Implement CUDA Helmholtz sum-factorization kernels

This is a preliminary implementation of the Helmholtz operator. The Helmholtz operator is build by using a combination of BwdTrans, PhysDeriv, IProductWRTBase, and IProductWRTDerivBas operators. Both a StdMat and a CUDA versions have been implemented. Currently, unity diffusion coefficient is assumed along with lambda = 1. However, provision for non-unity, but constant, diffusion coefficient and a non-unity lambda parameter have been-made. The various sub-operators required to perform the Helmholtz operator are instantiated in the constructor.

marked this merge request as draft

added 6 commits

• b0510908...554abc7b - 2 commits from branch nektar:master
• f77cce76 - Update comment
• 8cefb1f6 - Fix operator for tri, tet, pyr, and prism
• 2760a399 - Merge remote-tracking branch 'upstream/master' into iproductderivbase_sum_fac_cuda_kernels
• 9cece685 - Merge branch 'iproductderivbase_sum_fac_cuda_kernels' into helmholtz_sum_fac_cuda_kernels

Compare with previous version

added 1 commit

• f228a8ba - Add provision for scaling

Compare with previous version

marked this merge request as ready

assigned to @ccantwel

changed the description

added 1 commit

• 20024df6 - Some minor updates

Compare with previous version

added 1 commit

• 442ed4c3 - Add provision for non-unity lambda

Compare with previous version

added 1 commit

• c9186177 - Clang-format

Compare with previous version

added 1 commit

• 09753e38 - Further formatting

Compare with previous version

added 1 commit

• 9afd28f4 - Add provision for non-unity diffusion coefficient

Compare with previous version

resolved all threads

changed the description

changed the description

added 3 commits

• 9afd28f4...d05e90c5 - 2 commits from branch nektar:master
• 3c4eeec0 - Merge remote-tracking branch 'upstream/master' into helmholtz_sum_fac_cuda_kernels

Compare with previous version

added 7 commits

• 3c4eeec0...4360038b - 6 commits from branch nektar:master
• 6388487b - Merge remote-tracking branch 'upstream/master' into helmholtz_sum_fac_cuda_kernels

Compare with previous version

This looks reasonable to me, so will go ahead and merge. We could obviously do with some tests adding to ensure correctness, but probably best we sort them out for the constituent operators first.

merged

mentioned in commit 2f79a27b