Meinersbur (Michael Kruse)
User

Projects

User does not belong to any projects.

User Details

User Since
Jun 17 2015, 7:07 AM (174 w, 4 d)

Recent Activity

Thu, Oct 4

Meinersbur updated the diff for D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..
  • Rebase
Thu, Oct 4, 3:12 AM
Meinersbur updated the diff for D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..
Thu, Oct 4, 3:08 AM

Wed, Oct 3

Meinersbur added a comment to D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..

I actually already created a review for the Clang part which is D52117, but uploaded the wrong diff. Corrected. I'll wait for both being accepted before committing.

Wed, Oct 3, 4:09 AM
Meinersbur updated the diff for D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..
  • Upload diff for clang portion (instead of D52116)
Wed, Oct 3, 4:04 AM
Meinersbur added a comment to D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..

Thank you for your feedback. If you don'r mind, before I commit this, I will prepare a patch for clang generating this new kind of metadata (since atm nothing is generating it) and give others some time for feedback.

Wed, Oct 3, 1:29 AM

Mon, Oct 1

Meinersbur added a comment to D48348: [ADT] Add zip_longest iterators..

ping

Mon, Oct 1, 2:07 AM

Fri, Sep 28

Meinersbur updated the diff for D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
Fri, Sep 28, 4:24 AM
Meinersbur added inline comments to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
Fri, Sep 28, 4:23 AM

Thu, Sep 27

Herald added a reviewer for D33523: [Polly] [OpenMP] Update Subtree values for nested parallel loop: bollu.

Altough I fixed the related bug in r343212, this patch might still be useful in case we ever would want to allow nested parallelism.

Thu, Sep 27, 6:44 AM · Restricted Project
Meinersbur committed rL343212: [IslAst] Fix InParallelFor nesting..
[IslAst] Fix InParallelFor nesting.
Thu, Sep 27, 6:43 AM

Wed, Sep 26

Meinersbur committed rL343119: [doc] Fix HowToManuallyUseTheIndividualPiecesOfPolly.
[doc] Fix HowToManuallyUseTheIndividualPiecesOfPolly
Wed, Sep 26, 8:26 AM
Meinersbur committed rL343118: Move www/experiments to docs/experiments.
Move www/experiments to docs/experiments
Wed, Sep 26, 8:23 AM
Meinersbur updated the diff for D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..
  • Rebase
  • Use call access group if instruction's access group is not set
Wed, Sep 26, 4:04 AM

Mon, Sep 24

Meinersbur committed rL342861: Add inherited attributes before parsed attributes..
Add inherited attributes before parsed attributes.
Mon, Sep 24, 7:23 AM
Meinersbur committed rC342861: Add inherited attributes before parsed attributes..
Add inherited attributes before parsed attributes.
Mon, Sep 24, 7:22 AM
Meinersbur closed D50214: Add inherited attributes before parsed attributes..

Commited as rL342861.

Mon, Sep 24, 7:20 AM

Sep 19 2018

Meinersbur abandoned D50216: Pass IsInherited when merging attribute lists [RFC].
Sep 19 2018, 6:53 AM
Meinersbur abandoned D50215: Change attribute merge order [RFC].
Sep 19 2018, 6:52 AM
Meinersbur added a comment to D50214: Add inherited attributes before parsed attributes..

ping

Sep 19 2018, 6:52 AM

Sep 17 2018

Meinersbur added a comment to D51963: [Polly] Really fix Windows builds..

Currently, when I apply this, I get the following error (Windows and Linux):

CMake Error at cmake/modules/LLVM-Config.cmake:105 (target_link_libraries):
  Target "LLVMPolly" of type MODULE_LIBRARY may not be linked into another
  target.  One may link only to STATIC or SHARED libraries, or to executables
  with the ENABLE_EXPORTS property set.
Call Stack (most recent call first):
  cmake/modules/LLVM-Config.cmake:93 (explicit_llvm_config)
  cmake/modules/AddLLVM.cmake:774 (llvm_config)
  cmake/modules/AddLLVM.cmake:1108 (add_llvm_executable)
  tools/polly/unittests/CMakeLists.txt:9 (add_unittest)
  tools/polly/unittests/Isl/CMakeLists.txt:1 (add_polly_unittest)
Sep 17 2018, 2:26 PM
Meinersbur committed rL342406: [Loopinfo] Remove one latch-case in getLoopID. NFC..
[Loopinfo] Remove one latch-case in getLoopID. NFC.
Sep 17 2018, 11:44 AM
Meinersbur closed D52118: [Loopinfo] Remove one latch case in getLoopID. NFC..
Sep 17 2018, 11:44 AM
Meinersbur added a comment to D52118: [Loopinfo] Remove one latch case in getLoopID. NFC..

