Page MenuHomePhabricator
Feed Advanced Search

Mar 1 2016

llvm-commits added a comment to D17144: [AMDGPU] add AMDGPU target support to ELFObjectFile.h header.

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.

Mar 1 2016, 2:32 PM · Restricted Project

Feb 26 2016

llvm-commits added a comment to D15390: [ThinLTO] Launch importing backends in parallel threads from gold plugin.

Sure!

Feb 26 2016, 11:52 AM

Feb 25 2016

llvm-commits added a comment to D17475: MachineCopyPropagation: Catch copies of the form A<-B;A<-B.

That’ll work, though it should be a follow-up patch IMO.

Feb 25 2016, 11:45 AM

Feb 18 2016

llvm-commits added a comment to D16890: Fix PR25339: ARM Constant Island.

Hi Renato,

Feb 18 2016, 12:29 PM
llvm-commits added a comment to D17381: [Polly] [RFC] Fine-grain dependency analysis.

Build dependence information based on tagged access [1], e.g. for access:

MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 0]
    [N] -> { Stmt_for_body[i0] -> MemRef_b[6 + i0] };

we build tag it and get:

  • -> {[Stmt_for_body[i0] -> MemRef_a[]] -> MemRef_b[6 + i0] };

Why does it say MemRef_a and then MemRef_b ?

Feb 18 2016, 8:01 AM
llvm-commits added a comment to D14700: [mips][freebsd] Explicitly select PC-relative Personality and LSDA encodings..

It is unclear. Without knowing what is an actual gas/ld bug and what
is a reasonable limitation. Why is gas on linux rejecting indirect
pcrel? Is it at least acknowledge as a bug or are they aware of some
reason to reject it that we are not?

Feb 18 2016, 7:29 AM

Feb 15 2016

llvm-commits added a comment to D17241: [Polly] [Refactor] Move isl_ctx into Scop..

And if ScopInfo, DependencyInfo, IslAst are not free in order, it become a total mess ....

The pass manager should free IslAst, DependencyInfo, ... before ScopInfo
because they have a dependence on the ScopInfo results. Am I wrong?

Feb 15 2016, 7:28 AM

Feb 14 2016

llvm-commits added a comment to D16685: [tsan] Add a libc++ and lit testsuite for each ${TSAN_SUPPORTED_ARCH}..

Hi Hans,

Feb 14 2016, 4:45 PM

Feb 13 2016

llvm-commits added a comment to D17090: Fix LLVM's handling and detection of skylake and cannonlake CPUs.

We have
def : CannonlakeProc<"cannonlake">;
def : CannonlakeProc<"cnl">;

Feb 13 2016, 1:08 PM

Feb 9 2016

llvm-commits added a comment to D16885: [mips] Extend MipsAsmParser class to handle %got(sym + const) expressions.

LGTM

Feb 9 2016, 1:50 PM

Feb 2 2016

llvm-commits added a comment to D16835: InstCombine: Restrict computeKnownBits() on everyValue to OptLevel > 2.

Given that I don't have a single file in the whole llvm-testsuite where this transformation kicks in I wouldn't know how to start tuning it, so this patch simply pushes it to -O3 only.

Feb 2 2016, 6:46 PM

Jan 29 2016

llvm-commits added a comment to D16343: [BasicAA] Fix for missing must alias information.

Aside from the suspected UB (which I think Philip was right about,
see my comments inline below), this seems pretty obviously correct.

Jan 29 2016, 3:17 PM

Jan 25 2016

llvm-commits added a comment to D16530: [Polly] Introduce MemAccInst helper class; NFC.

+/ getValueOperand() and getPointerOperand() have different implementations in
+
/ LoadInst and StoreInst, but for most other members the implementation is
+/ identical. Those have been reproduced here to avoid useless branching to
+
/ identical implementations. If those implementations change in
+/// LoadInst/StoreInst, they must be changed here as well.

The last sentence alone guarantees problems. Duplication is (almost)
never a solution, especially not to save a call/jump/branch.

