Page MenuHomePhabricator

koriakin (Marcin Kościelnicki)Email Not Verified
User

Projects

User does not belong to any projects.

User Details

User Since
Feb 17 2016, 4:42 AM (191 w, 1 d)

Recent Activity

Feb 13 2017

koriakin added a comment to D29824: Fix AddressSanitizer on s390 (31-bit).
In D29824#674011, @kcc wrote:

Is this change testable?

In GCC testsuite yes. No idea how far is LLVM asan s390 31-bit port, on the asan side this was the only thing that was needed.

Feb 13 2017, 5:45 AM

Feb 9 2017

koriakin accepted D29735: s390x __tls_get_addr_internal vs. __tls_get_offset.

LGTM with a nit.

Feb 9 2017, 2:26 AM · Restricted Project
koriakin added inline comments to D29735: s390x __tls_get_addr_internal vs. __tls_get_offset.
Feb 9 2017, 12:27 AM · Restricted Project

Nov 29 2016

koriakin committed rL288116: Revert "Revert "[ASan] Enable on SystemZ."".
Revert "Revert "[ASan] Enable on SystemZ.""
Nov 29 2016, 12:35 AM

Nov 28 2016

koriakin committed rL288064: [sanitizers] Get the proper printf/scanf version when long double transition is….
[sanitizers] Get the proper printf/scanf version when long double transition is…
Nov 28 2016, 1:25 PM
koriakin closed D27118: [sanitizers] Get the proper printf/scanf version when long double transition is involved. by committing rL288064: [sanitizers] Get the proper printf/scanf version when long double transition is….
Nov 28 2016, 1:25 PM · Restricted Project
koriakin added a comment to D27118: [sanitizers] Get the proper printf/scanf version when long double transition is involved..

What about functions like __isoc99_scanf and snprintf_l?

LGTM

Nov 28 2016, 1:23 PM · Restricted Project

Nov 24 2016

koriakin retitled D27118: [sanitizers] Get the proper printf/scanf version when long double transition is involved. from to [sanitizers] Get the proper printf/scanf version when long double transition is involved..
Nov 24 2016, 3:54 PM · Restricted Project

Nov 23 2016

koriakin committed rL287767: Revert "[ASan] Enable on SystemZ.".
Revert "[ASan] Enable on SystemZ."
Nov 23 2016, 8:17 AM
koriakin committed rL287764: [ASan] Enable on SystemZ..
[ASan] Enable on SystemZ.
Nov 23 2016, 7:57 AM
koriakin closed D21831: [ASan] Enable on SystemZ. by committing rL287764: [ASan] Enable on SystemZ..
Nov 23 2016, 7:57 AM · Restricted Project
koriakin abandoned D21805: [PowerPC] Implement shouldExtI32* hooks..

Abandoning in favor of D21739.

Nov 23 2016, 6:16 AM
koriakin abandoned D21807: [Mips] Implement shouldSignExtI32Param hook..

Abandoning in favor of D21739.

Nov 23 2016, 6:16 AM
koriakin abandoned D21804: [SystemZ] Implement shouldExtI32* hooks..

Abandoning in favor of D21739.

Nov 23 2016, 6:16 AM
koriakin abandoned D21806: [SPARC] Implement shouldExtI32* hooks..

Abandoning in favor of D21739.

Nov 23 2016, 6:16 AM
koriakin committed rL287748: [profile] Enable on SystemZ..
[profile] Enable on SystemZ.
Nov 23 2016, 12:38 AM
koriakin closed D21738: [profile] Enable on SystemZ. by committing rL287748: [profile] Enable on SystemZ..
Nov 23 2016, 12:38 AM

Nov 22 2016

koriakin committed rL287689: Add test for D21736..
Add test for D21736.
Nov 22 2016, 12:13 PM
koriakin committed rL287688: [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..
[CodeGen] Insert TargetLibraryInfoWrapperPass before anything else.
Nov 22 2016, 12:13 PM
koriakin closed D21741: Add test for D21736 + D21737. by committing rL287689: Add test for D21736..
Nov 22 2016, 12:13 PM
koriakin closed D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else. by committing rL287688: [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..
Nov 22 2016, 12:13 PM

Nov 21 2016

koriakin committed rL287570: [TLI] Fix breakage introduced by D21739..
[TLI] Fix breakage introduced by D21739.
Nov 21 2016, 12:30 PM
koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

It seems like this revision causes failures on sanitizer-x86_64-linux-bootstrap, at least check-clang ubsan is unhappy:

/mnt/b/sanitizer-buildbot2/sanitizer-x86_64-linux-bootstrap/build/llvm/lib/Analysis/TargetLibraryInfo.cpp:453:59: runtime error: load of value 32, which is not a valid value for type 'bool'

The actual loaded value varies.