save an iteration over the loop's basic blocks (which is what getLoopLatch does)

I'm not sure this is true. getLoopLatch() in LoopInfoImpl.h
only traverses the children of the header in the inverse graph.
That should, I think, be similar to predecessors(Header) in case
of the IR CFG.

Sep 17 2018, 11:43 AM
Meinersbur updated the summary of D52118: [Loopinfo] Remove one latch case in getLoopID. NFC..
Sep 17 2018, 11:40 AM

Sep 14 2018

Meinersbur accepted D51836: [bugpoint] Revert r318459.

I talked with @hfinkel about what to do in such cases. He's ok with just reverting. So go ahead and commit the revert.

Sep 14 2018, 11:11 PM
Meinersbur created D52118: [Loopinfo] Remove one latch case in getLoopID. NFC..
Sep 14 2018, 12:56 PM
Meinersbur added inline comments to D49503: [test-suite] Add Image Processing Kernels Using Benchmark Library: Dither Algorithms.
Sep 14 2018, 12:51 PM
Meinersbur added a dependency for D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata.: D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..
Sep 14 2018, 12:06 PM
Meinersbur added a dependent revision for D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata.: D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..
Sep 14 2018, 12:06 PM
Meinersbur created D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..
Sep 14 2018, 12:06 PM
Meinersbur created D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..
Sep 14 2018, 12:06 PM

Sep 13 2018

Meinersbur updated subscribers of D51984: Fixes for `LLVM_LINK_LLVM_DYLIB` && Polly..

+ # LLVMPolly is the Polly loadable module target, the static archive is just Polly
It is unfortunate that Polly doesn't match the naming conventions of other LLVM components, but we have a lot of special case handling for this kind of thing anyways.

Sep 13 2018, 2:04 PM
Meinersbur added a comment to D51836: [bugpoint] Revert r318459.

Can you add a test case? Since I never contributed any code to bugpoint, I am not comfortable with just accepting it (usually one asks the original author, who, as you mentioned, is not active anymore). But it would obvious if there was a test case that illustrates the problem/ensure the bug does not reappear.

Sep 13 2018, 8:24 AM
Meinersbur added a comment to D51963: [Polly] Really fix Windows builds..

Does this fix the Windows or "non-MSVC" builds (as D51904)? For which configurations?

Sep 13 2018, 8:15 AM
Meinersbur added inline comments to D51984: Fixes for `LLVM_LINK_LLVM_DYLIB` && Polly..
Sep 13 2018, 7:02 AM

Sep 10 2018

Meinersbur added inline comments to D51836: [bugpoint] Revert r318459.
Sep 10 2018, 1:41 PM
Meinersbur added inline comments to D51836: [bugpoint] Revert r318459.
Sep 10 2018, 12:27 PM
Meinersbur added inline comments to D51836: [bugpoint] Revert r318459.
Sep 10 2018, 8:56 AM

Sep 8 2018

Meinersbur added a comment to D51836: [bugpoint] Revert r318459.

Sorry, this does not fix bug http://llvm.org/PR38390 for me. I still get:

$ bugpoint polly-timeout-grtestutils.ll -verify -opt-command=C:\Users\meinersbur\build\llvm\release\bin\opt.exe
Read input file      : 'polly-timeout-grtestutils.ll'
*** All input ok
Running selected passes on program to test for crash: C:\Users\meinersbur\build\llvm\release\bin\opt.exe: bugpoint-input-18caf6b.bc: error: Could not open input file: permission denied
Exited with error code '1'
Sep 8 2018, 11:36 PM

Sep 4 2018

Meinersbur added inline comments to D51634: Define PATH_MAX if missing..
Sep 4 2018, 9:21 AM

Aug 30 2018

Meinersbur added inline comments to D51080: litsupport/remote: Work without shared filesystem.
Aug 30 2018, 9:26 AM
Meinersbur accepted D51465: Revamp test-suite documentation.

