- User Since
- Jul 2 2015, 4:56 AM (171 w, 5 d)
- Mailing List
Wed, Oct 10
Ugh, I put absolute paths in the test. I can fix it in a few hours, i had
to leave tge office early. otherwise you’re free to revert, sorry :-/
Wed, Oct 3
Alex - any update on this?
Mon, Sep 24
Mon, Sep 17
Sep 14 2018
The symptoms of a collision are just going to be you can’t debug the
program, so not very severe imo, especially since it would almost certainly
be resolved on the next incremental build
Sep 11 2018
Sep 4 2018
I think this particular problem was fixed a long time ago, likely
before I started working on SCEV. I think the relevant bit is here:
Aug 24 2018
Aug 16 2018
Yea this approach is fine I guess, but I’m not really a fan of gigantic
headers, so if it’s possible to use runtime polymorphism that seems better
to me. Otoh, I’m not the maintainer of the itanium demangler, so feel free
to ignore :)
Yea, I agree it’s mostly bugfixes. My hypothesis is that bugfixes are, on
average, pretty small and isolated, so it shouldn’t be too hard to port one
between the two even if they evolve separately. I’m curious what Erik
Jul 30 2018
Jul 20 2018
Or just doing nothing. It’s a private header and is only included in 1 TU.
In that fine, cxxabi-config is already included first
Jul 19 2018
Il 18/07/2018 20:52, David Li via Phabricator ha scritto:
Jul 13 2018
Jun 25 2018
The default is actually true on Windows I think. I know I always have to
manually set false.
Jun 11 2018
Jun 6 2018
I attached my test cases to this email.
May 31 2018
Don't wait on me until June 10th. I'm just skimming emails.
May 17 2018
May 10 2018
Apr 18 2018
Currently llvm-mca doesn't know how to resolve variant scheduling classes.
This problem mostly affects the ARM target.
This has been reported here: https://bugs.llvm.org/show_bug.cgi?id=36672
Apr 12 2018
Sorry, i'll stare at this again.
Apr 4 2018
Apr 3 2018
The MIR-Canon changes I've got coming soon will test createVirtualRegister but not createGenericVirtualRegister. It should be good enough for now I think.
Mar 29 2018
Could this bit be in a subsequent patch?
Right now im not exactly sure what the right way to go about this is because you have a vreginfos but they are named in much the way physregs are named. The named physregs have the map in MIParser. Still thinking of the right way to do this.
Mar 26 2018
Thanks for fixing that!
Mar 8 2018
Scott Linder via Phabricator <email@example.com> writes:
Is this from a bot? A bit more info would be helpful.
Jul 8 2016
Committed in r274912.
Jul 7 2016
anna marked an inline comment as done.
anna added a comment.
I don't see a patch on the list. Is this a problem with how you created the Phab review?
Sent patch separately to @dexonsmith. Response on patch:
isa works on references as well as pointers, so this can just beif (isa<FenceInst>(*BBI))
Agree with Duncan, but the llvm ref states that this implicit conversion comes at a cost, and at some point in time, may be reverted to mean what standard iterators are. So, I'll leave it as-is to avoid any fix-ups required in future.
> From: Rui Ueyama [firstname.lastname@example.org]
> Sent: 06 July 2016 18:24
> To: reviews+D21779email@example.com
> Cc: Davide Italiano; Rafael Ávila de Espíndola; Daniel Sanders; George Rimar; Sean Silva; Peter Collingbourne; llvm-commits; Tim Amini Golling
> Subject: Re: [PATCH] D21779: [LTO] Infer EKind/EMachine from Bitcode files
> On Tue, Jul 5, 2016 at 6:33 AM, Daniel Sanders <firstname.lastname@example.org<mailto:email@example.com>> wrote:
> dsanders added a subscriber: dsanders.
