Page MenuHomePhabricator

bjope (Bjorn Pettersson)
User

Projects

User does not belong to any projects.

User Details

User Since
Sep 26 2016, 7:58 AM (226 w, 16 h)

Recent Activity

Fri, Jan 22

bjope committed rGea2cfda386f1: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC (authored by bjope).
[CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC
Fri, Jan 22, 12:15 PM
bjope committed rG72f863fd37c3: [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC (authored by bjope).
[CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC
Fri, Jan 22, 12:15 PM
bjope closed D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.
Fri, Jan 22, 12:15 PM · Restricted Project
bjope closed D94977: [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC.
Fri, Jan 22, 12:14 PM · Restricted Project

Thu, Jan 21

bjope added a comment to D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.

@lebedev.ri : I think I need your blessing as well on this, considering your earlier concerns. Is it still just confusing? (I doubt that we want/can replace all uses of getCharWidth/getIntWidth etc in clang with integers)

Thu, Jan 21, 9:43 AM · Restricted Project
bjope added a comment to D92309: [LegacyPM] Update InversedLastUser on the fly. NFC..

Still LGTM

Thu, Jan 21, 9:33 AM · Restricted Project
bjope retitled D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC from [CGExpr] Use getCharWidth() more consistently in ConstantAggregateBuilderUtils. NFC to [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.
Thu, Jan 21, 2:01 AM · Restricted Project
bjope updated the diff for D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.

Updated to add CharTy to the CodeGenTypeCache (based on review feedback).

Thu, Jan 21, 2:00 AM · Restricted Project

Wed, Jan 20

bjope added a comment to D92309: [LegacyPM] Update InversedLastUser on the fly. NFC..

LGTM

Thanks. This patch depends on D92308. Without D92308, this patch would add lots of extra lines to the output of opt -debug-pass=Structure, which breaks various tests in test/CodeGen and test/Other.

Wed, Jan 20, 6:49 AM · Restricted Project
bjope accepted D92309: [LegacyPM] Update InversedLastUser on the fly. NFC..
Wed, Jan 20, 4:59 AM · Restricted Project
bjope added a comment to D92309: [LegacyPM] Update InversedLastUser on the fly. NFC..

LGTM

Wed, Jan 20, 4:58 AM · Restricted Project
bjope committed rG985b9b7e421a: [PM] Avoid duplicates in the Used/Preserved/Required sets (authored by bjope).
[PM] Avoid duplicates in the Used/Preserved/Required sets
Wed, Jan 20, 4:56 AM
bjope closed D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets.
Wed, Jan 20, 4:56 AM · Restricted Project
bjope added reviewers for D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC: rjmccall, rsmith.
Wed, Jan 20, 4:51 AM · Restricted Project
bjope added reviewers for D94977: [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC: rjmccall, rsmith.
Wed, Jan 20, 4:51 AM · Restricted Project
bjope added inline comments to D94964: Describe vector layout in LangRef.
Wed, Jan 20, 2:54 AM · Restricted Project
bjope added inline comments to D92309: [LegacyPM] Update InversedLastUser on the fly. NFC..
Wed, Jan 20, 2:11 AM · Restricted Project

Tue, Jan 19

bjope retitled D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC from [CGExpr] Honor getCharWidth() in ConstantAggregateBuilderUtils to [CGExpr] Use getCharWidth() more consistently in ConstantAggregateBuilderUtils. NFC.
Tue, Jan 19, 11:10 AM · Restricted Project
bjope retitled D94977: [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC from [CodeGen] Honor getCharWidth() in CGRecordLowering to [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC.
Tue, Jan 19, 11:03 AM · Restricted Project
bjope added a comment to D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.

As far as i recall, all RFC's about LLVM support for non-8-bit char targets didn't end in favor of the proposal
I'd say that until there's general consensus/buy-in (including testing path), cleaning stuff up will only confuse things.
CC @jfb

Tue, Jan 19, 10:06 AM · Restricted Project
bjope requested review of D94979: [CGExpr] Use getCharWidth() more consistently in CCGExprConstant. NFC.
Tue, Jan 19, 9:20 AM · Restricted Project
bjope requested review of D94977: [CodeGen] Use getCharWidth() more consistently in CGRecordLowering. NFC.
Tue, Jan 19, 9:19 AM · Restricted Project