Thanks for investing time into the documentation

Aug 30 2018, 9:26 AM
Meinersbur added inline comments to D51080: litsupport/remote: Work without shared filesystem.
Aug 30 2018, 9:03 AM

Aug 29 2018

Meinersbur accepted D51080: litsupport/remote: Work without shared filesystem.

I'll bring up a separate patch for the test-suite docu about running remote (which lives in the llvm repository and is confusing/out of date anyway at the moment).

Aug 29 2018, 1:31 PM
Meinersbur added a reviewer for D51048: cmake: Specify reference outputs in llvm_test_data(): proton.
Aug 29 2018, 8:13 AM
Meinersbur added a comment to D51080: litsupport/remote: Work without shared filesystem.

ping

Aug 29 2018, 8:11 AM

Aug 22 2018

Meinersbur added a comment to D51080: litsupport/remote: Work without shared filesystem.
  1. cmake builddir for cross-compilation
  2. make/ninja
  3. rsync pushes test-suite builddir to target
  4. llvm-lit locale, but run every command on remote using ssh
  5. including fpmc/VALIDATE step in *.test (SPEC2017 also uses llvm_add_host_executable)

The VERIFY: steps in the scripts are run remotely as well.

Aug 22 2018, 3:30 PM
Meinersbur added a comment to D51080: litsupport/remote: Work without shared filesystem.

Can you outline the intended workflow? My guess:

Aug 22 2018, 2:08 PM

Aug 20 2018

Meinersbur added inline comments to D50209: cmake: Explicitely specify benchmark data.
Aug 20 2018, 5:29 PM
Meinersbur added inline comments to D50209: cmake: Explicitely specify benchmark data.
Aug 20 2018, 4:23 PM
Meinersbur accepted D50209: cmake: Explicitely specify benchmark data.

I am ok with committing this, but maybe we should have someone else's opinion as well?

Aug 20 2018, 10:38 AM
Meinersbur added inline comments to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
Aug 20 2018, 10:16 AM
Meinersbur added inline comments to D50698: [UnJ] Ensure unroll_and_jam metadata is removed once consumed..
Aug 20 2018, 9:39 AM

Aug 17 2018

Meinersbur added inline comments to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
Aug 17 2018, 4:24 PM
Meinersbur updated the diff for D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
  • Report unroll-and-jam as not applied even if unroll is present as well.
    • rename followup_cyclic/followup_noncyclic to followup_sequential/followup_coincident
    • Move hasUnrollAndJamTransformation in LoopUnrollAndJamPass to different place
    • Remove some unrelated whitespace changes made by clang-format
    • Extract followup attribute names into constant
Aug 17 2018, 4:23 PM
Meinersbur committed rL340072: [AST] Adapt Polly to AnalysisSetTracker changes. NFC..
[AST] Adapt Polly to AnalysisSetTracker changes. NFC.
Aug 17 2018, 12:32 PM

Aug 16 2018

Meinersbur added a comment to D50805: Don't warn on returning the address of a label from a statement expression.

If you added a new option -Wret-addr-label as suggested above (for a total patch of +2 lines), then is it accurate to say:

  • if -Wret-addr-label was enabled by default, we know of at least one codebase that would pass -Wno-ret-addr-label to their build
  • if -Wret-addr-label was disabled by default, we don't know of any codebases that would voluntarily enable it And if nobody would enable it voluntarily... might as well eliminate it, right?
Aug 16 2018, 8:47 AM · Restricted Project

Aug 14 2018

Meinersbur added inline comments to D50075: [UnJ] Improve explicit loop count checks.
Aug 14 2018, 3:56 PM
Meinersbur added inline comments to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
Aug 14 2018, 3:46 PM
Meinersbur added a comment to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..

I am thinking about adding a LoopMetadataTacker (sort of a combination of LoopVectorizeHints and AssumptionTracker) analysis pass which would centralize the interpretation of that metadata and avoid the linear search through the metadata list when looking up a specific attribute.

Aug 14 2018, 3:44 PM
Meinersbur added inline comments to D50698: [UnJ] Ensure unroll_and_jam metadata is removed once consumed..
Aug 14 2018, 3:04 PM

Aug 13 2018

Meinersbur added inline comments to D50214: Add inherited attributes before parsed attributes..
Aug 13 2018, 2:11 PM
Meinersbur added a reviewer for D50642: [IslExprBuilder] Allow 64-bit ints in hasLargeInts(): grosser.
Aug 13 2018, 11:35 AM
Meinersbur added a comment to D50642: [IslExprBuilder] Allow 64-bit ints in hasLargeInts().

