tree ed3376c4c9985c9fe04605516658e52d39069556
parent 488cb63b5b4d39860a9a264b997d9209c2f1c633
author Alexis Hetu <sugoi@google.com> 1657038029 -0400
committer swiftshader-scoped@luci-project-accounts.iam.gserviceaccount.com <swiftshader-scoped@luci-project-accounts.iam.gserviceaccount.com> 1657120763 +0000

Always perform depth test using floating point values

This CL removes a lot of duplicated code previously shared between
the 16-bit depth test and the 32-bit depth test. Instead of having
a separate path for 16-bit depth comparisons, all comparisons now
use floating point, which will make it more maintainable and easier
to convert to larger SIMD widths.

While the 16 bit depth comparison is now done using floating point
values, it is performed in the [0.0f, 0xFFFF] range for simplicity.

One of the issues was the use of convertFixed16() on depth values,
because depth value need to be rounded, whereas color values do not
require rounding when calling this function.

The increased accuracy also fixes 2 previously failing Vulkan tests.

Tests: dEQP-VK.fragment_operations.early_fragment.*
Bug: b/204919030
Bug: b/200826740
Change-Id: I87aefbd9aed36d5e0acbe4daebc213eaf3f77f9c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/66888
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Commit-Queue: Alexis Hétu <sugoi@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Sean Risser <srisser@google.com>