Page MenuHomePhabricator

jsji (Jinsong Ji)
User

Projects

User does not belong to any projects.

User Details

User Since
Dec 27 2017, 10:26 AM (181 w, 5 d)

Recent Activity

Today

jsji added inline comments to D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 21, 4:34 PM · Restricted Project
jsji committed rG3996311ee1b0: [DAGCombine] reassoc flag shouldn't enable contract (authored by jsji).
[DAGCombine] reassoc flag shouldn't enable contract
Mon, Jun 21, 2:17 PM
jsji closed D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 21, 2:16 PM · Restricted Project
jsji added a comment to D104247: [DAGCombine] reassoc flag shouldn't enable contract.

LGTM.
Since we are opting for a definition of reassoc that is not a superset of the other flags, it might be worth seeing if transforms related to arcp and afn are similarly affected.

Mon, Jun 21, 1:52 PM · Restricted Project
jsji added a comment to D102094: [AIX][PowerPC] Remove error when specifying mabi=vec-default on AIX.

The default Altivec ABI was implemented

Mon, Jun 21, 1:36 PM · Restricted Project
jsji added a reviewer for D102094: [AIX][PowerPC] Remove error when specifying mabi=vec-default on AIX: Restricted Project.
Mon, Jun 21, 1:29 PM · Restricted Project
jsji added inline comments to D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 21, 9:26 AM · Restricted Project
jsji updated the diff for D104247: [DAGCombine] reassoc flag shouldn't enable contract.

Address comments.

Mon, Jun 21, 9:17 AM · Restricted Project
jsji added a reviewer for D97184: [XCOFF] support DWARF for XCOFF for object output: sfertile.
Mon, Jun 21, 7:22 AM · Restricted Project
jsji added a reviewer for D97049: [XCOFF][NFC] add DWARF section support in XCOFF object writer: sfertile.
Mon, Jun 21, 7:20 AM · Restricted Project

Fri, Jun 18

jsji updated the diff for D104247: [DAGCombine] reassoc flag shouldn't enable contract.

Address comments -- updated testcases.

Fri, Jun 18, 9:05 PM · Restricted Project
jsji added a comment to D104247: [DAGCombine] reassoc flag shouldn't enable contract.

In code that I've looked at (mostly C compiled with -ffast-math), we always have contract when we have reassoc, so I don't see much practical difference.
Can you explain more how we could benefit from this change - in the example in the description, we would have 4 instructions rather than 3 if we use FMA - is that better?

Yes, you are right, the performance of having FMA should be better. However, we have quite some scenarios that users care about precision more than performance, they want to precise control of when FMA can be generated. So The major motivation of this is to ensure that we respect the IR semantics. For users that care about performance, we still can get them through default global option or emitting respect flag in IR.

AFAIU, the use of FMA doesn't decline the precision (but may improve a bit, which is also unexpected in some circumstance).

Fri, Jun 18, 8:24 PM · Restricted Project
jsji added a comment to D104247: [DAGCombine] reassoc flag shouldn't enable contract.

In code that I've looked at (mostly C compiled with -ffast-math), we always have contract when we have reassoc, so I don't see much practical difference.
Can you explain more how we could benefit from this change - in the example in the description, we would have 4 instructions rather than 3 if we use FMA - is that better?

Fri, Jun 18, 8:15 AM · Restricted Project

Thu, Jun 17

jsji added a comment to D103445: [PowerPC][AIX][RFC] Generate inlined quadword lock free atomic operations.

I would prefer we implementing it using AtomicExpand , not using AtomicExpansionKind::LLSC , but with AtomicExpansionKind::MaskedIntrinsic.

Thu, Jun 17, 2:42 PM · Restricted Project
jsji updated the diff for D104247: [DAGCombine] reassoc flag shouldn't enable contract.

Rebase to pick up Dave's D104255.

Thu, Jun 17, 2:02 PM · Restricted Project

Wed, Jun 16

jsji added a reviewer for D104396: [AIX] Use VSSRC/VSFRC Register classes for f32/f64 callee arguments on P8 and above: qiucf.
Wed, Jun 16, 12:59 PM · Restricted Project, Restricted Project
jsji added a reviewer for D104396: [AIX] Use VSSRC/VSFRC Register classes for f32/f64 callee arguments on P8 and above: Restricted Project.
Wed, Jun 16, 12:58 PM · Restricted Project, Restricted Project

Tue, Jun 15

jsji updated the diff for D104247: [DAGCombine] reassoc flag shouldn't enable contract.

Address review comments.

Tue, Jun 15, 8:33 PM · Restricted Project
jsji committed rG9ddb62589051: [NFC] Update renamed option in comments (authored by jsji).
[NFC] Update renamed option in comments
Tue, Jun 15, 12:45 PM
jsji updated the diff for D104247: [DAGCombine] reassoc flag shouldn't enable contract.

