Feed All Stories

Today

delena accepted D35772: [LV][X86] Support of AVX2 Gathers code generation and update the LV with this.
Sun, Nov 19, 1:29 PM
craig.topper created D40228: [Target] Keep the TargetOptions feature list sorted instead of sorting during CodeGen.
Sun, Nov 19, 1:05 PM
zvi added inline comments to D39952: [X86]: Adding full coverage of MC encoding for all X86 ISA Sets.NFC.
Sun, Nov 19, 12:41 PM
dorit added a comment to D38948: [LV] Support efficient vectorization of an induction with redundant casts.

Yes, IndVarSimpify wouldn't fix this issue, but I was thinking more of using the techniques there that use the SCEV expressions to find these cases instead of doing the pattern matching (see the inline comment).

Sun, Nov 19, 12:40 PM
zvi updated the diff for D38313: [InstCombine] Introducing Aggressive Instruction Combine pass.

Rebase on ToT. NFC in this revision.

Sun, Nov 19, 12:24 PM
majnemer added a comment to D40218: [Clang] Add __builtin_launder.

A test with restrict and __restrict might be interesting.

Sun, Nov 19, 12:21 PM
zvi updated the diff for D38313: [InstCombine] Introducing Aggressive Instruction Combine pass.

Address the last of Craig's comments:

  • Thanks, @lsaba, for porting the pass to the new PassManager.
  • Removed shrinkage of vector types until we sort out if it is generally allowed to shrink element types of vector operations.
  • Some minor fixes to comments.
Sun, Nov 19, 12:15 PM
mgrang added a comment to D40227: [Hexagon] Use stable sort for HexagonShuffler to remove non-deterministic ordering.

In HexagonShuffler.h, the lessCore and lessCVI comparators sort insts based on the number of units set. Now, two insts can have the same units set so their relative ordering is not defined.
We could add an index field as a tie-breaker but that would mean changes to function APIs which I felt could be an overkill. Hence, I went with the simpler approach to just use stable_sort instead.
Please let me know your thoughts on this.

Sun, Nov 19, 12:03 PM
mgrang created D40227: [Hexagon] Use stable sort for HexagonShuffler to remove non-deterministic ordering.
Sun, Nov 19, 11:58 AM
craig.topper created D40226: [CodeGen] Move Reciprocals option from TargetOptions to CodeGenOptions.
Sun, Nov 19, 11:28 AM
JDevlieghere added inline comments to D40200: [DWARF] Fix handling of extended line-number opcodes.
Sun, Nov 19, 11:22 AM · debug-info
jankratochvil committed rL318631: Remove 2 unused methods DWARFDebugInfo::Find and their FindCallbackString.
Remove 2 unused methods DWARFDebugInfo::Find and their FindCallbackString
Sun, Nov 19, 11:05 AM
jankratochvil closed D40216: #if 0 for DWARFDebugInfo::Find() as it is unused by committing rL318631: Remove 2 unused methods DWARFDebugInfo::Find and their FindCallbackString.
Sun, Nov 19, 11:05 AM
sbaranga added a comment to D38948: [LV] Support efficient vectorization of an induction with redundant casts.

IIUC, what IndVarSimply does is call the rewriter and then fix the users of the induction; This is similar in effect to what we do: We don't need to call the PSCEV rewriter again, we already have the nice AddRec for the induction phi (isInductionPhi() had already obtained it); The vectorization of the induction phi proceeds unchanged; The only thing we are adding is the def-use wiring so that in the vectorized loop, any users of the cast instructions will be users of the vectorized phi. We never vectorize the casts, and we never actively remove them -- they will end up dead code in the vectorized loop because they will not be used.

Sun, Nov 19, 10:55 AM
jankratochvil abandoned D40214: performance: Prevent needless DWARFCompileUnit::Clear() on freshly ctor-ed object.
Sun, Nov 19, 10:47 AM
craig.topper added a comment to D40222: [x86][icelake]BITALG.

If it supports masking we can't use the intrinsic in the tablegen as it would go against our normal lowering of intrinsics.

Sun, Nov 19, 10:47 AM
jankratochvil added a comment to D40214: performance: Prevent needless DWARFCompileUnit::Clear() on freshly ctor-ed object.

Thanks for the review but then it would become a performance regression, not the performance improvement I was trying to make.
Withdrawing this patch.

