Page MenuHomePhabricator
Feed All Stories

Today

cjdb added inline comments to D102037: [libcxx][views] Add drop_view..
Mon, Jun 21, 6:42 PM · Restricted Project
aqjune committed rT3e88eaf3f0f1: Fix SpecCPU2017's dependency from timeit-target to build-timeit-target (authored by aqjune).
Fix SpecCPU2017's dependency from timeit-target to build-timeit-target
Mon, Jun 21, 6:42 PM
aqjune closed D100734: Fix SpecCPU2017's dependency from timeit-target to build-timeit-target.
Mon, Jun 21, 6:42 PM
manas added inline comments to D103440: [WIP][analyzer] Introduce range-based reasoning for addition operator.
Mon, Jun 21, 6:41 PM · Restricted Project
rriddle added a comment to D104516: [mlir] Add a ThreadPool to MLIRContext and refactor MLIR threading usage.

did you consider shoving the existing threadpool stuff into a ManagedStatic? That is how we typically handle problems like this, they are destroyed on llvm_shutdown instead of at global deinit.

Mon, Jun 21, 6:40 PM · Restricted Project, Restricted Project
shchenz added a comment to D97049: [XCOFF][NFC] add DWARF section support in XCOFF object writer.

gentle ping

Mon, Jun 21, 6:40 PM · Restricted Project
shchenz added a comment to D97184: [XCOFF] support DWARF for XCOFF for object output.

gentle ping

Mon, Jun 21, 6:39 PM · Restricted Project
rriddle updated the diff for D104516: [mlir] Add a ThreadPool to MLIRContext and refactor MLIR threading usage.

update

Mon, Jun 21, 6:33 PM · Restricted Project, Restricted Project
bondhugula accepted D104614: [MLIR] Generalize detecting mods during slice computing.

LGTM - nice improvement and covers a useful case.

Mon, Jun 21, 6:31 PM · Restricted Project
hoy added inline comments to D104546: [CSSPGO][llvm-profgen] Handle reurn to external transition..
Mon, Jun 21, 6:29 PM · Restricted Project
rriddle added inline comments to D104516: [mlir] Add a ThreadPool to MLIRContext and refactor MLIR threading usage.
Mon, Jun 21, 6:28 PM · Restricted Project, Restricted Project
bondhugula requested changes to D104053: [MLIR] Correct memrefdataflow behavior in the presence of cast and other operations.
Mon, Jun 21, 6:27 PM · Restricted Project
estewart08 added reviewers for D104677: [OpenMP] Apply fix for isnan, isinf and isinfinite for amdgcn.: ronlieb, JonChesterfield, yaxunl, scchan, ashi1.
Mon, Jun 21, 6:25 PM · Restricted Project, Restricted Project
estewart08 requested review of D104677: [OpenMP] Apply fix for isnan, isinf and isinfinite for amdgcn..
Mon, Jun 21, 6:22 PM · Restricted Project, Restricted Project
wenlei added inline comments to D104546: [CSSPGO][llvm-profgen] Handle reurn to external transition..
Mon, Jun 21, 6:14 PM · Restricted Project
springerm requested review of D104676: [mlir][NFC] Move SubTensorOp and SubTensorInsertOp to TensorDialect.
Mon, Jun 21, 6:08 PM · Restricted Project
Quuxplusone added inline comments to D102037: [libcxx][views] Add drop_view..
Mon, Jun 21, 6:03 PM · Restricted Project
dblaikie added a comment to D104619: [clang] [WIP] Respect PrintingPolicy::FullyQualifiedName when printing a template-id.

Thanks for having a look!

This'll need a test case

Definitely. Do you have a suggestion for what test suite that should go into? I had a quick look but couldn't find anything that obviously exercised TypePrinter.

Mon, Jun 21, 6:02 PM · Restricted Project
hoy added inline comments to D104546: [CSSPGO][llvm-profgen] Handle reurn to external transition..
Mon, Jun 21, 5:59 PM · Restricted Project
MaskRay updated the diff for D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

Mention the one-time exception for .profraw compatibility

Mon, Jun 21, 5:57 PM · Restricted Project, Restricted Project, Restricted Project
nridge retitled D104619: [clang] [WIP] Respect PrintingPolicy::FullyQualifiedName when printing a template-id from [clang] [WIP] Fix for https://bugs.llvm.org/show_bug.cgi?id=50774 to [clang] [WIP] Respect PrintingPolicy::FullyQualifiedName when printing a template-id.
Mon, Jun 21, 5:56 PM · Restricted Project
nridge added a comment to D104619: [clang] [WIP] Respect PrintingPolicy::FullyQualifiedName when printing a template-id.