Address review comments.

Tue, Jun 15, 7:22 AM · Restricted Project

Mon, Jun 14

jsji added a comment to D99483: [clang][AIX] Define __STDC_NO_ATOMICS__ for c11 and above.

Done in https://reviews.llvm.org/D103707?

Mon, Jun 14, 1:32 PM · Restricted Project
jsji added inline comments to D104255: [InterleaveAccess] Copy fast math flags when adjusting binary operators in interleave access pass.
Mon, Jun 14, 12:44 PM · Restricted Project
jsji accepted D104255: [InterleaveAccess] Copy fast math flags when adjusting binary operators in interleave access pass.

LGTM

Mon, Jun 14, 12:39 PM · Restricted Project
jsji edited reviewers for D104247: [DAGCombine] reassoc flag shouldn't enable contract, added: dmgreen; removed: greened.
Mon, Jun 14, 11:49 AM · Restricted Project
jsji added inline comments to D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 14, 11:47 AM · Restricted Project
jsji updated the summary of D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 14, 11:41 AM · Restricted Project
jsji added a reviewer for D104247: [DAGCombine] reassoc flag shouldn't enable contract: Restricted Project.
Mon, Jun 14, 11:24 AM · Restricted Project
jsji requested review of D104247: [DAGCombine] reassoc flag shouldn't enable contract.
Mon, Jun 14, 11:18 AM · Restricted Project

Thu, Jun 10

jsji accepted D103010: [PowerPC] Export 16 byte load-store instructions.

LGTM with some nits.

Thu, Jun 10, 8:08 AM · Restricted Project

Wed, Jun 9

jsji committed rG4a89ed373cda: [AIX] Add traceback ssp canary bit support (authored by jsji).
[AIX] Add traceback ssp canary bit support
Wed, Jun 9, 7:58 PM
jsji closed D103202: [AIX] Add traceback ssp canary bit support.
Wed, Jun 9, 7:58 PM · Restricted Project
jsji added inline comments to D103202: [AIX] Add traceback ssp canary bit support.
Wed, Jun 9, 7:17 PM · Restricted Project
jsji added inline comments to D103010: [PowerPC] Export 16 byte load-store instructions.
Wed, Jun 9, 7:10 PM · Restricted Project
jsji added a comment to D103202: [AIX] Add traceback ssp canary bit support.

Ping ...

Wed, Jun 9, 2:38 PM · Restricted Project
jsji added inline comments to D103010: [PowerPC] Export 16 byte load-store instructions.
Wed, Jun 9, 2:29 PM · Restricted Project
jsji accepted D103786: [AIX][compiler-rt] Fix cmake build of libatomic for cmake-3.16+.

LGTM. Thanks.

Wed, Jun 9, 6:22 AM · Restricted Project

Tue, Jun 8

jsji added inline comments to D103786: [AIX][compiler-rt] Fix cmake build of libatomic for cmake-3.16+.
Tue, Jun 8, 1:34 PM · Restricted Project
jsji accepted D100290: [PowerPC] Make sure the first probe is full size or is the last probe when stack is realigned.

LGTM , with some comments on comments.

Tue, Jun 8, 9:04 AM · Restricted Project

Mon, Jun 7

jsji added a reviewer for D100375: [yaml2obj] Enable support for parsing 64-bit XCOFF.: sfertile.
Mon, Jun 7, 7:55 PM · Restricted Project
jsji added a comment to D103696: [XCOFF][AIX] Add support for XCOFF 64 bit Object files.

I don't think the YAML tools support generating 64-bit XCOFF objects yet, but we can add a precompiled object file as test input.

Mon, Jun 7, 10:45 AM · Restricted Project
jsji added a comment to D100290: [PowerPC] Make sure the first probe is full size or is the last probe when stack is realigned.

Looks mostly good to me, but I would prefer we split this into two fixes, one for the realign bp algorithm change, another for the other.

Mon, Jun 7, 10:38 AM · Restricted Project
jsji accepted D103761: [PowerPC][Dwarf] Assign MMA register's dwarf register number to negative value.

LGTM. Thanks for fixing.

Mon, Jun 7, 7:57 AM · Restricted Project

Fri, Jun 4

jsji added a comment to D100290: [PowerPC] Make sure the first probe is full size or is the last probe when stack is realigned.

This somehow was missed in my radar.. I will have a look .

Fri, Jun 4, 8:31 PM · Restricted Project
jsji added a reviewer for D103696: [XCOFF][AIX] Add support for XCOFF 64 bit Object files: Esme.
Fri, Jun 4, 7:10 AM · Restricted Project
jsji added reviewers for D103696: [XCOFF][AIX] Add support for XCOFF 64 bit Object files: jasonliu, sfertile.
Fri, Jun 4, 6:54 AM · Restricted Project

Thu, Jun 3

jsji committed rGcd9e1a020cf0: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue (authored by jsji).
[Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue
Thu, Jun 3, 1:31 PM
jsji closed D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.
Thu, Jun 3, 1:31 PM · Restricted Project
jsji updated the diff for D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.

