Page MenuHomePhabricator
Feed Advanced Search

Today

fhahn committed rG203f29b40ca3: [MemoryLocation] Use getForArgument in getForSource/getForDest. (NFC) (authored by fhahn).
[MemoryLocation] Use getForArgument in getForSource/getForDest. (NFC)
Sun, Dec 5, 3:13 AM

Yesterday

fhahn committed rGa9125792b3be: [MemoryLocation] Support missing atomic intrinsics in getForArg. (authored by fhahn).
[MemoryLocation] Support missing atomic intrinsics in getForArg.
Sat, Dec 4, 2:19 PM
fhahn committed rG89f0f2771abe: [BasicAA] Add atomic mem intrinsic tests. (authored by fhahn).
[BasicAA] Add atomic mem intrinsic tests.
Sat, Dec 4, 7:45 AM
fhahn added a comment to D115098: [NewPM] Fix MergeFunctions scheduling.

it would probably also good to have a pipeline ordering test that checks the placement of those extra passes

Sat, Dec 4, 7:10 AM · Restricted Project
fhahn accepted D115098: [NewPM] Fix MergeFunctions scheduling.

LGTM, this looks like an oversight and is closer to the position in the legacy pass manager.

Sat, Dec 4, 6:58 AM · Restricted Project

Fri, Dec 3

fhahn added a comment to D115052: [Passes] Only run extra vector passes if loops have been vectorized..

It would be good to mimic what the module equivalence check does
(when no changes were reported, the module should not have been changed),
and e.g. when in EXPENSIVE_CHECKS mode, if we weren't going to run those
extra vectorization passes because we predicted they wouldn't do anything,
to still run them, and check that they actually didn't do anything.

Fri, Dec 3, 1:33 PM · Restricted Project
fhahn updated the diff for D115052: [Passes] Only run extra vector passes if loops have been vectorized..

rebased on top of test changes, added TODO

Fri, Dec 3, 12:33 PM · Restricted Project
fhahn committed rG31413c4555ea: [Passes] Adjust SLPVectorizer placement in test. (authored by fhahn).
[Passes] Adjust SLPVectorizer placement in test.
Fri, Dec 3, 12:28 PM
fhahn committed rG5da920bf3a09: [Passes] Improve opt-pipeline-vector-passes.ll test. (authored by fhahn).
[Passes] Improve opt-pipeline-vector-passes.ll test.
Fri, Dec 3, 12:19 PM
fhahn added inline comments to D115052: [Passes] Only run extra vector passes if loops have been vectorized..
Fri, Dec 3, 12:08 PM · Restricted Project
fhahn requested review of D115052: [Passes] Only run extra vector passes if loops have been vectorized..
Fri, Dec 3, 10:39 AM · Restricted Project
fhahn committed rGead3979a92fc: [MemoryLocation] Move DSE intrinsic handling to MemoryLocation. (NFC) (authored by fhahn).
[MemoryLocation] Move DSE intrinsic handling to MemoryLocation. (NFC)
Fri, Dec 3, 8:01 AM
fhahn committed rGaf86aa798083: [MemoryLocation] Use None instead of {}. (NFC) (authored by fhahn).
[MemoryLocation] Use None instead of {}. (NFC)
Fri, Dec 3, 5:20 AM
fhahn added inline comments to D114526: [LV][NFC] New test case for compile time known trip count (TC).
Fri, Dec 3, 2:57 AM · Restricted Project
fhahn added inline comments to D114528: [LV] Make sure VF doesn't exceed compile time known TC.
Fri, Dec 3, 2:53 AM · Restricted Project
fhahn committed rGf078536f4659: [MemoryLocation] Move DSE's logic to new MemLoc::getForDest helper (NFC). (authored by fhahn).
[MemoryLocation] Move DSE's logic to new MemLoc::getForDest helper (NFC).
Fri, Dec 3, 1:12 AM
fhahn closed D114872: [MemoryLocation] Move DSE's logic to new MemLoc::getForDest helper (NFC)..
Fri, Dec 3, 1:12 AM · Restricted Project
fhahn committed rG829b29b619f3: [MemoryLocation] strcat/strncat/strcpy read/write after their args. (authored by fhahn).
[MemoryLocation] strcat/strncat/strcpy read/write after their args.
Fri, Dec 3, 12:48 AM
fhahn closed D114969: [MemoryLocation] strcat/strncat/strcpy read/write after their args..
Fri, Dec 3, 12:48 AM · Restricted Project

Thu, Dec 2

fhahn requested review of D114969: [MemoryLocation] strcat/strncat/strcpy read/write after their args..
Thu, Dec 2, 9:40 AM · Restricted Project
fhahn committed rG222442ec2d71: [BasicAA] Add tests for strcat/strncat/strcpy. (authored by fhahn).
[BasicAA] Add tests for strcat/strncat/strcpy.
Thu, Dec 2, 9:39 AM
fhahn committed rG68782a860dfa: [DSE] Read after strcpy test. (authored by fhahn).
[DSE] Read after strcpy test.
Thu, Dec 2, 9:38 AM
fhahn added a comment to D113544: [libFuzzer] Deflake entropic exec-time test..

