Feed All Stories

Today

pboettch created D38033: cmake: fix .git-dir detection with using git-rev-parse (regression since 4.0).
Tue, Sep 19, 4:42 AM · Restricted Project
bjope accepted D37254: [Sema] Disallow assigning record lvalues with nested const-qualified fields..

Nobody else seems to bother. At least no objections, so I think we can land this now.

Tue, Sep 19, 4:36 AM
jbhateja added inline comments to D36454: [X86] Changes to extract Horizontal addition operation for AVX-512..
Tue, Sep 19, 4:15 AM
jina.nahias committed rL313625: [x86] Lowering Mask Set1 intrinsics to LLVM IR.
[x86] Lowering Mask Set1 intrinsics to LLVM IR
Tue, Sep 19, 4:04 AM
jina.nahias closed D37669: [X86][intrinsics] lower _mm[256|512]_mask[z]_set1_epi[8|16|32|64] intrinsic to IR by committing rL313625: [x86] Lowering Mask Set1 intrinsics to LLVM IR.
Tue, Sep 19, 4:04 AM
jina.nahias committed rL313624: Lowering Mask Set1 intrinsics to LLVM IR.
Lowering Mask Set1 intrinsics to LLVM IR
Tue, Sep 19, 4:02 AM
jina.nahias closed D37668: [X86][intrinsics] lower _mm[256|512]_mask[z]_set1_epi[8|16|32|64] intrinsic to IR by committing rL313624: Lowering Mask Set1 intrinsics to LLVM IR.
Tue, Sep 19, 4:02 AM
avt77 updated the diff for D37262: The issues with X86 prefixes: step 2.

I fixed issues raised by Craig and Simon.

Tue, Sep 19, 3:58 AM
jonpa added inline comments to D37977: SystemZ: mischeduler enabled.
Tue, Sep 19, 3:46 AM
ilya-biryukov updated the diff for D38032: [clangd] Serialize onDiagnosticsReady callbacks for the same file..
  • Removed an incidental Mutex.unlock(), a leftover from previous drafts.
Tue, Sep 19, 3:46 AM
jonpa updated the diff for D37977: SystemZ: mischeduler enabled.

Tests further updated per review

Tue, Sep 19, 3:45 AM
avt77 abandoned D34056: Tail merge size.

It seems nobody wants to change the default value of the switch.

Tue, Sep 19, 3:45 AM
avt77 abandoned D36788: The issues with X86 prefixes.

D37262 covers all issues from this review plus some others that's why we don't need D36788 any more.

Tue, Sep 19, 3:43 AM
phschaad committed rL313623: [GPUJIT] Improved temporary file handling..
[GPUJIT] Improved temporary file handling.
Tue, Sep 19, 3:43 AM
Diffusion closed D37691: [GPUJIT] Improved temporary file handling. by committing rL313623: [GPUJIT] Improved temporary file handling..
Tue, Sep 19, 3:42 AM
ilya-biryukov updated the diff for D37972: [clangd] Introduced Logger interface..
  • Replaced JSONOutput in Protocol.h/Protocol.cpp with Logger.
Tue, Sep 19, 3:36 AM
alexfh added a comment to D37263: [clang-format] Ignore case when sorting using-declarations.

As discussed offline, the sorting should also be stable to avoid no-op replacements for identical using declarations and randomizing the order of using declarations differing only in case (not that I'd expect this to be a frequent thing, but it's better to handle it properly anyway).

Tue, Sep 19, 3:30 AM
RKSimon added a comment to D27846: [SLP] Support for horizontal min/max reduction.

Is there any chance that you can simplify the PR34635.ll test case that you committed? There's a lot of metadata/unnecessary code in there which is likely to make the testcase very brittle.

Tue, Sep 19, 3:21 AM
ilya-biryukov created D38032: [clangd] Serialize onDiagnosticsReady callbacks for the same file..
Tue, Sep 19, 3:20 AM
krasimir added a reviewer for D37263: [clang-format] Ignore case when sorting using-declarations: alexfh.
Tue, Sep 19, 3:19 AM
PhilippSchaad added a comment to D37691: [GPUJIT] Improved temporary file handling..