Sun, Nov 19, 10:46 AM
clayborg added a comment to D40212: refactor: Unify+simplify DWARFCompileUnit ctor+Clear() into in-class initializers.

Good change in the header file.

If you mean the in-class initializers they obviously cannot be used without the in-place construction+destruction as they would stay duplicate to the Clear() method.

OK if you are also uncomfortable with it I will keep Clear() as is and just call Clear() from the ctor to unify it at least a bit.

Sun, Nov 19, 10:44 AM
spatel updated the diff for D28314: Change sqrt partial inlining to depend on sqrt argument rather than result..

Patch updated:
Add a specialized TTI hook as suggested. So now this becomes a functional change only for x86 which overrides the default hook.

Sun, Nov 19, 10:34 AM
jankratochvil added a comment to D40212: refactor: Unify+simplify DWARFCompileUnit ctor+Clear() into in-class initializers.

Good change in the header file.

Sun, Nov 19, 10:34 AM
Ayal added inline comments to D38948: [LV] Support efficient vectorization of an induction with redundant casts.
Sun, Nov 19, 10:31 AM
spatel commandeered D28314: Change sqrt partial inlining to depend on sqrt argument rather than result..

Commandeering from the commandeer...er.

Sun, Nov 19, 10:29 AM
Meinersbur added a comment to D39971: Port ScopInfo to the isl cpp bindings.

Looks great (whatever you decide to to with the inline comment)

Sun, Nov 19, 10:25 AM
clayborg requested changes to D40216: #if 0 for DWARFDebugInfo::Find() as it is unused.

Feel free to remove any unused code. No need for review on dead code removal. So just remove the code, don't add #if 0

Sun, Nov 19, 10:23 AM
clayborg requested changes to D40214: performance: Prevent needless DWARFCompileUnit::Clear() on freshly ctor-ed object.

See inline comments.

Sun, Nov 19, 10:21 AM
clayborg added a comment to D40212: refactor: Unify+simplify DWARFCompileUnit ctor+Clear() into in-class initializers.

Good change in the header file. I am not sure I like the destruct this object in place and replace with new version... If this is commonly done and acceptable form of C++ I would be ok with it, but I agree with Pavel, it seems a little bit off the books.

Sun, Nov 19, 10:13 AM
rengolin accepted D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Great, LGTM now, thanks!

Sun, Nov 19, 9:49 AM
sepavloff updated the diff for D24933: Enable configuration files in clang.

Updated patch

Sun, Nov 19, 9:36 AM
spatel committed rL318630: [x86] add sqrt tests for partially-inline-libcalls (PR31455).
[x86] add sqrt tests for partially-inline-libcalls (PR31455)
Sun, Nov 19, 9:31 AM
kubamracek added a comment to D39114: [XRay][darwin] Initial XRay in Darwin Support.

Can we just not use clock_gettime on Darwin and instead use mach_absolute_time?

Sun, Nov 19, 9:24 AM
aaron.ballman created D40225: Add -std=c17 as a flag.
Sun, Nov 19, 9:19 AM
a.sidorin accepted D39886: [ASTImporter] Fix wrong conflict detections for unnamed structures.

Hello Takafumi,

Sun, Nov 19, 8:48 AM
spatel committed rL318629: [LibCallSimplifier] allow splat vectors for pow(x, 0.5) -> sqrt() transforms.
[LibCallSimplifier] allow splat vectors for pow(x, 0.5) -> sqrt() transforms
Sun, Nov 19, 8:42 AM
krytarowski planned changes to D40159: Make TLS/NetBSD handling more generic.

I will be back to it once I will fix other bugs, unrelated to TLS in TSan.

Sun, Nov 19, 8:41 AM · Restricted Project
gilr updated the diff for D38676: [LV] Model masking in VPlan, introducing VPInstructions.

Addressed review comments.

Sun, Nov 19, 8:18 AM
gilr added a comment to D38676: [LV] Model masking in VPlan, introducing VPInstructions.

I finished my review, and apart from my two final comments, everything looks fine.