Mon, Jan 18

bjope added a reviewer for D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets: chandlerc.
Mon, Jan 18, 6:55 AM · Restricted Project

Sat, Jan 16

bjope committed rG4f155567317d: [LegalizeDAG] Handle NeedInvert when expanding BR_CC (authored by bjope).
[LegalizeDAG] Handle NeedInvert when expanding BR_CC
Sat, Jan 16, 5:40 AM
bjope closed D94762: [LegalizeDAG] Handle NeedInvert when expanding BR_CC.
Sat, Jan 16, 5:40 AM · Restricted Project

Fri, Jan 15

bjope added inline comments to D94450: [LegalizeDAG][RISCV][PowerPC][AMDGPU][WebAssembly] Improve expansion of SETONE/SETUEQ on targets without SETO/SETUO..
Fri, Jan 15, 4:00 AM · Restricted Project
bjope requested review of D94762: [LegalizeDAG] Handle NeedInvert when expanding BR_CC.
Fri, Jan 15, 3:51 AM · Restricted Project

Thu, Jan 14

bjope added inline comments to D94450: [LegalizeDAG][RISCV][PowerPC][AMDGPU][WebAssembly] Improve expansion of SETONE/SETUEQ on targets without SETO/SETUO..
Thu, Jan 14, 12:11 PM · Restricted Project
bjope added inline comments to D94450: [LegalizeDAG][RISCV][PowerPC][AMDGPU][WebAssembly] Improve expansion of SETONE/SETUEQ on targets without SETO/SETUO..
Thu, Jan 14, 11:03 AM · Restricted Project
bjope committed rGd58512b2e31a: [SLP] Don't vectorize stores of non-packed types (like i1, i2) (authored by bjope).
[SLP] Don't vectorize stores of non-packed types (like i1, i2)
Thu, Jan 14, 2:31 AM
bjope closed D94446: [SLP] Don't vectorize stores of non-packed types (like i1, i2).
Thu, Jan 14, 2:31 AM · Restricted Project

Wed, Jan 13

bjope added a comment to D94446: [SLP] Don't vectorize stores of non-packed types (like i1, i2).

Looks good, but could you please precommit test (and rebase) to see actual output difference?

Wed, Jan 13, 1:08 PM · Restricted Project
bjope added a comment to D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets.

I've now tried to do some performance comparisons using perf stat -r 100 opt -O3 -o /dev/null --verify-dom-info -verify-assumption-cache -verify-loop-info.

Wed, Jan 13, 1:01 PM · Restricted Project

Tue, Jan 12

bjope added a comment to D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets.

I created this patch as my earlier patch (https://reviews.llvm.org/D94138) resulted in even more duplicates in the preserved sets. So the idea was to compensate a bit for a potential speed regression, when using verifiers, by simply getting rid of all duplicates in the sets.

Tue, Jan 12, 9:43 AM · Restricted Project
bjope added a comment to D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets.

Does this make any difference in practice? E.g. does the output of opt -O1 -debug-pass=Executions change, or can you measure any timing difference?

Tue, Jan 12, 9:36 AM · Restricted Project
bjope updated the diff for D94446: [SLP] Don't vectorize stores of non-packed types (like i1, i2).

Rebase after pre-committing the test case.

Tue, Jan 12, 7:37 AM · Restricted Project
bjope committed rGdd07d60ec339: [SLP] Add test case showing a bug when dealing with padded types (authored by bjope).
[SLP] Add test case showing a bug when dealing with padded types
Tue, Jan 12, 7:36 AM
bjope added a comment to D86079: [X86] Improved lowering for saturating float to int..

I landed this on behalf of @ebevhan

Tue, Jan 12, 6:46 AM · Restricted Project
bjope committed rG07605ea1f3c9: [X86] Improved lowering for saturating float to int. (authored by ebevhan).
[X86] Improved lowering for saturating float to int.
Tue, Jan 12, 6:45 AM
bjope closed D86079: [X86] Improved lowering for saturating float to int..
Tue, Jan 12, 6:45 AM · Restricted Project
bjope added a reviewer for D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets: foad.
Tue, Jan 12, 4:57 AM · Restricted Project
bjope added a comment to D86632: [Fixed Point] Add codegen for conversion between fixed-point and floating point..

I've just landed this on behalf of @ebevhan.

