Page MenuHomePhabricator
Feed All Stories

Yesterday

courbet added a comment to D67852: [Alignment] Get DataLayout::StackAlignment as Align.

Can you please explain why the switch from MaybeAlign to Align is fine ? ARMTargetLowering::getABIAlignmentForCallingConv can switch from non-zero to 0 it seems.

Sun, Sep 22, 11:55 PM · Restricted Project
craig.topper created D67902: [X86] Use TargetConstant for condition code on X86ISD::SETCC/CMOV/BRCOND nodes..
Sun, Sep 22, 11:55 PM · Restricted Project
mclow.lists committed rGab8f77a73e54: Revert "Extension: Mark the default constructor of chrono::duration as… (authored by mclow.lists).
Revert "Extension: Mark the default constructor of chrono::duration as…
Sun, Sep 22, 11:18 PM
marshall committed rL372546: Revert "Extension: Mark the default constructor of chrono::duration as….
Revert "Extension: Mark the default constructor of chrono::duration as…
Sun, Sep 22, 11:17 PM
mgorny requested changes to D67866: [lldb] Unify python site-packages path.

I do not believe this to be the right approach. Unconditionally installing Python modules to the wrong directory is a bug, not a feature. It should be fixed (as I proposed in D67890), not made more prominent by making everything hinge on it.

Sun, Sep 22, 11:03 PM · Restricted Project
craig.topper added a comment to D67898: [SDAG] Peek through bitcasts before commuting setcc operands.

I've removed the canonicalization in LowerIntVSETCC_AVX512 that was causing this in r372543. And I've moved it to X86ISelDAGToDAG in r372544. At the time LowerIntVSETCC_AVX512 we were using isel patterns for vptest and this was an attempt reduce the size of the table. We're now using custom isel in X86ISelDAGToDAG so the canonicalization is much cheaper to handle there.

Sun, Sep 22, 10:58 PM · Restricted Project
rsmith added a comment to D67897: Fix __is_signed builtin.

Thanks!

Sun, Sep 22, 10:58 PM · Restricted Project
jimlin committed rL372545: Request commit access for jimlin.
Request commit access for jimlin
Sun, Sep 22, 10:53 PM
craig.topper committed rG03b5a13ee32e: [X86] Canonicalize all zeroes vector to RHS in X86DAGToDAGISel::tryVPTESTM. (authored by craig.topper).
[X86] Canonicalize all zeroes vector to RHS in X86DAGToDAGISel::tryVPTESTM.
Sun, Sep 22, 10:36 PM
craig.topper committed rG5e26064c4061: [X86] Remove SETEQ/SETNE canonicalization code from LowerIntVSETCC_AVX512 to… (authored by craig.topper).
[X86] Remove SETEQ/SETNE canonicalization code from LowerIntVSETCC_AVX512 to…
Sun, Sep 22, 10:36 PM
craig.topper committed rG1f058538e0e4: [X86] Add 32-bit command line to avx512f-vec-test-testn.ll (authored by craig.topper).
[X86] Add 32-bit command line to avx512f-vec-test-testn.ll
Sun, Sep 22, 10:36 PM
ctopper committed rL372544: [X86] Canonicalize all zeroes vector to RHS in X86DAGToDAGISel::tryVPTESTM..
[X86] Canonicalize all zeroes vector to RHS in X86DAGToDAGISel::tryVPTESTM.
Sun, Sep 22, 10:35 PM
ctopper committed rL372543: [X86] Remove SETEQ/SETNE canonicalization code from LowerIntVSETCC_AVX512 to….
[X86] Remove SETEQ/SETNE canonicalization code from LowerIntVSETCC_AVX512 to…
Sun, Sep 22, 10:35 PM
ctopper committed rL372542: [X86] Add 32-bit command line to avx512f-vec-test-testn.ll.
[X86] Add 32-bit command line to avx512f-vec-test-testn.ll
Sun, Sep 22, 10:35 PM
owenpan added a comment to D67888: [clang-format] NFC clang-format the clang-format unit tests.

