Update to clang-format 12

.clang-format

@@ -5,18 +5,18 @@ DisableFormat:   false
 AccessModifierOffset: -4
 AlignAfterOpenBracket: AlwaysBreak
 AlignConsecutiveAssignments: false
-#AlignConsecutiveBitFields: false
+AlignConsecutiveBitFields: false
 AlignConsecutiveDeclarations: false
 AlignConsecutiveMacros: false
 AlignEscapedNewlines: Right
-AlignOperands: false #DontAlign
+AlignOperands: DontAlign
 AlignTrailingComments: false
 AllowAllArgumentsOnNextLine: false
 AllowAllConstructorInitializersOnNextLine: false
 AllowAllParametersOfDeclarationOnNextLine: false
 AllowShortBlocksOnASingleLine: Never
 AllowShortCaseLabelsOnASingleLine: false
-#AllowShortEnumsOnASingleLine: false
+AllowShortEnumsOnASingleLine: false
 AllowShortFunctionsOnASingleLine: None
 AllowShortIfStatementsOnASingleLine: Never
 AllowShortLambdasOnASingleLine: All
@@ -26,7 +26,7 @@ AlwaysBreakBeforeMultilineStrings: false
 AlwaysBreakTemplateDeclarations: Yes
 BinPackArguments: false
 BinPackParameters: false
-#BitFieldColonSpacing: Both
+BitFieldColonSpacing: Both
 BreakBeforeBinaryOperators: All
 BreakBeforeBraces: Allman
 BreakBeforeTernaryOperators: true
@@ -48,9 +48,9 @@ IncludeBlocks:   Regroup
 IncludeCategories:
 IncludeIsMainRegex: '(Test)?$'
 IncludeIsMainSourceRegex: ''
-#IndentCaseBlocks: true
+IndentCaseBlocks: true
 IndentCaseLabels: false
-#IndentExternBlock: AfterExternBlock
+IndentExternBlock: AfterExternBlock
 IndentGotoLabels: true
 IndentPPDirectives: AfterHash
 IndentWidth: 4
@@ -60,7 +60,6 @@ MacroBlockBegin: ''
 MacroBlockEnd:   ''
 MaxEmptyLinesToKeep: 2
 NamespaceIndentation: All
-#OperandAlignmentStyle: Align
 PenaltyBreakAssignment: 2
 PenaltyBreakBeforeFirstCallParameter: 19
 PenaltyBreakComment: 300

.github/workflows/ci.yaml

@@ -10,10 +10,10 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-    - uses: DoozyX/clang-format-lint-action@v0.11
+    - uses: DoozyX/clang-format-lint-action@v0.12
       with:
         exclude: './thirdparty'
-        clangFormatVersion: 11
+        clangFormatVersion: 12
 
   amalgamation:
     runs-on: ubuntu-latest

examples/alpaka/asyncblur/asyncblur.cpp

@@ -79,7 +79,8 @@ struct BlurKernel
     {
         const auto ti = alpaka::getIdx<alpaka::Grid, alpaka::Threads>(acc);
 
-        [[maybe_unused]] auto sharedView = [&] {
+        [[maybe_unused]] auto sharedView = [&]
+        {
             if constexpr (SHARED)
             {
                 // Using SoA for the shared memory

examples/alpaka/nbody/nbody.cpp

@@ -149,13 +149,15 @@ struct UpdateKernel
     template <typename Acc, typename View>
     LLAMA_FN_HOST_ACC_INLINE void operator()(const Acc& acc, View particles) const
     {
-        auto sharedView = [&] {
+        auto sharedView = [&]
+        {
             // if there is only 1 thread per block, use stack instead of shared memory
             if constexpr (BlockSize == 1)
                 return llama::allocViewStack<View::ArrayDomain::rank, typename View::DatumDomain>();
             else
             {
-                constexpr auto sharedMapping = [] {
+                constexpr auto sharedMapping = []
+                {
                     constexpr auto arrayDomain = llama::ArrayDomain{BlockSize};
                     if constexpr (MappingSM == AoS)
                         return llama::mapping::AoS{arrayDomain, Particle{}};
@@ -176,7 +178,8 @@ struct UpdateKernel
         const auto tbi = alpaka::getIdx<alpaka::Block, alpaka::Threads>(acc)[0];
 
         // TODO: we could optimize here, because only velocity is ever updated
-        auto pi = [&] {
+        auto pi = [&]
+        {
             constexpr auto arrayDomain = llama::ArrayDomain{Elems};
             constexpr auto mapping
                 = llama::mapping::SoA<typename View::ArrayDomain, typename View::DatumDomain, false>{arrayDomain};
@@ -242,7 +245,8 @@ void run(std::ostream& plotFile)
     using PltfAcc = alpaka::Pltf<DevAcc>;
     using Queue = alpaka::Queue<DevAcc, alpaka::Blocking>;
 
