Page MenuHomePhabricator

jedilyn (Kewen Lin)
User

Projects

User does not belong to any projects.

User Details

User Since
Feb 28 2018, 10:18 PM (89 w, 20 h)

Recent Activity

Sep 4 2019

jedilyn committed rL370907: Request commit access for jedilyn.
Request commit access for jedilyn
Sep 4 2019, 7:24 AM

Aug 12 2019

Herald added a project to D27366: [PowerPC][WIP] Provide context-sensitive cost to the Greedy Allocator to favour splitting over CSR first use: Restricted Project.
Aug 12 2019, 11:51 PM · Restricted Project

Mar 12 2019

jedilyn accepted D47332: [PowerPC] Exploit the vector min/max instructions.

LGTM. Thanks for your time!

Mar 12 2019, 8:44 PM · Restricted Project

Dec 18 2018

jedilyn committed rL349599: [PowerPC]Exploit P9 vabsdu for unsigned vselect patterns.
[PowerPC]Exploit P9 vabsdu for unsigned vselect patterns
Dec 18 2018, 7:07 PM
jedilyn closed D55812: [PowerPC]Exploit P9 vabsdu for some unsigned integer vselect patterns.
Dec 18 2018, 7:07 PM
jedilyn created D55812: [PowerPC]Exploit P9 vabsdu for some unsigned integer vselect patterns.
Dec 18 2018, 12:30 AM
jedilyn committed rL349446: [PowerPC][NFC]Update vabsd cases with vselect test cases.
[PowerPC][NFC]Update vabsd cases with vselect test cases
Dec 18 2018, 12:14 AM

Dec 17 2018

jedilyn committed rL349445: [PowerPC] Exploit power9 new instruction setb.
[PowerPC] Exploit power9 new instruction setb
Dec 17 2018, 11:56 PM
jedilyn closed D53275: [PowerPC] Exploit power9 new instruction setb.
Dec 17 2018, 11:56 PM
jedilyn committed rL349437: [PowerPC] Improve vec_abs on P9.
[PowerPC] Improve vec_abs on P9
Dec 17 2018, 7:20 PM
jedilyn closed D54783: [PowerPC] suboptimal vec_abs for some cases on P9.
Dec 17 2018, 7:20 PM
jedilyn updated the diff for D54783: [PowerPC] suboptimal vec_abs for some cases on P9.

Addressed the comments of Jinsong and Hiroshi, will submit VSELECT related exploitation patch once this gets approved.
The irrelevant test case updates have been done by https://reviews.llvm.org/rL349325 and https://reviews.llvm.org/rL349329.

Dec 17 2018, 12:08 AM

Dec 16 2018

jedilyn committed rL349329: [Power9][NFC]update vabsd case for better dumping.
[Power9][NFC]update vabsd case for better dumping
Dec 16 2018, 10:35 PM
jedilyn committed rL349325: [Power9][NFC]Make pre-inc-disable case more robust.
[Power9][NFC]Make pre-inc-disable case more robust
Dec 16 2018, 7:19 PM
jedilyn updated the diff for D53275: [PowerPC] Exploit power9 new instruction setb.

Addressed Jinsong's comments.

Dec 16 2018, 5:46 PM

Dec 14 2018

jedilyn committed rL349251: [Power9][NFC] add setb exploitation test case.
[Power9][NFC] add setb exploitation test case
Dec 14 2018, 8:43 PM
jedilyn closed D55696: [Power9][NFC] add setb exploitation test case.
Dec 14 2018, 8:42 PM
jedilyn created D55696: [Power9][NFC] add setb exploitation test case.
Dec 14 2018, 1:09 AM
jedilyn added inline comments to D54783: [PowerPC] suboptimal vec_abs for some cases on P9.
Dec 14 2018, 1:07 AM
jedilyn added inline comments to D54783: [PowerPC] suboptimal vec_abs for some cases on P9.
Dec 14 2018, 12:41 AM
jedilyn added a comment to D54783: [PowerPC] suboptimal vec_abs for some cases on P9.

Hi Jinsong @jsji, thanks for your time! I'll break down it to some separate patches and add more comments later.