Thanks for having a look!

Mon, Jun 21, 5:55 PM · Restricted Project
hoy added inline comments to D104546: [CSSPGO][llvm-profgen] Handle reurn to external transition..
Mon, Jun 21, 5:50 PM · Restricted Project
dblaikie added a comment to D104619: [clang] [WIP] Respect PrintingPolicy::FullyQualifiedName when printing a template-id.

This'll need a test case & does the change pass all existing tests?
Also, the patch description could use more detail - it can refer to the bug for more context, but there should be enough detail in the patch title/description to understand the general purpose, etc. (and you can shorten the bug reference to "PR50774" rather than the whole bugs.llvm.org URL)

Mon, Jun 21, 5:49 PM · Restricted Project
dblaikie added a comment to D104601: [Preprocessor] Implement -fnormalize-whitespace..

One of the concerns I'd have, for instance (have you done some broad testing of these patches on sizable code bases?) is that it wouldn't surprise me if clang had some scalability bugs/issues with very long source lines - so it might be necessary to introduce some (arbitrary?) newlines to break up the code. Though I'm not sure - no need to do that pre-emptively, but might be good to have some data that indicates whether this might be a problem or not.

Mon, Jun 21, 5:47 PM · Restricted Project
dblaikie added a comment to D104663: [OpaquePtr] Remove checking pointee type for byval/preallocated type.

"arguably incorrect IR" isn't a phrase that is especially comforting. It is or isn't, and should have well defined behavior (including being explicitly undefined, if that's the case). Any chance we can get a clearer statement one way or the other?

Is this an alternative to that patch that kept getting committed/reverted that I think was related to trying to unify call site and function declaration parameter attributes? Might be worth mentioning that one in this patch description for history/context?

There are other attributes that don't seem to do this kind of "try the call instruction or the function declaration", do they? so I'd be a bit hesitant about this direction. Would like to get @rnk's thoughts on this too, perhaps.

As someone who has all the context, the commit message makes sense to me. In the change you allude to, we tried to stand on the principle that a call instruction should carry all of the ABI attributes it needs, but in practice, we found that IR producers, in particular instrumentation passes, usually don't set ABI attributes on call sites. It should be a semantics-preserving transform to rinse the target of a call instruction through an identity function, but in practice, that would break all these IR producers. It would make most direct calls indirect, and the ABI attributes would no longer be correct.

Mon, Jun 21, 5:46 PM · Restricted Project
dblaikie added a comment to D104466: NOT FOR REVIEW: proof-of-concept for building lib/linux/libclang_rt.profile-x86_64.a on a mac in the gn build.

If you create a review using arc's --draft (I think that's the name/spelling) it'll avoid sending email to the *-dev list, so you can create it and share it around without adding more email to the busy *-dev lists until you're ready to send it for review, btw.

Mon, Jun 21, 5:43 PM · Restricted Project
bondhugula accepted D104260: [mlir] Fix invalid handling of AllocOp symbolOperands by SimplifyAllocConst..

Thanks for fixing this. Some minor comments.

Mon, Jun 21, 5:42 PM · Restricted Project
dblaikie added a comment to D104551: Delay initialization of OptBisect.

What situation are you dealing with where OptBisect is created during global constructors? OptBisect uses ManagedStatic which only creates the object on first access, right? So maybe it'd be suitable to change the code so it doesn't try to access OptBisect during global construction?

Mon, Jun 21, 5:42 PM · Restricted Project
davidxl updated subscribers of D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

One time exception sounds fine to me.

Mon, Jun 21, 5:34 PM · Restricted Project, Restricted Project, Restricted Project
wenlei added inline comments to D104546: [CSSPGO][llvm-profgen] Handle reurn to external transition..
Mon, Jun 21, 5:29 PM · Restricted Project
holland11 requested review of D104675: [MCA] [In-order pipeline] Fix for 0 latency instruction causing assertion to fail..
Mon, Jun 21, 5:21 PM · Restricted Project
thegameg accepted D104078: [Remarks] Make memsize remarks report as an analysis, not a missed opportunity..