Tue, Jan 12, 4:55 AM · Restricted Project, Restricted Project
bjope committed rGc4944a6f53f6: [Fixed Point] Add codegen for conversion between fixed-point and floating point. (authored by ebevhan).
[Fixed Point] Add codegen for conversion between fixed-point and floating point.
Tue, Jan 12, 4:53 AM
bjope closed D86632: [Fixed Point] Add codegen for conversion between fixed-point and floating point..
Tue, Jan 12, 4:53 AM · Restricted Project, Restricted Project

Mon, Jan 11

bjope added inline comments to D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.
Mon, Jan 11, 4:26 PM · Restricted Project
bjope added inline comments to D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.
Mon, Jan 11, 3:48 PM · Restricted Project
bjope requested review of D94446: [SLP] Don't vectorize stores of non-packed types (like i1, i2).
Mon, Jan 11, 1:23 PM · Restricted Project
bjope committed rG32c073acb320: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT (authored by bjope).
[GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT
Mon, Jan 11, 12:54 PM
bjope closed D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.
Mon, Jan 11, 12:54 PM · Restricted Project
bjope requested review of D94416: [PM] Avoid duplicates in the Used/Preserved/Required sets.
Mon, Jan 11, 7:15 AM · Restricted Project
bjope added a comment to D54749: Saturating float to int casts..

@bjope There are two more revisions based on this one, D86079 implements improved X86 lowerings and is already accepted, and D86078 implements improved AArch64 lowerings. Do you plan to land these as well?

Mon, Jan 11, 6:27 AM · Restricted Project
bjope committed rG675be6510624: Require chained analyses in BasicAA and AAResults to be transitive (authored by bjope).
Require chained analyses in BasicAA and AAResults to be transitive
Mon, Jan 11, 2:51 AM
bjope closed D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Mon, Jan 11, 2:51 AM · Restricted Project
bjope added inline comments to D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Mon, Jan 11, 1:49 AM · Restricted Project

Sat, Jan 9

bjope added a reviewer for D94138: Require chained analyses in BasicAA and AAResults to be transitive: foad.
Sat, Jan 9, 4:27 PM · Restricted Project

Thu, Jan 7

bjope added inline comments to D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Thu, Jan 7, 5:48 AM · Restricted Project

Wed, Jan 6

bjope added inline comments to D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Wed, Jan 6, 10:08 AM · Restricted Project

Tue, Jan 5

bjope added reviewers for D94138: Require chained analyses in BasicAA and AAResults to be transitive: nikic, fhahn, hfinkel.
Tue, Jan 5, 4:22 PM · Restricted Project
bjope updated subscribers of D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Tue, Jan 5, 4:20 PM · Restricted Project
bjope requested review of D94138: Require chained analyses in BasicAA and AAResults to be transitive.
Tue, Jan 5, 4:09 PM · Restricted Project

Mon, Jan 4

bjope added a comment to D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.

Ping!

Mon, Jan 4, 11:41 PM · Restricted Project
bjope updated subscribers of D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.
Mon, Jan 4, 11:41 PM · Restricted Project
bjope added a comment to rGb21840751278: [ValueTracking] Handle more non-trivial conditions in isKnownNonZero().

There is a bunch of lit tests that fail if adding a DT->verify() check in isKnownNonZero.

Mon, Jan 4, 11:16 AM

Wed, Dec 30

bjope added a comment to rGb21840751278: [ValueTracking] Handle more non-trivial conditions in isKnownNonZero().

Hi @nikic,

I've seen a miscompile for my OOT target with this patch. Unfortunately I haven't gone to the bottom with it yet (and I don't think I will until after new years), so I'm not sure what the problem is exactly, but I think that AA comes to the conclusion that a read and a write to the exact same location doesn't alias, and then things go wrong. Are you aware of any problems?

Wed, Dec 30, 5:12 PM

Dec 18 2020

bjope committed rGa89d751fb401: Add intrinsics for saturating float to int casts (authored by bjope).
Add intrinsics for saturating float to int casts
Dec 18 2020, 2:10 AM
bjope closed D54749: Saturating float to int casts..
Dec 18 2020, 2:10 AM · Restricted Project

Dec 16 2020

bjope requested review of D93416: [GlobalISel] Map extractelt to G_EXTRACT_VECTOR_ELT.
Dec 16 2020, 12:41 PM · Restricted Project
bjope added a comment to D54749: Saturating float to int casts..

