Reformulate index computation and remove conversion warnings

Author: Alexander Strack

include/plssvm/backends/HPX/kernel/cg_explicit/kernel_matrix_assembly.hpp

@@ -61,8 +61,8 @@ void device_kernel_assembly(const std::vector<real_type> &q, std::vector<real_ty
 
  ::hpx::for_each(::hpx::execution::par_unseq, range.begin(), range.end(), [&](const std::size_t idx) {
         // calculate the indices used in the current thread
-        const std::size_t col = static_cast<std::size_t>(0.5 * (2 * blocked_dept + 1 - std::sqrt(4 * blocked_dept * blocked_dept + 4 * blocked_dept - 8 * idx + 1)));
-        const std::size_t row = static_cast<std::size_t>(idx - (col * blocked_dept - 0.5 * col * col - 0.5 * col));
+        const std::size_t col = blocked_dept + static_cast<std::size_t>(0.5 - 0.5 * std::sqrt(4 * (blocked_dept * blocked_dept + blocked_dept) - 8 * idx + 1));
+        const std::size_t row = static_cast<std::size_t>(0.5 * static_cast<double>(2 * (idx - col * blocked_dept) + col * col + col));
 
         const std::size_t row_idx = row * INTERNAL_BLOCK_SIZE_uz;
         const std::size_t col_idx = col * INTERNAL_BLOCK_SIZE_uz;

include/plssvm/backends/HPX/kernel/cg_implicit/kernel_matrix_assembly_blas.hpp

@@ -75,8 +75,8 @@ inline void device_kernel_assembly_symm(const real_type alpha, const std::vector
 
     ::hpx::for_each(::hpx::execution::par_unseq, range.begin(), range.end(), [&](const std::size_t idx) {
         // calculate the indices used in the current thread
-        const std::size_t col = static_cast<std::size_t>(0.5 * (2 * blocked_dept + 1 - std::sqrt(4 * blocked_dept * blocked_dept + 4 * blocked_dept - 8 * idx + 1)));
-        const std::size_t row = static_cast<std::size_t>(idx - (col * blocked_dept - 0.5 * col * col - 0.5 * col));
+        const std::size_t col = blocked_dept + static_cast<std::size_t>(0.5 - 0.5 * std::sqrt(4 * (blocked_dept * blocked_dept + blocked_dept) - 8 * idx + 1));
+        const std::size_t row = static_cast<std::size_t>(0.5 * static_cast<double>(2 * (idx - col * blocked_dept) + col * col + col));
 
         const std::size_t row_idx = row * INTERNAL_BLOCK_SIZE_uz;
         const std::size_t col_idx = col * INTERNAL_BLOCK_SIZE_uz;