- User Since
- Apr 30 2013, 5:34 PM (199 w, 4 d)
I believe you :)
In the meantime, this seems like a strict improvement to me.
Fri, Feb 24
At least the usage of xargs seems correct now!
You talked about "value semantic" earlier, but we have smart pointer, and you mentioned reference counting. Overall I have no idea what's going on if I were to use these class.
I assume that's fine, as long as people that are familiar with ISL are happy with it, but I'd hardly consider this a "C++ API" at this point, seeing how much it seems very intrinsically tied to "the ISL way" of handling its internal and does not seem to be expressed in a way that would match common C++ ownership management idioms.
What big difference with std::unique_ptr (and a custom deleter) justify to rewrite so much code for your smart pointer?
S.copy() returns a raw pointer refering to a copy of S.
Thu, Feb 23
I'd add also that isSafeToSpeculativelyExecute may be conservative right now, but it should improve with https://reviews.llvm.org/D20116
So it is even unclear to me which cases GVN hoist would be able to catch that are not in the realm of isSafeToSpeculativelyExecute.
So it seems that GVN-hoist tries to be "smarter" and hoist things for which isSafeToSpeculativelyExecute is returning false.
This is a nice aggressive optimization, but if it isn't safe (it seems it is not right now considering there are bugs in 4.0), we should *disable* this behavior in clang-4.0 and GVN-hoist should rely *only* on isSafeToSpeculativelyExecute as a legality criteria.
fix test to fail the verifier
Wed, Feb 22
Maybe the testing is obvious to someone use to COFF, in which case I'm just out-of-place :)
Otherwise I'd rather have comment explaining what every check is doing.
Tue, Feb 21
Mon, Feb 20
Sun, Feb 19
Sat, Feb 18
I think some documentation in the dev guide would be welcome!
Fri, Feb 17
The IR part (LangRef) seems OK to me.
The ELF part is out of my scope, if rafael or pcc already checked it, that's fine.
Can this metadata be freely dropped? Clarify this in LangRef please.
Thu, Feb 16
You need to update: http://llvm.org/docs/BranchWeightMetadata.html
Wed, Feb 15
Make sense. LGTM.
Tue, Feb 14
Mon, Feb 13
Adding @bruno to check on the module story here, I'm still not sure what is or isn't reasonable to do with modules!
Also note that @chandlerc in r290398 made clang adding "noinline" on every function at O0 by default, which seems very similar to what I'm doing here.
Sun, Feb 12
Sat, Feb 11
Do you have commit access?
LGTM overall. I trust you to have tested the windows part.
Fri, Feb 10
Thanks for the update. Good to know.
Thu, Feb 9
OK, thanks for the cleanup! I hate these infinitely long files :)
(I wouldn't bother with the svn attributes, but if it is trivial *shrug*)
cfg-commits not subscribed apparently? Is this supposed to be up for review?
Yeah ignore me please :)
Wed, Feb 8
Aren't we blocked on clang? Or did it already stop emitting the null attr for memcpy and cie.
Awesome! Thanks for running it, this is not surprising at all :)
Ping? This is hitting 4.0
Great: at least we increased the coverage :)
I thought after discussing this on IRC this had landed two weeks ago! Apparently I didn't click approve here...