Nov 21 2016, 11:41 AM
koriakin committed rL287534: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32….
[InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32…
Nov 21 2016, 4:07 AM
koriakin closed D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate. by committing rL287534: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32….
Nov 21 2016, 4:07 AM
koriakin committed rL287533: [TLI] Add functions determining if int parameters/returns should be….
[TLI] Add functions determining if int parameters/returns should be…
Nov 21 2016, 4:07 AM
koriakin closed D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext. by committing rL287533: [TLI] Add functions determining if int parameters/returns should be….
Nov 21 2016, 4:07 AM

Nov 20 2016

koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Changed to composite getExtAttrForI32Param/Return functions as for D21739.

Nov 20 2016, 4:58 PM
koriakin added inline comments to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..
Nov 20 2016, 4:57 PM
koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Changed to composite getExtAttrForI32Param/Return functions.

Nov 20 2016, 4:56 PM

Nov 9 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Is related TLI change upstreamed?

Nov 9 2016, 8:46 PM
koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Readded missing test.

Nov 9 2016, 8:43 PM
koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..
Nov 9 2016, 8:42 PM

Nov 4 2016

koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

It seems I have accidentally swapped the diffs of this revision and D21736. Here comes the correct diff.

Nov 4 2016, 11:07 AM
koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

It seems I have accidentally swapped the diffs of this revision and D21739. Here comes the correct diff.

Nov 4 2016, 11:07 AM

Nov 1 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Ping?

Nov 1 2016, 3:48 PM

Oct 25 2016

koriakin added inline comments to D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..
Oct 25 2016, 3:17 AM

Oct 23 2016

koriakin updated the diff for D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..

Added a comment explaining the reason.

Oct 23 2016, 12:15 PM
koriakin added a comment to D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..

Missing test.

Oct 23 2016, 12:10 PM
koriakin retitled D21741: Add test for D21736 + D21737. from Add test for D21736. to Add test for D21736 + D21737..
Oct 23 2016, 12:08 PM
koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Moved the functions to TLI, as per ML: http://lists.llvm.org/pipermail/llvm-dev/2016-September/104641.html.

Oct 23 2016, 11:54 AM
koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Moved the functions to TLI, as per ML: http://lists.llvm.org/pipermail/llvm-dev/2016-September/104641.html.

Oct 23 2016, 11:51 AM

Oct 5 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Hi Marcin,

This work got mentioned in an internal bug report, do you have any intention of reviving this patch series?

Thanks,
Simon

Oct 5 2016, 4:27 AM

Jul 25 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

@hfinkel Ping?

Jul 25 2016, 1:43 AM

Jul 17 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Ping?

Jul 17 2016, 11:45 PM

Jul 10 2016

koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Ping?

Jul 10 2016, 1:15 PM
koriakin committed rL275016: [SystemZ] Utilize Test Data Class instructions..
[SystemZ] Utilize Test Data Class instructions.
Jul 10 2016, 7:48 AM
koriakin closed D21949: [SystemZ] Utilize Test Data Class instructions. by committing rL275016: [SystemZ] Utilize Test Data Class instructions..
Jul 10 2016, 7:48 AM
koriakin updated the diff for D21949: [SystemZ] Utilize Test Data Class instructions..

Minor fix for old gcc - use make_tuple instead of initializer list.

Jul 10 2016, 6:18 AM
koriakin added a comment to D21949: [SystemZ] Utilize Test Data Class instructions..

Would it also make sense to support the VECTOR FP TEST DATA CLASS IMMEDIATE instruction?

Jul 10 2016, 5:13 AM
koriakin updated the diff for D21949: [SystemZ] Utilize Test Data Class instructions..

Demagicked the masks; the only remaining magic constant is the shift right by 1 in fabs handling, which cannot be helped much.

Jul 10 2016, 5:07 AM

Jul 8 2016

koriakin accepted D22142: [SystemZ] Add support for the .word directive..

LGTM

Jul 8 2016, 9:10 AM
koriakin added inline comments to D21949: [SystemZ] Utilize Test Data Class instructions..
Jul 8 2016, 8:44 AM

Jul 7 2016

koriakin updated the diff for D21949: [SystemZ] Utilize Test Data Class instructions..

