hans (Hans Wennborg)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 18 2013, 6:48 AM (239 w, 2 d)

Recent Activity

Yesterday

hans committed rL318578: Fix some -Wunused-variable warnings.
Fix some -Wunused-variable warnings
Fri, Nov 17, 4:49 PM

Wed, Nov 15

hans accepted D40109: [MS] Apply adjustments after storing 'this'.

lgtm

Wed, Nov 15, 5:17 PM
hans added a comment to D39994: Loosen MSVC 2017 path requirements.

I think the patch is fine, but Zach should probably sign off on it.

Wed, Nov 15, 5:01 PM
hans committed rL318314: Try to fix test/SemaCXX/deleted-operator.cpp after r318309.
Try to fix test/SemaCXX/deleted-operator.cpp after r318309
Wed, Nov 15, 9:48 AM
hans committed rL318309: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on….
BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on…
Wed, Nov 15, 9:13 AM
hans closed D39579: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on the target (PR35174) by committing rL318309: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on….
Wed, Nov 15, 9:13 AM
hans added inline comments to D39579: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on the target (PR35174).
Wed, Nov 15, 9:13 AM

Tue, Nov 14

hans committed rL318239: Revert r318193 "[SLPVectorizer] Failure to beneficially vectorize 'copyable'….
Revert r318193 "[SLPVectorizer] Failure to beneficially vectorize 'copyable'…
Tue, Nov 14, 4:39 PM
hans committed rL318220: Fix switch-lower-peel-top-case.ll isel pass is not registered error.
Fix switch-lower-peel-top-case.ll isel pass is not registered error
Tue, Nov 14, 3:32 PM
hans committed rL318214: Try to fix the instrument-functions tests.
Try to fix the instrument-functions tests
Tue, Nov 14, 3:11 PM
hans committed rL318209: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option.
CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option
Tue, Nov 14, 2:33 PM
hans closed D40050: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option by committing rL318209: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option.
Tue, Nov 14, 2:32 PM
hans added inline comments to D40050: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option.
Tue, Nov 14, 2:32 PM
hans added a comment to D40050: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option.

I tested locally that if there's a current LLVM_ENABLE_LIBXML2:BOOL in CMakeCache.txt, that doesn't break things. There is no warning or error, it just seems to ignore the new string option and use the cached bool one.

Tue, Nov 14, 2:25 PM
hans created D40050: CMake: Turn LLVM_ENABLE_LIBXML2 into a tri-state option.
Tue, Nov 14, 2:21 PM
hans committed rL318199: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter….
Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter…
Tue, Nov 14, 1:15 PM
hans closed D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes by committing rL318199: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter….
Tue, Nov 14, 1:15 PM
hans committed rL318195: Rename CountingFunctionInserter and use for both mcount and cygprofile calls….
Rename CountingFunctionInserter and use for both mcount and cygprofile calls…
Tue, Nov 14, 1:10 PM
hans closed D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining by committing rL318195: Rename CountingFunctionInserter and use for both mcount and cygprofile calls….
Tue, Nov 14, 1:10 PM
hans added inline comments to D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.
Tue, Nov 14, 11:53 AM
hans updated the diff for D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.

Renaming the new flag and making it sufficient to turn on instrumentation by itself.

Tue, Nov 14, 11:53 AM
hans updated the diff for D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.

Minor updates to the diff, updating the description to match the new approach.

Tue, Nov 14, 10:58 AM
hans added inline comments to D39995: [SelectionDAG] Make sorting predicate stronger to remove non-deterministic ordering.
Tue, Nov 14, 10:11 AM

Mon, Nov 13

hans committed rL318115: Update some code.google.com links.
Update some code.google.com links
Mon, Nov 13, 3:48 PM
hans committed rL318110: Update link to protobuf.
Update link to protobuf
Mon, Nov 13, 3:28 PM
hans committed rL318109: Update a link to the old code.google.com bug tracker.
Update a link to the old code.google.com bug tracker
Mon, Nov 13, 3:28 PM
hans committed rL318108: Update link to the Chromium Clang page.
Update link to the Chromium Clang page
Mon, Nov 13, 3:28 PM
hans added a comment to D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining.
In D39287#923553, @rnk wrote:
In D39287#922534, @hans wrote:

Rebasing and addressing comments.

I've also added the pass in PassManagerBuilder::populateFunctionPassManager() because I realized that's how Clang sets up the passes.