Unfortunately it seems like this test still fails quite frequently., in particular for some of the sanitizer bots :(

Thu, Dec 2, 7:45 AM · Restricted Project
fhahn committed rG89d645dd3a60: [Clang] Fix LTO pipeline test after 770a50b28c00211f9a. (authored by fhahn).
[Clang] Fix LTO pipeline test after 770a50b28c00211f9a.
Thu, Dec 2, 7:44 AM
fhahn committed rG770a50b28c00: [AnnotationRemarks] Support generating annotation remarks with -O0. (authored by fhahn).
[AnnotationRemarks] Support generating annotation remarks with -O0.
Thu, Dec 2, 7:02 AM
fhahn committed rG639a78a4bf9d: [MemoryLocation] Support strncpy in getForArgument. (authored by fhahn).
[MemoryLocation] Support strncpy in getForArgument.
Thu, Dec 2, 6:18 AM
fhahn closed D114871: [MemoryLocation] Support strncpy in getForArgument..
Thu, Dec 2, 6:18 AM · Restricted Project
fhahn added inline comments to D114870: [MemoryLocation] Support memset_chk in getForArgument..
Thu, Dec 2, 5:51 AM · Restricted Project
fhahn committed rG9f9e8ba114ce: [MemoryLocation] Support memset_chk in getForArgument. (authored by fhahn).
[MemoryLocation] Support memset_chk in getForArgument.
Thu, Dec 2, 5:46 AM
fhahn closed D114870: [MemoryLocation] Support memset_chk in getForArgument..
Thu, Dec 2, 5:46 AM · Restricted Project
fhahn added a comment to D114586: [VPlan] Verify plan entry and exit blocks, set correct exit block..

ping :)

Thu, Dec 2, 3:55 AM · Restricted Project
fhahn committed rG47616c8855fd: [BasicAA] Add tests for memset_pattern{4,8,16}. (authored by fhahn).
[BasicAA] Add tests for memset_pattern{4,8,16}.
Thu, Dec 2, 3:51 AM
fhahn committed rG2de5f39e546d: [BuildLibCalls] Add support for memset_pattern{4,8}. (authored by fhahn).
[BuildLibCalls] Add support for memset_pattern{4,8}.
Thu, Dec 2, 3:05 AM
fhahn closed D114883: [BuildLibCalls] Add support for memset_pattern{4,8}..
Thu, Dec 2, 3:04 AM · Restricted Project
fhahn updated the diff for D114904: [BuildLibCalls] Add nounwind,willreturn to memset_pattern{4,8,16}..

rebase on top of precommitted test

Thu, Dec 2, 2:17 AM · Restricted Project
fhahn updated the diff for D114883: [BuildLibCalls] Add support for memset_pattern{4,8}..

rebase on top of precommitted tests

Thu, Dec 2, 2:16 AM · Restricted Project
fhahn committed rGd9bb9bd4bf7a: [InferAttrs] Add memset_pattern{4,8} declarations to test. (authored by fhahn).
[InferAttrs] Add memset_pattern{4,8} declarations to test.
Thu, Dec 2, 2:12 AM
fhahn committed rG0496edad4993: [BuildLibCalls] Add additional attrs to memcpy_chk. (authored by fhahn).
[BuildLibCalls] Add additional attrs to memcpy_chk.
Thu, Dec 2, 1:51 AM
fhahn closed D114863: [BuildLibCalls] Add additional attrs to memcpy_chk..
Thu, Dec 2, 1:50 AM · Restricted Project

