Page MenuHomePhabricator

flyingforyou (Junmo Park)
User

Projects

User does not belong to any projects.

User Details

User Since
Aug 6 2015, 11:49 PM (197 w, 6 d)

Recent Activity

May 29 2018

flyingforyou added a comment to D46884: [AArch64] Cortex-A55 scheduler model.

Hi Javed.

May 29 2018, 4:47 PM

May 28 2018

flyingforyou added a comment to D46884: [AArch64] Cortex-A55 scheduler model.

When I tried to test this patch, it showed some improvements on several benchmarks like Spec2000/2006.
But other benchmarks like dhrystone, commercial benchmark's sub workloads show performance degradation.

May 28 2018, 2:50 PM

May 16 2018

flyingforyou added a comment to D46884: [AArch64] Cortex-A55 scheduler model.

I will update benchmark results until next week.
benchmark list : dhrystone, spec2000, spec2006, one commercial benchmark.

May 16 2018, 4:15 PM

Apr 26 2018

flyingforyou added inline comments to D46109: [ARM,AArch64] Add intrinsics for dot product instructions.
Apr 26 2018, 5:00 PM
flyingforyou added inline comments to D46109: [ARM,AArch64] Add intrinsics for dot product instructions.
Apr 26 2018, 4:50 PM

Apr 17 2018

flyingforyou committed rC330229: [AAch64] Add the __ARM_FEATURE_DOTPROD macro definition.
[AAch64] Add the __ARM_FEATURE_DOTPROD macro definition
Apr 17 2018, 3:42 PM
flyingforyou committed rL330229: [AAch64] Add the __ARM_FEATURE_DOTPROD macro definition.
[AAch64] Add the __ARM_FEATURE_DOTPROD macro definition
Apr 17 2018, 3:41 PM
flyingforyou closed D45544: [AAch64] Add the __ARM_FEATURE_DOTPROD macro definition.
Apr 17 2018, 3:41 PM

Apr 12 2018

flyingforyou updated the diff for D45544: [AAch64] Add the __ARM_FEATURE_DOTPROD macro definition.

Addressed SjoerdMeijer' new comments. Thanks.

Apr 12 2018, 2:57 PM

Apr 11 2018

flyingforyou created D45544: [AAch64] Add the __ARM_FEATURE_DOTPROD macro definition.
Apr 11 2018, 4:34 PM

Sep 26 2017

flyingforyou abandoned D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..
Sep 26 2017, 4:00 PM

Mar 27 2017

flyingforyou added a comment to D30367: CodeGen : Check LLVM_ENABLE_DUMP definition for dumpMachineInstrRangeWithSlotIndex..

Thanks Benjamin.
Committed in r298895.

Mar 27 2017, 9:28 PM
flyingforyou committed rL298895: CodeGen : Check LLVM_ENABLE_DUMP definition for….
CodeGen : Check LLVM_ENABLE_DUMP definition for…
Mar 27 2017, 9:26 PM
flyingforyou closed D30367: CodeGen : Check LLVM_ENABLE_DUMP definition for dumpMachineInstrRangeWithSlotIndex. by committing rL298895: CodeGen : Check LLVM_ENABLE_DUMP definition for….
Mar 27 2017, 9:26 PM

Mar 6 2017

flyingforyou added a comment to D30364: AArch64 : Add FastCSEL feature..

@evandro
This is code review about the case "Some targets using CSEL is more prefer than CSINV, CSINC.".

Mar 6 2017, 3:46 PM

Mar 5 2017

flyingforyou updated the diff for D30364: AArch64 : Add FastCSEL feature..

Change the feature name PreferCSEL to FastCSEL.

Mar 5 2017, 4:49 PM
flyingforyou updated the diff for D30364: AArch64 : Add FastCSEL feature..

Addressing Renato, James's comments.
Thanks.

Mar 5 2017, 4:25 PM
flyingforyou updated the diff for D30367: CodeGen : Check LLVM_ENABLE_DUMP definition for dumpMachineInstrRangeWithSlotIndex..

Addressing Benjamin's comment.
Thanks.

Mar 5 2017, 3:25 PM

Mar 2 2017

flyingforyou added a comment to D30364: AArch64 : Add FastCSEL feature..

Thanks for your opinion & example Renato.

Mar 2 2017, 8:25 PM

Mar 1 2017

flyingforyou updated the diff for D30364: AArch64 : Add FastCSEL feature..

