by glsl types during spirv_to_nir + +# stable Explicitly 19.0 nomination, seemingly a performance patch +5ef2b8f1f2ebcdb4ffe5c98b3f4f48e584cb4b22 nir: Add a pass for lowering IO back to vector when possible + +# fixes Temporary block perf. fix depending on previous opt. pass +6d5d89d25a0a4299dbfcbfeca71b6c7e65ef3d45 intel/nir: Vectorize all IO