Thanks for the hard work! :)

Sun, Nov 19, 8:15 AM
spatel committed rL318628: [LibCallSimplifier] partly fix pow(x, 0.5) -> sqrt() transforms.
[LibCallSimplifier] partly fix pow(x, 0.5) -> sqrt() transforms
Sun, Nov 19, 8:15 AM
spatel closed D40150: [LibCallSimplifier] fix pow(x, 0.5) -> sqrt() transforms by committing rL318628: [LibCallSimplifier] partly fix pow(x, 0.5) -> sqrt() transforms.
Sun, Nov 19, 8:15 AM
joerg added a comment to D40159: Make TLS/NetBSD handling more generic.

The public interface for obtaining the TLS storage is the combination of reading the DTV vector of a thread in combination with dl_iterate_phdr to find the size of the TLS block of a specific module. That gives you all that you need to know. It is important to keep in mind that the vector can be initialized lazily, so __tls_get_addr and friends will have to be intercepted to update the global view.

Sun, Nov 19, 7:51 AM · Restricted Project
spatel committed rL318627: [InstSimplify] fold and/or of fcmp ord/uno when operand is known nnan.
[InstSimplify] fold and/or of fcmp ord/uno when operand is known nnan
Sun, Nov 19, 7:34 AM
spatel closed D40130: [InstSimplify] fold and/or of fcmp ord/uno when operand is known nnan by committing rL318627: [InstSimplify] fold and/or of fcmp ord/uno when operand is known nnan.
Sun, Nov 19, 7:34 AM
jankratochvil committed rL318626: Add comments to DWARFCompileUnit length fields/methods.
Add comments to DWARFCompileUnit length fields/methods
Sun, Nov 19, 6:35 AM
jankratochvil closed D40211: Add comments to DWARFCompileUnit length fields/methods by committing rL318626: Add comments to DWARFCompileUnit length fields/methods.
Sun, Nov 19, 6:35 AM
RKSimon added a reviewer for D40222: [x86][icelake]BITALG: RKSimon.

Disassembler tests?

Sun, Nov 19, 5:48 AM
oren_ben_simhon created D40224: [X86] Control-Flow Enforcement Technology - Shadow Stack and Indirect Branch Tracking support (Clang side).
Sun, Nov 19, 5:29 AM
dorit added a comment to D38948: [LV] Support efficient vectorization of an induction with redundant casts.

Hi Silviu,

Sun, Nov 19, 5:24 AM
oren_ben_simhon created D40223: [X86] Control-Flow Enforcement Technology - Shadow Stack support (LLVM side).
Sun, Nov 19, 5:15 AM
jklaehn updated the diff for D35181: Defer addition of keywords to identifier table when loading AST.

Thanks for taking a look! I removed the constructor argument as suggested; keywords are now added in PP.Initialize.

Sun, Nov 19, 4:58 AM
RKSimon added a reviewer for D40215: [X86][AVX512] Use PACKSS/PACKUS for vXi16->vXi8 truncations without BWI.: pcordes.

This patch came about as possible tweak prior to adding support AVX512BW PACKUS/PACKSS for PR34871, which was concerned about port5 load from double the number of truncations. But yes register limits is likely to be an issue. I still need to finish PR34773 first though so will come back to this in a while.

Sun, Nov 19, 4:51 AM
mstorsjo added inline comments to D39673: Toolchain: Normalize dwarf, sjlj and seh eh.
Sun, Nov 19, 4:16 AM
JonasToth added inline comments to D39027: [docs][refactor] Add a new tutorial that talks about how one can implement refactoring actions.
Sun, Nov 19, 4:08 AM
dberris reopened D39114: [XRay][darwin] Initial XRay in Darwin Support.

Hi @kubamracek -- do you know how to work around the lack of clock_gettime from the 10.11 SDK?

Sun, Nov 19, 2:30 AM
labath added a comment to D40212: refactor: Unify+simplify DWARFCompileUnit ctor+Clear() into in-class initializers.

You're using a fairly novel (to this codebase at least) simplification technique, so I think we should discuss that first. The way we have normally done these things is to just have the constructor call the Clear() function.

Sun, Nov 19, 2:08 AM
jankratochvil updated subscribers of D40216: #if 0 for DWARFDebugInfo::Find() as it is unused.
Sun, Nov 19, 2:04 AM
jankratochvil updated subscribers of D40214: performance: Prevent needless DWARFCompileUnit::Clear() on freshly ctor-ed object.
Sun, Nov 19, 2:03 AM
jankratochvil updated subscribers of D40212: refactor: Unify+simplify DWARFCompileUnit ctor+Clear() into in-class initializers.
Sun, Nov 19, 2:03 AM
labath accepted D40211: Add comments to DWARFCompileUnit length fields/methods.