I'm not entirely sure what you're still waiting for, really. It's a big patch with a lot of diffuse responsibilities, but you've gotten sign-off from individual people across at least most of it. Do you feel like there's something significant that hasn't been reviewed?

Dec 16 2020, 12:22 PM · Restricted Project

Dec 14 2020

bjope added a comment to D86203: [GlobalISel][TableGen] Add handling of unannotated dst pattern ops.

But if I understand this patch correctly it has tried to derive the type from "some_isnt" when it is omitted, rather than reusing the type from the source pattern. Would it make more sense to infer the type from the source pattern instead of from the target instruction when it is omitted? Is that possible?

The problem isn't the type, it's the register class. It can be ambiguous to go from the target instruction back to the source bank if the target register class supports multiple banks

Dec 14 2020, 3:37 PM · Restricted Project
bjope added a comment to D54749: Saturating float to int casts..

Gentle ping again.

Dec 14 2020, 4:19 AM · Restricted Project

Dec 10 2020

bjope added a comment to D86203: [GlobalISel][TableGen] Add handling of unannotated dst pattern ops.

But if I understand this patch correctly it has tried to derive the type from "some_isnt" when it is omitted, rather than reusing the type from the source pattern. Would it make more sense to infer the type from the source pattern instead of from the target instruction when it is omitted? Is that possible?

Dec 10 2020, 1:55 AM · Restricted Project

Dec 9 2020

bjope added a comment to D91358: [ARM][RegAlloc] Add t2LoopEndDec.

@SjoerdMeijer : I've reviewed the non-target-specific changes now. And that part looks good to me now.

Dec 9 2020, 7:07 AM · Restricted Project

Dec 7 2020

bjope added inline comments to D91358: [ARM][RegAlloc] Add t2LoopEndDec.
Dec 7 2020, 12:24 PM · Restricted Project

Nov 27 2020

bjope added inline comments to D91358: [ARM][RegAlloc] Add t2LoopEndDec.
Nov 27 2020, 6:05 AM · Restricted Project

Nov 25 2020

bjope added inline comments to D91358: [ARM][RegAlloc] Add t2LoopEndDec.
Nov 25 2020, 1:06 PM · Restricted Project

Nov 20 2020

bjope added a comment to D91425: [DebugInfo] Skip dbg.value+derefs in RemoveRedundantDbgInstrs forward scan [3/3].
Nov 20 2020, 7:02 AM · debug-info, Restricted Project

Nov 18 2020

bjope added inline comments to D91425: [DebugInfo] Skip dbg.value+derefs in RemoveRedundantDbgInstrs forward scan [3/3].
Nov 18 2020, 6:33 AM · debug-info, Restricted Project
bjope added inline comments to D91425: [DebugInfo] Skip dbg.value+derefs in RemoveRedundantDbgInstrs forward scan [3/3].
Nov 18 2020, 6:18 AM · debug-info, Restricted Project

Nov 16 2020

bjope added inline comments to D90848: llvmbuildectomy - part 2.
Nov 16 2020, 3:57 AM · Restricted Project
bjope added inline comments to D90848: llvmbuildectomy - part 2.
Nov 16 2020, 1:41 AM · Restricted Project

Nov 12 2020

bjope added inline comments to D90829: [GlobalISel][TableGen] Fix seg fault for zero instruction.
Nov 12 2020, 5:16 AM · Restricted Project

Nov 11 2020

bjope added inline comments to D90829: [GlobalISel][TableGen] Fix seg fault for zero instruction.
Nov 11 2020, 1:39 PM · Restricted Project
bjope updated the diff for D54749: Saturating float to int casts..

Add back test cases (they were accidentally removed when commandeering this patch and rebasing earlier).

Nov 11 2020, 1:16 PM · Restricted Project

Nov 3 2020

bjope added inline comments to D90610: [Inline] Fix in handling of ptrtoint in InlineCost.
Nov 3 2020, 5:34 AM · Restricted Project

Nov 2 2020

bjope added inline comments to D90610: [Inline] Fix in handling of ptrtoint in InlineCost.
Nov 2 2020, 9:57 AM · Restricted Project

Oct 30 2020

bjope commandeered D54749: Saturating float to int casts..

Big thanks to @ebevhan for doing lots of work moving forward with the fixed-point number (embedded-c) support.

Oct 30 2020, 5:16 AM · Restricted Project
bjope updated the diff for D54749: Saturating float to int casts..

