- User Since
- May 5 2014, 7:26 AM (349 w, 5 d)
@myhsu Please can you raise a bug covering the CodeBeads vs TSFlags refactor options and make it a blocker against a second bug for making the m68k backend non-experimental?
add ult/ugt handling
The impact on existing targets is pretty minimal - I suppose the logical operand mappings code is the most exposed part? The CodeBeads code is relatively self-contained.
Thu, Jan 14
FWIW I still think we're better off keeping (m68k only) CodeBeads rather than refactoring a lot of code that will affect mainstream targets. After m68k has been pushed, we can reinvestigate whether to keep CodeBeads (and whether other targets would benefit from using it), or moving m68k to a refactored TSFlag mechanism - we can make either a pre-requisite for it losing its experimental status if necessary.
LGTM - cheers!
@foad Please can you rebase? I think I've replaced the dodgy test with something useful now
Anyone have any final comments - otherwise I'll commit this tomorrow.
This was a direct move from SelectionDAG's implementation as I wanted to avoid introducing any subtle behaviour changes - I've added the exhaustive tests (which has already exposed one issue I'm fixing now) and I'll start cleaning up the KnownBits::sextInReg implementation after that.
LGTM - cheers
@jplayer-nv It looks like this patch has broken the ppc buildbots - please can you take a look? http://lab.llvm.org:8011/#/builders/93/builds/1435
LGTM but a PPC-guru needs to signoff
Also, sorry for hijacking but if you're looking at std::move cleanup, this scary static analysis warning has been around for a while: https://llvm.org/reports/scan-build/report-BitcodeReader.cpp-BitcodeReader-0-1.html#EndPath
I'm not certain but I thought some of these moves had to be kept to appease old versions of gcc - but I can't remember any details.
Wed, Jan 13
@foad Can you raise a patch please?
Tue, Jan 12
I've been working on PR40111 to make most x86 ISD::USUBSAT code generic and moved into DAGCombine - I'll look at this as well.
Mon, Jan 11
@craig.topper Whatever happened to this patch? It still looks useful.
Sat, Jan 9
This all needs properly documenting in the release notes, and possibly in the clang docs as well explaining that mmx/3dnow is supported only inside assembly,
Fri, Jan 8
@kariddi Are you still looking at this, otherwise abandon it?
Thu, Jan 7
Wed, Jan 6
Add better comment suggested by @spatel
Add commuted pattern handling
rebase (still WIP though)
Improve scalar test coverage
LGTM - cheers
Tue, Jan 5
Improve handling of AND/OR(OR(X2,SHL(Y2,BW/2)) , OR(X1,SHL(Y1,BW/2)) ) style patterns to fix remaining movmsk patterns
LGTM with one minor