diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
index 96569404e..6890903a5 100644
--- a/.github/workflows/ci.yaml
+++ b/.github/workflows/ci.yaml
@@ -4,8 +4,6 @@ on:
   pull_request:
   schedule:
     - cron:  '0 3 * * *'
-env:
-  build_type: Release
 jobs:
   #clang-format:
   #  runs-on: ubuntu-latest
@@ -20,34 +18,27 @@ jobs:
     runs-on: ubuntu-latest
     env:
       dashboard_model: Experimental
+      build_type: ${{ matrix.build_type }}
       NUMBER_OF_PROCESSORS: 2
-      CXX: ${{ matrix.CXX }}
-    name: ${{ matrix.name }}
+      CXX: ${{ matrix.cxx }}
     strategy:
       fail-fast: false
       matrix:
+        build_type: [Debug, Release]
+        cxx: [g++-9, g++-10, g++-11, clang++-10, clang++-11, clang++-12, icpc]
         include:
-          - name: build-ubuntu-gcc9
-            CXX: g++-9
-          - name: build-ubuntu-gcc10
-            CXX: g++-10
-          - name: build-ubuntu-gcc11
-            CXX: g++-11
+          - cxx: g++-11
             INSTALL_EXTRA: g++-11
-          - name: build-ubuntu-clang10
-            CXX: clang++-10
-          - name: build-ubuntu-clang11
-            CXX: clang++-11
+          - cxx: clang++-11
             INSTALL_EXTRA: clang-11
-          - name: build-ubuntu-clang12
-            CXX: clang++-12
+          - cxx: clang++-12
             INSTALL_EXTRA: clang-12
-          - name: build-ubuntu-icpc
-            CXX: icpc
+          - cxx: icpc
             INSTALL_ONEAPI: true
-          #- name: build-ubuntu-icpx
-          #  CXX: icpx
-          #  INSTALL_ONEAPI: true
+        exclude:
+          # icpc in debug mode runs out of memory in CI
+          - cxx: icpc
+            build_type: Debug
     steps:
       - uses: actions/checkout@v2
         with:
@@ -77,10 +68,12 @@ jobs:
 
   build-windows:
     runs-on: ${{ matrix.os }}
-    name: build-${{ matrix.os }}
+    env:
+      build_type: ${{ matrix.build_type }}
     strategy:
       fail-fast: false
       matrix:
+        build_type: [Debug, Release]
         os: [windows-2019]
     steps:
     - uses: actions/checkout@v2
diff --git a/src/avx_sorthelper.cpp b/src/avx_sorthelper.cpp
index fba2480d0..78d2ad503 100644
--- a/src/avx_sorthelper.cpp
+++ b/src/avx_sorthelper.cpp
@@ -71,71 +71,71 @@ Vc_CONST AVX2::short_v sorted<CurrentImplementation::current()>(AVX2::short_v x_
     // sort pairs (one min/max)
     auto x = AVX::lo128(x_.data());
     auto y = AVX::hi128(x_.data());
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     auto l = _mm_min_epi16(x, y);
     auto h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
 
     // merge left & right quads (two min/max)
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(h, l);
-    Vc_DEBUG << "8x2 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "8x2 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteLo<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h));
     y = Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
 
     // merge quads into octs (three min/max)
     x = _mm_unpacklo_epi16(h, l);
     y = _mm_unpackhi_epi16(l, h);
-    Vc_DEBUG << "4x4 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "4x4 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteLo<X2, X3, X0, X1>(Mem::blend<X0, X1, Y2, Y3, X4, X5, Y6, Y7>(h, l));
     y = Mem::permuteHi<X6, X7, X4, X5>(Mem::blend<X0, X1, Y2, Y3, X4, X5, Y6, Y7>(l, h));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h));
     y = Mem::permuteLo<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h) << " done?";
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h) << " done?";
 
