[Trifocal+P2Pt] alignment priming

CMakeLists.txt

@@ -2,7 +2,7 @@
 #
 
 cmake_minimum_required(VERSION 3.10.2 FATAL_ERROR)
-set(CMAKE_CXX_STANDARD "11")
+set(CMAKE_CXX_STANDARD "17")
 if(NOT CMAKE_CXX_STANDARD_REQUIRED)
   set(CMAKE_CXX_STANDARD_REQUIRED ON)
 endif()

minus/chicago14a.hxx

@@ -3607,8 +3607,8 @@ minus<chicago14a, F>::solve(
   if (!io::point_tangents2params(p, tgt, id_tgt0, id_tgt1, params))
     return false;
 
-  typename M::solution solutions[M::nsols];
-  typename M::track_settings settings = M::DEFAULT;
+  alignas(64) typename M::solution solutions[M::nsols];
+  alignas(64) typename M::track_settings settings = M::DEFAULT;
 
   unsigned npaths_per_thread = M::nsols/nthreads;
   assert(M::nsols % nthreads == 0);

minus/minus.h

@@ -85,7 +85,7 @@ class minus_core { // fully static, not to be instantiated - just used for templ
 
   struct solution
   {
-    alignas(sizeof(C<F>)) C<F> x[f::nve];    // array of n coordinates
+    alignas(64) C<F> x[f::nve];    // array of n coordinates
     F t;               // last value of parameter t used
     unsigned num_steps;  // number of steps taken along the path
     solution_status status;

minus/minus.hxx

@@ -12,7 +12,7 @@
 #include "minus.h"
 #include "internal-util.hxx"
 
-#include "Eigen/Core"
+#include "Eigen-latest/Core"
 
 namespace MiNuS {
 
@@ -30,8 +30,11 @@ using namespace Eigen; // only used for linear solve
 // 
 template <problem P, typename F> void 
 minus_core<P, F>::
-track(const track_settings &s, const C<F> s_sols[f::nve*f::nsols], const C<F> params[2*f::nparams], solution raw_solutions[f::nsols], unsigned sol_min, unsigned sol_max)
+track(const track_settings &s, const C<F> s_sols_u[f::nve*f::nsols], const C<F> params_u[2*f::nparams], solution raw_solutions_u[f::nsols], unsigned sol_min, unsigned sol_max)
 {
+  const C<F> *s_sols = reinterpret_cast<C<F> *> (__builtin_assume_aligned(s_sols_u,64));
+  const C<F> *params = reinterpret_cast<C<F> *> (__builtin_assume_aligned(params_u,64));
+  solution *raw_solutions = reinterpret_cast<solution *> (__builtin_assume_aligned(raw_solutions_u,64));
   assert(sol_min <= sol_max && sol_max <= f::nsols);
   alignas(64) C<F> Hxt[NVEPLUS1 * f::nve]; 
   alignas(64) C<F> x0t0[NVEPLUS1];

scripts/minus-disassemble

@@ -1,5 +1,5 @@
 #/opt/local/bin/c++  -DMINUS_WARN_DEPRECATED -DMINUS_WARN_DEPRECATED_ONCE -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus-bin -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus  -fPIC -O3 -DNDEBUG -march=native -ffast-math -g  -Wno-undefined-var-template -std=gnu++11 -o minus.s -S -c /Users/rfabbri/cprg/vxlprg/lemsvpe/minus/Templates/minus-chicago-alltypes-allformulations+double.cxx
 
 
- /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++ -DMINUS_WARN_DEPRECATED -DMINUS_WARN_DEPRECATED_ONCE -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus-bin -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus-bin/minus -fPIC -O3 -DNDEBUG -march=x86-64-v4 -ffast-math -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX13.1.sdk -mmacosx-version-min=12.6   -Wno-undefined-var-template -std=gnu++11 -o m.s -S -c /Users/rfabbri/cprg/vxlprg/lemsvpe/minus/minus/Templates/minus-chicago-alltypes-allformulations+double.cxx
+ /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++ -DMINUS_WARN_DEPRECATED -DMINUS_WARN_DEPRECATED_ONCE -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus-bin -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus -I/Users/rfabbri/cprg/vxlprg/lemsvpe/minus-bin/minus -fPIC -O3 -DNDEBUG -march=x86-64-v4 -ffast-math -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX13.1.sdk -mmacosx-version-min=12.6   -Wno-undefined-var-template -std=gnu++17 -o m.s -S  -fomit-frame-pointer -fno-exception -c /Users/rfabbri/cprg/vxlprg/lemsvpe/minus/minus/Templates/minus-chicago-alltypes-allformulations+double.cxx