Thank you both for the review!

Tue, Sep 19, 3:19 AM
avt77 updated the diff for D37292: 'into' instruction should not be decoded as valid in 64-bit mode.

I re-implemented the fix accordingly to Craig suggest.

Tue, Sep 19, 3:17 AM
klimek added a comment to D37980: [clang-format] Better parsing of lambda captures with initializer expressions..

Thanks for the patch and the discussion around this. I fixed this in r313622 in what I think is a more principled approach that also works for nested lambdas (and gets rid of a lot of now-obsolete code).
The big problem with this code was that it evolved a bit to the point where we were able to generalize it, but we never did. Let me know if you find any problems with the aforementioned change.

Tue, Sep 19, 3:02 AM
klimek committed rL313622: Fix formatting of lambda introducers with initializers..
Fix formatting of lambda introducers with initializers.
Tue, Sep 19, 3:01 AM
grimar updated the diff for D37059: [ELF] - LTO: do not optimize away symbols accessed from linkerscript..
  • Addressed review comments.
Tue, Sep 19, 2:59 AM
curdeius updated the diff for D37980: [clang-format] Better parsing of lambda captures with initializer expressions..

Minor: use FormatToken::isNot instead of !FormatToken::is.

Tue, Sep 19, 2:59 AM
aymanmus created D38031: Adding a shufflevector and select LLVM IR instructions fuzz tool.
Tue, Sep 19, 2:52 AM
curdeius added a comment to D35743: [clang-format] Adjust space around &/&& of structured bindings.

There is one big missing thing here: PointerAlignment. Actually only PAS_Left is taken into account.
There are 3 possible options:
Left: auto& [a, b] = f();
Middle: auto & [a, b] = f();
Right: auto &[a, b] = f();

Tue, Sep 19, 2:48 AM · Restricted Project
RKSimon added a comment to D36454: [X86] Changes to extract Horizontal addition operation for AVX-512..

Would we be better off focussing on developing the @llvm.experimental.vector.reduce.add.* intrinsics? Or looking to get slpvectorizer to do extract_subvector style shuffles as part of the reduction?

Tue, Sep 19, 2:43 AM
grimar committed rL313621: [ELF] - Do not merge sections from SHT_GROUP when -relocatable.
[ELF] - Do not merge sections from SHT_GROUP when -relocatable
Tue, Sep 19, 2:42 AM
grimar closed D37574: [ELF] - Do not merge sections from SHT_GROUP when -relocatable by committing rL313621: [ELF] - Do not merge sections from SHT_GROUP when -relocatable.
Tue, Sep 19, 2:42 AM
grimar added inline comments to D37574: [ELF] - Do not merge sections from SHT_GROUP when -relocatable.
Tue, Sep 19, 2:41 AM
grosser added a comment to D37691: [GPUJIT] Improved temporary file handling..

No worries. Thank you for reviewing!

Tue, Sep 19, 2:38 AM
rayslava updated the diff for D38026: Make LSan compliant with recovery mode when running on top of ASan.

clang-format applied

Tue, Sep 19, 2:36 AM
klimek added a comment to D35743: [clang-format] Adjust space around &/&& of structured bindings.

So we actually didn't need to change the UnwrappedLineParser? I assume we still incorrectly assume the structured bindings are labmdas then?

Tue, Sep 19, 2:34 AM · Restricted Project
RKSimon added a comment to D38023: [X86] Prefer MOVSS/SD over BLENDI during legalization. Remove BLENDI versions of scalar arithmetic patterns.