Jan 25 2016, 6:58 AM · Restricted Project

Jan 20 2016

llvm-commits added a comment to D15983: [GlobalISel] Add the proper cmake plumbing..

I believe r258379 should fix the problem.

Jan 20 2016, 5:59 PM

Jan 19 2016

llvm-commits added a comment to D16170: Registry initialization and linkage interactions.

This was already reverted due to those build failures.

Jan 19 2016, 12:00 PM

Jan 14 2016

llvm-commits added a comment to D16196: [test-suite] Add HOME to test env under lit.

We have found that some benchmarks will differ their performance based on the sizes of the environment variables. KS is a good example of this. Is there a better way to do this than populating the environment?

Jan 14 2016, 11:30 AM

Jan 11 2016

llvm-commits added a comment to D15703: [AVR] Add instruction definitions.

Ping.

Jan 11 2016, 12:13 AM

Jan 7 2016

llvm-commits added a comment to D15792: [AArch64] Turn off PredictableSelectIsExpensive on the Cortex-A57.

As far as I know, there is a mixup in reference results (rdar://23577619 rdar://23577619 for apple people). About the state of the LNTBased/—with-spec-ref you best ask Adam (CCed now).

Jan 7 2016, 1:37 PM
llvm-commits added a comment to D15798: Fix for Bug 24852 (crash with -debug -instcombine).

This is what I was fearing, it needs a real metadata attached.
Replacing the setDebugLoc with

Jan 7 2016, 10:00 AM

Jan 5 2016

llvm-commits added a comment to D15564: [ELF] - Optimize .eh_frame section: remove CIE if all FDEs referencing it were removed..

This patch does not particularly seem that it has a negative impact on
performance. It is reasonable to see only noises when comparing
performances before and after this patch. I believe Rafael left the note
for "just in case".

Jan 5 2016, 9:03 AM

Dec 24 2015

llvm-commits added a comment to D15480: [ELF][MIPS] Handle R_MIPS_HI16/LO16 relocations against _gp_disp symbol.

I thinking about that place last time.
May be we should add A argument to Target->relocateOne() ?
That can help to move that calculation to mips target, since it looks like absence of A inside relocateOne() is the only stoppable for doing that.

Currently S+A is a single parameter SA. Are you suggesting split that as two parameters, S and A?

I don't know if that makes code simpler, but it may worth trying.

Dec 24 2015, 3:22 AM · lld

Dec 22 2015

llvm-commits added a comment to D15693: [Polly] Take a PHI's incoming block as value user.
Dec 22 2015, 5:26 PM · Restricted Project

Dec 21 2015

llvm-commits added a comment to D15706: [Polly] Follow uses to create value MemoryAccesses.

+; TODO: Because of the phi with 3 incoming blocks, the loop for => outofscop is
+; not recognized as a loop, but as a non-affine subregion.

My guess would be this is because ScalarEvolution cannot create an
AddRec for %j. PHIs with multiple incoming blocks are generally not a
problem.

Dec 21 2015, 5:39 PM · Restricted Project

Dec 14 2015

llvm-commits added a comment to D14715: [Polly] Loop versioning after code generation (WIP).

Polly has to become a SCC pass anyway in order to make parallelization
valid (and allow inter-procedural analysis), thus this should not be a
concern. However, it seems odd and expensive to create a function only
to inline it again, so we might not want to do that.

Dec 14 2015, 8:23 AM · Restricted Project
llvm-commits added a comment to D14715: [Polly] Loop versioning after code generation (WIP).

To bad, I thought this was a nice solution.

Dec 14 2015, 7:25 AM · Restricted Project

Dec 10 2015

llvm-commits added a comment to D15328: [x86][avx512] more changes in intrinsics to be align with gcc format.

The stance I had always taken was that if clang and gcc both have a builtin
for something we match them. If clang doesn't need a builtin and gcc does
then we just don't match.

Dec 10 2015, 1:41 PM
llvm-commits added a comment to D15328: [x86][avx512] more changes in intrinsics to be align with gcc format.

To be compatible with GCC headers.