> Comment at: lld/trunk/ELF/InputFiles.cpp:559-562
> @@ +558,6 @@
> + Triple TheTriple(TripleStr);
> + bool Is64Bits = TheTriple.isArch64Bit();
> + if (TheTriple.isLittleEndian())
> + return Is64Bits ? ELF64LEKind : ELF32LEKind;
> + return Is64Bits ? ELF64BEKind : ELF32BEKind;
> Sorry for the late comment but Triple::isArch64Bit() doesn't correspond to ELF64 or ELF32 on MIPS. In particular, the triple used for the
> N32 ABI will return true for isArch64Bit() but this ABI uses ELF32 objects. If lld doesn't support the N32 ABI yet then the current code is
> ok for the remaining two ABI's (O32 and N64).
Ah, interesting. So N32 is a 32-bit ABI on 64-bit MIPS just like x32 ABI is on x86-64. Do you know how to determine whether it is ELF32 or ELF64 for such triple?
Jul 6 2016
You can mark the dependency in phabricator explicitly.
Jul 1 2016
Jun 27 2016
Jun 23 2016
If you don't mind, I'll add the test as a separate patch - we don't only want
>this specific test, we need tests for a bunch of sexts/zexts, like ARM has. >
[Demikhovsky, Elena] yes, you can add tests later. You can look at this revision as a test reference for X86
Intel Israel (74) Limited
Should we add overloads of the UTF8 conversion functions that accept
Jun 22 2016
lemon is mostly stable for me as well, but from time to time it has huge outliers (just speculating but maybe iOS throttles too much forking in some circumstances and that benchmark is at the edge of such throttling...)
Jun 13 2016
Sounds reasonable, but why not return a SubtargetFeatures?
dsanders created this revision.
dsanders added a subscriber: llvm-commits.
Herald added subscribers: sdardis, dsanders.
Right, my idea was to make loop-rotation more generic, thus enabling it on more complicated loops. It has some interesting corner cases though: for instance you can end up with several nested loops after rotating a single loop, but that's details, which we'll figure out in the process:)
Jun 10 2016
Sorry, though i clicked accept, i'll do so.
Jun 9 2016
On second thought i think i can actually make that work. I was thinking
you'd want a full blown interval tree, but it could be done using the same
walk I'm doing for writing. Since this is still a rare case, I'm fine
adding this, or not adding it, depending on what people prefer
Jun 8 2016
llvm-commits mailing list
Jun 6 2016
The renaming also makes the diff hard to read which I don't like either. I
will revert the renaming part.
Jun 2 2016
To elaborate a bit here: what's happening is that the update the list sees
is that you added the list and some reviewers, rather than an initial email
that contains the patch and your description of it so that people can scan
and look at patches without needing to go to phabricator.
I wonder if that's the right long term solution?
Jun 1 2016
llvm-commits mailing list
Although I like the change, I proposed something similar in http://lists.llvm.org/pipermail/llvm-dev/2016-May/099752.html and it was rejected.
Yeah, i don't either.
At this point, why don't we just have it own the caching walker and see
what it breaks.
May 31 2016
I know some of the issues are more a matter of personal preference than there being one true way. I just felt I should point out how I feel about them before people start writing them down as the one true way in the coding conventions :) I will of course not block the patches if we end up with a different style here.
May 30 2016
May 24 2016
You should probably note somewhere that assume is, in fact, readnone, we
just don't model it that way right now because it hasn't been changed to
use a real control dependency instead of a fake memory one.
This seems strange to me. For example, it breaks the otherwise very convenient:
$ clang -flto t.c -mllvm -some-internal-option-for-cc1
May 23 2016
llvm-commits mailing list
May 20 2016
Commit message typo: "depricated" should be "deprecated". Probably also
missing a "we" before [or use "it is" instead of] "are" in the last line.
May 14 2016
Can you investigate? So that we can have a correct implementation to propose instead.
May 11 2016
Thanks for taking on this work!
May 10 2016
I am fine with the change as is now after realizing that it does not change the behaviour of the subregister def case. So this change is fine!
May 5 2016
May 4 2016
I have no problem with the suggestion for the body, but please leave the
PR# bit in the subject. The PR is the least useful part of any subject
to me and I don't want to devote more space to it.
