++

Author: PhilipDeegan

.github/workflows/build_win.yml

@@ -27,4 +27,4 @@ jobs:
       run: | # /bin/link interferes with cl/link.exe
         bash -c "rm /bin/link"
         bash -c "curl -Lo mkn.exe https://github.com/mkn/mkn/releases/download/latest/mkn.exe"
-        bash -c 'KLOG=3 ./mkn clean build test run -p test,b0,b1 -OtKda "-arch:AVX2 -std:c++20 -fPIC -DKUL_FORCE_TRACE" -l -pthread -g 0'
+        bash -c 'KLOG=3 ./mkn clean build test run -p test,b0,b1 -OtKda "-arch:AVX2 -std:c++20 -DKUL_FORCE_TRACE" -g 0'

inc/mkn/avx/lazy.hpp

@@ -260,7 +260,7 @@ struct LazyEvaluator
 #include <cstdint>
 
 template<typename E>
-using AVXVec = std::vector<E, mkn::kul::AlignedAllocator<E, 32>>;
+using AVXVec = std::vector<E, mkn::kul::AlignedAllocator<E, Options::ALIGN()>>;
 )";
 
         std::string const funcheader = R"(
@@ -346,7 +346,7 @@ void exec(LazyVal_t const& t, T* const ret){
     std::vector<std::string> fn_strs{"add", "sub", "mul", "div"};
 
     template<typename E>
-    using AVXVec = std::vector<E, mkn::kul::AlignedAllocator<E, 32>>;
+    using AVXVec = std::vector<E, mkn::kul::AlignedAllocator<E, Options::ALIGN()>>;
     static inline thread_local AVXVec<std::array<T, N>> tmps{};
 };
 

test/test_lazy.cpp

@@ -8,7 +8,7 @@ constexpr static std::size_t N = 1e6 + 5;
 
 void add()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2);
     auto [l0, l1] = lazy(a0, a1);
     auto r        = eval(l0 + l1 + l0);
@@ -19,7 +19,7 @@ void add()
 
 void mul()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2);
     auto [l0, l1] = lazy(a0, a1);
     auto r        = eval(l0 * l1 * l0);
@@ -30,7 +30,7 @@ void mul()
 void fma3()
 {
     {
-        using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+        using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
         DV a0(N, 1), a1(N, 2), a2(N, 3);
         auto [l0, l1, l2] = lazy(a0, a1, a2);
         auto lz           = l0 * l1 + l2;
@@ -39,7 +39,7 @@ void fma3()
         assert(r.front() == 5 and r.back() == 5);
     }
     {
-        using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+        using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
         DV a0(N, 1), a1(N, 2), a2(N, 3);
         auto [l0, l1, l2] = lazy(a0, a1, a2);
         auto lz           = l0 + l1 * l2;
@@ -51,7 +51,7 @@ void fma3()
 
 void fma0()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2), a2(N, 3);
     auto [l0, l1, l2] = lazy(a0, a1, a2);
     auto lz           = l0 * l1 + l2 + l0;
@@ -62,7 +62,7 @@ void fma0()
 
 void fma1()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2), a2(N, 3), a3(N, 4), a4(N, 5);
     auto [l0, l1, l2, l3, l4] = lazy(a0, a1, a2, a3, a4);
     auto r                    = eval(l0 * l1 + l2 * l3 + l4 * l1 + l2 * l3 + l4 * l1);
@@ -71,7 +71,7 @@ void fma1()
 }
 void fma2()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2), a2(N, 3), a3(N, 4), a4(N, 5), a5(N, 6);
     auto [l0, l1, l2, l3, l4, l5] = lazy(a0, a1, a2, a3, a4, a5);
     auto r                        = eval(l0 * l1 + l2 + l3 + l4 * l5);
@@ -84,7 +84,7 @@ void fma() {}
 
 void fn0()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2), a2(N, 3), a3(N, 4), a4(N, 5);
     assert(a0.data() and a4.data());
     auto [l0, l1, l2, l3, l4] = lazy(a0, a1, a2, a3, a4);
@@ -96,7 +96,7 @@ void fn0()
 
 void fn1()
 {
-    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, 32>>;
+    using DV = std::vector<double, mkn::kul::AlignedAllocator<double, Options::ALIGN()>>;
     DV a0(N, 1), a1(N, 2), a2(N, 3), a3(N, 4), a4(N, 5);
     assert(a0.data() and a4.data());
     auto [l0, l1, l2, l3, l4] = lazy(a0, a1, a2, a3, a4);