Reduce expected precision of RGB10A2 filtering

The code in the YCbCr tests says that it was copied from the texture
explicit lod tests, but those tests give rgb10a2 a 'low' precision, not
'fp16'. Reduce the precision expected in the YCbCr tests.

Components: Vulkan
VK-GL-CTS Issue: 1731
Affects: dEQP-VK.ycbcr.*
         dEQP-VK.protected_memory.interaction.ycbcr.*

Change-Id: Ib0ae233fc271ded2f9987fe448dcd351e509a0c4
(cherry picked from commit 5deebf4709ed3e2c3aaad3b8bb8703f236309b97)

diff --git a/external/vulkancts/modules/vulkan/ycbcr/vktYCbCrUtil.cpp b/external/vulkancts/modules/vulkan/ycbcr/vktYCbCrUtil.cpp
index 7413376..fedeae6 100644 (file)
--- a/external/vulkancts/modules/vulkan/ycbcr/vktYCbCrUtil.cpp
+++ b/external/vulkancts/modules/vulkan/ycbcr/vktYCbCrUtil.cpp
@@ -1170,10 +1170,10 @@ tcu::FloatFormat getYCbCrFilteringPrecision (vk::VkFormat format)
                case vk::VK_FORMAT_G8_B8_R8_3PLANE_422_UNORM:
                case vk::VK_FORMAT_G8_B8R8_2PLANE_422_UNORM:
                case vk::VK_FORMAT_G8_B8_R8_3PLANE_444_UNORM:
-                       return low;
-
                case vk::VK_FORMAT_A2R10G10B10_UNORM_PACK32:
                case vk::VK_FORMAT_A2B10G10R10_UNORM_PACK32:
+                       return low;
+
                case vk::VK_FORMAT_R16G16B16_UNORM:
                case vk::VK_FORMAT_R16G16B16A16_UNORM:
                case vk::VK_FORMAT_R10X6_UNORM_PACK16: