Cleanup - Lapack to tile

Keep going on with the cleanup of the flush and getoncpu.

I start by the basic function of lapack to tile and tile to lapack. The idea is to remove all workspaces from pz*.c files if possible.

In this MR:

• Remove the asynchronous interface to change the data format. People cans till do it by copy/paste the synchronous one in their own code and using the asynchronous version of zlacpy.
• Remove the pztile.c file
  • Data format converion uses zlacpy
  • ztile_to_zero is a zlaset so no need for duplicated code

changed the description

added 13 commits

• 63273913 - Remove references to morse_pzlapack_to_tile and morse_pztile_to_lapack
• 8f7daef4 - Move zpotrf to the new conversion functions
• 03b4a34f - Replace all zooplap2tile by new zlap2tile
• 4fd458c7 - Keep converting the lapack interface function
• 35fe8c53 - Indent and whitespace cleanup
• 6a341c23 - Replace the tile2lap
• e1129800 - Add cleanup
• bafbfed5 - Cleanup
• b86e73ef - Correct Dptr calls
• 579dc09e - Explode descriptor one per line
• ecbee0df - Add descriptor declarations
• 313aeb54 - Replace the descriptors in the calls
• 3ef551c6 - Fix returned values

Compare with previous version

added 2 commits

• 06546b16 - Use uplo when possible
• 9f07583e - Add in,out,inout mode to descriptor to avoid useless layout conversions

Compare with previous version

added 1 commit

• 588c3b26 - Try to limit the number of tests running in parallel

Compare with previous version

added 1 commit

• 2b62f08f - Fix mistake in zposv

Compare with previous version

Ok. Now that everything works. @thibault, @furmento, @agullo, @fpruvost is some of you could have a quick look. That would be nice. Thanks.

added 1 commit

• e7fb36f3 - Remove references to RUNTIME interface in z*.c files

Compare with previous version

added 4 commits

• cbc1b247 - Some cleanup for sonarqube
• c27ce41a - Minor
• 83d965ee - More for sonarqube
• dc4b62db - Check that all flushes are now in z*.c files with a script

Compare with previous version

added 1 commit

• cec2d4a6 - Factorize code

Compare with previous version

added 2 commits

• 373bc62c - Change inout to make sure its a composition of other two
• 2313a45c - Remove unused macro

Compare with previous version

added 3 commits

• a780d6bf - Change inout to make sure its a composition of other two
• bfcf8ea8 - Remove unused macro
• dde7c55d - Factorize code

Compare with previous version

added 1 commit

• a29ceeb8 - Add extra flush to norm computations for parsec

Compare with previous version

added 1 commit

• b33153a7 - Fix issue with desc created/destroyed in mat_alloc/free

Compare with previous version

Mathieu, could you give us permission on your repository? https://gitlab.inria.fr/faverge/chameleon gives us 404.

I mean, the diff in the web interface is unreadable with all the whitespace changes, so I'd rather read a git diff -w or even --word-diff

I'll do that, but you know uyou can disable the white diff online :)

Well, I already know how to use git, I don't really want to spend time on diving into a plane cabine-like interface whose fanciness will probably fade within the next years, for another fancy web interface.

I also already sent that link that explain how to configure your git to get access to PR even if you don't have access to the original repository of the PR.

https://gitlab.inria.fr/help/user/project/merge_requests/index.md#checkout-merge-requests-locally

What I mean, is that I don't want to force you to use the web interface. But they are plenty of solutions to do it, even when we know how to use git ;)

added 2 commits

• 1663a7cd - Add a missing destroy
• ba9be25f - Factorize some code

Compare with previous version

Thanks, that's much more usable in vim that in a webbrowser :)

The principle of the patch looks good to me, and it's clearly useful factorization :D

Too bad it doesn't want to work anymore :P

added 1 commit

• 8fda2448 - Temporary fix for missing flush/wait in QR functions

Compare with previous version

Thanks much @faverge. Hard to check visually such an amount of changes but looks good. In the long term, it might be nice to think replacing explicit flushes with implicit memory-aware mechanisms.

added 1 commit

• e2236009 - I hope this time I did not forget any

Compare with previous version

merged

mentioned in commit 1ff90ef1