Use isExactlyValue() for all floating-point types.

Thu, Jun 3, 1:03 PM · Restricted Project
jsji retitled D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue from [BitcodeWriter][PowerPC] Avoid clearing lower bits for ppc_fp128 NullValues to [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.
Thu, Jun 3, 12:38 PM · Restricted Project
jsji updated the diff for D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.

Fix isNullValue for ppc_fp128.

Thu, Jun 3, 12:35 PM · Restricted Project
jsji added a comment to D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.

Constant::isNullValue() is supposed to mean "is this value all zero bits"; various parts of the code use it that way. It shouldn't care whether the value is semantically equal to zero. Anything doing floating-point arithmetic should be using the appropriate ConstantFP/APFloat methods.

Thu, Jun 3, 11:33 AM · Restricted Project
jsji added a comment to D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.

If Constant::isNullValue() isn't returning the right result, please fix it directly. Getting it wrong could have other unexpected consequences.

Thu, Jun 3, 11:17 AM · Restricted Project
jsji retitled D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue from [BitcodeWriter][PowerPC] Avoid clearing lower bits for NullValues to [BitcodeWriter][PowerPC] Avoid clearing lower bits for ppc_fp128 NullValues.
Thu, Jun 3, 10:47 AM · Restricted Project
jsji added a reviewer for D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue: timshen.
Thu, Jun 3, 10:45 AM · Restricted Project
jsji updated the summary of D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.
Thu, Jun 3, 10:45 AM · Restricted Project
jsji requested review of D103634: [Constants][PowerPC] Check exactlyValue for ppc_fp128 in isNullValue.
Thu, Jun 3, 10:45 AM · Restricted Project
jsji committed rGc8279e8acd30: add ppc_fp128 bitcode writer testcase (authored by jsji).
add ppc_fp128 bitcode writer testcase
Thu, Jun 3, 10:18 AM
jsji added a comment to D103202: [AIX] Add traceback ssp canary bit support.

Ping...

Thu, Jun 3, 9:48 AM · Restricted Project
jsji added a comment to D103614: [PowerPC][AIX][RFC] Generate inlined quadword lock free atomic operations via AtomicExpand.

Looks like this is better than D103445, I would recommend go with this unless there is other comments. Thanks!

Thu, Jun 3, 6:44 AM · Restricted Project

Tue, Jun 1

jsji added a reviewer for D103445: [PowerPC][AIX][RFC] Generate inlined quadword lock free atomic operations: xingxue.
Tue, Jun 1, 6:55 AM · Restricted Project

Fri, May 28

jsji added a comment to D99901: [Driver][test] Test intended target only.
In D99901#2787034, @ted wrote:

Why would we not want to support this on other targets, like Arm? Fixes for AIX shouldn't impact other things.

Fri, May 28, 7:45 AM · Restricted Project

Thu, May 27

jsji committed rGb2581196eb03: [AIX] Enable stackprotect feature (authored by jsji).
[AIX] Enable stackprotect feature
Thu, May 27, 7:47 PM
jsji closed D103100: [AIX] Enable stackprotect feature.
Thu, May 27, 7:47 PM · Restricted Project
jsji updated the diff for D103202: [AIX] Add traceback ssp canary bit support.

Add option control to enable this feature.

Thu, May 27, 7:16 PM · Restricted Project
jsji added a comment to D103202: [AIX] Add traceback ssp canary bit support.

I respectfully disagree with the requirement of mandating the canary word position.

Thu, May 27, 11:06 AM · Restricted Project
jsji added inline comments to D103100: [AIX] Enable stackprotect feature.
Thu, May 27, 7:16 AM · Restricted Project
jsji updated the diff for D103100: [AIX] Enable stackprotect feature.

Address comments.

Thu, May 27, 7:16 AM · Restricted Project
jsji added inline comments to D103202: [AIX] Add traceback ssp canary bit support.
Thu, May 27, 6:58 AM · Restricted Project
jsji added inline comments to D103202: [AIX] Add traceback ssp canary bit support.
Thu, May 27, 6:53 AM · Restricted Project

Wed, May 26

jsji requested review of D103202: [AIX] Add traceback ssp canary bit support.
Wed, May 26, 1:32 PM · Restricted Project
jsji added a reviewer for D103100: [AIX] Enable stackprotect feature: hubert.reinterpretcast.
Wed, May 26, 8:09 AM · Restricted Project

Tue, May 25

jsji added inline comments to D103100: [AIX] Enable stackprotect feature.
Tue, May 25, 8:07 PM · Restricted Project
jsji updated the diff for D103100: [AIX] Enable stackprotect feature.

