Page MenuHomePhabricator
Feed Advanced Search

Fri, Nov 20

jdoerfert added a comment to D91746: [libomptarget] Fail on missing symbols in device image.

Manually, yeah - jury rig clang to elide one symbol, rename the one in deviceRTL. Automatically? Not sure it's worth the setup.

Fri, Nov 20, 6:53 AM · Restricted Project
jdoerfert added a comment to D91746: [libomptarget] Fail on missing symbols in device image.

This makes sense, can we test this?

Fri, Nov 20, 5:54 AM · Restricted Project

Thu, Nov 19

jdoerfert accepted D91644: [OPENMP]Honor constantness of captured variables..
Thu, Nov 19, 6:26 PM · Restricted Project
jdoerfert added inline comments to D83281: [OpenMP] Allow traits for the OpenMP context selector `isa`.
Thu, Nov 19, 4:55 PM · Restricted Project, Restricted Project
jdoerfert accepted D91805: [OPENMP]Use the real pointer value as base, not indexed value..
Thu, Nov 19, 11:17 AM · Restricted Project
jdoerfert requested changes to D91556: Add capturing of parameters to pass to omp::parallel.

It doesn't make sense to do this here. The OpenMPIRBuilder is used in other places that have the same problem, especially OpenMPOpt. The logic needs to be part of the OpenMPIRBuilder.

Thu, Nov 19, 7:58 AM · Restricted Project
jdoerfert accepted D87946: [OpenMP] Add Location Fields to Libomptarget Runtime for Debugging.

Still LGTM

Thu, Nov 19, 7:54 AM · Restricted Project, Restricted Project, Restricted Project

Tue, Nov 17

jdoerfert added a comment to D91649: [AA] Split up LocationSize::unknown().

The test changes in cs-cs.ll in particular illustrate a previously clearly incorrect AA result: We were effectively assuming that argmemonly functions were only allowed to access their arguments after the passed pointer, but not before it. I'm pretty sure that this was not intentional, and it's certainly not specified by LangRef that way.

Tue, Nov 17, 5:27 PM · Restricted Project
jdoerfert added inline comments to D17993: [CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments..
Tue, Nov 17, 3:49 PM · Restricted Project, Restricted Project
jdoerfert added inline comments to D17993: [CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments..
Tue, Nov 17, 3:18 PM · Restricted Project, Restricted Project
jdoerfert accepted D91660: [OPENMP]Fix PR48174: compile-time crash with target enter data on a global struct..

LGTM

Tue, Nov 17, 2:25 PM · Restricted Project
jdoerfert added a comment to D89898: changing OMP rtl to use shared memory instead of env variable .

I wonder if due to abnormal termination of the process some files might not have a chance to be removed? In that case I understand that the 0666 in shm_open(..., O_CREAT, ...) will still honour umask (commonly it is 0022 so the files are created with rw-r--r--). Was the intent of this change to create a rw-rw-rw- file instead? That would avoid the EACCES, in case another user hits the same file because it wasn't removed. Not sure if this is the intended behaviour here.

Tue, Nov 17, 10:44 AM · Restricted Project
jdoerfert accepted D91627: [OPENMP] Fix PR47999: correctly map implicit firstprivates in outer tasks..

LGTM

Tue, Nov 17, 8:20 AM · Restricted Project

Mon, Nov 16

jdoerfert accepted D17993: [CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments..

LGTM, thx!

Mon, Nov 16, 3:49 PM · Restricted Project, Restricted Project
jdoerfert added a comment to D17993: [CodeGen] Apply 'nonnull' and 'dereferenceable(N)' to 'this' pointer arguments..

Please modify the commit subject and add a proper message.

Mon, Nov 16, 3:31 PM · Restricted Project, Restricted Project
jdoerfert resigned from D37821: [SimplifyCfg] Don't sink loads/stores to geps of allocas.
Mon, Nov 16, 12:10 PM
jdoerfert requested changes to D91428: Add support for multiple program address spaces.

I'll be on the lookout for the RFC. There, and in an updated commit message, you have to provide more details.