After internal(Samsung only) discussion, revert patch about Exynos-M3 part.

Mar 1 2017, 5:14 PM

Feb 27 2017

flyingforyou added a comment to D30364: AArch64 : Add FastCSEL feature..

Thanks for comment Renato.

Feb 27 2017, 3:39 PM
flyingforyou added a reviewer for D30367: CodeGen : Check LLVM_ENABLE_DUMP definition for dumpMachineInstrRangeWithSlotIndex.: bkramer.
Feb 27 2017, 3:18 PM

Feb 24 2017

flyingforyou created D30367: CodeGen : Check LLVM_ENABLE_DUMP definition for dumpMachineInstrRangeWithSlotIndex..
Feb 24 2017, 6:13 PM
flyingforyou committed rL296222: Minor code cleanup. NFC..
Minor code cleanup. NFC.
Feb 24 2017, 6:02 PM
flyingforyou updated the diff for D30364: AArch64 : Add FastCSEL feature..

Addressing Matthias's comment.
Thanks.

Feb 24 2017, 5:52 PM
flyingforyou created D30364: AArch64 : Add FastCSEL feature..
Feb 24 2017, 5:00 PM
flyingforyou committed rL296207: Minor code cleanup. NFC..
Minor code cleanup. NFC.
Feb 24 2017, 4:20 PM

Sep 28 2016

flyingforyou added inline comments to D21571: [AArch64] Avoid generating indexed vector instructions for Exynos.
Sep 28 2016, 5:49 PM

Sep 27 2016

flyingforyou updated subscribers of D24833: [LoopDataPrefetch/AArch64] Allow selective prefetching of irregular symbolic strided accesses.
Sep 27 2016, 3:58 PM

Sep 8 2016

flyingforyou updated subscribers of D24366: [MachineCombiner] Support for floating-point FMA under the FPOpFusion Fast flag on ARM64.
Sep 8 2016, 4:50 PM

Aug 18 2016

flyingforyou updated subscribers of D23583: [AArch64] Add feature has-fast-fma.
Aug 18 2016, 3:17 PM

Aug 15 2016

flyingforyou abandoned D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

Thanks. Hal, Tim.

Aug 15 2016, 2:47 PM

Aug 8 2016

flyingforyou updated subscribers of D23270: TII: Add branch relaxation target hooks.
Aug 8 2016, 3:46 PM
flyingforyou updated subscribers of D23269: AArch64: Move remaining target specific BranchRelaxation bits to TII.
Aug 8 2016, 3:46 PM
flyingforyou updated subscribers of D23271: Move AArch64BranchRelaxation to generic code.
Aug 8 2016, 3:38 PM

Aug 7 2016

flyingforyou updated subscribers of D22942: [SCEV] Fix runtime error caused by ValueOffsetPair.
Aug 7 2016, 4:27 PM

Aug 2 2016

flyingforyou added a comment to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

OK Tim.

Aug 2 2016, 5:39 PM
flyingforyou added a comment to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

Or how about give up inserting prefetch intrinsic when the loop has inline asm?
I don't think users are stupid that they don't know prefetch is necessary or not when they use inline asm.

Aug 2 2016, 4:36 PM
flyingforyou added a comment to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

Thanks for the comment Tim.

Aug 2 2016, 3:47 PM

Aug 1 2016

flyingforyou committed rL277415: Minor code cleanups. NFC..
Minor code cleanups. NFC.
Aug 1 2016, 9:46 PM
flyingforyou added a comment to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

ping2.

Aug 1 2016, 5:29 PM

Jul 29 2016

flyingforyou updated subscribers of D22975: [DAG Combiner] Fix the native computation of the Newton series for reciprocals.
Jul 29 2016, 6:16 PM

Jul 26 2016

flyingforyou updated subscribers of D22630: Loop rotation.
Jul 26 2016, 3:09 PM

Jul 24 2016

flyingforyou updated subscribers of D22377: [SCEV] trip count calculation for loops with unknown stride.
Jul 24 2016, 4:48 PM
flyingforyou added a comment to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

ping?

Jul 24 2016, 3:21 PM

Jul 18 2016

flyingforyou added inline comments to D22448: [AArch64] Cache line size and PredictableSelectIsExpensive for Vulcan.
Jul 18 2016, 3:39 PM
flyingforyou updated subscribers of D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

Is Diff 64111 too much modification?