Handle Internalize.

Tue, May 25, 8:06 PM · Restricted Project
jsji updated the diff for D103100: [AIX] Enable stackprotect feature.

Address comments.

Tue, May 25, 5:27 PM · Restricted Project
jsji added inline comments to D103100: [AIX] Enable stackprotect feature.
Tue, May 25, 2:23 PM · Restricted Project
jsji added inline comments to D103100: [AIX] Enable stackprotect feature.
Tue, May 25, 12:39 PM · Restricted Project
jsji updated the diff for D103100: [AIX] Enable stackprotect feature.

Address comments.

Tue, May 25, 12:36 PM · Restricted Project
jsji requested review of D103100: [AIX] Enable stackprotect feature.
Tue, May 25, 9:48 AM · Restricted Project
jsji committed rG882e4cbd7477: [AIX][AsmPrinter] Print Symbol in comments for TOC load (authored by jsji).
[AIX][AsmPrinter] Print Symbol in comments for TOC load
Tue, May 25, 9:38 AM
jsji closed D102949: [AIX][AsmPrinter] Print Symbol in comments for TOC load.
Tue, May 25, 9:37 AM · Restricted Project

Mon, May 24

jsji committed rGbec6b0225211: [compiler-rt][scudo] Fix sign-compare warnings (authored by jsji).
[compiler-rt][scudo] Fix sign-compare warnings
Mon, May 24, 2:42 PM
jsji closed D103029: [compiler-rt][scudo] Fix sign-compare warnings.
Mon, May 24, 2:42 PM · Restricted Project
jsji updated the diff for D103029: [compiler-rt][scudo] Fix sign-compare warnings.

Update type of N to size_t.

Mon, May 24, 2:07 PM · Restricted Project
jsji added inline comments to D103029: [compiler-rt][scudo] Fix sign-compare warnings.
Mon, May 24, 2:00 PM · Restricted Project
jsji requested review of D103029: [compiler-rt][scudo] Fix sign-compare warnings.
Mon, May 24, 8:07 AM · Restricted Project

May 21 2021

jsji updated the diff for D102949: [AIX][AsmPrinter] Print Symbol in comments for TOC load.

Update update_llc_test tests as well.

May 21 2021, 7:09 PM · Restricted Project
jsji requested review of D102949: [AIX][AsmPrinter] Print Symbol in comments for TOC load.
May 21 2021, 3:07 PM · Restricted Project
jsji committed rGf8bb0d97cb99: [PowerPC] Add stack guard tests (authored by jsji).
[PowerPC] Add stack guard tests
May 21 2021, 9:16 AM

May 20 2021

jsji committed rGedf4d69d3888: [AIX] Print printable byte list as quoted string (authored by jsji).
[AIX] Print printable byte list as quoted string
May 20 2021, 7:38 PM
jsji closed D102814: [AIX] Print printable byte list as quoted string.
May 20 2021, 7:38 PM · Restricted Project, Restricted Project
jsji updated the diff for D102814: [AIX] Print printable byte list as quoted string.

Address comments.

May 20 2021, 7:12 PM · Restricted Project, Restricted Project
jsji added inline comments to D102814: [AIX] Print printable byte list as quoted string.
May 20 2021, 12:52 PM · Restricted Project, Restricted Project
jsji updated the diff for D102814: [AIX] Print printable byte list as quoted string.

Update .string handling.

May 20 2021, 12:51 PM · Restricted Project, Restricted Project
jsji added inline comments to D102814: [AIX] Print printable byte list as quoted string.
May 20 2021, 11:31 AM · Restricted Project, Restricted Project
jsji updated the diff for D102814: [AIX] Print printable byte list as quoted string.

Address comments.

May 20 2021, 10:37 AM · Restricted Project, Restricted Project
jsji added reviewers for D102844: XFAIL align.pass.pass.cpp on aix regarding atomic alignment: lkail, Restricted Project.
May 20 2021, 7:12 AM

May 19 2021

jsji added a comment to D102724: Revert "[AIX] Avoid structor alias; die before bad alias codegen".

Good idea, fyi, this is the patch: https://reviews.llvm.org/D83252

May 19 2021, 2:51 PM · Restricted Project, Restricted Project
jsji requested review of D102814: [AIX] Print printable byte list as quoted string.
May 19 2021, 2:25 PM · Restricted Project, Restricted Project
jsji added a comment to D102724: Revert "[AIX] Avoid structor alias; die before bad alias codegen".

because AIX now has an alias implementation.

May 19 2021, 2:19 PM · Restricted Project, Restricted Project

May 18 2021

jsji added a reviewer for D102724: Revert "[AIX] Avoid structor alias; die before bad alias codegen": Restricted Project.
May 18 2021, 2:13 PM · Restricted Project, Restricted Project