- User Since
- Feb 2 2017, 2:24 AM (321 w, 17 h)
can this land independently of its parents in the stack ?
I don't recall when/ why this was added but thanks for the cleanup!
Tue, Mar 28
Fantastic, thank you!
Looks good but is this really NFC ?
Mon, Mar 27
Add negative tests
Sun, Mar 26
Sat, Mar 25
Fri, Mar 24
sorry clicked too soon, added some comments
thanks for adding vector.transfer too, LGTM!
Update commit message
Drop untested transforms for now
Missed a spot
Get rid of TestVectorTransferFullPartialSplitPatterns
Get rid of TestVectorTransferLoweringPatterns
Get rid of https://reviews.llvm.org/D146755
Reorder transforms alphabetically.
Add declareDependentDialect<LLVM::LLVMDialect>(); and fix test
Thu, Mar 23
Sorry, this is actually a WIP that I wanted to disseminate, not yet ready
@ftynse @qcolombet what is the magic incantation to auto-declare produced dialects without having to declare dependent dialects ?
Revert spurious formatting changes
Get rid of TestVectorTransposeLowering
Get rid of TestVectorContractionLowering
Drop one-off test pass option for multi-reduction lowering.
Bazel build fixes
Thanks, LGTM, make sense to keep the interface work for later.
Update commit message.
Wed, Mar 22
@springerm please note the comments around deprecating canonicalization patterns, these have impacts on bufferization and I expect we'll need to test and relayer things properly to avoid spurious copies.
Does this update change the way this is supposed to interact with HoistPadding ?
Or in other words, how do you see this interacting with HoistPadding ?
Tue, Mar 21
Address and rebase
Note the automatic d -> s promotion due to affine scoping rules.
I would personally prefer to trigger the promotion explicitly rather than have it apply everywhere all the time but this has not been a hard blocker recently.
thanks for picking up the slack @herhut , won't have time to look deeper atm so please go with this