I suppose the .clang-format file you used only had BasedOnStyle: LLVM in it. Did you run clang-format the second time to ensure that the formatted file was stable?

Sun, Sep 22, 10:33 PM · Restricted Project, Restricted Project
Richard Smith <richard-llvm@metafoo.co.uk> committed rGdca01801bb84: For P0784R7: add further testing of requirements on constexpr destructors. (authored by Richard Smith <richard-llvm@metafoo.co.uk>).
For P0784R7: add further testing of requirements on constexpr destructors.
Sun, Sep 22, 10:08 PM
rsmith committed rL372541: For P0784R7: add further testing of requirements on constexpr.
For P0784R7: add further testing of requirements on constexpr
Sun, Sep 22, 10:08 PM
davezarzycki committed rGa7a515cb7738: Prefer AVX512 memcpy when applicable (authored by davezarzycki).
Prefer AVX512 memcpy when applicable
Sun, Sep 22, 10:03 PM
davezarzycki closed D67874: Prefer AVX512 memcpy when applicable.

r372540

Sun, Sep 22, 10:03 PM · Restricted Project
davezarzycki committed rL372540: Prefer AVX512 memcpy when applicable.
Prefer AVX512 memcpy when applicable
Sun, Sep 22, 9:59 PM
nridge created D67901: [clangd] Improve semantic highlighting in dependent contexts (fixes #154).
Sun, Sep 22, 9:29 PM · Restricted Project
zoecarver updated the diff for D67897: Fix __is_signed builtin.
  • fix behavior when passed an enumeration type
Sun, Sep 22, 9:21 PM · Restricted Project
plotfi added inline comments to D63978: Clang Interface Stubs merger plumbing for Driver.
Sun, Sep 22, 9:21 PM · Restricted Project
mclow.lists committed rGd8ac51ab8fde: Extension: Mark the default constructor of chrono::duration as conditionally… (authored by mclow.lists).
Extension: Mark the default constructor of chrono::duration as conditionally…
Sun, Sep 22, 9:20 PM
marshall committed rL372539: Extension: Mark the default constructor of chrono::duration as conditionally….
Extension: Mark the default constructor of chrono::duration as conditionally…
Sun, Sep 22, 9:19 PM
plotfi updated the diff for D63978: Clang Interface Stubs merger plumbing for Driver.

Adding back -c. Fixing driver for .ifs files

Sun, Sep 22, 9:19 PM · Restricted Project
Ayal added a comment to D67764: [LV] Forced vectorization with runtime checks and OptForSize.

So starting from D65197 or so, a loop that is forced to vectorize in a function that is OptForSize, is expected to get vectorized even if it involves replicating the loop.
Another assert which presumably requires similar update is the "Cannot SCEV check stride or overflow when optimizing for size" one.

Sun, Sep 22, 9:19 PM · Restricted Project
Richard Smith <richard-llvm@metafoo.co.uk> committed rG457226e02a6e: For P0784R7: add support for constexpr destructors, and call them as… (authored by Richard Smith <richard-llvm@metafoo.co.uk>).
For P0784R7: add support for constexpr destructors, and call them as…
Sun, Sep 22, 8:50 PM
Hui added inline comments to D67689: [llvm-objcopy] Add support for --gap-fill and --pad-to options.
Sun, Sep 22, 8:50 PM · Restricted Project
rsmith committed rL372538: For P0784R7: add support for constexpr destructors, and call them as.
For P0784R7: add support for constexpr destructors, and call them as
Sun, Sep 22, 8:47 PM
rsmith accepted D67899: Fix __is_fundamental to accept nullptr_t.
Sun, Sep 22, 8:45 PM · Restricted Project
hhb updated the diff for D67866: [lldb] Unify python site-packages path.

Fix symbol linker

Sun, Sep 22, 8:40 PM · Restricted Project
rsmith accepted D67897: Fix __is_signed builtin.