Wed, Dec 1

fhahn committed rGad88a37ceadb: [TLI] Add memset_pattern4, memset_pattern8 lib functions. (authored by fhahn).
[TLI] Add memset_pattern4, memset_pattern8 lib functions.
Wed, Dec 1, 1:19 PM
fhahn closed D114881: [TLI] Add memset_pattern4, memset_pattern8 lib functions..
Wed, Dec 1, 1:19 PM · Restricted Project
fhahn retitled D114905: [MemoryLocation] Support memset_pattern{4,8} in getForArgument. from [MemoryLocation] support for memset_pattern{4,8} in getForDest. to [MemoryLocation] Support memset_pattern{4,8} in getForArgument..
Wed, Dec 1, 12:55 PM · Restricted Project
fhahn retitled D114905: [MemoryLocation] Support memset_pattern{4,8} in getForArgument. from [MemoryLocation] Add support for memset_pattern{4,8}. to [MemoryLocation] support for memset_pattern{4,8} in getForDest..
Wed, Dec 1, 12:54 PM · Restricted Project
fhahn requested review of D114906: [MemoryLocation] Handle memset_pattern{4,8,16} in getForDest..
Wed, Dec 1, 12:53 PM · Restricted Project
fhahn requested review of D114905: [MemoryLocation] Support memset_pattern{4,8} in getForArgument..
Wed, Dec 1, 12:52 PM · Restricted Project
fhahn requested review of D114904: [BuildLibCalls] Add nounwind,willreturn to memset_pattern{4,8,16}..
Wed, Dec 1, 12:51 PM · Restricted Project
fhahn updated the diff for D114881: [TLI] Add memset_pattern4, memset_pattern8 lib functions..

fix formatting and /llvm-tli-checker::ps4-tli-check.s test failure

Wed, Dec 1, 12:42 PM · Restricted Project
fhahn added inline comments to D114872: [MemoryLocation] Move DSE's logic to new MemLoc::getForDest helper (NFC)..
Wed, Dec 1, 12:33 PM · Restricted Project
fhahn committed rG5fe151f98f02: [DSE] Add libcall tests for functions only available on Darwin. (authored by fhahn).
[DSE] Add libcall tests for functions only available on Darwin.
Wed, Dec 1, 12:30 PM
fhahn added inline comments to D114881: [TLI] Add memset_pattern4, memset_pattern8 lib functions..
Wed, Dec 1, 12:17 PM · Restricted Project
fhahn added a comment to D114894: [unroll] Fix a functional change in an NFC patch.

Do you have a test case by any chance that would show the functional change?

Wed, Dec 1, 12:09 PM · Restricted Project
fhahn requested review of D114883: [BuildLibCalls] Add support for memset_pattern{4,8}..
Wed, Dec 1, 8:40 AM · Restricted Project
fhahn requested review of D114881: [TLI] Add memset_pattern4, memset_pattern8 lib functions..
Wed, Dec 1, 8:30 AM · Restricted Project
fhahn added inline comments to D114688: [Clang] Add __builtin_elementwise_ceil.
Wed, Dec 1, 8:19 AM · Restricted Project
fhahn added inline comments to D114688: [Clang] Add __builtin_elementwise_ceil.
Wed, Dec 1, 7:39 AM · Restricted Project
fhahn committed rGe44298a8f8af: [LV] Move code from vectorizeMemoryInstruction to recipe's execute(). (authored by fhahn).
[LV] Move code from vectorizeMemoryInstruction to recipe's execute().
Wed, Dec 1, 7:16 AM
fhahn closed D114324: [LV] Move code from vectorizeMemoryInstruction to recipe's execute()..
Wed, Dec 1, 7:15 AM · Restricted Project
fhahn requested review of D114872: [MemoryLocation] Move DSE's logic to new MemLoc::getForDest helper (NFC)..
Wed, Dec 1, 6:20 AM · Restricted Project
fhahn requested review of D114871: [MemoryLocation] Support strncpy in getForArgument..
Wed, Dec 1, 6:18 AM · Restricted Project
fhahn requested review of D114870: [MemoryLocation] Support memset_chk in getForArgument..
Wed, Dec 1, 6:17 AM · Restricted Project
fhahn committed rG524ad6babb8e: [BasicAA] Add memset_chk libfunc tests. (authored by fhahn).
[BasicAA] Add memset_chk libfunc tests.
Wed, Dec 1, 6:16 AM
fhahn committed rGc6bd63803f62: [BasicAA] Add strncpy libfunc tests. (authored by fhahn).
[BasicAA] Add strncpy libfunc tests.
Wed, Dec 1, 6:16 AM
fhahn added inline comments to D114793: [BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk..
Wed, Dec 1, 4:00 AM · Restricted Project
fhahn requested review of D114863: [BuildLibCalls] Add additional attrs to memcpy_chk..
Wed, Dec 1, 3:59 AM · Restricted Project
fhahn added inline comments to D114688: [Clang] Add __builtin_elementwise_ceil.
Wed, Dec 1, 3:57 AM · Restricted Project
fhahn committed rG6a5e29d13f8e: [BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk. (authored by fhahn).
[BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk.
Wed, Dec 1, 2:10 AM
fhahn closed D114793: [BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk..
Wed, Dec 1, 2:10 AM · Restricted Project
fhahn committed rG7de410440d46: [DSE] Allow DSE to optimize MemorySSA by default. (authored by fhahn).
[DSE] Allow DSE to optimize MemorySSA by default.
Wed, Dec 1, 12:29 AM
fhahn closed D113712: [DSE] Allow DSE to optimize MemorySSA by default..
Wed, Dec 1, 12:29 AM · Restricted Project