LGTM

Mon, Jun 21, 5:18 PM · Restricted Project
taolq added inline comments to D104654: [llvm][Inliner] Make PriorityInlineOrder lazily updated.
Mon, Jun 21, 5:17 PM · Restricted Project
tcanens added inline comments to D102037: [libcxx][views] Add drop_view..
Mon, Jun 21, 5:13 PM · Restricted Project
kazu added inline comments to D104654: [llvm][Inliner] Make PriorityInlineOrder lazily updated.
Mon, Jun 21, 5:08 PM · Restricted Project
efriedma added a comment to D104648: [Mem2Reg] Use poison instead of undef for read of uninitialized memory.

This is a big change because AFAIK clang intentionally encodes the value of uninitialized variables in a benign way.
For example: https://godbolt.org/z/EYjasxM1M is optimized to 0 even if it is undef ^ undef = undef. If poison is used, this won't work.

Mon, Jun 21, 5:05 PM · Restricted Project
rnk added a comment to D104663: [OpaquePtr] Remove checking pointee type for byval/preallocated type.

"arguably incorrect IR" isn't a phrase that is especially comforting. It is or isn't, and should have well defined behavior (including being explicitly undefined, if that's the case). Any chance we can get a clearer statement one way or the other?

Is this an alternative to that patch that kept getting committed/reverted that I think was related to trying to unify call site and function declaration parameter attributes? Might be worth mentioning that one in this patch description for history/context?

There are other attributes that don't seem to do this kind of "try the call instruction or the function declaration", do they? so I'd be a bit hesitant about this direction. Would like to get @rnk's thoughts on this too, perhaps.

Mon, Jun 21, 5:03 PM · Restricted Project
cjdb accepted D102037: [libcxx][views] Add drop_view..

Per offline discussion, I'm okay with this going ahead, provided we return to add __non_propagating_cache once I've landed it.

Mon, Jun 21, 5:01 PM · Restricted Project
arsenm added inline comments to D103408: Teach peephole optimizer to not emit sub-register defs.
Mon, Jun 21, 5:00 PM · Restricted Project
mehdi_amini added a comment to D95339: [RFC][test] Adapt debug-info lit framework for more general purposes - part 1.

Seems like there are only a few line changed other than the renaming? What kind of review is needed here? Just an agreement on the principle of moving these? I think it should be someone having a stake in the debug-info tests to give some sort of approval here.

Mon, Jun 21, 5:00 PM · Restricted Project
Meinersbur updated the summary of D104601: [Preprocessor] Implement -fnormalize-whitespace..
Mon, Jun 21, 4:53 PM · Restricted Project
MaskRay added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

Not keeping version compatibility for the raw profile format would actually make my life easier.
But I was CCed on the Linux kernel news https://www.phoronix.com/scan.php?page=news_item&px=Clang-PGO-For-Linux-Next so I paid additional care in this patch.

Mon, Jun 21, 4:53 PM · Restricted Project, Restricted Project, Restricted Project
Meinersbur added a comment to D104601: [Preprocessor] Implement -fnormalize-whitespace..

Some bikeshedding: Calls to AvoidConcat could be avoided by always inserting a space between tokens at the cost of making the output larger. In the current form, the flag could also be named -fminimize-whitespace.

Mon, Jun 21, 4:51 PM · Restricted Project
efriedma committed rG74909e4b6e9b: Rename MachineMemOperand::getOrdering -> getSuccessOrdering. (authored by efriedma).
Rename MachineMemOperand::getOrdering -> getSuccessOrdering.
Mon, Jun 21, 4:50 PM
efriedma closed D103338: Rename MachineMemOperand::getOrdering -> getSuccessOrdering..
Mon, Jun 21, 4:50 PM · Restricted Project
nickdesaulniers updated subscribers of D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.
Mon, Jun 21, 4:48 PM · Restricted Project, Restricted Project, Restricted Project
nickdesaulniers accepted D104667: Improve the diagnostic of DiagnosticInfoResourceLimit (and warn-stack-size in particular).
Mon, Jun 21, 4:48 PM · Restricted Project, Restricted Project
Meinersbur added a comment to D104601: [Preprocessor] Implement -fnormalize-whitespace..

This is probably more @aaron.ballman 's wheelhouse, but for my money this seems pretty problematic - will make quoted text in compiler diagnostics weird/difficult to read, etc.

Mon, Jun 21, 4:46 PM · Restricted Project
Quuxplusone added a comment to D102037: [libcxx][views] Add drop_view..