Jul 18 2016, 12:39 AM

Jul 15 2016

flyingforyou committed rL275637: Minor code cleanups. NFC..
Minor code cleanups. NFC.
Jul 15 2016, 3:50 PM
flyingforyou updated the diff for D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..

Addressed Hal's comments.

Jul 15 2016, 2:29 AM

Jul 14 2016

flyingforyou added inline comments to D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..
Jul 14 2016, 11:26 PM
flyingforyou retitled D22399: [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch. from to [LoopDataPrefetch/AArch64] Don't add prefetch intrinsic, when the loop already has InlineAsm prefetch..
Jul 14 2016, 10:16 PM

Jul 12 2016

flyingforyou updated subscribers of D22267: [AArch64] Properly validate the reciprocal estimation.
Jul 12 2016, 3:02 PM
flyingforyou updated subscribers of D22256: [AArch64] Set FMOVS0 and FMOVD0 as isAsCheapAsAMove when needed..
Jul 12 2016, 1:07 AM

Jul 6 2016

flyingforyou committed rL274704: fix documentation comment. NFC..
fix documentation comment. NFC.
Jul 6 2016, 4:26 PM
flyingforyou committed rL274702: Minor code cleanup. NFC..
Minor code cleanup. NFC.
Jul 6 2016, 4:22 PM

Jun 21 2016

flyingforyou added inline comments to D21571: [AArch64] Avoid generating indexed vector instructions for Exynos.
Jun 21 2016, 3:18 PM
flyingforyou committed rL273255: [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC..
[TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC.
Jun 21 2016, 1:17 AM
flyingforyou closed D21548: [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC. by committing rL273255: [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC..
Jun 21 2016, 1:17 AM
flyingforyou added a comment to D21548: [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC..

Thanks for review Hal.!

Jun 21 2016, 1:16 AM
flyingforyou retitled D21548: [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC. from to [TargetSchedule] Use 'isOutOfOrder' as possible to avoid magic number. NFC..
Jun 21 2016, 1:01 AM

Jun 20 2016

flyingforyou updated subscribers of D21535: [Kryo] Enable loop prefetcher..
Jun 20 2016, 10:47 PM

Jun 16 2016

flyingforyou updated subscribers of D21299: [Codegen Prepare] Swap commutative binops before splitting branch condition..
Jun 16 2016, 2:34 AM

Jun 13 2016

flyingforyou updated subscribers of D21291: [SimplifyCFG] Range reduce switches.
Jun 13 2016, 3:11 PM

Jun 9 2016

flyingforyou updated subscribers of D21203: [AArch64] Add preferred alignments for Exynos M1.
Jun 9 2016, 3:05 PM

Jun 7 2016

flyingforyou updated subscribers of D21103: optimized xsgetn for better performance.
Jun 7 2016, 4:33 PM

Jun 2 2016

flyingforyou updated subscribers of D20865: Adding Cortex-A73 as an available target..
Jun 2 2016, 2:09 AM

May 8 2016

flyingforyou committed rL268888: Minor code cleanups. NFC..
Minor code cleanups. NFC.
May 8 2016, 4:29 PM

Apr 26 2016

flyingforyou accepted D19488: [CodeGenPrepare] use branch weight metadata to decide if a select should be turned into a branch.

LGTM.

Apr 26 2016, 3:26 AM

Apr 25 2016

flyingforyou added a comment to D19474: Remove MinLatency in SchedMachineModel. NFC..

Commited in r267502.

Apr 25 2016, 5:45 PM
flyingforyou committed rL267502: Remove MinLatency in SchedMachineModel. NFC..
Remove MinLatency in SchedMachineModel. NFC.
Apr 25 2016, 5:43 PM
flyingforyou closed D19474: Remove MinLatency in SchedMachineModel. NFC. by committing rL267502: Remove MinLatency in SchedMachineModel. NFC..
Apr 25 2016, 5:43 PM

Apr 24 2016

flyingforyou updated subscribers of D19474: Remove MinLatency in SchedMachineModel. NFC..
Apr 24 2016, 10:49 PM
flyingforyou retitled D19474: Remove MinLatency in SchedMachineModel. NFC. from to Remove MinLatency in SchedMachineModel. NFC..
Apr 24 2016, 6:51 PM
flyingforyou committed rL267375: Minor code cleanups. NFC..
Minor code cleanups. NFC.
Apr 24 2016, 6:46 PM

Apr 22 2016

flyingforyou updated subscribers of D19426: [AArch64] Use the reciprocal estimation machinery.
Apr 22 2016, 11:27 PM
flyingforyou updated subscribers of D19426: [AArch64] Use the reciprocal estimation machinery.
Apr 22 2016, 11:27 PM

Apr 13 2016

flyingforyou added a comment to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..