Tue, Nov 30

fhahn added inline comments to D114793: [BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk..
Tue, Nov 30, 1:57 PM · Restricted Project
fhahn added a comment to D114688: [Clang] Add __builtin_elementwise_ceil.

Thanks for the patch! I left a suggestion inline.

Tue, Nov 30, 12:57 PM · Restricted Project
fhahn added a comment to D114650: [SCEV] Construct SCEV iteratively (WIP)..

I want to make sure I understand the problem to be solved. The description and discussion confused me a bit, so I may in fact be wildly off here.

Is the basis issue being solved that calling createSCEV on some IR instruction can require recursing through a very deep chain of instructions which have not yet had SCEVs formed for them? More specifically, is that the *only* deep recursion we're concerned by? The discussion about new invariants for creating a SCEV seemed to indicate it might be much wider, but that discussion didn't parse for me.

Tue, Nov 30, 12:54 PM · Restricted Project
fhahn closed D114183: [ThreadPool] Support returning futures with results..
Tue, Nov 30, 10:11 AM · Restricted Project, Restricted Project
fhahn added a comment to D114163: Use VersionTuple for parsing versions in Triple. This makes it possible to distinguish between "16" and "16.0" after parsing, which previously was not possible..

It is also breaking building llvm-project/llvm/unittests/Support/Host.cpp on macOS: https://green.lab.llvm.org/green/job/clang-stage1-cmake-RA-incremental/25781/consoleFull

Tue, Nov 30, 9:56 AM · Restricted Project, Restricted Project
fhahn committed rGc9ad356266f3: [DSE] Use optimized access if available for redundant store elimination. (authored by fhahn).
[DSE] Use optimized access if available for redundant store elimination.
Tue, Nov 30, 7:41 AM
fhahn closed D112315: [DSE] Use optimized access if available for redundant store elimination..
Tue, Nov 30, 7:41 AM · Restricted Project
fhahn requested review of D114793: [BuildLibCalls] Add argmemonly, writeonly, nounwind to memset_chk..
Tue, Nov 30, 6:00 AM · Restricted Project
fhahn committed rG41d59a36459e: [DSE] Add memset_chk tests. (authored by fhahn).
[DSE] Add memset_chk tests.
Tue, Nov 30, 5:50 AM
fhahn committed rG6f4dd1b2dca8: [BuildLibCalls] Add memset_chk test. (authored by fhahn).
[BuildLibCalls] Add memset_chk test.
Tue, Nov 30, 5:50 AM
fhahn added inline comments to D114324: [LV] Move code from vectorizeMemoryInstruction to recipe's execute()..
Tue, Nov 30, 3:41 AM · Restricted Project
fhahn updated the diff for D114324: [LV] Move code from vectorizeMemoryInstruction to recipe's execute()..

remove commented out code

Tue, Nov 30, 3:41 AM · Restricted Project
fhahn committed rGdab776dd0fb4: [LV] Move code from widenSelectInstruction to VPWidenSelectRecipe. (NFC) (authored by fhahn).
[LV] Move code from widenSelectInstruction to VPWidenSelectRecipe. (NFC)
Tue, Nov 30, 2:33 AM
fhahn closed D114323: [LV] Move code from widenSelectInstruction to VPWidenSelectRecipe. (NFC).
Tue, Nov 30, 2:33 AM · Restricted Project

Mon, Nov 29

fhahn added a comment to D108699: [LAA] Analyze pointers forked by a select.

LGTM!

Hi @huntergr, thanks for making all the changes. I think the patch looks good to
go for now. It's still worth looking into the MemAccessInfo approach as a follow-up,
but the patch has been sat in review for long enough (3 months) without any
fundamental objections so I'd prefer we got something merged now to get the
functionality defended and unblock future work.

