Increase tolerance in mat3 determinant compute test

The determinant operation does not have strict precision requirements
and a recent change in Mesa broke this test for RADV. A slight increase
in tolerance fixes it.

Affected tests:
dEQP-VK.glsl.builtin.precision_double.determinant.compute.mat3

VK-GL-CTS issue: 3571
Components: Vulkan

Change-Id: I5185bc4d45d848cbad11798f743ae65fbb798657

diff --git a/external/vulkancts/data/vulkan/amber/builtin/precision/square_matrix/determinant_compute_fp64_mat_3x3.amber b/external/vulkancts/data/vulkan/amber/builtin/precision/square_matrix/determinant_compute_fp64_mat_3x3.amber
index 9503e36..1ed071a 100644 (file)
--- a/external/vulkancts/data/vulkan/amber/builtin/precision/square_matrix/determinant_compute_fp64_mat_3x3.amber
+++ b/external/vulkancts/data/vulkan/amber/builtin/precision/square_matrix/determinant_compute_fp64_mat_3x3.amber
@@ -106,7 +106,7 @@ EXPECT bufOut IDX  8 TOLERANCE 0.0000000000000002220446   EQ              -1
 EXPECT bufOut IDX 16 TOLERANCE 0.000000000000000222044605 EQ              -0.085873487391295
 
 # Decimals between 8 and 16: 1.776356839e-15
-EXPECT bufOut IDX 24 TOLERANCE 0.0000000000000017763568   EQ               9.5173150336492971682352
+EXPECT bufOut IDX 24 TOLERANCE 0.0000000000000017763569   EQ               9.5173150336492971682352
 
 # Decimals between 4096 and 8192: 9.094947018e-13
 EXPECT bufOut IDX 32 TOLERANCE 0.000000000001818989403546 EQ           -4572.819874970912419629875268