More tests

Jul 7 2016, 11:55 AM

Jul 6 2016

koriakin accepted D22054: [SystemZ] Fix regression when handling conditional calls.

A testcase would be nice, otherwise LGTM.

Jul 6 2016, 12:32 PM
koriakin added inline comments to D22050: [SystemZ] Add support for missing instructions.
Jul 6 2016, 11:49 AM
koriakin added a comment to D22054: [SystemZ] Fix regression when handling conditional calls.

Thanks for fixing this! Looks OK, but ISTM the whole thing could be avoided by just doing addOperand instead of the more specific variants?

Jul 6 2016, 11:18 AM

Jul 4 2016

koriakin added a comment to D21949: [SystemZ] Utilize Test Data Class instructions..

I haven't looked into the detailed implementation yet. However, one important feature to verify would be that use of the TEST DATA CLASS instruction is not only triggered by those specific IR sequences you test for on the LLVM level, but more importantly, triggered by whatever code clang generates for the isinf/isnan/signbit/fpclassify etc. builtins, since those are what users (hopefully) typically use ... [ This might actually involve changing clang as well, not sure about that. ]

Jul 4 2016, 11:41 AM

Jul 2 2016

koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Updated for new revision of D21739 - now aborts if knownTarget() is false.

Jul 2 2016, 8:33 AM
koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Added knownTarget() hook - it returns true if the TTI was created for a specific target, false if it was created for an unknown target.

Jul 2 2016, 8:33 AM

Jul 1 2016

koriakin added a comment to D21949: [SystemZ] Utilize Test Data Class instructions..

This is my attempt at utilizing the test data class instructions. I've tried
and failed to do this during the SelectionDAG phase, via DAG combiners.
The problems are:

Jul 1 2016, 7:29 PM
koriakin retitled D21949: [SystemZ] Utilize Test Data Class instructions. from to [SystemZ] Utilize Test Data Class instructions..
Jul 1 2016, 7:29 PM
koriakin committed rL274417: [SystemZ] Move misplaced SystemZ::TDC to non-memory opcode range..
[SystemZ] Move misplaced SystemZ::TDC to non-memory opcode range.
Jul 1 2016, 7:27 PM

Jun 29 2016

koriakin closed D21837: [SystemZ] Split up PerformDAGCombine. [NFC].
Jun 29 2016, 5:16 PM
koriakin committed rL274191: [SystemZ] Split up PerformDAGCombine. [NFC].
[SystemZ] Split up PerformDAGCombine. [NFC]
Jun 29 2016, 5:16 PM
koriakin added a comment to D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Looks reasonable to me. Added Hal for a second look.

If a given target is not compiled in, then the defaults TTI provides might be wrong for the target. I think you need to add some what for the transformation to detect whether or not it is getting an accurate answer so it can abort the transformation in that case. We shouldn't miscompile IR just because we don't have a particular target available.

Jun 29 2016, 3:03 PM
koriakin added a comment to D21809: [SystemZ] Add support for the .insn directive.

From my understanding, gas only supports some of the possible s390 instruction formats when using ".insn". For example, the "ri" format in the insn directive corresponds to the RI-a format, which uses an immediate field and not a relative-immediate field. Therefore, for the .insn directive specifically:

ri = RI-a
rie = RIE-d
ril = RIL-a

Jun 29 2016, 2:46 PM
koriakin retitled D21837: [SystemZ] Split up PerformDAGCombine. [NFC] from to [SystemZ] Split up PerformDAGCombine. [NFC].
Jun 29 2016, 6:30 AM
koriakin added inline comments to D21809: [SystemZ] Add support for the .insn directive.
Jun 29 2016, 2:43 AM
koriakin added a comment to D21809: [SystemZ] Add support for the .insn directive.

Oh, and the .word change should be split off (it'll likely be a one-liner using the alias functionality)

Jun 29 2016, 2:36 AM
koriakin requested changes to D21809: [SystemZ] Add support for the .insn directive.

See inline comments. Also, this might not be the good level to have .insn handling - for one, we forfeit any hope of handling fixups in immediate operands (which is necessary to handle pc-relative forms correctly). It might be better to represent these directives as special machine instructions, one per form.