Address review comments from RKSimon.

Oct 30 2020, 5:12 AM · Restricted Project
bjope updated the diff for D54749: Saturating float to int casts..

Rebase

Oct 30 2020, 4:52 AM · Restricted Project

Oct 29 2020

bjope added inline comments to D48803: Place the BlockAddress type in the address space of the containing function.
Oct 29 2020, 3:11 PM · Restricted Project

Oct 27 2020

bjope added inline comments to D48803: Place the BlockAddress type in the address space of the containing function.
Oct 27 2020, 5:28 AM · Restricted Project

Oct 23 2020

bjope added a comment to rG529ac33197f6: [libc++abi] Fix the standalone build after the __config_site change.

Our "master integration" also stopped after this commit, and we hit the LIBCXXABI_LIBCXX_INCLUDES= is not a valid directory. error. We've never provided any LIBCXXABI_LIBCXX_INCLUDES so I'm not sure what I should point it to.

Oct 23 2020, 5:21 AM

Oct 21 2020

bjope added a comment to D89818: [Compile Time] Make it possible to skip redundant debuginst removal.

As I see it this solution focuses on the problem seen in LoopUnroll. So I agree, it is conservative and simple so if it solves the PR I think it is just fine.

Oct 21 2020, 3:36 AM · Restricted Project

Oct 19 2020

bjope added inline comments to D89640: [Sanitizers] Remove OpenBSD support.
Oct 19 2020, 11:01 AM · Restricted Project

Oct 7 2020

bjope added a comment to D88928: [Utils] Skip RemoveRedundantDbgInstrs in MergeBlockIntoPredecessor (PR47746).

So: skip the call to RemoveRedundantDbgInstrs. There's surprisingly little fallout from this, and most of it can be addressed by doing RemoveRedundantDbgInstrs later.

As an alternative, would it make sense to have a dedicated pass to eliminate redundant debug instructions, that's run late in the pipeline? not sure if running it once, but for every function will be better or worse in the end. It should be easy to skip for functions without debug info though.

Oct 7 2020, 2:18 AM · Restricted Project

Oct 5 2020

bjope added a comment to D86203: [GlobalISel][TableGen] Add handling of unannotated dst pattern ops.

One thing that I've not understood is how GlobalISel is different from the legacy ISel here. Apparently it isn't necessary to annotate things with the legacy ISel today? So is legacy ISel doing the same thing that this patch suggests to do also for GlobalISel, or why do we suddenly need to update all patterns now (I mean, somehow it has worked fine in the past, right)?

Oct 5 2020, 5:33 AM · Restricted Project
bjope added inline comments to D48803: Place the BlockAddress type in the address space of the containing function.
Oct 5 2020, 4:42 AM · Restricted Project

Oct 2 2020

bjope added a comment to D88454: [CMake] Use -isystem flag to access libc++ headers.

I think I figured it out, D88756 should address the issue.

Oct 2 2020, 3:16 PM · Restricted Project, Restricted Project
bjope added a comment to D88454: [CMake] Use -isystem flag to access libc++ headers.

Looking at build.ninja we used to get

Oct 2 2020, 12:45 PM · Restricted Project, Restricted Project
bjope accepted D87494: Improve LSR debug-info.

I've checked the late "better safe than sorry type checks" and they LGTM.

Oct 2 2020, 6:36 AM · Restricted Project, debug-info
bjope added a comment to D88454: [CMake] Use -isystem flag to access libc++ headers.

Hi!
I don't understand why, but with this commit I start seeing several warnings like this when compiling builtins:

/data/repo/master/compiler-rt/lib/sanitizer_common/sanitizer_flag_parser.h:45:7: error: '__sanitizer::FlagHandler<bool>' has virtual functions but non-virtual destructor [-Werror,-Wnon-virtual-dtor]
class FlagHandler : public FlagHandlerBase {
      ^
/data/repo/master/compiler-rt/lib/sanitizer_common/sanitizer_flag_parser.h:71:13: note: in instantiation of template class '__sanitizer::FlagHandler<bool>' requested here
inline bool FlagHandler<bool>::Parse(const char *value) {
            ^

That's a bit surprising. What's your build configuration? Do you use runtimes build? I haven't managed to reproduce it locally so far.

Oct 2 2020, 4:52 AM · Restricted Project, Restricted Project