-    auto mappingName = [](int m) -> std::string {
+    auto mappingName = [](int m) -> std::string
+    {
         if (m == 0)
             return "AoS";
         if (m == 1)
@@ -258,7 +262,8 @@ void run(std::ostream& plotFile)
     const DevHost devHost(alpaka::getDevByIdx<PltfHost>(0u));
     Queue queue(devAcc);
 
-    auto mapping = [] {
+    auto mapping = []
+    {
         const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
         if constexpr (MappingGM == AoS)
             return llama::mapping::AoS{arrayDomain, Particle{}};

examples/alpaka/vectoradd/vectoradd.cpp

@@ -82,7 +82,8 @@ try
     // LLAMA
     const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
 
-    const auto mapping = [&] {
+    const auto mapping = [&]
+    {
         if constexpr (MAPPING == 0)
             return llama::mapping::AoS{arrayDomain, Vector{}};
         if constexpr (MAPPING == 1)

examples/bufferguard/bufferguard.cpp

@@ -228,7 +228,8 @@ void run(const std::string& mappingName)
     std::cout << "\nView 2:\n";
     printView(view2, rows, cols);
 
-    auto copyBlobs = [&](auto& srcView, auto& dstView, auto srcBlobs, auto dstBlobs) {
+    auto copyBlobs = [&](auto& srcView, auto& dstView, auto srcBlobs, auto dstBlobs)
+    {
         static_assert(srcBlobs.size() == dstBlobs.size());
         for (auto i = 0; i < srcBlobs.size(); i++)
         {

examples/heatequation/heatequation.cpp

@@ -113,7 +113,8 @@ try
     auto uNext = llama::allocView(mapping);
     auto uCurr = llama::allocView(mapping);
 
-    auto run = [&](std::string_view updateName, auto update) {
+    auto run = [&](std::string_view updateName, auto update)
+    {
         // init
         for (uint32_t i = 0; i < extent; i++)
             uCurr[i] = exactSolution(i * dx, 0.0);

examples/nbody/nbody.cpp

@@ -109,7 +109,8 @@ namespace usellama
     template <int Mapping, bool UseAccumulator, std::size_t AoSoALanes = 8 /*AVX2*/>
     auto main(std::ostream& plotFile) -> int
     {
-        auto mappingName = [](int m) -> std::string {
+        auto mappingName = [](int m) -> std::string
+        {
             if (m == 0)
                 return "AoS";
             if (m == 1)
@@ -127,7 +128,8 @@ namespace usellama
             title += " Acc";
         std::cout << title << "\n";
         Stopwatch watch;
-        auto mapping = [&] {
+        auto mapping = [&]
+        {
             const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
             if constexpr (Mapping == 0)
                 return llama::mapping::AoS{arrayDomain, Particle{}};
@@ -149,14 +151,16 @@ namespace usellama
         if constexpr (DUMP_MAPPING)
             std::ofstream(title + ".svg") << llama::toSvg(mapping);
 
-        auto tmapping = [&] {
+        auto tmapping = [&]
+        {
             if constexpr (TRACE)
                 return llama::mapping::Trace{std::move(mapping)};
             else
                 return std::move(mapping);
         }();
 
-        auto hmapping = [&] {
+        auto hmapping = [&]
+        {
             if constexpr (HEATMAP)
                 return llama::mapping::Heatmap{std::move(tmapping)};
             else
@@ -730,7 +734,8 @@ namespace manualAoSoA_manualAVX
         const __m256 distSqr
             = _mm256_add_ps(_mm256_add_ps(_mm256_add_ps(vEPS2, xdistanceSqr), ydistanceSqr), zdistanceSqr);
         const __m256 distSixth = _mm256_mul_ps(_mm256_mul_ps(distSqr, distSqr), distSqr);
-        const __m256 invDistCube = [&] {
+        const __m256 invDistCube = [&]
+        {
             if constexpr (ALLOW_RSQRT)
             {
                 const __m256 r = _mm256_rsqrt_ps(distSixth);
@@ -1019,7 +1024,8 @@ namespace manualAoSoA_Vc
         const vec zdistanceSqr = zdistance * zdistance;
         const vec distSqr = EPS2 + xdistanceSqr + ydistanceSqr + zdistanceSqr;
         const vec distSixth = distSqr * distSqr * distSqr;
-        const vec invDistCube = [&] {
+        const vec invDistCube = [&]
+        {
             if constexpr (ALLOW_RSQRT)
             {
                 const vec r = Vc::rsqrt(distSixth);
@@ -1383,25 +1389,33 @@ try
 