Mon, Nov 16, 8:26 AM · Restricted Project, Restricted Project

Fri, Nov 13

jdoerfert accepted D90473: [OpenMP][Tool] Add Archer option to disable data race analysis for sequential part.
Fri, Nov 13, 2:12 PM · Restricted Project
jdoerfert added inline comments to D91417: [IRGen] Add !annotation metadata for auto-init stores..
Fri, Nov 13, 1:57 PM · Restricted Project, Restricted Project

Thu, Nov 12

jdoerfert added inline comments to D91383: [BasicAA] Make alias GEP positive offset handling symmetric.
Thu, Nov 12, 1:40 PM · Restricted Project
jdoerfert added inline comments to D91383: [BasicAA] Make alias GEP positive offset handling symmetric.
Thu, Nov 12, 1:13 PM · Restricted Project
jdoerfert accepted D91376: [OPENMP]Fix PR47790: segfault in frontend while parsing Objective-C with OpenMP..

LGTM

Thu, Nov 12, 12:35 PM · Restricted Project
jdoerfert added a reviewer for D91370: [OPENMP]Fix PR48076: Check map types array before accessing its front.: jhuber6.
Thu, Nov 12, 11:31 AM · Restricted Project
jdoerfert accepted D91188: Add !annotation metadata and remarks pass..

LGTM

Thu, Nov 12, 9:34 AM · Restricted Project
jdoerfert accepted D90758: [OpenMP][Tool] Update archer to accept new OpenMP 5.1 enum values.

LGTM, thanks!

Thu, Nov 12, 6:47 AM · Restricted Project

Wed, Nov 11

jdoerfert accepted D91085: [OpenMP] Begin Adding OpenMP Tool to Gather OpenMP Information.

LG

Wed, Nov 11, 4:59 PM · Restricted Project
jdoerfert accepted D91085: [OpenMP] Begin Adding OpenMP Tool to Gather OpenMP Information.

I think this is a good start, we can improve on it as we go. Please check if we need license information in the python files, e.g., check other python files. Also, one nit below.

Wed, Nov 11, 4:21 PM · Restricted Project
jdoerfert accepted D91280: [OpenMP][Tests] Fix compiler warnings in OpenMP tests.

LGTM thx!

Wed, Nov 11, 9:58 AM · Restricted Project

Tue, Nov 10

jdoerfert added inline comments to D91188: Add !annotation metadata and remarks pass..
Tue, Nov 10, 5:24 PM · Restricted Project
jdoerfert added a comment to D91188: Add !annotation metadata and remarks pass..

Makes sense to me.

Tue, Nov 10, 2:38 PM · Restricted Project
jdoerfert added a comment to D77609: [OpenMP] Added the support for unshackled task in RTL.

Some variables are only defined when the MACRO is enabled. I have changed some code to make it more readable and less complex.

As I said before, I don't see the point in omitting declarations. It just increases our testing surface for no real benefit. If you don't use this but have two more functions and a few declarations, all of which you don't use, you really don't pay a price in the big scheme of things.

What is a right way to detect whether a CMake variable or macro is defined?

In C/C++ (#ifdef) or in CMake (idk)?

The point is, our test cases are not run by CMake, so it cannot detect whether we define any variable.

Tue, Nov 10, 9:38 AM · Restricted Project
jdoerfert added a comment to D90758: [OpenMP][Tool] Update archer to accept new OpenMP 5.1 enum values.

What do you think about adding assertions or warnings for the cases that "should not happen"?

Tue, Nov 10, 7:25 AM · Restricted Project
jdoerfert accepted D91117: [Flang][Docs] Fix warnings when building docs..

LGTM

Tue, Nov 10, 7:22 AM · Restricted Project

Mon, Nov 9

jdoerfert added inline comments to D91117: [Flang][Docs] Fix warnings when building docs..
Mon, Nov 9, 8:52 PM · Restricted Project
jdoerfert accepted D91010: [BasicAA] Remove checks for GEP decomposition limit reached.

LGTM

Mon, Nov 9, 12:58 PM · Restricted Project
jdoerfert requested changes to D90290: [LoopInterchange] Prevent Loop Interchange for non-affine value store to affine access.

I don't understand the reasoning here, even if I assume for a moment the underlying idea is sound, the patch doesn't seem to be.
For example, it will only check stores of a certain shape, why can we store anything to non-affine accesses? What about non-store accesses?
This looks too much like ad-hoc pattern matching.

Mon, Nov 9, 12:51 PM · Restricted Project
jdoerfert added a comment to D91027: [BasicAA] Generalize base offset modulus handling.

Yes, now I get it. I think I was also missing the "V2 has offset 0" precondition.

Mon, Nov 9, 10:17 AM · Restricted Project
jdoerfert added a comment to D91010: [BasicAA] Remove checks for GEP decomposition limit reached.

