- User Since
- Jan 23 2017, 12:28 AM (201 w, 3 d)
Tue, Dec 1
Thanks for your feedback so far!
Mon, Nov 30
A helpful comment could explain why these instructions should not be used. Nits, else LGTM.
Fri, Nov 27
I understand this is necessary for VE ABI compliance. Is there anything in LLVM that says a truncate has to zero-out all leading bits?
It's about making the code more reusable - this is not a stopper for this patch, however.
Thu, Nov 26
- Make all tests use the same run command
NFC. Improved tests - also checking the sdnode operand list and types
Added one test for every VP SDNode type.
Wed, Nov 25
NFC. Fixed formatting, tidiness.
Mon, Nov 23
NFC. Attaching this here because the rebase required manual merging.
NFC, Adressed comments
Removed immediate variants in isel and tests. Added comment for future immediate folding.
- Simplified macros
- Keeping the vector inst+imm isel patterns and tests for now. When vector-immediate folding is upstream, we can remove that code
Fri, Nov 20
LGTM. Did this actually overflow in the wild? A test would be nice.. but also hard to come by given that you'd need to manufacture such a large frame.
Thu, Nov 19
The current versions seems a good middle ground between computing a tighter bound and abolishing the assertion altogether. Trying to be smart in assertions is usually a bad idea because people might pull the carpet (underlying assumptions) from under you some day.
Followed @fhahn 's suggestion to test against the widest vector register width instead.
Added i64,f64 immediate patterns and tests
Wed, Nov 18
Rebased onto "speculatable" VP intrinsic patch.
- Add isel and tests for vbrd-with-immediate.
- Added <128 x i16>, <256 x i16>, <128 x i32> broadcast tests to show promotion, widening and scalarization.
- Added immediate patterns and tests
- Addressed comments
- Adressed comments
- Added override options to override the expansion strategy.
- Added expansion tests for fixed and scalable types and all possible expansion strategies.
Tue, Nov 17
A good share of VE vector intrinsics have been upstreamed (see test/CodeGen/VE/VELIntrinsics/). No need to keep this reference patch around any longer.
Here is that prior commit that changed this - i guess, we can just commit our change, if this looks good to you.
Mon, Nov 16
We should upstream the utils/update_llc_test_checks.py changes that were used to generate these tests.
- added vector param tests that trigger explicit vreg copies.
- Rebased onto renaming patch (OC -> OPC)
Fri, Nov 13
Linking this to the SelectionDAG patch: D91441
Split up into:
Merged into one fastcc vreg-to-vreg patch
Largely supserseeded by D91416 - the LVLGen patch.
I'll shrink this down/rename to just the vreg-to-vreg move change.
- Add a test case for non-changing VL after initial setting.
Any comments for this one? Is this good to go?
Thu, Nov 12
Wed, Nov 11
Taking this off the review queues - we will use the ext_vector_type attribute instead (D88905).