@ldionne wrote:

Here's what I ended up with that is not dependent on non_propagating_cache (we can go back afterwards and perhaps simplify some of this if we have non_propagating_cache)

Mon, Jun 21, 4:46 PM · Restricted Project
cjdb accepted D102809: [libcxx][ranges] Add `ranges::iter_swap`..
Mon, Jun 21, 4:41 PM · Restricted Project
mehdi_amini edited reviewers for D104260: [mlir] Fix invalid handling of AllocOp symbolOperands by SimplifyAllocConst., added: bondhugula; removed: mehdi_amini.
Mon, Jun 21, 4:40 PM · Restricted Project
vitalybuka committed rGac15a128d875: [NFC] Add getUnderlyingObjects test (authored by vitalybuka).
[NFC] Add getUnderlyingObjects test
Mon, Jun 21, 4:37 PM
vitalybuka closed D104585: [NFC] Add getUnderlyingObjects test.
Mon, Jun 21, 4:37 PM · Restricted Project
lei updated the diff for D104664: [PowerPC][NFC] Clean up builtin sema checks.

remove extra ";"

Mon, Jun 21, 4:35 PM · Restricted Project
jsji added inline comments to D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 21, 4:34 PM · Restricted Project
nickdesaulniers added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

I don't have much experience with the Linux build system, so apologies if this is naive, but why is it exactly that it can't statically link libprofile? This is something we do for the Darwin kernel (there's a stripped-down runtime target called libclang_rt.cc_kext_x86_64_osx.a the kernel links).

Mon, Jun 21, 4:33 PM · Restricted Project, Restricted Project, Restricted Project
rsmith added inline comments to D103088: [clang] pre-0388 array parm list initialization.
Mon, Jun 21, 4:33 PM
zoecarver updated the diff for D102809: [libcxx][ranges] Add `ranges::iter_swap`..

Add a modulemap entry.

Mon, Jun 21, 4:31 PM · Restricted Project
abidmalikwaterloo abandoned D104674: Made changes as per comments from the reviewers.
Mon, Jun 21, 4:29 PM · Restricted Project
naromero77 updated the diff for D104150: [test-suite] NIST Fortran Compiler Validation Suite..
  • More explicit comment about the purpose and usage of driver_run script.
Mon, Jun 21, 4:28 PM
Quuxplusone added inline comments to D102037: [libcxx][views] Add drop_view..
Mon, Jun 21, 4:28 PM · Restricted Project
efriedma committed rG8f3d16905d75: [ScalarEvolution] Ensure backedge-taken counts are not pointers. (authored by efriedma).
[ScalarEvolution] Ensure backedge-taken counts are not pointers.
Mon, Jun 21, 4:25 PM
efriedma closed D103656: [ScalarEvolution] Ensure backedge-taken counts are not pointers..
Mon, Jun 21, 4:25 PM · Restricted Project
jingham added a comment to D104488: Create synthetic symbol names on demand to improve memory consumption and startup times..

Are the Symbol ID's for unnamed symbols the same each time you read in a symbol file? While the unnamed_symbol symbol names are not significant, it would be good if you were crashing in __lldb_unnamed_symbol111 on one lldb run, you would also crash in the same unnamed symbol when you crashed again.

Yes, symbol IDs are consistent as they encode the UserID of the symbol as the number which will be the same on each run as long as the binary doesn't change. The UserID for synthetic symbols always start with the last valid actual symbol index from the main symbol table. So the numbers are just as good as they are before, they just don't start at 1 anymore, the start at the size of the actual symbol table.

Mon, Jun 21, 4:25 PM · Restricted Project
vsk added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Why/how does the Linux PGO patch "care" about the .profraw format?

Because the Linux kernel cannot statically link against compiler-rt; we MUST emulate the interfaces provided by compiler-rt within the Linux kernel's runtime.

One way to solve this is to use a matching version of llvm-profdata tool and export indexed profile data only.

I don't have much experience with the Linux build system, so apologies if this is naive, but why is it exactly that it can't statically link libprofile? This is something we do for the Darwin kernel (there's a stripped-down runtime target called libclang_rt.cc_kext_x86_64_osx.a the kernel links).

Does it use buffer API? It does not support value profiling.

There may be other reasons.

Mon, Jun 21, 4:25 PM · Restricted Project, Restricted Project, Restricted Project
dexonsmith added a comment to D104663: [OpaquePtr] Remove checking pointee type for byval/preallocated type.

I agree with @dblaikie that it'd be nice to get a clear answer for the behaviour (and documenting it in LangRef). Also, I wonder about ignoring the instruction attribute when the function declaration exists... seems like that is a step toward removing the instruction attribute entirely (maybe that's the right thing to do (not sure it is), but if so then it should be a high-level bit, not a side effect).