To add some details: The original failure was external/skia/src/gpu/GrTestUtils.cpp of the aosp buildbot. It expects a library function __muloti4 which was lowered from the intrinsic llvm.smul.with.overflow.i128. The next smaller llvm.smul.with.overflow.i64 should be able to handle 64 bit values (not just 63 bits).

Aug 13 2018, 9:56 AM
Meinersbur accepted D50529: [test-suite] Add Bilateral Filter kernel using benchmark library.
Aug 13 2018, 9:17 AM

Aug 10 2018

Meinersbur added a comment to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..

The added paragraph looks good to me on the implementation side specification. Looking forward to see the programmers (i.e., compiler users, not compiler writers) side pragma description, but that will not gate my review of this patch. There is a difference between specification forcing one behavior versus implementation choice ends up in the same behavior. I wanted the former, not the latter.

Aug 10 2018, 7:43 PM
Meinersbur committed rL339484: [DepInfo] Use isl++ in Dependences::isValidSchedule. NFC..
[DepInfo] Use isl++ in Dependences::isValidSchedule. NFC.
Aug 10 2018, 3:34 PM

Aug 9 2018

Meinersbur added a comment to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..

I'd like to see us explicitly saying that any subsequent explicit transformation metadata will be ignored for the given loop ---- if that's what we'll agree on, or be explicit about something else we'll agree on in the terms that can be clearly explainable to the programmers. "Compiler will skip all remaining transformations after the first failed transform" is pretty straightforward to the programmers. If anyone is proposing other behaviors, I'd like to also see how to explain those behaviors to the programmers.

Aug 9 2018, 8:43 PM
Meinersbur updated the diff for D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..
  • Explicitly document followup of not applied transformations to be ignored
  • Unroll/UnrollAndJam: Interpret enable/count/full as forced
  • Unroll/UnrollAndJam: Add tests for disable_nonforced combined with enable/count/full
  • Reduce size of unroll-pragmas_transform.ll
Aug 9 2018, 8:42 PM
Meinersbur added a comment to D50529: [test-suite] Add Bilateral Filter kernel using benchmark library.

Should the total execution time of benchmarks be around second or less?
Maybe reduce the number of tested parameters to reduce the total runtime?

Aug 9 2018, 3:22 PM
Meinersbur accepted D50063: [UnJ] Pull code out into a separate function NFC.

Ah, sorry, I now see HasConsistentIterationCounts has been suggested in that other ticket. Oh well, as I said, I was bike shedding names (but still think 'invariant' is better :-)) Anyway, please ignore if you disagree.

Aug 9 2018, 2:41 PM
Meinersbur accepted D50075: [UnJ] Improve explicit loop count checks.

There are other pragma tests in pragma.ll, which check combinations of unroll and unroll_and_jam pragmas. The current behaviour if there is both unroll metadata and unroll_and_jam metadata isn't currently very refined. I would expect, at least in the default pipeline, for the unroll metadata to be handled first in one of the early unroll passes.

Aug 9 2018, 2:27 PM
Meinersbur accepted D49883: [test-suite] Added Image Processing Kernels Using Benchmark Library: Dilate Kernel.

LGTM

Aug 9 2018, 11:59 AM
Meinersbur added a comment to D50322: Bitcode tests failure because of the change in r338787.

Could you close the review if it's not needed anymore?

Aug 9 2018, 8:54 AM

Aug 7 2018

Meinersbur edited reviewers for D50425: The fix for c++11 flag issue was fixed in svn r339154 but for one Blur benchmark., added: proton; removed: proton0001.
Aug 7 2018, 10:34 PM
Meinersbur closed D50379: New ImageProcessing tests were failing becuase c++ flags are used in c flags..

Seems that @proton already fixed this as part of rL339154. Closing this review. In any case, I want to thank you for opening it!

Aug 7 2018, 12:31 PM
Meinersbur added a comment to D50322: Bitcode tests failure because of the change in r338787.

Looks like an obvious change, but r338787 is by @MatzeB I'd wait for his ok.