Mon, Nov 29, 2:07 PM · Restricted Project
fhahn added reviewers for D114487: [LAA] Support runtime checks for select GEP base pointers. (WIP): Ayal, anemet, david-arm, huntergr.
Mon, Nov 29, 1:58 PM · Restricted Project
fhahn updated the summary of D114487: [LAA] Support runtime checks for select GEP base pointers. (WIP).
Mon, Nov 29, 1:58 PM · Restricted Project
fhahn added a comment to D113183: [LV] Patch up induction phis after VPlan execution..

Unfortunately VPWidenIntOrFpInductionRecipe is already a multi-def in some cases (due to handling casts as well).

wonder how R->addOperand(IncR->getVPSingleValue()); currently works in fixHeaderPhis - is a testcase of a casted induction feeding a reduction is needed?

Mon, Nov 29, 1:57 PM · Restricted Project
fhahn added a comment to D114183: [ThreadPool] Support returning futures with results..

Could you close the review again if done?

Mon, Nov 29, 1:55 PM · Restricted Project, Restricted Project
fhahn accepted D114183: [ThreadPool] Support returning futures with results..
Mon, Nov 29, 1:54 PM · Restricted Project, Restricted Project
fhahn requested review of D114739: [LV] Always record recipe for optimized cast..
Mon, Nov 29, 1:54 PM · Restricted Project
fhahn requested review of D114720: [LV] Fix crash when optimized phi is feeding first-order recurrence..
Mon, Nov 29, 9:44 AM · Restricted Project
fhahn updated the diff for D114487: [LAA] Support runtime checks for select GEP base pointers. (WIP).

rebased on top of precommitted tests, simplified code

Mon, Nov 29, 7:25 AM · Restricted Project
fhahn added inline comments to D108371: [LAA] Add Memory dependence and unknown bounds remarks..
Mon, Nov 29, 3:20 AM · Restricted Project
fhahn added a comment to D113497: [IPSCCP] Support unfeasible default dests for switch..

LG.

If SCCP runs again, will it then again create a new unreachable block
for the default that is already pointing at the unreachable block?
Should it not do such redundant work?

Mon, Nov 29, 3:09 AM · Restricted Project
fhahn added a comment to D114650: [SCEV] Construct SCEV iteratively (WIP)..

>>>! In D114650#3157263, @lebedev.ri wrote:

To ask the obvious question: Can we get away with simply adding a getSCEV recursion limit instead? Is there reason to believe that a sensible recursion cutoff (similar to the many SCEV already has) would adversely affect analysis quality in practice? I'd rather avoid the additional complexity if we can.

Recursion itself is rarely the right solution in the first place, let alone working around it by introducing random cut-offs.

I think one issue with a cut-off for the recursion is that such a cut-off would make SCEV instantiations depend on the instantiation order. I am not sure how much this would be an issue in practice, but among other things it may make verification of SCEV harder (e.g. if we compute SCEVs from scratch to the cached version we would have to make sure to use the same evaluation order as when computing the original SCEV).

Mon, Nov 29, 3:01 AM · Restricted Project
fhahn added a comment to D114650: [SCEV] Construct SCEV iteratively (WIP)..

Thanks for looking into this!
I previously looked into this: https://github.com/LebedevRI/llvm-project/commit/6b006aa21caf018aa0f280828899d510274c8444
... but as it is apparent, never posted the diff. Nonetheless, it may be interesting to look at.

The most interesting question is, do we really want to avoid (indirect self-) recursion (through getSCEV)?
Then the api of ScalarEvolution needs more changes than what the current diff contains,
basically everything that can end up creating new SCEV expressions from IR
needs to be able to return "in-progress, requery me later" status.

Mon, Nov 29, 2:55 AM · Restricted Project
fhahn added a comment to D112315: [DSE] Use optimized access if available for redundant store elimination..

ping :)

Mon, Nov 29, 2:26 AM · Restricted Project
fhahn added inline comments to D114253: [LV] Check VPValue operand instead of Cost::isUniformAfterVec (NFC)..
Mon, Nov 29, 2:21 AM · Restricted Project
fhahn updated the diff for D114253: [LV] Check VPValue operand instead of Cost::isUniformAfterVec (NFC)..

Move logic to VPlan::isUniformAfterVectorization, use it

Mon, Nov 29, 2:20 AM · Restricted Project
fhahn added inline comments to D114322: [LV] Move code from widenInstruction to VPWidenRecipe. (NFC).
Mon, Nov 29, 1:12 AM · Restricted Project