If you are confident that the behavior you are documenting is correct, I think you can submit patches like these without approval. If you do end up sending something up for review, please be sure to add "lldb-commits" as a subscriber.

Sun, Nov 19, 1:56 AM
EricWF committed rL318625: Attempt to fix template depth test failures on older Clang versions.
Attempt to fix template depth test failures on older Clang versions
Sun, Nov 19, 1:46 AM
tk1012 added inline comments to D39886: [ASTImporter] Fix wrong conflict detections for unnamed structures.
Sun, Nov 19, 1:45 AM
tk1012 added a comment to D39886: [ASTImporter] Fix wrong conflict detections for unnamed structures.

Fix the broken indentation.

Sun, Nov 19, 1:40 AM
tk1012 updated the diff for D39886: [ASTImporter] Fix wrong conflict detections for unnamed structures.

I update the diff to solve the below thing.

Sun, Nov 19, 1:38 AM
coby added a dependent revision for D40213: [x86][icelake]BITALG: D40222: [x86][icelake]BITALG.
Sun, Nov 19, 1:06 AM
coby added a dependency for D40222: [x86][icelake]BITALG: D40213: [x86][icelake]BITALG.
Sun, Nov 19, 1:06 AM
coby created D40222: [x86][icelake]BITALG.
Sun, Nov 19, 1:05 AM
zvi added inline comments to D38313: [InstCombine] Introducing Aggressive Instruction Combine pass.
Sun, Nov 19, 12:51 AM
krytarowski added a comment to D40159: Make TLS/NetBSD handling more generic.

It seems like Linux uses __libc_memalign() interceptor to catch dynamically allocated TLS blocks.

Sun, Nov 19, 12:08 AM · Restricted Project

Yesterday

krytarowski added a comment to D40159: Make TLS/NetBSD handling more generic.

It looks like the proper TLS vector should be in a form of std::vector<std::pair<uptr,uptr>>, as there might be multiple regions mapped (true for DSO).

Sat, Nov 18, 11:43 PM · Restricted Project
delena added inline comments to D40215: [X86][AVX512] Use PACKSS/PACKUS for vXi16->vXi8 truncations without BWI..
Sat, Nov 18, 10:44 PM
ctopper committed rL318624: [X86] Add test cases for rndscaless/sd intrinsics..
[X86] Add test cases for rndscaless/sd intrinsics.
Sat, Nov 18, 10:24 PM
ctopper committed rL318623: [X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using….
[X86] Improve load folding of scalar rcp28 and rsqrt28 instructions using…
Sat, Nov 18, 9:43 PM
EricWF committed rL318622: Fix min/max usage in variant.
Fix min/max usage in variant
Sat, Nov 18, 8:57 PM
EricWF committed rL318621: [libc++] Shrink variant's index type when possible.
[libc++] Shrink variant's index type when possible
Sat, Nov 18, 8:19 PM
EricWF closed D40210: [libc++] Shrink variant's index type when possible.
Sat, Nov 18, 8:19 PM
mpark accepted D40210: [libc++] Shrink variant's index type when possible.
Sat, Nov 18, 8:18 PM
ctopper committed rL318620: [X86] Make sure 'knm' is accepted by -target-cpu.
[X86] Make sure 'knm' is accepted by -target-cpu
Sat, Nov 18, 8:13 PM
ctopper committed rL318619: [X86] Make sure 'knm' and 'cannonlake' are accepted by builtin_cpu_is.
[X86] Make sure 'knm' and 'cannonlake' are accepted by builtin_cpu_is
Sat, Nov 18, 8:12 PM
EricWF committed rL318618: Fix nodiscard test when modules are enabled.
Fix nodiscard test when modules are enabled
Sat, Nov 18, 7:52 PM
EricWF updated the diff for D40210: [libc++] Shrink variant's index type when possible.
  • Address inline comments.
Sat, Nov 18, 7:48 PM
EricWF added inline comments to D40210: [libc++] Shrink variant's index type when possible.
Sat, Nov 18, 7:47 PM
mpark added inline comments to D40210: [libc++] Shrink variant's index type when possible.
Sat, Nov 18, 7:39 PM
strager updated the summary of D40221: [clang-format] Parse blocks in braced lists.
Sat, Nov 18, 7:23 PM
strager updated the summary of D40221: [clang-format] Parse blocks in braced lists.
Sat, Nov 18, 7:22 PM
strager created D40221: [clang-format] Parse blocks in braced lists.
Sat, Nov 18, 7:21 PM
craig.topper abandoned D38824: [X86] Synchronize the existing CPU predefined macros with the cases that gcc defines them.

