User Details
- User Since
- May 5 2014, 7:26 AM (464 w, 1 d)
Today
LGTM - cheers
Yesterday
please can you add alive2 links?
@foad MSVC builds seem to be breaking due to this:
https://lab.llvm.org/buildbot/#/builders/86/builds/53509
LLVMARMAsmParser.lib(ARMAsmParser.cpp.obj) : error LNK2001: unresolved external symbol "class llvm::MCInstrDesc const * const llvm::ARMDescs" (?ARMDescs@llvm@@3QBVMCInstrDesc@1@B) bin\llc.exe : fatal error LNK1120: 1 unresolved externals ninja: build stopped: subcommand failed.
LGTM - cheers
Sun, Mar 26
LGTM
Fri, Mar 24
LGTM - was this by manual inspection or did an analyzer spot it?
Thu, Mar 23
unblocking this - thanks @kazu
Wed, Mar 22
LGTM - cheers
LGTM - cheers!
The <3 x i64> is interesting - but I meant other legal SSE vector types likes <16 x i8> / <8 x i16> / <4 x i32>
@goldstein.w.n reverse-ping - whats the plan for this patch?
Tue, Mar 21
What regressions are you seeing for ULE vector cases with this fold? Manual optimization looks OK - https://llvm.godbolt.org/z/n9ezae5En
Please can you add remaining icmp predicates - I think we still need sge/uge/eq/ne?
Thanks for taking a look at this!
please can you add vector test coverage?
please can you add vector test coverage?
LGTM with one very optional minor
Mon, Mar 20
address regressions in foldBinOpIntoSelect when handling shift(x, trunc/zext(y)) patterns
Fri, Mar 17
rebase
(temporarily commandeering to rebase patch) @deadalnix please take this back when you're about
Thu, Mar 16
Looks like we're very close to finally getting this in - @kazu @goldstein.w.n do you recognize any of the remaining regressions?
@goldstein.w.n What is happening with this patch? After they leave my "Ready to Review" list I tend to lose track......
Wed, Mar 15
rebase again (and actually update the changed test this time....)
rebase
LGTM
drop useless vector handling
rebase
LGTM - a reduction cost is almost impossible without the type anyhow
Tue, Mar 14
LGTM - was originally added in D20601 when we had much a weaker cost model
rebase + address feedback
Consistently use logicop
LGTM - cheers
Mon, Mar 13
ping - apart from the x86 multiuse cmp issues does anyone have anymore comments?
Patch itself LGTM but the summary needs a cleanup - the defm section is irrelevant (and confusing) - just explain that ATOMIC_LOGIC_OP_RM shouldn't be using the *i8 instruction encodings
LGTM
LGTM - cheers