One thing that worries me is that this set-up is pretty fragile against how the passes are run. I think it will currently work, but if the pass were to somehow run more than once, the instrumentation could get inserted multiple times. One way to avoid that would be to remove the attribute (or probably faster, give it an empty value?) when the pass "consumes" the attribute. Does that sound reasonable?

What do other instrumentation passes do to defend against this? I guess they just don't have this problem because they don't appear twice in the pipeline?

Mon, Nov 13, 2:00 PM
hans added inline comments to D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.
Mon, Nov 13, 1:47 PM

Fri, Nov 10

hans updated the diff for D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining.

Rebasing and addressing comments.

Fri, Nov 10, 4:51 PM
hans updated the diff for D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.

Updating to match the LLVM-side: https://reviews.llvm.org/D39287#inline-348143

Fri, Nov 10, 4:51 PM
hans committed rL317944: Remove declaration of EmitMCountInstrumentation(). NFC.
Remove declaration of EmitMCountInstrumentation(). NFC
Fri, Nov 10, 2:35 PM

Thu, Nov 9

hans updated the diff for D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining.

Here's a new approach which renames and extends the existing CountingFunctionInserter to run both before and after inlining (with the old PM it's two passes actually) and handle both cygprofile and mcount calls as controlled by function attributes.

Thu, Nov 9, 4:29 PM
hans added a comment to D39761: [Support] Make llvm::Error faster.

r317824

Thu, Nov 9, 11:36 AM

Tue, Nov 7

hans accepted D39761: [Support] Make llvm::Error faster.

Maybe mention llvm::Expected in the commit message too.

Tue, Nov 7, 8:01 PM
hans added inline comments to D39761: [Support] Make llvm::Error faster.
Tue, Nov 7, 2:36 PM
hans removed a reviewer for D39761: [Support] Make llvm::Error faster: hansw.
Tue, Nov 7, 2:14 PM
hans added inline comments to D39761: [Support] Make llvm::Error faster.
Tue, Nov 7, 2:13 PM
hans accepted D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.

Thanks for working on this!

Tue, Nov 7, 10:07 AM

Mon, Nov 6

hans updated the diff for D39579: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on the target (PR35174).

Addressing comments and changing the test to exercise Sema more directly.

Mon, Nov 6, 5:47 PM
hans added inline comments to D39579: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on the target (PR35174).
Mon, Nov 6, 5:43 PM
hans added inline comments to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.
Mon, Nov 6, 4:20 PM
hans committed rL317518: Revert r317510 "[InstCombine] Pull shifts through a select plus binop with….
Revert r317510 "[InstCombine] Pull shifts through a select plus binop with…
Mon, Nov 6, 2:28 PM
hans committed rL317517: Revert r316064 "Fix the incorrect detection of ICONV_LIBRARY_PATH".
Revert r316064 "Fix the incorrect detection of ICONV_LIBRARY_PATH"
Mon, Nov 6, 2:18 PM

Fri, Nov 3

hans added inline comments to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.
Fri, Nov 3, 8:55 AM

Thu, Nov 2

hans created D39579: BuiltinOperatorOverloadBuilder: Don't consider types that are unavailable on the target (PR35174).
Thu, Nov 2, 4:32 PM
hans committed rL317263: Fix llvm-dsymutil test in -DLLVM_ENABLE_THREADS=OFF mode.
Fix llvm-dsymutil test in -DLLVM_ENABLE_THREADS=OFF mode
Thu, Nov 2, 1:22 PM

Wed, Nov 1

hans added inline comments to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.
Wed, Nov 1, 4:17 PM
hans committed rL317131: Publication: Devirtualization in LLVM.
Publication: Devirtualization in LLVM
Wed, Nov 1, 3:04 PM
hans added a comment to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.

Starting to look really good.

Wed, Nov 1, 2:24 PM
hans added a comment to D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.
In D39052#900899, @hans wrote:

When we ran into the test failures last time, in PR33859, it turned out there were no upstream buildbots running the tests, so we didn't find out they were failing until we started the release process. Are the tests run by any buildbots now?

Good point, let me find that out and come back to you.

Turned out they were not (yet). See linked revision that adapts the buildbot configuration to change that.

Hi Hans,

the following bots are now configured to test libomptarget:
http://lab.llvm.org:8011/builders/openmp-clang-x86_64-linux-debian
http://lab.llvm.org:8011/builders/openmp-clang-ppc64le-linux-debian
And this one will just build libomptarget with GCC:
http://lab.llvm.org:8011/builders/openmp-gcc-x86_64-linux-debian

Unfortunately, there is still an error with the arguments passed to lit that I just fixed but the master needs to be restarted...