(Can I interest you in fixing the misbehaviour for enumeration types too?)

Certainly. You mean that __is_signed should return false for enumeration types?

Sun, Sep 22, 8:40 PM · Restricted Project
yubing updated the diff for D67212: [x86] Adding support for some missing intrinsics: _castf32_u32, _castf64_u64, _castu32_f32, _castu64_f64.
Sun, Sep 22, 8:37 PM · Restricted Project
zoecarver created D67900: [libc++] Use builtin type traits whenever possible.
Sun, Sep 22, 8:36 PM · Restricted Project
zoecarver updated the diff for D67897: Fix __is_signed builtin.
  • fix docs
Sun, Sep 22, 8:21 PM · Restricted Project
zoecarver added a comment to D67897: Fix __is_signed builtin.

(Can I interest you in fixing the misbehaviour for enumeration types too?)

Sun, Sep 22, 8:21 PM · Restricted Project
zoecarver created D67899: Fix __is_fundamental to accept nullptr_t.
Sun, Sep 22, 8:19 PM · Restricted Project
lkail added a comment to D66991: [PowerPC] Fix SH field overflow issue.

I think the test can be simplified more, kinda like

# RUN: llc -O3 -mtriple=powerpc64le-unknown-linux-gnu -stop-after=ppc-pre-emit-peephole -verify-machineinstrs %s -o - | FileCheck %s
---                                                                                                                                                                           
name: foo                                                                                                                                                                   
alignment: 4                                                                                                                                                                  
tracksRegLiveness: true                                                                                                                                                       
body: |                                                                                                                                                                       
  bb.0.entry:                                                                                                                                                                 
    liveins: $r3                                                                                                                                                              
    renamable $r4 = LI 0                                                                                                                                                      
    renamable $r5 = SRW renamable $r3, renamable killed $r4                                                                                                                   
    $r3 = COPY renamable killed $r5                                                                                                                                           
    BLR implicit $lr, implicit $rm, implicit $r3                                                                                                                              
...

And a NFC patch to pre-commit the test would also be preferable.

Sun, Sep 22, 8:19 PM · Restricted Project
pengfei updated the diff for D67898: [SDAG] Peek through bitcasts before commuting setcc operands.

Add testcase. Thanks @xbolva00

Sun, Sep 22, 7:10 PM · Restricted Project
xbolva00 added a comment to D67898: [SDAG] Peek through bitcasts before commuting setcc operands.

Please add a testcase

Sun, Sep 22, 6:54 PM · Restricted Project
pengfei created D67898: [SDAG] Peek through bitcasts before commuting setcc operands.
Sun, Sep 22, 6:40 PM · Restricted Project
craig.topper committed rGda4a4707d2ed: [X86] Convert to Constant arguments to MMX shift by i32 intrinsics to… (authored by craig.topper).
[X86] Convert to Constant arguments to MMX shift by i32 intrinsics to…
Sun, Sep 22, 6:21 PM
craig.topper committed rG5efc928ab278: [X86] Remove stale FIXME. (authored by craig.topper).
[X86] Remove stale FIXME.
Sun, Sep 22, 6:21 PM
ctopper committed rL372537: [X86] Convert to Constant arguments to MMX shift by i32 intrinsics to….
[X86] Convert to Constant arguments to MMX shift by i32 intrinsics to…
Sun, Sep 22, 6:20 PM
ctopper committed rL372536: [X86] Remove stale FIXME..
[X86] Remove stale FIXME.
Sun, Sep 22, 6:20 PM
yaxunl added a comment to D67509: [CUDA][HIP] Fix hostness of defaulted constructor.
In D67509#1677528, @tra wrote:

I am taking a look.

I can reproduce similar asserts locally. It seems the assertion I added assert(!(HasD || HasH) || (NeedsD == HasD && NeedsH == HasH)); is not always true. Since we do not have this assert before, I removed it. I will study what causes it to assert and post it later.

