Fix #56 - Issue in CUDA_zparfb kernels for TT case

• Fix issue #56 (closed) in CUDA_zparfb kernel that was not handling the pentagonal V of TT kernels
• Remove zt[st][mqr|mlq|qrt|lqt] codelets that were useless and were duplicating ztpxxx ones.

added 9 commits

• 9bf29e37...16165abb - 2 commits from branch solverstack:master
• 5ed9ade8 - Minor on cuda_zlarfb to directly call ztrmm
• f343a4fa - Add missing tpmlqt kernel for GPU, and fix the parfb kernel to handle the TT cases
• 2f9dd95a - Small changes to fix documentation and/or code
• c69e5f0a - Update the workspace sizes to call ztpmXXt kernels
• 5715a0a8 - Update StarPU codelets
• 14482041 - Remove all TS/TT codelets for QR/LQ kernels
• d3492f22 - Fix columnwise/rowise issue to apply any unmqr on GPU

Compare with previous version

@ltaief and @sukkarde : This should fix the issue you had with the qdwh branch :). So as soon as it's merged. I will rebase the qdwh branch on top of it.

@fpruvost, @agullo it's ready for review.

added 1 commit

• 99fecefb - Remove references to tsmqr/ttmqr/tsmlq/ttmlq/...

Compare with previous version

added 2 commits

• 16954d45 - Fix mispell
• f39c61b6 - Fix mispell

Compare with previous version

added 1 commit

• cc2a3552 - Silent warning for unused variables

Compare with previous version

Let's discuss the documentation later this week, and update it in a later PR.

merged

mentioned in commit 55938b98