projects / platform / upstream / VK-GL-CTS.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: cc64df6) | patch
Avoid 32-bit precision issues in ShaderRender tests
authorMaciej Jesionowski <maciej.jesionowski@amd.com>
Wed, 4 Oct 2017 17:08:12 +0000 (19:08 +0200)
committerAlexander Galazin <Alexander.Galazin@arm.com>
Fri, 6 Oct 2017 21:24:47 +0000 (17:24 -0400)
commite68883ec831b98c5fb4eddf6239e0eee9a376d2f
treea03c3bb1c8de862a595c67283e8d2ac84205e665tree | snapshot
parentcc64df67764cfb4cad8a2ac622c809303d3ca5a8commit | diff
Avoid 32-bit precision issues in ShaderRender tests

These tests are very sensitive to floating-point precision. The default
GRID_SIZE of 90 units is problematic in 32-bit mode. The error is
visible only in Release builds and affects CPU-generated reference
image.

Affects:
* dEQP-VK.glsl.texture_functions.*
* dEQP-VK.glsl.operator.*

VK-GL-CTS issue: 751
Components: Vulkan

Change-Id: I0ca501e180339cd0e1430cab3eb08a14f68f9802
external/vulkancts/modules/vulkan/shaderrender/vktShaderRenderOperatorTests.cpp diff | blob | history
external/vulkancts/modules/vulkan/shaderrender/vktShaderRenderTextureFunctionTests.cpp diff | blob | history
Domain: Graphics System / GL;