Dec 10 2015, 4:14 AM

Dec 2 2015

llvm-commits added a comment to D14687: Macro support in LLVM IR.

Sorry for the delay, was away all last week and haven't made it back
through my inbox yet. I just saw your ping to Alexey.

Dec 2 2015, 10:06 AM
llvm-commits added a comment to D15110: Fixed a bug in FP logic operation lowering on AVX-512.

I can't enable FABS before I fix ANDPS.
I can fix the ANDPS, but I don't know how to test it without FABS.

Dec 2 2015, 1:11 AM

Dec 1 2015

llvm-commits added a comment to D14473: [Verifier] Improve error for cross-module references.

Yup, LGTM.

Dec 1 2015, 9:23 AM

Nov 25 2015

llvm-commits added a comment to D14909: [X86][FMA] Optimize FNEG(FMUL) Patterns.

[I added this to Phab over an hour ago, but it hasn't made it to the list
afaict, so repeating with an email.]

Nov 25 2015, 10:01 AM

Nov 22 2015

llvm-commits added a comment to D14815: AVX-512: INSERT_SUBVECTOR optimization for i1 vectors.

Thank you for the review. I applied your comments.

What are the consequences of removing the HasVLX predicate?

VK legal types cover predicates. (If VK4 of VK2 are legal, the VLX is on)

Nov 22 2015, 11:14 AM

Nov 18 2015

llvm-commits added a comment to D14788: Fix bug 24715.

Looking a related code, it looks like we're failing to consider the
return value of the same function at it's other call site. Any interest
in fixing that too?

Nov 18 2015, 2:22 PM

Nov 17 2015

llvm-commits added a comment to D14480: Improve running time of getAsmSrcLocInfo.

David, did you also have a clang-format issue where it wasn't breaking
short if's? In the end how did you fix it? Yunlian seems to be having a
similar problem.

Nov 17 2015, 7:42 PM · Restricted Project

Nov 16 2015

llvm-commits added a comment to D14699: [GlobalOpt] Allow constant globals to be SRA'd.
Nov 16 2015, 8:08 AM

Nov 11 2015

llvm-commits added a comment to D14477: Remove unnecessary call to getAllocatableRegClass.

Hi Fiona,

Nov 11 2015, 5:08 PM

Nov 4 2015

llvm-commits added a comment to D14227: Add a new attribute: norecurse.

This should probably be raised on llvm-dev for boarder visibility.

Nov 4 2015, 9:46 AM

Oct 29 2015

llvm-commits added a comment to D13989: This removes the eating of the error in Archive::Child::getSize() when the characters in the size field in the archive header for the member is not a number..
Oct 29 2015, 5:11 PM

Oct 27 2015

llvm-commits added a comment to rL250085: [x86] Fix wrong lowering of vsetcc nodes (PR25080)..

Yes, this is OK.

Oct 27 2015, 10:12 AM · Restricted Project

Oct 22 2015

llvm-commits added a comment to D13708: AArch64: Disable the latency heuristic.

Thanks for benchmarking. TL;DR: I will change the policy only for Cyclone CPUs.

Oct 22 2015, 10:59 AM

Oct 21 2015

llvm-commits added a comment to D13949: [SLP] Treat SelectInsts as reduction values..

The patch looks okay. The problem of adding new instructions to the list of ‘reductions’ is that it increases the compile time of the SLP-vectorizer. Do you know how effective is this pattern? Do you know how many times it hits when you compile the llvm test suite?

Oct 21 2015, 10:16 AM

Oct 20 2015

llvm-commits added a comment to D13815: [ELF2] Add support for Gnu Hash section.

Rebased patch attached. I did a merge to review the patch.

Oct 20 2015, 10:56 AM · lld
llvm-commits added a comment to D13904: [ELF2] .shstrtab section implemented.

