From 2221cfc09d4d447137fefdd359dfbcb273fac646 Mon Sep 17 00:00:00 2001
From: Ryan McConnell <rammcconnell@gmail.com>
Date: Thu, 30 Jul 2026 12:15:51 -0400
Subject: [PATCH 1/2] avx2-patch

---
 src/pixie/simd/avx2.nim | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/pixie/simd/avx2.nim b/src/pixie/simd/avx2.nim
index 1800e583..b2405569 100644
--- a/src/pixie/simd/avx2.nim
+++ b/src/pixie/simd/avx2.nim
@@ -695,7 +695,7 @@ proc blendLineCoverageMaskAvx2*(
       15, -1, 15, -1, 11, -1, 11, -1, 7, -1, 7, -1, 3, -1, 3, -1,
       15, -1, 15, -1, 11, -1, 11, -1, 7, -1, 7, -1, 3, -1, 3, -1
     )
-  while i < len - 16:
+  while i < len - 32:
     let
       coverage = mm256_loadu_si256(coverages[i].addr)
       eqZero = mm256_cmpeq_epi8(coverage, vecZero)

From 8f06f1b4a817d008584b71942fbdbd0a2ec7dd56 Mon Sep 17 00:00:00 2001
From: treeform <starplant@gmail.com>
Date: Sun, 13 Sep 2026 09:11:14 -0700
Subject: [PATCH 2/2] Keep full AVX2 mask blocks vectorized and add coverage
 tests

---
 src/pixie/simd/avx2.nim |  2 +-
 tests/test_simd.nim     | 51 +++++++++++++++++++++++++++++++++++++++++
 tests/tests.nim         |  1 +
 3 files changed, 53 insertions(+), 1 deletion(-)
 create mode 100644 tests/test_simd.nim

diff --git a/src/pixie/simd/avx2.nim b/src/pixie/simd/avx2.nim
index b2405569..d467c32c 100644
--- a/src/pixie/simd/avx2.nim
+++ b/src/pixie/simd/avx2.nim
@@ -695,7 +695,7 @@ proc blendLineCoverageMaskAvx2*(
       15, -1, 15, -1, 11, -1, 11, -1, 7, -1, 7, -1, 3, -1, 3, -1,
       15, -1, 15, -1, 11, -1, 11, -1, 7, -1, 7, -1, 3, -1, 3, -1
     )
-  while i < len - 32:
+  while i <= len - 32:
     let
       coverage = mm256_loadu_si256(coverages[i].addr)
       eqZero = mm256_cmpeq_epi8(coverage, vecZero)
diff --git a/tests/test_simd.nim b/tests/test_simd.nim
new file mode 100644
index 00000000..5b4535a4
--- /dev/null
+++ b/tests/test_simd.nim
@@ -0,0 +1,51 @@
+import pixie/simd
+
+when allowSimd and defined(amd64) and not defined(pixieNoAvx):
+  import chroma, pixie/blends, pixie/common
+
+  if cpuHasAvx2:
+    let
+      original = rgbx(120, 80, 40, 255)
+      mask = rgbx(255, 255, 255, 255)
+    var cases: int
+    for length in 0 .. 160:
+      for shift in 0 .. 7:
+        for pattern in 0 .. 3:
+          var pixels = newSeq[ColorRGBX](length + 96)
+          for pixel in pixels.mitems:
+            pixel = original
+
+          # Exercise every pixel alignment, with sentinels on both sides.
+          var start: int
+          while (cast[uint](pixels[start].addr) and 31) != 0:
+            inc start
+          start += 16 + shift
+
+          # Padding makes the old overread safe while its extra stores still
+          # corrupt the sentinels, independently of the allocator's layout.
+          var coverages = newSeq[uint8](length + 32)
+          for i in 0 ..< length:
+            coverages[i] = case pattern
+              of 0: 0'u8
+              of 1: 255'u8
+              of 2: 128'u8
+              else: ((i * 73 + 19) mod 256).uint8
+
+          blendLineCoverageMaskAvx2(
+            cast[ptr UncheckedArray[ColorRGBX]](pixels[start].addr),
+            cast[ptr UncheckedArray[uint8]](coverages[0].addr),
+            mask,
+            length
+          )
+
+          for i, pixel in pixels:
+            let expected =
+              if i >= start and i < start + length:
+                blendMask(original, mask * coverages[i - start])
+              else:
+                original
+            doAssert pixel == expected,
+              "length=" & $length & " shift=" & $shift &
+              " pattern=" & $pattern & " pixel=" & $(i - start)
+          inc cases
+    echo "AVX2 coverage mask: ", cases, " cases passed"
diff --git a/tests/tests.nim b/tests/tests.nim
index 002bfeeb..ef6bc496 100644
--- a/tests/tests.nim
+++ b/tests/tests.nim
@@ -12,6 +12,7 @@ import
   test_png,
   test_ppm,
   test_qoi,
+  test_simd,
   test_webp,
   test_svg,
   xrays