All skylake-avx512 and cannonlake now set corei7 as of r318616. Abandoning this.

Sat, Nov 18, 6:59 PM
ctopper committed rL318617: [X86] Add icelake CPU support for -march..
[X86] Add icelake CPU support for -march.
Sat, Nov 18, 6:56 PM
ctopper committed rL318616: [X86] Set __corei7__ preprocessor defines for skylake server and cannonlake..
[X86] Set __corei7__ preprocessor defines for skylake server and cannonlake.
Sat, Nov 18, 6:56 PM
ast committed rL318615: [bpf] remove unused variable.
[bpf] remove unused variable
Sat, Nov 18, 6:43 PM
compnerd closed D40115: Driver: remove `SupportsObjCGC`.

SVN r318609

Sat, Nov 18, 6:34 PM
compnerd accepted D40115: Driver: remove `SupportsObjCGC`.
Sat, Nov 18, 6:33 PM
strager abandoned D33814: CodeGen: Fix ARM cmpxchg64 register fragmentation in fast-regalloc.

I don't plan to work on this anymore.

Sat, Nov 18, 6:33 PM
compnerd accepted D40127: [Driver][ARM] For assembler files recognize -Xassembler or -Wa, -mthumb.

Would be nice to rename the variable prior to commit.

Sat, Nov 18, 6:33 PM
phosek added a comment to D39930: [CMake] Use libc++ and compiler-rt as default libraries in Fuchsia toolchain.

This is something I was already thinking about. Ideally, in our toolchain, we would build runtimes for all host and target platforms we support, i.e. on every host (Linux, macOS and Windows) we would build runtimes for i386, x86_64 and aarch64 Linux, x86_64 macOS, x86_64 Windows, and x86_64 and aarch64 Fuchsia. This means that the toolchain on every platform can target every other platform as long as you provide the right sysroot. However, there are two issues I'm aware of:

Sat, Nov 18, 6:16 PM
ast committed rL318614: [bpf] allow direct and indirect calls.
[bpf] allow direct and indirect calls
Sat, Nov 18, 5:35 PM
dcoughlin accepted D40073: [Analyzer] Non-determinism: don't sort indirect goto LabelDecl's by addresses.

Thanks for finding and fixing this!

Sat, Nov 18, 5:35 PM
ctopper committed rL318613: [X86] Switch cannonlake to use the SkylakeServer scheduling model instead of….
[X86] Switch cannonlake to use the SkylakeServer scheduling model instead of…
Sat, Nov 18, 5:26 PM
aaron.ballman added a comment to D39863: [clang-tidy] Relax the way misc-move-const-arg treats trivially copyable types.
In D39863#929572, @oleg wrote:

I've just rebased the patch over the top of LLVM/Clang/tools-extra master branches and tests pass: make check-clang-tools.

Could one of the reviewers commit this please?

Sat, Nov 18, 5:17 PM · Restricted Project
ctopper committed rL318612: [X86] Add skeleton support for icelake CPU..
[X86] Add skeleton support for icelake CPU.
Sat, Nov 18, 5:13 PM
ctopper committed rL318611: [X86] Fix 80 column violation and remove trailing whitespace. NFC.
[X86] Fix 80 column violation and remove trailing whitespace. NFC
Sat, Nov 18, 5:12 PM
oleg added a comment to D39863: [clang-tidy] Relax the way misc-move-const-arg treats trivially copyable types.

I've just rebased the patch over the top of LLVM/Clang/tools-extra master branches and tests pass: make check-clang-tools.

Sat, Nov 18, 5:04 PM · Restricted Project
ctopper committed rL318610: [X86] Add cpu detection for cannonlake..
[X86] Add cpu detection for cannonlake.
Sat, Nov 18, 4:46 PM