Wed, Nov 1, 10:56 AM
hans added a comment to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.

Thanks! I think that's looking better.

Wed, Nov 1, 10:39 AM

Fri, Oct 27

hans added a comment to D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.

The request for -fno-cygprofile-args was from another user. You raise a good point, it sounds like mcount() might be perfect for them :-)

Fri, Oct 27, 2:14 AM

Thu, Oct 26

hans created D39331: Switch -mcount and -finstrument-functions to emit EnterExitInstrumenter attributes.
Thu, Oct 26, 6:31 AM
hans committed rL316644: Tidy up CountingFunctionInserter a little. NFC..
Tidy up CountingFunctionInserter a little. NFC.
Thu, Oct 26, 1:29 AM
hans added a comment to D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining.

If you only want these calls post-inlining (and otherwise wish to minimize the effect on the rest of the optimizer),

Thu, Oct 26, 1:22 AM

Wed, Oct 25

hans created D39287: Use CountingFunctionInserter both mcount and cygprofile calls, before and after inlining.
Wed, Oct 25, 7:08 AM
hans added a comment to D39262: [CodeGen] Peel off the dominant case in switch statement in lowering.

Thanks! I think doing it here rather than in SimplifyCFG is much better.

Wed, Oct 25, 3:31 AM

Mon, Oct 23

hans committed rL316312: Win snapshot: r316086 + separate clang-format binary.
Win snapshot: r316086 + separate clang-format binary
Mon, Oct 23, 12:54 AM

Sun, Oct 22

hans updated subscribers of rL310905: Avoid PointerIntPair of constexpr EvalInfo structs.

+Tom who manages the 5.0.1 release.

Sun, Oct 22, 11:58 PM

Oct 18 2017

hans added a comment to D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.

When we ran into the test failures last time, in PR33859, it turned out there were no upstream buildbots running the tests, so we didn't find out they were failing until we started the release process. Are the tests run by any buildbots now?

Oct 18 2017, 7:44 AM

Oct 17 2017

hans accepted D39011: [SimplifyCFG] try harder to forward switch condition to phi (PR34471).

lgtm

Oct 17 2017, 4:16 PM

Oct 16 2017

hans updated subscribers of D38972: Add the /v option which prints out the LLD version string..

Nico, this addresses your needs, right?

Oct 16 2017, 1:57 PM
hans added a comment to D38972: Add the /v option which prints out the LLD version string..

Would it be easy to print the version number as part of /? instead? Maybe that's safer.

Oct 16 2017, 1:44 PM

Oct 12 2017

hans committed rL315597: Work around lack of Wine support for SetFileInformationByHandle harder.
Work around lack of Wine support for SetFileInformationByHandle harder
Oct 12 2017, 10:38 AM

Oct 11 2017

hans committed rL315554: Revert r315545 "llvm-isel-fuzzer: Handle a subset of backend flags in the….
Revert r315545 "llvm-isel-fuzzer: Handle a subset of backend flags in the…
Oct 11 2017, 8:32 PM
hans committed rL315553: Revert r315533 "Reland "[sanitizer] Introduce ReservedAddressRange to….
Revert r315533 "Reland "[sanitizer] Introduce ReservedAddressRange to…
Oct 11 2017, 8:23 PM
hans committed rL315520: Support: Work around missing SetFileInformationByHandle on Wine.
Support: Work around missing SetFileInformationByHandle on Wine
Oct 11 2017, 3:04 PM
hans closed D38817: Support: Work around missing SetFileInformationByHandle on Wine by committing rL315520: Support: Work around missing SetFileInformationByHandle on Wine.
Oct 11 2017, 3:04 PM
hans added inline comments to D38817: Support: Work around missing SetFileInformationByHandle on Wine.
Oct 11 2017, 2:37 PM
hans updated the diff for D38817: Support: Work around missing SetFileInformationByHandle on Wine.

Check for ERROR_CALL_NOT_IMPLEMENTED instead.