OK, this sounds sane to me. Can we create a test though?

Mon, Nov 9, 9:10 AM · Restricted Project
jdoerfert added a comment to D90103: Add OpenMP for optimization.

reverse ping. Also, I left comments in the original review that have not been addressed.

Mon, Nov 9, 8:24 AM · Restricted Project, Restricted Project
jdoerfert added a comment to D90473: [OpenMP][Tool] Add Archer option to disable data race analysis for sequential part.

Can we somehow test this?

Mon, Nov 9, 8:22 AM · Restricted Project
jdoerfert added inline comments to D90758: [OpenMP][Tool] Update archer to accept new OpenMP 5.1 enum values.
Mon, Nov 9, 8:20 AM · Restricted Project
jdoerfert accepted D90752: [OpenMP][OMPT] Update the omp-tools header file to reflect 5.1 changes.

LGTM, one nit below.

Mon, Nov 9, 8:19 AM · Restricted Project
jdoerfert added a comment to D91085: [OpenMP] Begin Adding OpenMP Tool to Gather OpenMP Information.

Do we need the ++ version, it looks the same.

Mon, Nov 9, 8:18 AM · Restricted Project
jdoerfert added inline comments to D90909: [OpenMPOpt][WIP] Expand parallel region merging.
Mon, Nov 9, 8:10 AM · Restricted Project
jdoerfert added a comment to D90943: Fix for Bug25668. Clang is asserting when lookup argument is a class that hasn't been instantiated..

FWIW, as with the code, the test should probably live in /OpenMP instead.

Mon, Nov 9, 8:07 AM
jdoerfert added inline comments to D91038: [LoopIdiom] Introduce 'left-shift until bittest' idiom.
Mon, Nov 9, 8:02 AM · Restricted Project
jdoerfert added a comment to D91010: [BasicAA] Remove checks for GEP decomposition limit reached.

I would assume the "leftover" parts could just as well be part of the base, right? We'd still run the same logic, wouldn't we?

Mon, Nov 9, 7:53 AM · Restricted Project
jdoerfert updated subscribers of D91002: [OpenMP] Fix asm for arm64-darwin.

@kiranchandramohan @DavidTruby @richard.barton.arm Could you test this patch and D88252 as described above? I don't have proper setup but you should be able to test all system of interest I assume.

Mon, Nov 9, 7:51 AM · Restricted Project
jdoerfert accepted D90830: [OpenMPIRBuilder] Implement CreateCanonicalLoop..

LGTM. We will make progress based on this as we go. Thanks :)

Mon, Nov 9, 7:47 AM · Restricted Project
jdoerfert added inline comments to D91027: [BasicAA] Generalize base offset modulus handling.
Mon, Nov 9, 7:43 AM · Restricted Project
jdoerfert added a comment to D90670: Simplifying memory globalization from the front end to move optimizations to the middle end..

Let's start by adding an updated test to this so we can see how the result looks.

Mon, Nov 9, 7:00 AM · Restricted Project, Restricted Project, Restricted Project
jdoerfert accepted D91075: [clang] Fix ForStmt mustprogress handling.

LGTM

Mon, Nov 9, 6:57 AM · Restricted Project

Fri, Nov 6

jdoerfert added a comment to D90962: [OpenMP] Fix possible NULL dereferences.

So, we had asserts, right? And this code is "copied" a few times all over. What I try to say is, I'm not sure this is a "fix" and it is sufficient.

KMP_DEBUG_ASSERT is no-op in release build. Thus this patch adds missed checks in order to make static code analyzers happy.
In real codes we haven't encountered broken source location string, but in theory this might happen and then NULL pointer could be dereferenced.

I don't think we should generally "make the static code analyzers happy" by swapping out reasonable asserts with some if-then-else logic for a case we don't expect to happen.
It also doesn't help us in any of the other copies of this code.

I am still not sure I got your comment, let me clarify.

What do you mean by "reasonable asserts"? As I pointed out the KMP_DEBUG_ASSERT is no-op in release build used by most people.
So there were no asserts in this code in the release build at all, which is potentially dangerous, as any call to strchr can return NULL.
We replaced possible NULL dereferencing crash with giving a bit lesser info to the tool, that sounds pretty reasonable to me.
What's wrong here?

Fri, Nov 6, 3:35 PM · Restricted Project
jdoerfert added a comment to D90962: [OpenMP] Fix possible NULL dereferences.