Making a note here as I think PR33079 was having problems with the MOVSS<->BLENDPS style commutation

Tue, Sep 19, 2:32 AM
grimar committed rL313620: [ELF] - Don't crash when --emit-relocs is used with --gc-sections.
[ELF] - Don't crash when --emit-relocs is used with --gc-sections
Tue, Sep 19, 2:29 AM
grimar closed D37561: [ELF] - Don't crash when --emit-relocs is used with --gc-sections by committing rL313620: [ELF] - Don't crash when --emit-relocs is used with --gc-sections.
Tue, Sep 19, 2:29 AM
JDevlieghere added inline comments to D37971: [dwarfdump] Add verbose output for .debug-line section.
Tue, Sep 19, 2:26 AM
JDevlieghere updated the diff for D37971: [dwarfdump] Add verbose output for .debug-line section.
  • Moved printing outside switch.
  • Un-verbosed test cases.
Tue, Sep 19, 2:23 AM
grimar committed rL313619: [ELF] - Introduce std::vector<InputFile *> global arrays..
[ELF] - Introduce std::vector<InputFile *> global arrays.
Tue, Sep 19, 2:22 AM
grimar closed D35987: [ELF] - Introduce std::vector<InputFile *> global arrays. by committing rL313619: [ELF] - Introduce std::vector<InputFile *> global arrays..
Tue, Sep 19, 2:22 AM
philip.pfaffe added a comment to D37691: [GPUJIT] Improved temporary file handling..

My bad, didn't realize this was in GPUJIT and assumed it was related to D37429.

Tue, Sep 19, 2:13 AM
grimar added a comment to D37517: [ELF] - Report orphan sections if -verbose given..
In D37517#874133, @ruiu wrote:

log() is a regular function, so its argument is evaluated first and then its value is passed to log(). That means toString(S) is always executed even if it's not going to be written. Is it slow? (I do not request you update this patch, but I'm wondering whether it's negligible or not.)

Tue, Sep 19, 2:10 AM
grosser added a comment to D37691: [GPUJIT] Improved temporary file handling..

This is a run-time library, we do not (yet) link with LLVM libraries.

Tue, Sep 19, 2:09 AM
m.ostapenko updated subscribers of D38026: Make LSan compliant with recovery mode when running on top of ASan.
Tue, Sep 19, 2:09 AM
rogfer01 committed rL313618: [ARM] Use ADDCARRY / SUBCARRY.
[ARM] Use ADDCARRY / SUBCARRY
Tue, Sep 19, 2:07 AM
rogfer01 closed D35192: [ARM] Use ADDCARRY / SUBCARRY by committing rL313618: [ARM] Use ADDCARRY / SUBCARRY.
Tue, Sep 19, 2:07 AM
philip.pfaffe added a comment to D37691: [GPUJIT] Improved temporary file handling..

Instead of POSIX, you should use the OS-independent abstractions for interfacing with the system in the llvm::sys and llvm::sys::fs namespaces.

Tue, Sep 19, 2:02 AM
xiangzhai added a comment to D38029: [AVR] Override ParseDirective.

Dear Dylan,

Tue, Sep 19, 2:02 AM
philip.pfaffe added a comment to D37429: [GPGPU][PPCGCodeGen][GPUJIT] Added AMD support to GPGPU code generation.

I don't fully understand the reasoning behind this, and I dislike the LLD hack for a couple of reasons:

  • Why invoke a program instead of linking against lld?
  • What if lld is not installed? Why not default to ld?
  • Generally I feel like this is a huge violation of layering.
Tue, Sep 19, 2:01 AM
yrouban created D38030: Fix APFloat from string conversion for Inf.
Tue, Sep 19, 2:00 AM
jhenderson accepted D38017: [llvm-objcopy] Add test to check that architecture specific values are not used on wrong architecture..

LGTM.

Tue, Sep 19, 1:56 AM
jhenderson added a comment to D37915: [llvm-objcopy] Add support for dynamic relocations.

Yes, the SectionTable refactor definitely has improved things. Once the other change is ready, I'll approve this one.

Tue, Sep 19, 1:55 AM
xiangzhai created D38029: [AVR] Override ParseDirective.
Tue, Sep 19, 1:52 AM
jhenderson added a comment to D38008: [llvm-objcopy] Refactor code to include initialize method.

Just to point out, there's no diff context here! Aside from that, the overall change here looks good. I'm slightly surprised to see the SectionTable being copied everywhere, as I'd expect there to only be one, but I guess it doesn't matter since it's just a wrapper around a reference to an array of pointers.