Mon, Jun 21, 4:23 PM · Restricted Project
davidxl added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Why/how does the Linux PGO patch "care" about the .profraw format?

Because the Linux kernel cannot statically link against compiler-rt; we MUST emulate the interfaces provided by compiler-rt within the Linux kernel's runtime.

One way to solve this is to use a matching version of llvm-profdata tool and export indexed profile data only.

I don't have much experience with the Linux build system, so apologies if this is naive, but why is it exactly that it can't statically link libprofile? This is something we do for the Darwin kernel (there's a stripped-down runtime target called libclang_rt.cc_kext_x86_64_osx.a the kernel links).

Mon, Jun 21, 4:23 PM · Restricted Project, Restricted Project, Restricted Project
aeubanks accepted D104625: [NewPM] Print passes with params when using "opt -print-passes".

lgtm

Mon, Jun 21, 4:21 PM · Restricted Project
abidmalikwaterloo added a comment to D104674: Made changes as per comments from the reviewers.

How can I remove this patch?

Mon, Jun 21, 4:21 PM · Restricted Project
abidmalikwaterloo updated subscribers of D104674: Made changes as per comments from the reviewers.

Yes, I just realized this as well.

Mon, Jun 21, 4:20 PM · Restricted Project
rsuderman committed rGad1a9d629b75: [mlir][tosa] Enable tosa.div for TosaMakeBroadcastable (authored by rsuderman).
[mlir][tosa] Enable tosa.div for TosaMakeBroadcastable
Mon, Jun 21, 4:19 PM
rsuderman closed D104157: [mlir][tosa] Enable tosa.div for TosaMakeBroadcastable.
Mon, Jun 21, 4:19 PM · Restricted Project
mehdi_amini added a comment to D104674: Made changes as per comments from the reviewers.

You need to keep everything in the same revision, you probably didn't amend the git commit locally and ended up creating a new revision.

Mon, Jun 21, 4:18 PM · Restricted Project
vsk added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Why/how does the Linux PGO patch "care" about the .profraw format?

Because the Linux kernel cannot statically link against compiler-rt; we MUST emulate the interfaces provided by compiler-rt within the Linux kernel's runtime.

Mon, Jun 21, 4:16 PM · Restricted Project, Restricted Project, Restricted Project
MaskRay updated the diff for D104667: Improve the diagnostic of DiagnosticInfoResourceLimit (and warn-stack-size in particular).

rebase

Mon, Jun 21, 4:15 PM · Restricted Project, Restricted Project
davidxl added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Why/how does the Linux PGO patch "care" about the .profraw format?

Because the Linux kernel cannot statically link against compiler-rt; we MUST emulate the interfaces provided by compiler-rt within the Linux kernel's runtime.

Mon, Jun 21, 4:14 PM · Restricted Project, Restricted Project, Restricted Project
abidmalikwaterloo requested review of D104674: Made changes as per comments from the reviewers.
Mon, Jun 21, 4:12 PM · Restricted Project
clayborg committed rG4181bfe6888f: Clarify the "env" launch configuration setting. (authored by clayborg).
Clarify the "env" launch configuration setting.
Mon, Jun 21, 4:11 PM
clayborg closed D104578: Clarify the "env" launch configuration setting..
Mon, Jun 21, 4:11 PM · Restricted Project
clayborg added a comment to D104488: Create synthetic symbol names on demand to improve memory consumption and startup times..

Are the Symbol ID's for unnamed symbols the same each time you read in a symbol file? While the unnamed_symbol symbol names are not significant, it would be good if you were crashing in __lldb_unnamed_symbol111 on one lldb run, you would also crash in the same unnamed symbol when you crashed again.

Mon, Jun 21, 4:08 PM · Restricted Project
shafik added inline comments to D104488: Create synthetic symbol names on demand to improve memory consumption and startup times..
Mon, Jun 21, 4:06 PM · Restricted Project
tra added a comment to D104505: [HIP] Defer operator overloading errors.

However, this does cause source level incompatibilities, i.e. CUDA code that passes nvcc does not pass clang. This patch somehow addresses that without compromising clang's more sophisticated __host__/__device__ overloading resolution capabilities.

Mon, Jun 21, 4:05 PM
nickdesaulniers accepted D104667: Improve the diagnostic of DiagnosticInfoResourceLimit (and warn-stack-size in particular).
Mon, Jun 21, 4:01 PM · Restricted Project, Restricted Project
NoQ accepted D104640: [clang][Analyzer] Track null stream argument in alpha.unix.Stream ..