So, we had asserts, right? And this code is "copied" a few times all over. What I try to say is, I'm not sure this is a "fix" and it is sufficient.

KMP_DEBUG_ASSERT is no-op in release build. Thus this patch adds missed checks in order to make static code analyzers happy.
In real codes we haven't encountered broken source location string, but in theory this might happen and then NULL pointer could be dereferenced.

Fri, Nov 6, 12:15 PM · Restricted Project
jdoerfert added a comment to D90962: [OpenMP] Fix possible NULL dereferences.

So, we had asserts, right? And this code is "copied" a few times all over. What I try to say is, I'm not sure this is a "fix" and it is sufficient.

Fri, Nov 6, 11:48 AM · Restricted Project
jdoerfert accepted D90954: [OpenMP] Fix potential division by 0.

LGTM

Fri, Nov 6, 9:14 AM · Restricted Project
jdoerfert added inline comments to D90909: [OpenMPOpt][WIP] Expand parallel region merging.
Fri, Nov 6, 9:12 AM · Restricted Project

Thu, Nov 5

jdoerfert added a comment to D90830: [OpenMPIRBuilder] Implement CreateCanonicalLoop..

Some minor comments, overall this looks reasonable. I'll let someone else take a look.

Thu, Nov 5, 2:31 PM · Restricted Project
jdoerfert added inline comments to D89959: UBSAN: emit distinctive traps in trapping mode.
Thu, Nov 5, 1:58 PM · Restricted Project
jdoerfert accepted D90802: [OpenMP] [DOCS] Update OMP5.1 feature status table [NFC].

LGTM

Thu, Nov 5, 1:53 PM · Restricted Project
jdoerfert added a comment to D90688: [CaptureTracking] Avoid overly restrictive dominates check.

One thing I thought about: Do we have a test where the "nocapture use" has a second operand that captures?

Thu, Nov 5, 12:26 PM · Restricted Project

Wed, Nov 4

jdoerfert added inline comments to D90802: [OpenMP] [DOCS] Update OMP5.1 feature status table [NFC].
Wed, Nov 4, 3:36 PM · Restricted Project

Tue, Nov 3

jdoerfert added a comment to D90704: [OpenMP] target nested `use_device_ptr() if()` and is_device_ptr trigger asserts.

I do not understand the commit message. Can you try to make it clearer?

Tue, Nov 3, 12:06 PM · Restricted Project, Restricted Project
jdoerfert added a comment to D90275: [clang][IR] Add support for leaf attribute.

Hi,

Naming is a hard thing... I have no special preference. :/

However, I'd like to understand the details of this attribute.

Would LTO be affected because leaf is guaranteed to untouch the current translation unit only?

// a.c
int x;
void f1() {
  f2();
}
void g() { x = 3; }

// b.c
void f2() {
  leaf();
}

// leaf.c
attribute((leaf)) void leaf() {
  g();
}

IIUC this program is okay because g() and the caller of leaf() are in different TUs.
But, let's assume that a.c and b.c are LTO-ed, and leaf.c is separately compiled.
If LTO merges a.c and b.c into the same module, the two TUs cannot be distinguished anymore; either leaf should be dropped, or LTO should somehow conceptually keep two TUs.
Would it be a valid concern? Then I think it should be mentioned.

Tue, Nov 3, 11:20 AM · Restricted Project, Restricted Project
jdoerfert accepted D90688: [CaptureTracking] Avoid overly restrictive dominates check.

LGTM. Attributor does the same with an "optimistic" version of reachability via AAReachability. I guess that is why no test there changed.

Tue, Nov 3, 9:01 AM · Restricted Project
jdoerfert updated subscribers of D90275: [clang][IR] Add support for leaf attribute.

The more I think about it, the more I think we should never create a leaf/nocallback definition. Only declarations should carry that attribute.

Tue, Nov 3, 8:28 AM · Restricted Project, Restricted Project

Mon, Nov 2

jdoerfert accepted D86841: [clang] Add mustprogress and llvm.loop.mustprogress attribute deduction.

LGTM, some nits

Mon, Nov 2, 5:18 PM · Restricted Project
jdoerfert added inline comments to D90275: [clang][IR] Add support for leaf attribute.
Mon, Nov 2, 2:03 PM · Restricted Project, Restricted Project
jdoerfert updated subscribers of D90529: Allow nonnull attribute to accept poison.