От: davide.italiano@gmail.com [davide.italiano@gmail.com] от имени Davide Italiano [davide@freebsd.org]
Отправлено: 20 октября 2015 г. 21:21
Кому: reviews+D13904+public+03561865476a15e1@reviews.llvm.org; George Rimar
Копия: Rafael Ávila de Espíndola; Davide Italiano; Rui Ueyama; llvm-commits
Тема: Re: [PATCH] D13904: [ELF2] .shstrtab section implemented

Oct 20 2015, 10:37 AM
llvm-commits added a comment to D13904: [ELF2] .shstrtab section implemented.

Oh yes, sorry, I am talking right about it !


От: davide.italiano@gmail.com [davide.italiano@gmail.com] от имени Davide Italiano [davide@freebsd.org]
Отправлено: 20 октября 2015 г. 21:01
Кому: reviews+D13904+public+03561865476a15e1@reviews.llvm.org; George Rimar
Копия: Rui Ueyama; Rafael Ávila de Espíndola; llvm-commits
Тема: Re: [PATCH] D13904: [ELF2] .shstrtab section implemented

Oct 20 2015, 10:09 AM

Oct 13 2015

llvm-commits added a comment to D13676: [Polly] Do not store scalar accesses in InstructionToAccess.

Meinersbur added a comment.

In http://reviews.llvm.org/D13676#265724, @grosser wrote:

> the patch looks generally fine. I am only slightly concerned your removal of the instruction to multiple read/write accesses. As it seems to not be strictly necessary, I wonder if we should rather leave the multi read/write accesses in?

This was one of the things I was most happy to simplify. No more memory management of lists in lists, or the difficulty to understand when a MemoryAccess is connected to an instruction.

With this patch but multiple accesses per instruction we would still
simplify both. All MemoryAccesses connected to an instruction would be
real (explicit) memory accesses (maybe) caused by that instruction.

Oct 13 2015, 6:14 AM · Restricted Project

Oct 9 2015

llvm-commits added a comment to D13547: Clear SectionSymbols in MCContext::Reset.

LGTM

Oct 9 2015, 7:20 AM

Oct 6 2015

llvm-commits added a comment to D13447: [AArch64]Fix bug in DAG combine for ADDV.

Yes. This occurs for all other ISD nodes, see what happens for example if
you try a v8i32 ADD.

Oct 6 2015, 9:01 AM

Oct 2 2015

llvm-commits added a comment to D13341: [Polly] Earlier creation of ScopStmt objects.
Oct 2 2015, 8:19 AM · Restricted Project

Sep 29 2015

llvm-commits added a comment to D13233: Combining invariant.group metadata bugfix.

If the GVN example is reasonable small, that would be fine. In general,
even complex scenarios can be reduced if care is taken to eliminate all
the redundant pieces.

Sep 29 2015, 10:41 AM

Sep 16 2015

llvm-commits added a comment to D12882: [SimplifyCFG] do not speculate fdiv by default; it's expensive (PR24818).

Can you give some context on specifically which speculative
optimizations you're hoping to avoid here? Are you only worried about
if-then-else-to-select from SimplifyCFG? What about cases like LICM?
Do you think those should hoist or not?

Sep 16 2015, 7:34 PM

Sep 14 2015

llvm-commits added a comment to D12843: [Polly] Merge IRAccess into MemoryAccess.

Meinersbur added a comment.
If you give me a definition of what the order is supposed to be, I can try to preserve it.

The order of non-scalar memory accesses in a statement should be the same as
the order of the corresponding memory access instructions in the BB. If
I understand you correctly that will not change, thus I'm good.

Sep 14 2015, 1:38 PM · Restricted Project
llvm-commits added a comment to D12843: [Polly] Merge IRAccess into MemoryAccess.

Meinersbur added a comment.

In http://reviews.llvm.org/D12843#245239, @llvm-commits wrote:

> > but their order might be different now.
>
>
> I always believed the order of (non-scalar) memory accesses in the SCoP
> statement corresponded to the order of memory instructions in the basic
> block. Was this true would not be anymore? If so I would like to keep the
> old order if possible. If it wasn't true before, this comment is void.