Hi Chad.

Apr 13 2016, 8:23 PM

Apr 11 2016

flyingforyou committed rL266045: Minor code cleanup. NFC..
Minor code cleanup. NFC.
Apr 11 2016, 11:07 PM
flyingforyou added a comment to D18949: [AArch64] Cannonicalize commutative operands based on LSLFast.

I think this optimization will affect all ARM Architectures. Is this optimization also good for cortex-a57?

Apr 11 2016, 12:21 AM

Apr 8 2016

flyingforyou updated subscribers of D18890: [AArch64] add SSA Load Store optimization pass.
Apr 8 2016, 2:40 AM

Apr 7 2016

flyingforyou updated the diff for D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..

Addressed Gerolf's comments.

Apr 7 2016, 5:12 PM

Apr 5 2016

flyingforyou added inline comments to D18751: [MachineCombiner] Support for floating-point FMA on ARM64.
Apr 5 2016, 6:34 PM
flyingforyou updated subscribers of D18523: [TTI] Add hook for vector extract with extension.
Apr 5 2016, 2:59 PM
flyingforyou committed rL265468: Minor code cleanups. NFC..
Minor code cleanups. NFC.
Apr 5 2016, 2:20 PM
flyingforyou updated subscribers of D18712: [LoopUnroll] Fix the way we update DT after complete unrolling..
Apr 5 2016, 2:17 PM
flyingforyou updated subscribers of D18804: Provide loop info to inliner.
Apr 5 2016, 2:16 PM

Apr 4 2016

flyingforyou added a comment to D18751: [MachineCombiner] Support for floating-point FMA on ARM64.

sure, sorry I missed that. I looked at this too long, I guess :-). It is principally the same ‘better ILP' story as for integers. The prototypical idea is this: imagine two fmul operands feeding the fadd. When the two fmul can execute in parallel it can be faster to issue fmul, fmul, fadd rather than fmul, fmadd.

Apr 4 2016, 9:01 PM
flyingforyou updated subscribers of D18751: [MachineCombiner] Support for floating-point FMA on ARM64.
Apr 4 2016, 4:11 PM

Apr 3 2016

flyingforyou added inline comments to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..
Apr 3 2016, 11:10 PM

Apr 1 2016

flyingforyou updated subscribers of D13363: [DeadStoreElimination] Add support for non-local DSE.
Apr 1 2016, 5:57 PM

Mar 31 2016

flyingforyou added inline comments to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..
Mar 31 2016, 8:42 PM

Mar 30 2016

flyingforyou added a comment to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..

Hi Gerolf.

Mar 30 2016, 9:18 PM
flyingforyou updated subscribers of D18546: Prevent X86IselLowering from merging volatile loads.
Mar 30 2016, 6:14 PM
flyingforyou added a comment to D18570: [AArch64] Handle missing store pair opportunity.

LGTM, too. with a few minor nits.

Mar 30 2016, 5:40 PM
flyingforyou added a comment to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..

Hi Gerolf.

Mar 30 2016, 4:54 PM

Mar 29 2016

flyingforyou updated subscribers of D18572: [AArch64] Relax branches by fusing compare with conditional branch when we can infer that source register is zero/non-zero..
Mar 29 2016, 6:27 PM
flyingforyou committed rL264706: fix CHECK_NOT -> CHECK-NOT.
fix CHECK_NOT -> CHECK-NOT
Mar 29 2016, 12:58 AM
flyingforyou committed rL264704: fixed typo - CHECK-LABEL.
fixed typo - CHECK-LABEL
Mar 29 2016, 12:08 AM

Mar 28 2016

flyingforyou added a comment to D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive..

Evandro, we still wait your reply over 2weeks.
Could you please answer or review about our questions?

Mar 28 2016, 4:40 PM
flyingforyou added a comment to D18507: [AArch64] Do not lower scalar sdiv/udiv to a shifts + mul sequence when optimizing for minsize.

LGTM! too.

Mar 28 2016, 3:03 AM