Fair. Though, I think we want to produce poison for one set of attributes for which the name "value attribute" was not well chosen.
So far, the things I think should produce poison not UB are:

(pure) value attributes:

  • nonnull
  • align
  • [used_bits] (not existing yet)

(context) value attributes:

  • dereferenceable
  • dereferenceable_or_null
  • [object_size] (as proposed on the list)

WDYT?

In case of dereferenceable, my opinion is still slightly different: it represents the property of the memory at the point.
If a memory block is freed, a same pointer value won't be dereferenceable after the deallocation.

Mon, Nov 2, 1:56 PM · Restricted Project
jdoerfert accepted D90632: [OpenMP] Add ident_t flags for compiler OpenMP version.

LGTM

Mon, Nov 2, 1:30 PM · Restricted Project
jdoerfert added a comment to D89671: [LLVM][OpenMP] Adding support for OpenMP sections construct in OpenMPIRBuilder.

The diff is against the previous version, not against a base version in LLVM. I usually squash all changes into a single commit but there are other ways too.

Mon, Nov 2, 7:34 AM · Restricted Project, Restricted Project
jdoerfert added a comment to D90529: Allow nonnull attribute to accept poison.

Agreed. So let's go with 1) and we change all the "value attribute" semantics to produce poison on violation.
We have noundef for the UB case, that was the reason I wanted it ;)

Thank you, glad to hear that moving towards the change!

Mon, Nov 2, 7:32 AM · Restricted Project

Sun, Nov 1

jdoerfert added a comment to D90566: [LazyCallGraph] Build SCCs of the reference graph in order.

I stumbled over the difference before, this makes sense to me.

Sun, Nov 1, 10:51 AM · Restricted Project
jdoerfert added inline comments to D90519: [Libomptarget][NFC] Move global Libomptarget state to a struct.
Sun, Nov 1, 10:26 AM · Restricted Project
jdoerfert added a comment to D90529: Allow nonnull attribute to accept poison.

I think we have two choices here:

  1. Don't raise UB when "value attributes" are passed a "wrong value", e.g., null for a nonnull attribute, but make the value poison. Use nonull + noundef to make it UB.
  2. Make all "value attributes" accept poison without raising UB.

If we don't do 1), we should talk about 2) before we make nonnull special.

It seems these two choices are entangled.
If f(nonnull poison) is okay (in other words, not UB), then f(nonnull null) shouldn't be UB as well.
The reason is that poison can be folded into null in any time.
For example, when inbounds is stripped from v = gep inbounds ..., v can be transformed from poison to null pointer.

Sun, Nov 1, 10:25 AM · Restricted Project

Sat, Oct 31

jdoerfert requested changes to D90529: Allow nonnull attribute to accept poison.

I think we have two choices here:

  1. Don't raise UB when "value attributes" are passed a "wrong value", e.g., null for a nonnull attribute, but make the value poison. Use nonull + noundef to make it UB.
  2. Make all "value attributes" accept poison without raising UB.
Sat, Oct 31, 3:35 PM · Restricted Project

Fri, Oct 30

jdoerfert accepted D90487: [CFG] Replace hardcoded max BBs explored as CL option. NFC..

LGTM, one nit though

Fri, Oct 30, 11:15 AM · Restricted Project
jdoerfert accepted D90391: [gvn] PRE needs to skip convergent intrinsics/calls..

LGTM

Fri, Oct 30, 8:14 AM · Restricted Project
jdoerfert accepted D90281: [FileCheck] Report missing prefixes when more than one is provided..

LGTM

Fri, Oct 30, 7:09 AM · Restricted Project

Thu, Oct 29

jdoerfert committed rG30e818db91a4: [OpenMP][Docs] Structure and content for the OpenMP documentation (authored by jdoerfert).
[OpenMP][Docs] Structure and content for the OpenMP documentation
Thu, Oct 29, 11:32 PM
jdoerfert committed rG2c58fa54e8fa: [OpenMP][NFC] Clang format ParseOpenMP (authored by jdoerfert).
[OpenMP][NFC] Clang format ParseOpenMP
Thu, Oct 29, 11:32 PM
jdoerfert closed D90256: [OpenMP][Docs] Structure and content for the OpenMP documentation.
Thu, Oct 29, 11:32 PM · Restricted Project
jdoerfert closed D90440: [OpenMP][NFC] Clang format ParseOpenMP.
Thu, Oct 29, 11:32 PM · Restricted Project
jdoerfert added inline comments to D90391: [gvn] PRE needs to skip convergent intrinsics/calls..
Thu, Oct 29, 10:31 PM · Restricted Project
jdoerfert added a comment to D90012: [PruneEH] Pin tests to legacy PM.