Tue, Sep 19, 1:44 AM
aymanmus added a dependency for D38028: [X86][TableGen] Recommitting the X86 memory folding tables TableGen backend while disabling it by default.: D38027: [X86] Add new attribute to X86 instructions to enable marking them as "not memory foldable".
Tue, Sep 19, 1:41 AM
aymanmus added a dependent revision for D38027: [X86] Add new attribute to X86 instructions to enable marking them as "not memory foldable": D38028: [X86][TableGen] Recommitting the X86 memory folding tables TableGen backend while disabling it by default..
Tue, Sep 19, 1:41 AM
aymanmus created D38028: [X86][TableGen] Recommitting the X86 memory folding tables TableGen backend while disabling it by default..
Tue, Sep 19, 1:39 AM
jhenderson accepted D36560: [llvm-objcopy] Add support for .dynamic, .dynsym, and .dynstr.

LGTM with the suggested argument name changes.

Tue, Sep 19, 1:35 AM
aymanmus created D38027: [X86] Add new attribute to X86 instructions to enable marking them as "not memory foldable".
Tue, Sep 19, 1:28 AM
jhenderson accepted D36558: [llvm-objcopy] Add support for nested and overlapping segments.

LGTM.

Tue, Sep 19, 1:25 AM
nemanjai accepted D37851: [Power9] Add missing Power9 instructions..

LGTM.

Tue, Sep 19, 1:13 AM
Typz updated the diff for D37813: clang-format: better handle namespace macros.

rebase

Tue, Sep 19, 1:09 AM
Typz added inline comments to D33440: clang-format: better handle statement macros.
Tue, Sep 19, 1:04 AM
Typz updated the diff for D33440: clang-format: better handle statement macros.

Add tests.
Replace sorted list with hashtable.

Tue, Sep 19, 1:03 AM
a.elovikov committed rL313617: Test commit..
Test commit.
Tue, Sep 19, 12:58 AM
delena accepted D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.
Tue, Sep 19, 12:54 AM
coby added a comment to D37461: [X86][AsmParser] re-introduce 'offset' operator.

sure, will update shortly

Tue, Sep 19, 12:50 AM
rayslava created D38026: Make LSan compliant with recovery mode when running on top of ASan.
Tue, Sep 19, 12:48 AM
craig.topper updated the diff for D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

Remove from X86InstrFragments.td

Tue, Sep 19, 12:47 AM
aymanmus created D38025: [X86] Change the Format attribute for TEST*rr instruction from the default MRMDestReg to MRMSrcReg.
Tue, Sep 19, 12:43 AM
delena added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

I did remove the usages from the td patterns. It wasn't used in very many places.

I meant X86InstrFragmentsSIMD.td.

Tue, Sep 19, 12:42 AM
ashahid added a comment to D36130: [SLP] Vectorize jumbled memory loads..

Fixed the issue to consider the respective operand while recording the mask for user tree entry. Added the test which you have used to show this issue.

Tue, Sep 19, 12:41 AM
arsenm committed rL313616: AMDGPU: Run internalize symbols at -O0.
AMDGPU: Run internalize symbols at -O0
Tue, Sep 19, 12:41 AM
arsenm closed D37887: AMDGPU: Run internalize symbols at -O0.

r313616

Tue, Sep 19, 12:41 AM
craig.topper added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

I did remove the usages from the td patterns. It wasn't used in very many places.

Tue, Sep 19, 12:37 AM
ashahid updated the diff for D36130: [SLP] Vectorize jumbled memory loads..
  • Review comments updated accordingly
  • Added a TODO for sortLoadAccesses API
  • Modified the TODO for sortLoadAccesses API
  • Review comment update for using OpdNum to insert the mask in respective location
Tue, Sep 19, 12:34 AM
craig.topper added a comment to D36454: [X86] Changes to extract Horizontal addition operation for AVX-512..

Why this be solved by just combining extract_subvectors through things like binops and onto their inputs.

