From ab93ea1f2230fcfaece83bb1782d4000ad915242 Mon Sep 17 00:00:00 2001
From: Patrick Walton <pcwalton@mimiga.net>
Date: Tue, 29 Jan 2019 11:27:25 -0800
Subject: [PATCH] Fix bogus thin culling that was causing artefacts

---
 demo3/shaders/fill.vs.glsl |  5 +----
 geometry/src/simd.rs       | 40 --------------------------------------
 renderer/src/gpu_data.rs   | 14 ++++++-------
 3 files changed, 7 insertions(+), 52 deletions(-)

diff --git a/demo3/shaders/fill.vs.glsl b/demo3/shaders/fill.vs.glsl
index b841116d..dbe46795 100644
--- a/demo3/shaders/fill.vs.glsl
+++ b/demo3/shaders/fill.vs.glsl
@@ -50,8 +50,5 @@ void main() {
     vFrom = from - position;
     vTo = to - position;
 
-    if (!(abs(from.x - to.x) > 0.1) || !(abs(uTileSize.y - min(from.y, to.y)) > 0.1))
-        gl_Position = vec4(0.0);
-    else
-        gl_Position = vec4((tileOrigin + position) / uFramebufferSize * 2.0 - 1.0, 0.0, 1.0);
+    gl_Position = vec4((tileOrigin + position) / uFramebufferSize * 2.0 - 1.0, 0.0, 1.0);
 }
diff --git a/geometry/src/simd.rs b/geometry/src/simd.rs
index f98c73d9..5895609f 100644
--- a/geometry/src/simd.rs
+++ b/geometry/src/simd.rs
@@ -645,46 +645,6 @@ mod x86 {
         }
     }
 
-    /*
-    // Two pairs of 32-bit floats
-
-    #[derive(Clone, Copy)]
-    pub struct F64x2x2(pub __m128d);
-
-    impl F64x2 {
-        // Shuffles
-
-        #[inline]
-        pub fn interleave(self, other: F64x2) -> (F64x2, F64x2) {
-            unsafe {
-                (
-                    F64x2(x86_64::_mm_unpacklo_pd(self.0, other.0)),
-                    F64x2(x86_64::_mm_unpackhi_pd(self.0, other.0)),
-                )
-            }
-        }
-
-        // Creates `<self[0], self[1], other[2], other[3]>`.
-        #[inline]
-        pub fn combine_low_high(self, other: F64x2) -> F64x2 {
-            unsafe {
-                F64x2(x86_64::_mm_shuffle_pd(self.0, other.0, 0b10))
-            }
-        }
-
-        // Casts these packed floats to 32-bit floats.
-        //
-        // NB: This is a pure bitcast and does no actual conversion; only use this if you know what
-        // you're doing.
-        #[inline]
-        pub fn as_f32x4(self) -> F32x4 {
-            unsafe {
-                F32x4(x86_64::_mm_castpd_ps(self.0))
-            }
-        }
-    }
-    */
-
     // 32-bit signed integers
 
     #[derive(Clone, Copy)]
diff --git a/renderer/src/gpu_data.rs b/renderer/src/gpu_data.rs
index 01e92119..23cb34fa 100644
--- a/renderer/src/gpu_data.rs
+++ b/renderer/src/gpu_data.rs
@@ -136,19 +136,17 @@ impl BuiltObject {
             .shuffle(shuffle_mask.as_u8x16())
             .as_i32x4();
 
+        // Unpack whole and fractional pixels.
         let px = LineSegmentU4((segment[1] | (segment[1] >> 12)) as u16);
         let subpx = LineSegmentU8(segment[0] as u32);
 
-        let tile_index = self.tile_coords_to_index(tile_x, tile_y);
-
-        /*
-        // TODO(pcwalton): Cull degenerate fills again.
         // Cull degenerate fills.
-        let (from_px, to_px) = (from.to_u8(), to.to_u8());
-        if from_px.x == to_px.x && from_subpx.x == to_subpx.x {
-            return
+        if (px.0 & 0xf) as u8 == ((px.0 >> 8) & 0xf) as u8 &&
+                (subpx.0 & 0xff) as u8 == ((subpx.0 >> 16) & 0xff) as u8 {
+            return;
         }
-        */
+
+        let tile_index = self.tile_coords_to_index(tile_x, tile_y);
 
         self.fills.push(FillObjectPrimitive {
             px,