May 3 2016
The cpio command was dropped, but the format is still documented. The pax
command has to support it.
Comment at: ELF/DriverUtils.cpp:130
@@ +129,3 @@
+ OS << "000000"; c_dev
+ OS << "000000"; c_ino FIXME: should be unique
+ OS << "100664"; c_mode: C_ISREG | rw-rw-r--
If you are not going to fix it, it is not a fixme. Please change the comment to say that this should be unique, but since no one cares, we don't bother to set a unique number.
Apr 30 2016
Looking again, sorry for the long delay.
Apr 29 2016
I link Polly into my tools anyway, thus I'm in favor of changing the
Apr 27 2016
Apr 26 2016
Yes, I had already kicked off a build with your patches. Unfortunately it does not (at least no significantly) fix the issue and the test case is still running (for about 30 min now). My fix brings CT down to 2 min on an RA compiler.
Apr 25 2016
the only code change is
I'm still vaguely thinking we should be doing this in the frontend instead.
Apr 21 2016
This looks fine to me.
Apr 19 2016
If at all possible my preference would be:
- Require PRODIVE to be in an unambiguous location: inside an
- Produce DefinedSynthetic symbols for them so that the section is set
Apr 18 2016
I like this. I think you're missing updates to ValueMapper.cpp, so
llvm-link will crash if you leave this as is. For the code there,
it would be best to treat DITypeIndex exactly like MDString.
Apr 15 2016
Apr 14 2016
What about the subtargets themselves?
Apr 12 2016
It should already be fixed with r266151
Apr 7 2016
Apr 4 2016
I'll get to the rest of it, but for this Kit's question about O2: It's
preferred to test such things in the back end.
Mar 29 2016
Mar 27 2016
waiting for some one to accept :) Next time will push when I got a LGTM
Mar 24 2016
llvm-objdump wants to print the relocationOffset, so it calls RelocationRef::relocationOffset which calls ELFObjectFile::getRelocationOffset
I plan on adding this to sys::fs::file_status indeed, hopefully it is not too hard and this patch can wait for it.
This is copy-pasted-adapted from module pruning. I plan to refactor module pruning to use this at some point.
Mar 22 2016
Ah, this is why. LGTM.
We can make it conditional (with a clang flag?), but the desired behavior is to be able to decide on enabling/disabling the caching at link time, so any bitcode static archive distributed would need to have a hash embedded. The user friendly choice would be to enabled it by *default* when -flto=thing, and an explicit option to disable it.
Mar 21 2016
I would really ask the reviewers then if they will have a time to take a look on current aarch64 patches.
As updating them in every day lld changing world is really painfull sometimes.
Mar 15 2016
I'm concerned that fixing up GlobalOpt::deleteIfDead isn't enough
to avoid affecting CodeGen. There are other places in the
pipeline that will check Value::use_empty and/or have logic like
(or use directly) Function::isDefTriviallyDead. I think it's
important that the call to GV.removeDeadConstantUsers() has the
same effect on use-lists whether or not there's any debug info.
Moreover, GlobalOpt::deleteIfDead should have the same effect on
the entire module (including use-lists) regardless of debug info.
Mar 10 2016
In other words, this is a Elf_Rel file? Why do you need a new extension?
Mar 3 2016
Mar 2 2016
Static checking should be more robust long-term than relying on the dynamic checks with the assert. This however means that instructions not supported by a CPU should be explicitely marked as unsupported. I just pushed a commit which lets you mark InstRW expressions as unsupported (on top of the existing support to mark scheduling resources as unsupported or whole instructions as not appearing in the scheduler). So it should be easy to get the list from tablegen and exclude the missing instructions with something like:
Mar 1 2016
Can you fix it and then add code to LLVM? Adding code to LLVM allows people
to start creating object files in the odd "ELF" format. If a fix is not
made in a timely manner, it can be too late, and there's a chance that you
will have no choice other than supporting the old broken format.
Feb 26 2016
Feb 25 2016
That’ll work, though it should be a follow-up patch IMO.