Tue, Sep 19, 12:33 AM
dvyukov added inline comments to D38019: [asan] Fix nested error detection.
Tue, Sep 19, 12:32 AM
delena added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

Yes, I saw this comment. We create a custom node in order to avoid an additional "entrance" to LowerSELECT(), right?
In general, this approach is generic - create a custom node to stop the lowering process.
If we'll keep a table of custom-to-common last-minute replacement, it'd be, probably, more robust.
It may be implemented later.

Tue, Sep 19, 12:26 AM
jbhateja updated the diff for D36454: [X86] Changes to extract Horizontal addition operation for AVX-512..
Tue, Sep 19, 12:10 AM
jonpa updated the diff for D37899: [SystemZ] Implement shouldCoalesce() to help regalloc to avoid running out of registers with GR128 regs.

Updated per review.

Tue, Sep 19, 12:08 AM
jonpa added inline comments to D37899: [SystemZ] Implement shouldCoalesce() to help regalloc to avoid running out of registers with GR128 regs.
Tue, Sep 19, 12:05 AM

Yesterday

Ka-Ka created D38024: [RegisterCoalescer] Move debug value after rematerialize trivial def.
Mon, Sep 18, 11:56 PM
craig.topper added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

Here's the comment from getVectorMaskingNode

Mon, Sep 18, 11:56 PM
delena added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

Ok. This is NFC. Why we can't generate (or stay with) ISD::SELECT from the beginning?

Mon, Sep 18, 11:49 PM
vedantk committed rL313615: [ubsan-minimal] Test exported symbol set against RTUBsan.
[ubsan-minimal] Test exported symbol set against RTUBsan
Mon, Sep 18, 11:49 PM
vedantk committed rL313614: [ubsan-minimal] Make the interface more compatible with RTUBSan.
[ubsan-minimal] Make the interface more compatible with RTUBSan
Mon, Sep 18, 11:49 PM
grosser accepted D37691: [GPUJIT] Improved temporary file handling..

Typo: "improve"

Mon, Sep 18, 11:42 PM
craig.topper added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

What sort of test are you looking for? The truncate intrinsic tests should already cover this.

Mon, Sep 18, 11:34 PM
craig.topper created D38023: [X86] Prefer MOVSS/SD over BLENDI during legalization. Remove BLENDI versions of scalar arithmetic patterns.
Mon, Sep 18, 11:32 PM
delena added a comment to D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.

Please add a test.

Mon, Sep 18, 11:26 PM
grosser accepted D37984: [Polly][ForwardOpTree] Allow out-of-quota in examination part of forwardTree..

LGTM. Thank you!

Mon, Sep 18, 11:25 PM · Restricted Project
gadi.haber committed rL313613: [X86][Skylake] Adding the scheduling information for the SkylakeClient target.
[X86][Skylake] Adding the scheduling information for the SkylakeClient target
Mon, Sep 18, 11:21 PM
gadi.haber closed D37294: [X86][Skylake] Adding the scheduling information for the SkylakeClient target by committing rL313613: [X86][Skylake] Adding the scheduling information for the SkylakeClient target.
Mon, Sep 18, 11:21 PM
vitalybuka added a comment to D37631: [libFuzzer] Support using libc++.

Can we just rely on LLVM_ENABLE_LIBCXX instead of introducing COMPILER_RT_ENABLE_LIBCXX?

Mon, Sep 18, 10:55 PM
ctopper committed rL313612: [X86] Remove some unnecessary patterns for truncate with X86ISD::SELECT and….
[X86] Remove some unnecessary patterns for truncate with X86ISD::SELECT and…
Mon, Sep 18, 10:34 PM
craig.topper created D38022: [X86] Convert X86ISD::SELECT to ISD::VSELECT just before instruction selection to avoid duplicate patterns.
Mon, Sep 18, 10:30 PM
craig.topper retitled D38001: [X86] Don't select anyext GR32->GR64 to SUBREG_TO_REG. Use INSERT_SUBREG instead. from [X86] Don't select any extend of anyext GR32->GR64 to SUBREG_TO_REG. Use INSERT_SUBREG instead. to [X86] Don't select anyext GR32->GR64 to SUBREG_TO_REG. Use INSERT_SUBREG instead..
Mon, Sep 18, 10:27 PM