Aug 7 2018, 9:36 AM
Meinersbur added inline comments to D50055: Update the coding standard about NFC changes and whitespace.
Aug 7 2018, 8:31 AM
Meinersbur added a comment to D50379: New ImageProcessing tests were failing becuase c++ flags are used in c flags..

@santosh Do you need someone to commit?

Aug 7 2018, 8:30 AM
Meinersbur accepted D50379: New ImageProcessing tests were failing becuase c++ flags are used in c flags..

Thank you. LGTM.

Aug 7 2018, 8:28 AM

Aug 6 2018

Meinersbur accepted D50345: [test-suite] Add Interpolation kernels Using Benchmark Library.

LGTM

Aug 6 2018, 7:15 PM
Meinersbur accepted D48046: [test-suite] Backprop kernel from Rodinia Benchmark.

LGTM

Aug 6 2018, 12:23 PM
Meinersbur added a comment to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..

What I'm not seeing from this RFC/patch is that, if the programmer specifies transformation behavior A -> B -> C, what happens if transformation A does not kick-in? Should we just warn that "A did not happen" and stop processing the request B and C?

Aug 6 2018, 12:10 PM
Meinersbur accepted D49341: [test-suite] Added Image Processing Kernels Using Benchmark Library: Blur Algorithms.

LGTM

Aug 6 2018, 8:32 AM

Aug 4 2018

Meinersbur added inline comments to D50055: Update the coding standard about NFC changes and whitespace.
Aug 4 2018, 7:48 PM

Aug 3 2018

Meinersbur added a comment to D50214: Add inherited attributes before parsed attributes..

For this patch the goal is to have the attributes in the AST in an order that is less surprising to consumers (including out-of-tree). If we change it now, new/revised code/diagnostics will be written to match this order.

Aug 3 2018, 8:07 PM
Meinersbur added a comment to D49281: [Unroll/UnrollAndJam/Vectorizer/Distribute] Add followup loop attributes..

Some extra tests for nonforced + a pragma would be good to see.

Aug 3 2018, 7:47 PM
Meinersbur added inline comments to D49341: [test-suite] Added Image Processing Kernels Using Benchmark Library: Blur Algorithms.
Aug 3 2018, 11:32 AM
Meinersbur accepted D49886: [test-suite] Add pathfinder kernel from Rodinia Benchmark.

LGTM

Aug 3 2018, 9:15 AM

Aug 2 2018

Meinersbur committed rL338810: [CodeGen] Emit parallel_loop_access for each loop in the loop stack..
[CodeGen] Emit parallel_loop_access for each loop in the loop stack.
Aug 2 2018, 9:43 PM
Meinersbur committed rC338810: [CodeGen] Emit parallel_loop_access for each loop in the loop stack..
[CodeGen] Emit parallel_loop_access for each loop in the loop stack.
Aug 2 2018, 9:43 PM
Meinersbur closed D48808: [CodeGen] Emit parallel_loop_access for each loop in the loop stack..
Aug 2 2018, 9:43 PM
Meinersbur added a comment to D48348: [ADT] Add zip_longest iterators..

Getting the size may require iterating the sequence in advance, which I tried to avoid in D48100.

Yeah, it's a tradeoff there, to be sure. Not sure if anyone else has thoughts on the design tradeoffs here - would hope other folks might chime in.

Aug 2 2018, 8:55 PM
Meinersbur added a comment to D48100: Append new attributes to the end of an AttributeList..

I have two approaches to tackle the wrong marker order: D50215 and D50216. IMHO both are too invasive to be justified for the small issue.

Aug 2 2018, 8:04 PM
Meinersbur added a dependency for D50216: Pass IsInherited when merging attribute lists [RFC]: D50214: Add inherited attributes before parsed attributes..
Aug 2 2018, 8:01 PM
Meinersbur added a dependent revision for D50214: Add inherited attributes before parsed attributes.: D50216: Pass IsInherited when merging attribute lists [RFC].
Aug 2 2018, 8:01 PM
Meinersbur created D50216: Pass IsInherited when merging attribute lists [RFC].
Aug 2 2018, 8:00 PM
Meinersbur created D50215: Change attribute merge order [RFC].
Aug 2 2018, 7:37 PM
Meinersbur created D50214: Add inherited attributes before parsed attributes..
Aug 2 2018, 7:23 PM
Meinersbur committed rL338800: Append new attributes to the end of an AttributeList..
Append new attributes to the end of an AttributeList.
Aug 2 2018, 6:22 PM