Page MenuHomePhabricator

[CUDA] Add explicit mapping from sm_XX to compute_YY.
ClosedPublic

Authored by jlebar on Jan 11 2016, 4:38 PM.

Diff Detail

Repository
rL LLVM

Event Timeline

jlebar updated this revision to Diff 44580.Jan 11 2016, 4:38 PM
jlebar retitled this revision from to [CUDA] Add explicit mapping from sm_XX to compute_YY..
jlebar updated this object.
jlebar added a reviewer: tra.
jlebar added subscribers: echristo, jhen, cfe-commits.
jlebar updated this revision to Diff 44585.Jan 11 2016, 5:09 PM

Update to latest set of architectures, according to ptxas --help from cuda 7.5.

Seems reasonable. Why remove all of the tests though?

Seems reasonable. Why remove all of the tests though?

Previously we were checking that the arch was of the form /^sm_\d+$/ -- now we're checking that it's one of some list of acceptable things. I figured that the edge-casey tests aren't as useful in that latter case, but I can add them back if you like, not a big deal.

As long as the list of BAD things is somewhat conclusive I don't mind either way. Just making sure that this was the reason :)

-eric

echristo accepted this revision.Jan 12 2016, 1:54 PM
echristo added a reviewer: echristo.
This revision is now accepted and ready to land.Jan 12 2016, 1:54 PM
jlebar updated this revision to Diff 44673.Jan 12 2016, 2:19 PM
jlebar edited edge metadata.

Use llvm::StringSwitch instead of a sequence of if statements.

Still good. much nicer now.

This revision was automatically updated to reflect the committed changes.