Skip to content
Snippets Groups Projects

p2p: define isrc, itgt in block kernel

Merged Alexandru Fikl requested to merge fikl2/sumpy:fix-p2p-block-kernel into master
1 unresolved thread

I fixed this in the QBX block matrix generator a while back, but forgot about this one.

Merge request reports

Loading
Loading

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
319 319 loopy_knl = lp.make_kernel(
320 320 "{[imat, idim]: 0 <= imat < nresult and 0 <= idim < dim}",
321 321 self.get_kernel_scaling_assignments()
322 # NOTE: itgt, isrc need to always be defined in case a statement
323 # in loopy_insns or kernel_exprs needs them (e.g. hardcoded in
324 # places like get_kernel_exprs)
  • Good catch. Also: This certainly isn't so immensely elegant...

  • The hardcoding in get_kernel_exprs isn't really a problem here because it only shows up when there's a strength (although that should probably be fixed too).

    The one that caused me problems seems to be in here. Any idea how to fix it?

  • As some small relief, loopy!271 (closed) teaches loopy how to do functions, which I am hoping will let us replace the current hot-glue-based approach to combining kernels/translations and iteration structures and make it clearer what variables are and aren't intended to be exposed.

    Beyond that, I haven't yet been able to come up with anything that's substantially cleaner. I'm open to suggestions though.

  • Please register or sign in to reply
  • Andreas Klöckner enabled an automatic merge when the pipeline for 4cb0f33e succeeds

    enabled an automatic merge when the pipeline for 4cb0f33e succeeds

  • Thanks for working on this!

  • mentioned in commit 63cfd0a0

  • Please register or sign in to reply
    Loading