I think the order is undefined. Former TempScopInfo added IRAccesses to AccFuncMap (apart from the Functions set in buildAccessFunctions) depending on the order of iteration through the BBs, which is I think also undefined (but deterministic). Also its name is AccFuncSetType, indicating that it was never intended to make order matter. If this is wring, it should be documented.

Iterating over a BB has a defined order! I think we should not abandon
this for no good reason. At the latest the invariant hoisting patch will
make use of this (important for multiple ptr indirections).

Sep 14 2015, 10:29 AM · Restricted Project
llvm-commits added a comment to D12843: [Polly] Merge IRAccess into MemoryAccess.

but their order might be different now.

I always believed the order of (non-scalar) memory accesses in the SCoP
statement corresponded to the order of memory instructions in the basic
block. Was this true would not be anymore? If so I would like to keep the
old order if possible. If it wasn't true before, this comment is void.

Sep 14 2015, 9:18 AM · Restricted Project

Sep 10 2015

llvm-commits added a comment to D12758: Allow general loops with one latch.

I agree with most comments but I am kinda busy right now. I will come
back to this though.

Sep 10 2015, 10:31 AM

Sep 8 2015

llvm-commits added a comment to D12694: Merge TempScopInfo into ScopInfo.

LGTM

Sep 8 2015, 8:07 AM · Restricted Project

Sep 2 2015

llvm-commits added a comment to D12559: Fix IRBuilder CreateBitOrPointerCast for vector types.

Thanks, Evgeniy! Here is IR testcase:

Sep 2 2015, 3:22 PM

Aug 27 2015

llvm-commits added a comment to D12142: [PM/AA] Flesh out statistics to cover all the primary use cases of the old AliasAnalysisCounter pass..

I'm OK working on doing that (I actually started there), but it looked like
I'd need to do a fair amount of work and it will slow me down... How badly
do you want this now, and how much do you want to put it on my critical
path to moving forward? I don't have strong opinions here, I'm jsut looking
for the fastest way to make forward progress.

Aug 27 2015, 6:37 PM
llvm-commits added a comment to D12341: add unpredictable metadata type for control flow.
  • Original Message -----

From: "Sanjay Patel" <spatel@rotateright.com>
To: spatel@rotateright.com, chandlerc@gmail.com, kparzysz@codeaurora.org, hfinkel@anl.gov, "bob wilson"
<bob.wilson@apple.com>, kubastaszak@gmail.com
Cc: "kevin b smith" <kevin.b.smith@intel.com>, llvm-commits@lists.llvm.org
Sent: Thursday, August 27, 2015 6:38:39 PM
Subject: Re: [PATCH] D12341: add llvm.unpredictable intrinsic and lower it to metadata

spatel added a comment.

In http://reviews.llvm.org/D12341#234428, @hfinkel wrote:

> The problem with the intrinsic is that ... it confuses information
> about a value with information about a branch based on that value.

Do you see this being different than __builtin_expect() in Clang? If
it's information about the branch itself, do we have:

__builtin_unpredictable_if( MyCondition ) { ... }

If it's about the condition, then it looks like the existing
__builtin_expect:

if ( __builtin_unpredictable( MyCondition ) { ... }

...but we'll have to propagate the metadata from the condition value
to branches that use it in Clang?

Aug 27 2015, 4:56 PM
llvm-commits added a comment to D11678: [CodeGen] Fixes *absdiff* intrinsic: LangRef doc/test case improvement and corresponding code change.

Hi Shahid,

Aug 27 2015, 2:18 AM

Aug 26 2015

llvm-commits added a comment to D11612: [lld][ELF2] Apply relocations..

Rebased to trunk with all tests passing.

Aug 26 2015, 2:38 PM
llvm-commits added a comment to D12285: [LV] Switch to using canonical induction variables..

Sure. If we have a single, non canonical phi, and we create a canonical
one, then we'll end up with one phi with a constant add inside the loop.

Aug 26 2015, 11:22 AM
llvm-commits added a comment to D12166: Assign weights to edges to jump table when lowering switch statement..