Yes please!

Mon, Jun 21, 4:01 PM · Restricted Project
haowei updated the diff for D100139: [ifs][elfabi] Merge llvm-ifs/elfabi tools.
Mon, Jun 21, 4:00 PM · Restricted Project, Restricted Project
nickdesaulniers added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Why/how does the Linux PGO patch "care" about the .profraw format?

Mon, Jun 21, 4:00 PM · Restricted Project, Restricted Project, Restricted Project
MaskRay added inline comments to D104667: Improve the diagnostic of DiagnosticInfoResourceLimit (and warn-stack-size in particular).
Mon, Jun 21, 3:55 PM · Restricted Project, Restricted Project
zoecarver updated the diff for D102037: [libcxx][views] Add drop_view..

Apply a few remaining review comments.

Mon, Jun 21, 3:54 PM · Restricted Project
zoecarver added inline comments to D102037: [libcxx][views] Add drop_view..
Mon, Jun 21, 3:50 PM · Restricted Project
mcberg2017 added inline comments to D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 21, 3:48 PM · Restricted Project
MaskRay added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

On COFF, it would be more traditional to make these image-relative 32-bit offsets, so they would come out as .long __profc_foo@IMGREL. If you grep for IMGREL in test/, you can see various exception handling and RTTI tables use that structure.

I seem to recall this was also an issue for swift, which wanted label difference relocations.

The absolute function address (used by llvm-profdata to collect indirect call
targets) can be converted to relative as well, but is not done in this patch.

If we're doing a format change, we should think about this a bit. Previously I was considering removing this field altogether and trying to store this information in a new section, but I don't think it's worth it after D102818.

Mon, Jun 21, 3:46 PM · Restricted Project, Restricted Project, Restricted Project
eugenis added inline comments to D104463: [hwasan] print exact mismatch offset for short granules..
Mon, Jun 21, 3:45 PM · Restricted Project
jeroen.dobbelaere updated the diff for D69542: Full Restrict Support - single patch.
NOTE: (June 21, 2021) - full restrict is now again enabled by default - this update also contains a number of bugfixes (triggered by csmith) - llvm.noalias.copy.guard now uses a representation based on offsets for knowing where in memory noalias pointers are stored - the separate patches are not yet in sync with this version
Mon, Jun 21, 3:45 PM · Restricted Project
vkmr requested review of D104673: [LV, VP] Add CLI option to support predicated vectorization.
Mon, Jun 21, 3:43 PM · Restricted Project, Restricted Project
zoecarver updated the diff for D102037: [libcxx][views] Add drop_view..
  • Split up tests.
  • Use std::optional as a non-propagating cache (and add a test).
Mon, Jun 21, 3:43 PM · Restricted Project
davidxl added inline comments to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.
Mon, Jun 21, 3:40 PM · Restricted Project, Restricted Project, Restricted Project
vsk added a comment to D104556: [InstrProfiling] Make CountersPtr in __profd_ relative.

llvm has had a long-standing policy of keeping details of the .profraw format private between libprofile and InstrProfReader, opting to only commit to backwards-compat for the indexed .profdata format. I'm concerned that breaking with that practice would send the wrong message and commit llvm to maintaining an extra compat layer for a long time.

Mon, Jun 21, 3:33 PM · Restricted Project, Restricted Project, Restricted Project
dblaikie added a comment to D104668: [OpaquePtr] Handle addrspacecasts in InstCombine.

Any thoughts on the approach here? Is the !isOpaque() ? getElementType() : nullptr pattern something we should add a helper function for? Not sure how common this is going to be outside these cast transforms.

Or possibly PointerType::hasSameElementTypeAs(PointerType *) would make sense here.

I think PointerType::hasSameElementTypeAs(PointerType *) is good, we already have weird PointerType methods that are only for the opaque pointer transition, like PointerType::getWithSamePointeeType(), that aren't used in a lot of places and are to be deleted after.

Mon, Jun 21, 3:27 PM · Restricted Project
aeubanks added a comment to D104668: [OpaquePtr] Handle addrspacecasts in InstCombine.

Any thoughts on the approach here? Is the !isOpaque() ? getElementType() : nullptr pattern something we should add a helper function for? Not sure how common this is going to be outside these cast transforms.

Or possibly PointerType::hasSameElementTypeAs(PointerType *) would make sense here.

Mon, Jun 21, 3:21 PM · Restricted Project