Jun 29 2016, 2:25 AM
koriakin added inline comments to D21809: [SystemZ] Add support for the .insn directive.
Jun 29 2016, 1:52 AM
koriakin closed D21793: [SystemZ] Add floating-point test data class instructions..
Jun 29 2016, 1:10 AM
koriakin added inline comments to D21809: [SystemZ] Add support for the .insn directive.
Jun 29 2016, 1:04 AM
koriakin added inline comments to D21809: [SystemZ] Add support for the .insn directive.
Jun 29 2016, 12:56 AM
koriakin removed a parent revision for D21738: [profile] Enable on SystemZ.: D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else..
Jun 29 2016, 12:46 AM
koriakin removed a child revision for D21737: [PATCH] [CodeGen] Insert TargetLibraryInfoWrapperPass before anything else.: D21738: [profile] Enable on SystemZ..
Jun 29 2016, 12:46 AM
koriakin added a parent revision for D21738: [profile] Enable on SystemZ.: D21804: [SystemZ] Implement shouldExtI32* hooks..
Jun 29 2016, 12:45 AM
koriakin added a child revision for D21804: [SystemZ] Implement shouldExtI32* hooks.: D21738: [profile] Enable on SystemZ..
Jun 29 2016, 12:45 AM
koriakin retitled D21831: [ASan] Enable on SystemZ. from to [ASan] Enable on SystemZ..
Jun 29 2016, 12:43 AM · Restricted Project
koriakin committed rL274108: [SystemZ] Add floating-point test data class instructions..
[SystemZ] Add floating-point test data class instructions.
Jun 29 2016, 12:36 AM

Jun 28 2016

koriakin added parent revisions for D21807: [Mips] Implement shouldSignExtI32Param hook.: D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext., D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Depends on D21736, D21739.

Jun 28 2016, 11:10 AM
koriakin added a child revision for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate.: D21807: [Mips] Implement shouldSignExtI32Param hook..
Jun 28 2016, 11:10 AM
koriakin added parent revisions for D21806: [SPARC] Implement shouldExtI32* hooks.: D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext., D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Depends on D21736, D21739.

Jun 28 2016, 11:10 AM
koriakin added a child revision for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate.: D21806: [SPARC] Implement shouldExtI32* hooks..
Jun 28 2016, 11:10 AM
koriakin added a child revision for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext.: D21805: [PowerPC] Implement shouldExtI32* hooks..
Jun 28 2016, 11:10 AM
koriakin added a child revision for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate.: D21805: [PowerPC] Implement shouldExtI32* hooks..
Jun 28 2016, 11:10 AM
koriakin added parent revisions for D21805: [PowerPC] Implement shouldExtI32* hooks.: D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext., D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Depends on D21736, D21739.

Jun 28 2016, 11:10 AM
koriakin added a child revision for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate.: D21804: [SystemZ] Implement shouldExtI32* hooks..
Jun 28 2016, 11:10 AM
koriakin added parent revisions for D21804: [SystemZ] Implement shouldExtI32* hooks.: D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate., D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Depends on D21736, D21739.

Jun 28 2016, 11:10 AM
koriakin added a child revision for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext.: D21804: [SystemZ] Implement shouldExtI32* hooks..
Jun 28 2016, 11:10 AM
koriakin retitled D21807: [Mips] Implement shouldSignExtI32Param hook. from to [Mips] Implement shouldSignExtI32Param hook..
Jun 28 2016, 11:10 AM
koriakin retitled D21806: [SPARC] Implement shouldExtI32* hooks. from to [SPARC] Implement shouldExtI32* hooks..
Jun 28 2016, 11:09 AM
koriakin retitled D21805: [PowerPC] Implement shouldExtI32* hooks. from to [PowerPC] Implement shouldExtI32* hooks..
Jun 28 2016, 11:08 AM
koriakin retitled D21804: [SystemZ] Implement shouldExtI32* hooks. from to [SystemZ] Implement shouldExtI32* hooks..
Jun 28 2016, 11:06 AM
koriakin updated the diff for D21736: [InstrProfiling] Mark __llvm_profile_instrument_target last parameter as i32 zeroext or signext if appropriate..

Split off the tests (they'll come together with target-specific implementations for D21739).

Jun 28 2016, 11:05 AM
koriakin updated the diff for D21739: [TLI] Add functions determining if int parameters/returns should be zeroext/signext..

Split off the target-specific parts.

Jun 28 2016, 11:04 AM
koriakin added a comment to D21793: [SystemZ] Add floating-point test data class instructions..
Jun 28 2016, 8:56 AM
koriakin updated the diff for D21793: [SystemZ] Add floating-point test data class instructions..
Jun 28 2016, 8:56 AM
koriakin retitled D21793: [SystemZ] Add floating-point test data class instructions. from to [SystemZ] Add floating-point test data class instructions..
Jun 28 2016, 7:27 AM