- User Since
- Sep 12 2013, 6:57 AM (454 w, 1 d)
Tue, May 24
- Move fast math flag requirements from fcmp to sqrt.
- Don't require 'fast' flag on sqrt, only 'nnan && (reassoc || afn)'
Mon, May 23
- Be stricter around fast math flags and require 'fast' not just 'nnan'.
- Propagate fast math flags through combine.
Thu, May 19
Wed, May 18
Ah, a quick check suggests your patch also works for our case. I'll abandon this in favour of that one. Thanks!
Tue, May 17
It should be noted that I'm not 100% certain this patch is sound so would very much like some guidance here, I'm looking to address the discussion from here as we have hit this case: https://reviews.llvm.org/D105942#inline-1007492. It's not clear to me whether more needs to be done to cope with the fact that we are dealing with a poison value here.
Mon, May 16
- Use m_NonNegative instead of manually checking for splats and their values
Wed, May 11
- Fix typos in comments, SQSHL -> SRSHL.
- Rearrange tests to have positive cases first.
Mon, May 9
Wed, May 4
Tue, May 3
Fri, Apr 29
- Rework patch to use same mechanism as stack realignment, rather than using base pointer.
- Only restore from base pointer when SVE CSRs are present
Thu, Apr 28
Apr 27 2022
- Add testcase protecting original issue
- Add explict false return for invalid frame index
- Update test to use update_llc_test_checks.py script
Apr 26 2022
Apr 20 2022
Apr 19 2022
Is there any update on this change? It provides some nice improvements to various workloads we have been looking at and would be good to get in.
Apr 8 2022
Mar 4 2022
This all looks reasonable to me, but I'll let @peterwaller-arm have the final say.
Feb 24 2022
Feb 11 2022
Feb 10 2022
- Tighten check around type sizes
Feb 9 2022
- Tighten type size check
- Move test to already existing file
- Add rudimentary check lines to test
Feb 8 2022
Feb 3 2022
LGTM with minor nits.
LGTM, as long as Sander agrees.
Feb 1 2022
Jan 31 2022
Dec 1 2021
Nov 30 2021
- Only perform duplicated dag combines before legalization
Nov 29 2021
Nov 26 2021
Nov 25 2021
- Change approach again to be between the last two
- Go back to matching asrd nodes in instruction selection
- Create the asrd nodes during lowering rather than BuildSDIVPow2
- This makes the patch cleaner whilst still supporting larger than legal types
- Intrinsics will be handled in a separate patch in instcombine.
- Fix incorrect type used in addTypeForFixedLengthSVE
- Cleanup conditions in LowerFixedLengthVectorLoadToSVE
Nov 24 2021
Nov 23 2021
Nov 22 2021
Nov 18 2021
The tests in this are broken for windows, which I've fixed in 45e102a173680fd3c90def79a7f0766ed2786ff0.
- Rework patch to use instruction selection to match (sdiv (dup <pow2>)) nodes rather than generating asrd nodes
- This allows both larger than legal fixed types to work correctly, as well as sdiv intrinsics.
Nov 17 2021
- Use more brute force approach to ensure ordering is accounted for
- This massively simplifies things and removes what was becoming very confusing logic
- Add tests for missing cases
Nov 16 2021
Nov 15 2021
- Fix duplicate arch feature in unit test
- Use enum class instead of plain enum with typedef
Nov 12 2021
Nov 10 2021
Nov 5 2021
- Make semantics of SRAD_PRED match the other *_PRED nodes
Oct 29 2021
Apologies, this should be fixed in bf72a469ba32f786b26d314241e8eb7dfd42f807