     int r = 0;
     using namespace boost::mp11;
-    mp_for_each<mp_iota_c<5>>([&](auto i) {
-        // only AoSoA (3) needs lanes
-        using Lanes
-            = std::conditional_t<decltype(i)::value == 3, mp_list_c<std::size_t, 8, 16>, mp_list_c<std::size_t, 0>>;
-        mp_for_each<Lanes>([&, i](auto lanes) {
-            mp_for_each<mp_list_c<bool, false, true>>([&, i](auto useAccumulator) {
-                r += usellama::main<decltype(i)::value, decltype(useAccumulator)::value, decltype(lanes)::value>(
-                    plotFile);
-            });
+    mp_for_each<mp_iota_c<5>>(
+        [&](auto i)
+        {
+            // only AoSoA (3) needs lanes
+            using Lanes
+                = std::conditional_t<decltype(i)::value == 3, mp_list_c<std::size_t, 8, 16>, mp_list_c<std::size_t, 0>>;
+            mp_for_each<Lanes>(
+                [&, i](auto lanes)
+                {
+                    mp_for_each<mp_list_c<bool, false, true>>(
+                        [&, i](auto useAccumulator) {
+                            r += usellama::
+                                main<decltype(i)::value, decltype(useAccumulator)::value, decltype(lanes)::value>(
+                                    plotFile);
+                        });
+                });
         });
-    });
     r += manualAoS::main<false>(plotFile);
     r += manualAoS::main<true>(plotFile);
     r += manualSoA::main<false>(plotFile);
     r += manualSoA::main<true>(plotFile);
-    mp_for_each<mp_list_c<std::size_t, 8, 16>>([&](auto lanes) {
-        r += manualAoSoA::main<false, false, decltype(lanes)::value>(plotFile);
-        r += manualAoSoA::main<true, false, decltype(lanes)::value>(plotFile);
-    });
+    mp_for_each<mp_list_c<std::size_t, 8, 16>>(
+        [&](auto lanes)
+        {
+            r += manualAoSoA::main<false, false, decltype(lanes)::value>(plotFile);
+            r += manualAoSoA::main<true, false, decltype(lanes)::value>(plotFile);
+        });
     // r += manualAoSoA::main<false, true>(plotFile);
     // r += manualAoSoA::main<true, true>(plotFile);
 #ifdef __AVX2__
@@ -1417,9 +1431,14 @@ try
             {
                 if (useUpdate1 && tiled)
                     continue;
-                mp_for_each<mp_list_c<bool, false, true>>([&](auto useAccumulator) {
-                    r += manualAoSoA_Vc::main<decltype(useAccumulator)::value>(plotFile, threads, useUpdate1, tiled);
-                });
+                mp_for_each<mp_list_c<bool, false, true>>(
+                    [&](auto useAccumulator) {
+                        r += manualAoSoA_Vc::main<decltype(useAccumulator)::value>(
+                            plotFile,
+                            threads,
+                            useUpdate1,
+                            tiled);
+                    });
             }
 #endif
 

examples/nbody_benchmark/nbody.cpp

@@ -83,7 +83,8 @@ void run(std::ostream& plotFile)
 
     constexpr FP ts = 0.0001f;
 
-    auto mapping = [&] {
+    auto mapping = [&]
+    {
         const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
         if constexpr (Mapping == 0)
             return llama::mapping::AoS{arrayDomain, Particle{}};
@@ -155,13 +156,17 @@ try
     plotFile.exceptions(std::ios::badbit | std::ios::failbit);
     plotFile << "\"alignment\"\t\"AoS\"\t\"SoA\"\t\"SoA MB\"\n";
 
-    mp_for_each<mp_iota_c<28>>([&](auto ae) {
-        mp_for_each<mp_list_c<std::size_t, 0, 1, 2>>([&](auto m) {
-            constexpr auto mapping = decltype(m)::value;
-            constexpr auto alignment = std::size_t{1} << decltype(ae)::value;
-            run<mapping, alignment>(plotFile);
+    mp_for_each<mp_iota_c<28>>(
+        [&](auto ae)
+        {
+            mp_for_each<mp_list_c<std::size_t, 0, 1, 2>>(
+                [&](auto m)
+                {
+                    constexpr auto mapping = decltype(m)::value;
+                    constexpr auto alignment = std::size_t{1} << decltype(ae)::value;
+                    run<mapping, alignment>(plotFile);
+                });
         });
-    });
 
     std::cout << "Plot with: ./nbody.sh\n";
     std::ofstream{"nbody.sh"} << fmt::format(

examples/vectoradd/vectoradd.cpp

@@ -46,7 +46,8 @@ namespace usellama
         std::cout << "\nLLAMA\n";
         Stopwatch watch;
 
-        const auto mapping = [&] {
+        const auto mapping = [&]
+        {
             const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
             if constexpr (MAPPING == 0)
                 return llama::mapping::AoS{arrayDomain, Vector{}};