Oct 11 2017, 2:26 PM
hans added inline comments to D38817: Support: Work around missing SetFileInformationByHandle on Wine.
Oct 11 2017, 2:25 PM
hans accepted D38776: [codeview] Implement FPO data assembler directives.

looks amazing to me

Oct 11 2017, 2:05 PM
hans created D38817: Support: Work around missing SetFileInformationByHandle on Wine.
Oct 11 2017, 1:09 PM

Oct 10 2017

hans committed rL315374: XFAIL ubsan/TestCases/TypeCheck/Function/function.cpp on Windows.
XFAIL ubsan/TestCases/TypeCheck/Function/function.cpp on Windows
Oct 10 2017, 3:11 PM
hans committed rL315330: For dllexport class templates, export specializations of member functions….
For dllexport class templates, export specializations of member functions…
Oct 10 2017, 9:53 AM

Oct 5 2017

hans committed rL315025: For dllexport class templates, export specializations of member functions….
For dllexport class templates, export specializations of member functions…
Oct 5 2017, 2:47 PM

Oct 4 2017

hans committed rL314936: Fix a -Wparentheses warning. NFC..
Fix a -Wparentheses warning. NFC.
Oct 4 2017, 2:15 PM
hans committed rL314919: Revert r314886 "[X86] Improvement in CodeGen instruction selection for LEAs (re….
Revert r314886 "[X86] Improvement in CodeGen instruction selection for LEAs (re…
Oct 4 2017, 10:56 AM

Oct 3 2017

hans committed rL314827: Update test expectation after r314821.
Update test expectation after r314821
Oct 3 2017, 11:48 AM
hans committed rL314826: Fix -Wcovered-switch-default warnings from r314821.
Fix -Wcovered-switch-default warnings from r314821
Oct 3 2017, 11:46 AM
hans committed rL314825: Revert r314817 "[dwarfdump] Add -lookup option".
Revert r314817 "[dwarfdump] Add -lookup option"
Oct 3 2017, 11:41 AM
hans committed rL314824: Revert r314806 "[SLP] Vectorize jumbled memory loads.".
Revert r314806 "[SLP] Vectorize jumbled memory loads."
Oct 3 2017, 11:34 AM
hans committed rL314821: CodeView: Provide a .def file with the register ids.
CodeView: Provide a .def file with the register ids
Oct 3 2017, 11:29 AM
hans closed D38480: CodeView: Provide a .def file with the register ids by committing rL314821: CodeView: Provide a .def file with the register ids.
Oct 3 2017, 11:29 AM
hans updated the diff for D38480: CodeView: Provide a .def file with the register ids.

Dropping the AMD64_ prefixes except for the YMM registers.

Oct 3 2017, 9:33 AM

Oct 2 2017

hans added inline comments to D38480: CodeView: Provide a .def file with the register ids.
Oct 2 2017, 5:18 PM
hans committed rL314737: Win snapshot: r314411..
Win snapshot: r314411.
Oct 2 2017, 4:18 PM
hans committed rL314736: Build fix: don't try to link in lldConfig.
Build fix: don't try to link in lldConfig
Oct 2 2017, 4:11 PM
hans created D38480: CodeView: Provide a .def file with the register ids.
Oct 2 2017, 3:57 PM
hans committed rL314709: Fix two header comments. NFC..
Fix two header comments. NFC.
Oct 2 2017, 12:50 PM
hans committed rL314690: CodeView symbol dumper: use symbolic names for registers.
CodeView symbol dumper: use symbolic names for registers
Oct 2 2017, 10:46 AM
hans added inline comments to D38469: CodeView symbol dumper: use symbolic names for registers.
Oct 2 2017, 10:44 AM
hans added inline comments to D38469: CodeView symbol dumper: use symbolic names for registers.
Oct 2 2017, 10:34 AM
hans created D38469: CodeView symbol dumper: use symbolic names for registers.
Oct 2 2017, 10:33 AM

Sep 28 2017

hans committed rL314420: Docs: fix link to Debugger intrinsic functions.
Docs: fix link to Debugger intrinsic functions
Sep 28 2017, 8:18 AM

Sep 25 2017

hans committed rL314033: clang-format plugin: Add missing NL (new line) at EOF (end of file).
clang-format plugin: Add missing NL (new line) at EOF (end of file)
Sep 25 2017, 5:06 AM
hans closed D37732: Add missing NL (new line) at EOF (end of file) by committing rL314033: clang-format plugin: Add missing NL (new line) at EOF (end of file).
Sep 25 2017, 5:06 AM · Restricted Project

Sep 22 2017

hans accepted D37732: Add missing NL (new line) at EOF (end of file).

This looks good to me. Do you have commit access, or would you like me to commit for you?

Sep 22 2017, 9:25 AM · Restricted Project

Sep 20 2017

hans committed rL313781: Revert r313771 "[SLP] Vectorize jumbled memory loads.".
Revert r313771 "[SLP] Vectorize jumbled memory loads."
Sep 20 2017, 11:02 AM