Draft: Develop

This MR adds two features to pystencils. First, the base pointer specification is revealed to the user which allows producing kernels with less register usage. Second, the summands insider the summation printer are printer recursively now which allows for more parallelism inside a single core.

added feature label

marked this merge request as draft

added 3 commits

• 5e092a56...30641109 - 2 commits from branch pycodegen:master
• e17e56dc - Merge branch 'master' of i10git.cs.fau.de:pycodegen/pystencils into develop

Compare with previous version

added 1 commit

• 7976ae1d - Fixed vectorization

Compare with previous version

added 1 commit

• ce97d891 - Remove time estimation

Compare with previous version

added 1 commit

• 29610801 - Use likwid-pin for benchmarking

Compare with previous version

added 5 commits

• 29610801...24dde405 - 4 commits from branch pycodegen:master
• 21b93c1b - Merge branch 'master' of i10git.cs.fau.de:pycodegen/pystencils into develop

Compare with previous version

added 1 commit

• 6482fab8 - Minor change

Compare with previous version

See !210 (closed) for some previous comments on this:

• base_pointer_spec should be chosen appropriately for the number of registers provided by the architecture
• the recursive summation should be used independently of the number of summands and the same scheme should be used for multiplication as well

added 1 commit

• af8eb9cc - Imporved benchmark app

Compare with previous version

removed review request for @kuron

added 1 commit

• 58f3d520 - Comment regarding likwid

Compare with previous version

closed