- User Since
- Jun 1 2015, 8:58 AM (314 w, 5 d)
Thu, Jun 10
LGTM other than my last question. Thanks!
Wed, Jun 9
LGTM other than the last few nits.
Thanks for the detailed explanation.
I'm not familiar enough with all the decisions this affects but I agree that it should be the same. Have you seen any changes introduced by this? I'm assuming you may have tried it downstream, for example.
- rebase on top of new MVT formatting
- adjust tests for new VType formatting
Note that I pushed rG292f4197249b to fix the failing tests introduced by this patch. Please check the test results before merging in the future, e.g. https://reviews.llvm.org/B107914 was showing this issue.
Tue, Jun 8
Awfully brave to add new non-pow2 types; I like it. Unfortunate that we've both filed patches to this same file at the same time (D103884). One of us will have a fun time.
- update code comments
Might as well throw my hat in here, LGTM.
LGTM, nice catch.
Mon, Jun 7
- slice out change to dyn_castNegVal
- update test with FIXME accordingly
- add test fixme comment
- normalize check order to fixed-length then (scalable) splat
- consistently remove "this->" from these methods
Fri, Jun 4
My browser's really chugging on this huge patch so my input has to be brief. Maybe we could hide the test changes for now?
Thu, Jun 3
- let the build bots have a go at it
- rebase & add a fixme
Wed, Jun 2
- early-exit for non-V functions
- update test checks
- update test checks
Tue, Jun 1
Ah that's interesting, thanks both for the input. I've actually done both of those things in other backends and wasn't a huge fan of either approach. I found the flag to be quite brittle and didn't like how we had to clean up the redundant addressing instructions to get the best code.
Ping @craig.topper, just making sure you're okay with this patch as-is before diving into the next stage.
Yep, seems reasonable. LGTM. It might be prudent to quote the spec where it says For the purpose of register group overlap constraints, mask elements have EEW=1, since it doesn't explicitly say that in the compares section.
- rebase to see how the buildbots get on
Mon, May 31
- catch non-constant splat_vectors
- update rvv test checks
- use MaybeAlign over std::max
- rebase to include D103262
Fri, May 28
Since @simoll's happy, LGTM.
- try to find good names for the various helper functions
Thu, May 27
- remove FIXMEs, on account of it being.. fixed
- rebase for test changes