Sun, Sep 22, 6:14 PM · Restricted Project
craig.topper committed rGa533e87792ca: [X86][SelectionDAGBuilder] Move the hack for handling MMX shift by i32… (authored by craig.topper).
[X86][SelectionDAGBuilder] Move the hack for handling MMX shift by i32…
Sun, Sep 22, 6:05 PM
ctopper committed rL372535: [X86][SelectionDAGBuilder] Move the hack for handling MMX shift by i32….
[X86][SelectionDAGBuilder] Move the hack for handling MMX shift by i32…
Sun, Sep 22, 6:03 PM
Quuxplusone added a comment to D67897: Fix __is_signed builtin.

But std::is_signed_v<float> needs to yield false.

It should yield true; the spec says "If is_­arithmetic_­v<T> is true, the same result as T(-1) < T(0); otherwise, false".

Sun, Sep 22, 5:16 PM · Restricted Project
rsmith added a comment to D67897: Fix __is_signed builtin.

But std::is_signed_v<float> needs to yield false.

Sun, Sep 22, 4:53 PM · Restricted Project
zoecarver added a comment to D67897: Fix __is_signed builtin.

But std::is_signed_v<float> needs to yield false. Isn't it cleaner to leave the compiler builtin matching the library type-trait, so that the library doesn't have to check for integral types separately?

Sun, Sep 22, 4:53 PM · Restricted Project
rsmith added a comment to D67897: Fix __is_signed builtin.

Looks good, but please also update http://clang.llvm.org/docs/LanguageExtensions.html#type-trait-primitives

Sun, Sep 22, 4:52 PM · Restricted Project
craig.topper committed rGe4c17651243e: [X86] Require last argument to LWPINS/LWPVAL builtins to be an ICE. Add ImmArg… (authored by craig.topper).
[X86] Require last argument to LWPINS/LWPVAL builtins to be an ICE. Add ImmArg…
Sun, Sep 22, 4:49 PM
Quuxplusone added a comment to D67897: Fix __is_signed builtin.

But std::is_signed_v<float> needs to yield false. Isn't it cleaner to leave the compiler builtin matching the library type-trait, so that the library doesn't have to check for integral types separately?

Sun, Sep 22, 4:49 PM · Restricted Project
ctopper committed rL372534: [X86] Require last argument to LWPINS/LWPVAL builtins to be an ICE. Add ImmArg….
[X86] Require last argument to LWPINS/LWPVAL builtins to be an ICE. Add ImmArg…
Sun, Sep 22, 4:47 PM
zoecarver created D67897: Fix __is_signed builtin.
Sun, Sep 22, 4:26 PM · Restricted Project
xbolva00 added a comment to D67728: Scrub FileCheck regex delimiters from test checks.
I could mock-up a testcase but where does it go?
Sun, Sep 22, 3:38 PM · Restricted Project
xbolva00 updated subscribers of rL372533: [NFC] Fixed failed test.