     // merge octs into hexa (four min/max)
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(h, l);
-    Vc_DEBUG << "2x8 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "2x8 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_unpacklo_epi64(l, h);
     y = _mm_unpackhi_epi64(l, h);
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_castps_si128(Mem::permute<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3>(_mm_castsi128_ps(h), _mm_castsi128_ps(l))));
     y = _mm_castps_si128(Mem::blend<X0, Y1, X2, Y3>(_mm_castsi128_ps(l), _mm_castsi128_ps(h)));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h);
     y = Mem::permuteLo<X1, X0, X3, X2>(
         Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l)));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epi16(x, y);
     h = _mm_max_epi16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(l, h);
     return AVX::concat(x, y);
@@ -147,71 +147,71 @@ Vc_CONST AVX2::ushort_v sorted<CurrentImplementation::current()>(AVX2::ushort_v
     // sort pairs (one min/max)
     auto x = AVX::lo128(x_.data());
     auto y = AVX::hi128(x_.data());
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     auto l = _mm_min_epu16(x, y);
     auto h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
 
     // merge left & right quads (two min/max)
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(h, l);
-    Vc_DEBUG << "8x2 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "8x2 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteLo<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h));
     y = Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
 
     // merge quads into octs (three min/max)
     x = _mm_unpacklo_epi16(h, l);
     y = _mm_unpackhi_epi16(l, h);
-    Vc_DEBUG << "4x4 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "4x4 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteLo<X2, X3, X0, X1>(Mem::blend<X0, X1, Y2, Y3, X4, X5, Y6, Y7>(h, l));
     y = Mem::permuteHi<X6, X7, X4, X5>(Mem::blend<X0, X1, Y2, Y3, X4, X5, Y6, Y7>(l, h));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h));
     y = Mem::permuteLo<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h) << " done?";
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h) << " done?";
 
     // merge octs into hexa (four min/max)
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(h, l);
-    Vc_DEBUG << "2x8 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "2x8 sorted xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_unpacklo_epi64(l, h);
     y = _mm_unpackhi_epi64(l, h);
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_castps_si128(Mem::permute<X1, X0, X3, X2>(Mem::blend<X0, Y1, X2, Y3>(_mm_castsi128_ps(h), _mm_castsi128_ps(l))));
     y = _mm_castps_si128(Mem::blend<X0, Y1, X2, Y3>(_mm_castsi128_ps(l), _mm_castsi128_ps(h)));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(l, h);
     y = Mem::permuteLo<X1, X0, X3, X2>(
         Mem::permuteHi<X5, X4, X7, X6>(Mem::blend<X0, Y1, X2, Y3, X4, Y5, X6, Y7>(h, l)));
-    Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
+    // Vc_DEBUG << "xy: " << AVX::addType<short>(x) << AVX::addType<short>(y);
     l = _mm_min_epu16(x, y);
     h = _mm_max_epu16(x, y);
-    Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
+    // Vc_DEBUG << "lh: " << AVX::addType<short>(l) << AVX::addType<short>(h);
     x = _mm_unpacklo_epi16(l, h);
     y = _mm_unpackhi_epi16(l, h);
     return AVX::concat(x, y);
diff --git a/tests/ulp.cpp b/tests/ulp.cpp
index a7fa3e436..3c1d6c89a 100644
--- a/tests/ulp.cpp
+++ b/tests/ulp.cpp
@@ -29,6 +29,13 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 TEST_TYPES(V, testUlpDiff, concat<RealVectors, RealSimdArrayList>)  //{{{1
 {
+    // MSVC takes too long in debug mode
+#if defined _MSC_VER && defined _DEBUG
+    const auto range = 1000;
+#else
+    const auto range = 10000;
+#endif
+
     typedef typename V::EntryType T;
 
     using vir::detail::ulpDiffToReference;
@@ -41,7 +48,7 @@ TEST_TYPES(V, testUlpDiff, concat<RealVectors, RealSimdArrayList>)  //{{{1
         frexp(base, &exp);
         const V eps = ldexp(V(std::numeric_limits<T>::epsilon()), exp - 1);
         //std::cout << base << ", " << exp << ", " << eps << std::endl;
-        for (int i = -10000; i <= 10000; ++i) {
+        for (int i = -range; i <= range; ++i) {
             const V i_v = V(T(i));
             const V diff = base + i_v * eps;