FWIW, once we enable the Attributor, prune-eh is obsolete either way.

Thu, Oct 29, 9:07 PM · Restricted Project
jdoerfert added a comment to D90415: [OpenMP] Use __OPENMP_NVPTX__ instead of _OPENMP in complex wrapper headers..

Thanks, will land it later.

For the bug see: https://bugs.llvm.org/show_bug.cgi?id=48014
Do I have to create a new phabricator review, too?
I'm currently building the release/11.x branch with the patch cherry-picked (which worked flawlessly).

Thu, Oct 29, 5:02 PM · Restricted Project
jdoerfert requested review of D90440: [OpenMP][NFC] Clang format ParseOpenMP.
Thu, Oct 29, 4:46 PM · Restricted Project
jdoerfert added a comment to D77609: [OpenMP] Added the support for unshackled task in RTL.

Some variables are only defined when the MACRO is enabled. I have changed some code to make it more readable and less complex.

Thu, Oct 29, 1:08 PM · Restricted Project
jdoerfert accepted D90415: [OpenMP] Use __OPENMP_NVPTX__ instead of _OPENMP in complex wrapper headers..

You need to define & undefine the macro around the includes of __clang_cuda_complex_builtins.h in clang/lib/Headers/openmp_wrappers/complex and clang/lib/Headers/openmp_wrappers/complex.h. (see also rG7f1e6fcff942) That should fix the tests. Assuming the tests pass, LGTM.

Thu, Oct 29, 1:06 PM · Restricted Project
jdoerfert accepted D81314: [OpenMP][Tool] Handle detached tasks in Archer.

This looks reasonable, not that I have much experience with ARCHER.

Thu, Oct 29, 1:01 PM · Restricted Project
jdoerfert added a comment to D90412: Fix hierarchical barrier deadlock for infinite blocktime.

Can we have a test for this?

Thu, Oct 29, 11:53 AM · Restricted Project
jdoerfert accepted D90403: [OpenMP] Add NULL check in dispatcher debug output.

LGTM

Thu, Oct 29, 9:33 AM · Restricted Project
jdoerfert requested changes to D90286: [Analysis] Improve EmitGEPOffset by avoiding summ with zero.

and there is a related test (test/Transforms/InstCombine/sub-gep.ll), which I believe is enough for this small improvement.

Thu, Oct 29, 9:02 AM · Restricted Project

Wed, Oct 28

jdoerfert committed rG92f7683d8a67: [Attributor][NFC] Rerun update test script (authored by jdoerfert).
[Attributor][NFC] Rerun update test script
Wed, Oct 28, 10:48 PM
jdoerfert committed rGd39f574dcc99: [Attributor][FIX] Properly promote arguments pointers to arrays (authored by jdoerfert).
[Attributor][FIX] Properly promote arguments pointers to arrays
Wed, Oct 28, 10:48 PM
jdoerfert added a comment to D90341: Use deref facts derived from minimum object size of allocations.

Use the nofree argument/function attribute to exclude this case.

Wed, Oct 28, 4:34 PM · Restricted Project
jdoerfert added a comment to D90281: [FileCheck] Report missing prefixes when more than one is provided..

The remaining question is how to stage this. I think there are 2 broad ways to do it:

  1. we enable it by default, and this patch goes and patches all failing tests with an opt-out.
  2. we start with it disabled. Then, people opt in explicitly.
Wed, Oct 28, 10:55 AM · Restricted Project
jdoerfert added a comment to D89971: [OpenMP][CUDA] Add missing overload for `remquo(float,float,int*)`.
In D89971#2359398, @tra wrote:

Never mind. I see that you've fixed it in 17c8251bca83072d2f3e00f936d6ce24500e6b02

Wed, Oct 28, 10:08 AM · Restricted Project
jdoerfert accepted D90248: [libomptarget][nvptx] Undef, weak shared variables.

LGTM, not totally convinced about weak linkage but ok

Wed, Oct 28, 7:04 AM · Restricted Project

Tue, Oct 27

jdoerfert added a comment to D90286: [Analysis] Improve EmitGEPOffset by avoiding summ with zero.

Drive by: Tests needed and we should see if we can avoid 3-5 line assignments.

Tue, Oct 27, 11:28 PM · Restricted Project