Hi @erik.pilkington, as the author of https://reviews.llvm.org/D62825,

Sun, Sep 22, 3:19 PM
xbolva00 committed rG7b4d40e8db36: [NFC] Fixed failed test (authored by xbolva00).
[NFC] Fixed failed test
Sun, Sep 22, 3:15 PM
xbolva00 committed rL372533: [NFC] Fixed failed test.
[NFC] Fixed failed test
Sun, Sep 22, 3:14 PM
lebedev.ri committed rG7c3d6f5a1bf6: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to… (authored by lebedev.ri).
[X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to…
Sun, Sep 22, 3:08 PM
lebedevri committed rL372532: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to….
[X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to…
Sun, Sep 22, 3:05 PM
Diffusion closed D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).
Sun, Sep 22, 3:05 PM · Restricted Project
xbolva00 committed rGfb218170b47e: [Diagnostics] Warn if ?: with integer constants always evaluates to true (authored by xbolva00).
[Diagnostics] Warn if ?: with integer constants always evaluates to true
Sun, Sep 22, 3:05 PM
xbolva00 committed rL372531: [Diagnostics] Warn if ?: with integer constants always evaluates to true.
[Diagnostics] Warn if ?: with integer constants always evaluates to true
Sun, Sep 22, 3:04 PM
Tyker committed rG914c4c306d12: [clang] fixing conditional explicit for out-of-line definition PR42980 (authored by Tyker).
[clang] fixing conditional explicit for out-of-line definition PR42980
Sun, Sep 22, 3:04 PM
Tyker committed rL372530: [clang] fixing conditional explicit for out-of-line definition PR42980.
[clang] fixing conditional explicit for out-of-line definition PR42980
Sun, Sep 22, 3:03 PM
Tyker closed D67889: [clang] fixing conditional explicit for out-of-line definition PR42980.
Sun, Sep 22, 3:02 PM · Restricted Project, Restricted Project
lebedev.ri added a comment to D67677: [InstCombine] dropRedundantMaskingOfLeftShiftInput(): pat. a/b with mask (PR42563).

@spatel will you want to review the sibling patch D67725, or should i "self-review" since it's rather identical to this one?

Sun, Sep 22, 3:02 PM · Restricted Project
lebedev.ri updated the diff for D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

Fixup comments.

Sun, Sep 22, 2:57 PM · Restricted Project
craig.topper accepted D67874: Prefer AVX512 memcpy when applicable.

LGTM

Sun, Sep 22, 2:55 PM · Restricted Project
craig.topper accepted D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

LGTM with those comment fixes.

Sun, Sep 22, 2:54 PM · Restricted Project
lebedev.ri requested review of D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).
Sun, Sep 22, 2:43 PM · Restricted Project
lebedev.ri added a comment to D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

What if we just do the larger than 32-bit mask? Its not clear that making BZHI just to fold a load is an improvement. You have to materialize an immediate instead so the total uops increased.

Actually, i think it's clearly still not an improvement: https://godbolt.org/z/1M1deG

Sun, Sep 22, 2:43 PM · Restricted Project
RKSimon committed rG638933acab3d: Verifier - silence static analyzer dyn_cast<VectorType> null dereference… (authored by RKSimon).
Verifier - silence static analyzer dyn_cast<VectorType> null dereference…
Sun, Sep 22, 2:14 PM
RKSimon committed rG557cee337b99: [AMDGPU] isSDNodeAlwaysUniform - silence static analyzer dyn_cast<LoadSDNode>… (authored by RKSimon).
[AMDGPU] isSDNodeAlwaysUniform - silence static analyzer dyn_cast<LoadSDNode>…
Sun, Sep 22, 2:12 PM
RKSimon committed rL372529: Verifier - silence static analyzer dyn_cast<VectorType> null dereference….
Verifier - silence static analyzer dyn_cast<VectorType> null dereference…
Sun, Sep 22, 2:00 PM
RKSimon committed rL372528: [AMDGPU] isSDNodeAlwaysUniform - silence static analyzer dyn_cast<LoadSDNode>….
[AMDGPU] isSDNodeAlwaysUniform - silence static analyzer dyn_cast<LoadSDNode>…
Sun, Sep 22, 2:00 PM
mgorny committed rG96f35266a5d6: [clang] [Basic] Enable __has_feature(leak_sanitizer) (authored by mgorny).
[clang] [Basic] Enable __has_feature(leak_sanitizer)
Sun, Sep 22, 1:55 PM
mgorny committed rL372527: [clang] [Basic] Enable __has_feature(leak_sanitizer).
[clang] [Basic] Enable __has_feature(leak_sanitizer)
Sun, Sep 22, 1:53 PM
mgorny closed D67719: [clang] [Basic] Enable __has_feature(leak_sanitizer).
Sun, Sep 22, 1:53 PM · Restricted Project
phosek accepted D67719: [clang] [Basic] Enable __has_feature(leak_sanitizer).