Hi Hans

Aug 26 2015, 10:53 AM
llvm-commits added a comment to D12149: [AArch64] Turn on by default interleaved access vectorization.

Hi Gerolf,

Aug 26 2015, 9:15 AM
llvm-commits added a comment to D12269: Add a pass to lift aggregate into allocas so SROA can get rid of them..

I'm really not sure this is the right approach, despite liking a number of
aspects and being optimistic early-on. But I've not yet had time to put my
thoughts here into a nice cohesive email. I'll do that as soon as I can to
help resolve this.

Aug 26 2015, 2:12 AM

Aug 25 2015

llvm-commits added a comment to D12282: [MachO] Introduce MinVersion API.

r245963.

Aug 25 2015, 11:29 AM

Aug 24 2015

llvm-commits added a comment to D12298: Revert two bad commits..

Generally we try not to send out code reviews unless there's an expectation
that review is required/desired.

Aug 24 2015, 4:37 PM

Aug 21 2015

llvm-commits added a comment to D12084: [llvm-readobj] Add support for MachO DataInCodeDataCommand.

I went forward and commited this as r245732.
If anybody has other opinions about the lexicon -- we can change it later.
I'll start converting macho-dump tests soon so this feature will get tested.

Aug 21 2015, 1:31 PM

Aug 20 2015

llvm-commits added a comment to D12213: [SLP] Propagate 'nontemporal' attribute into vectorized instructions..

LGTM!

Aug 20 2015, 1:48 PM
llvm-commits added a comment to D11710: [DSE] Enable removal of lifetime intrinsics in terminating blocks.

Interesting. Definitely not an interaction I saw coming. Do you have
any thoughts on how to address? I'm assuming we need to tweak the
inliner's logic here, but I'm not familiar enough with the code in
question to give good suggestions.

Aug 20 2015, 9:56 AM
llvm-commits added a comment to D11933: Extending debug ranges.
Aug 20 2015, 9:42 AM

Aug 19 2015

llvm-commits added a comment to D12149: [AArch64] Turn on by default interleaved access vectorization.

I’m not sure about this LG and have a number of questions:

Aug 19 2015, 1:02 PM
llvm-commits added a comment to D11933: Extending debug ranges.
Aug 19 2015, 9:15 AM

Aug 18 2015

llvm-commits added a comment to D10555: [X86] Replace avx2.pbroadcast intrinsics with native IR..

Hi all,

Aug 18 2015, 10:41 AM

Aug 17 2015

llvm-commits added a comment to D12029: [lld] LinkDriver, lld-link: introduce shim..

I don't think a pure aliasing or wrapping approach will work long term.
There are things like -rdynamic that we want to support but they have no
link.exe analogue.

Aug 17 2015, 8:28 PM
llvm-commits added a comment to D11994: Doxygen: add build option to use svg instead of png files for graphs.
Aug 17 2015, 6:24 PM
llvm-commits added a comment to D12029: [lld] LinkDriver, lld-link: introduce shim..
Aug 17 2015, 5:47 PM
llvm-commits added a comment to D11933: Extending debug ranges.
Aug 17 2015, 10:35 AM
llvm-commits added a comment to D12072: [FIX] Restrict the AST build with the assumed context.
Aug 17 2015, 2:33 AM

Aug 14 2015

llvm-commits added a comment to D11977: Use modulo semantic to generate non-integer-overflow assumptions.

Attached 2 json files, without_overflow_assumptions.json
and with_overflow_assumptions.json. The former is a recent upstream
commit with (I think) 4 lnt runs, the latter the current integer
wrapping patch applied to the upstream head with 2 lnt runs.

Aug 14 2015, 2:12 AM

Aug 13 2015

llvm-commits added a comment to D11612: [lld][ELF2] Apply relocations..

llvm-mc

This is testing explicit relocations. Assembly doesn't represent specific relocations, and it's not obvious what gets generated.

Aug 13 2015, 6:03 PM
llvm-commits added a comment to D11530: RFC: LoopEditor, a high-level loop transform toolkit.

Hi Adam,

Aug 13 2015, 12:18 AM