LGTM

Sun, Sep 22, 1:45 PM · Restricted Project
andreadb added inline comments to D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).
Sun, Sep 22, 1:36 PM · Restricted Project
Yi-Hong.Lyu updated the diff for D66991: [PowerPC] Fix SH field overflow issue.

Address Jinsong's comments

Sun, Sep 22, 1:31 PM · Restricted Project
lebedev.ri updated the diff for D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).
Sun, Sep 22, 1:27 PM · Restricted Project
Yi-Hong.Lyu added inline comments to D66991: [PowerPC] Fix SH field overflow issue.
Sun, Sep 22, 1:22 PM · Restricted Project
lebedev.ri added a comment to D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

What if we just do the larger than 32-bit mask? Its not clear that making BZHI just to fold a load is an improvement. You have to materialize an immediate instead so the total uops increased.

Sun, Sep 22, 1:19 PM · Restricted Project
craig.topper committed rGc4802de31b1a: [X86] Fix some VCVTPS2PH isel patterns where 'i32' was used instead of 'timm' (authored by craig.topper).
[X86] Fix some VCVTPS2PH isel patterns where 'i32' was used instead of 'timm'
Sun, Sep 22, 1:10 PM
ctopper committed rL372526: [X86] Fix some VCVTPS2PH isel patterns where 'i32' was used instead of 'timm'.
[X86] Fix some VCVTPS2PH isel patterns where 'i32' was used instead of 'timm'
Sun, Sep 22, 1:07 PM
craig.topper added a comment to D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

What if we just do the larger than 32-bit mask? Its not clear that making BZHI just to fold a load is an improvement. You have to materialize an immediate instead so the total uops increased.

Sun, Sep 22, 1:07 PM · Restricted Project
lebedev.ri updated the diff for D67875: [X86] X86DAGToDAGISel::matchBEXTRFromAndImm(): if can't use BEXTR, fallback to BZHI (PR43381).

Scaling down ambigiousness of the patch: if not using BEXTR, only use BZHI if either the mask is larger than 32-bit, or the load folding will happen.

Sun, Sep 22, 12:54 PM · Restricted Project
craig.topper committed rG80fda375b225: [X86][TableGen] Allow timm to appear in output patterns. Use it to remove… (authored by craig.topper).
[X86][TableGen] Allow timm to appear in output patterns. Use it to remove…
Sun, Sep 22, 12:49 PM
davezarzycki updated the diff for D67874: Prefer AVX512 memcpy when applicable.

This patch has now been run through update_llc_test_checks.

Sun, Sep 22, 12:48 PM · Restricted Project
ctopper committed rL372525: [X86][TableGen] Allow timm to appear in output patterns. Use it to remove….
[X86][TableGen] Allow timm to appear in output patterns. Use it to remove…
Sun, Sep 22, 12:48 PM
lebedev.ri committed rG24159592cac9: [NFC][X86] Add BEXTR test with load and 33-bit mask (PR43381 / D67875) (authored by lebedev.ri).
[NFC][X86] Add BEXTR test with load and 33-bit mask (PR43381 / D67875)
Sun, Sep 22, 12:38 PM
lebedevri committed rL372524: [NFC][X86] Add BEXTR test with load and 33-bit mask (PR43381 / D67875).
[NFC][X86] Add BEXTR test with load and 33-bit mask (PR43381 / D67875)
Sun, Sep 22, 12:38 PM
asmith added inline comments to D67689: [llvm-objcopy] Add support for --gap-fill and --pad-to options.
Sun, Sep 22, 12:37 PM · Restricted Project
RKSimon added a reviewer for D67874: Prefer AVX512 memcpy when applicable: RKSimon.
Sun, Sep 22, 12:33 PM · Restricted Project
asmith updated the diff for D67689: [llvm-objcopy] Add support for --gap-fill and --pad-to options.

Add new options to objcopy documentation

Sun, Sep 22, 12:33 PM · Restricted Project