Dec 14 2018, 12:35 AM

Dec 13 2018

jedilyn added inline comments to D53275: [PowerPC] Exploit power9 new instruction setb.
Dec 13 2018, 6:36 PM

Dec 6 2018

jedilyn added a comment to D53275: [PowerPC] Exploit power9 new instruction setb.

gentle ping.... this is a patch for Power9 setb exploitation

Dec 6 2018, 7:30 PM

Nov 21 2018

jedilyn added reviewers for D53275: [PowerPC] Exploit power9 new instruction setb: jsji, steven.zhang, wuzish.
Nov 21 2018, 6:40 PM
jedilyn added reviewers for D54783: [PowerPC] suboptimal vec_abs for some cases on P9: stefanp, steven.zhang, wuzish.
Nov 21 2018, 6:32 PM
jedilyn updated the summary of D54783: [PowerPC] suboptimal vec_abs for some cases on P9.
Nov 21 2018, 2:06 AM

Nov 20 2018

jedilyn created D54783: [PowerPC] suboptimal vec_abs for some cases on P9.
Nov 20 2018, 11:58 PM

Nov 18 2018

jedilyn added a comment to D53275: [PowerPC] Exploit power9 new instruction setb.

gentle ping...

Nov 18 2018, 10:07 PM

Nov 7 2018

jedilyn added a comment to D53275: [PowerPC] Exploit power9 new instruction setb.

gentle ping...

Nov 7 2018, 6:08 PM
jedilyn added inline comments to D54185: [PowerPC][llvm-exegesis] Add a PowerPC target.
Nov 7 2018, 6:04 PM

Oct 29 2018

jedilyn added inline comments to D53275: [PowerPC] Exploit power9 new instruction setb.
Oct 29 2018, 8:11 PM
jedilyn added inline comments to D53275: [PowerPC] Exploit power9 new instruction setb.
Oct 29 2018, 7:55 PM
jedilyn updated the diff for D53275: [PowerPC] Exploit power9 new instruction setb.

Addressed some comments from Nemanja.

Oct 29 2018, 7:55 PM

Oct 25 2018

jedilyn added a comment to D53383: [PowerPC] Make no-PIC default to match GCC - LLVM .

@jedilyn
You are correct that PIC is the only model on AIX. However, the object model (COFF) on AIX is different so the idea of PIC / no-PIC does not mean the same thing it does on Linux anyway. Also AIX is currently not supported for LLVM. If we decide to support it we can look at the options then. However, since AIX only uses one object model and there is no PIC/ no-PIC difference we will probably just end up ignoring the option anyway.

Thanks for clarification Stefan, it's a good idea just to ignore the PIC option on aix when it gets supported in future.

Oct 25 2018, 8:20 PM
jedilyn added a comment to D53383: [PowerPC] Make no-PIC default to match GCC - LLVM .

PPC as architecture is strongly geared towards PIC code"

Really some truth, as I know, AIX 64bit mode only supports PIC.

Oct 25 2018, 12:08 AM

Oct 24 2018

jedilyn added a comment to D53275: [PowerPC] Exploit power9 new instruction setb.

Thanks a lot for your time @nemanjai , such a good idea to use script to verify various combination. I'm sorry I can't actually get your point about simplifying the logic in the switch block. Each case check in the switch seems not too much and exactly follow the patterns in above comments nearing the case, people can easily map the checks to patterns there. Do I miss something there? Could you kindly clarify more on that? Thanks in advance!

Oct 24 2018, 11:59 PM

Oct 21 2018

jedilyn added a reviewer for D53275: [PowerPC] Exploit power9 new instruction setb: echristo.
Oct 21 2018, 7:48 PM

Oct 17 2018

Herald added a reviewer for D42600: [WIP][Shrink-wrap]split restore point: javed.absar.
Oct 17 2018, 6:54 PM

Oct 15 2018

jedilyn created D53275: [PowerPC] Exploit power9 new instruction setb.
Oct 15 2018, 2:39 AM

Aug 13 2018

jedilyn added inline comments to D49958: [MISC]Fix wrong usage of std::equal().
Aug 13 2018, 7:51 PM

Aug 12 2018

jedilyn added inline comments to D49958: [MISC]Fix wrong usage of std::equal().
Aug 12 2018, 11:31 PM

Aug 2 2018

jedilyn added a comment to D50004: [PowerPC] Emit xscpsgndp instead of xxlor when copying floating point scalar registers for P9.

+1, even for Power9 XSCPSGNDP makes pipeline busy longer than XXLOR, XXLOR is still better.

Can you clarify this please? Where is this information coming from? According to the UM, XXLOR takes up a whole superslice whereas XSCPSGNDP takes up a single slice so we can dispatch 2 of the former per cycle and 4 of the latter. And the "Pipe Busy Cycles" field for both is 1.

Aug 2 2018, 7:08 PM

Aug 1 2018

jedilyn added a comment to D50004: [PowerPC] Emit xscpsgndp instead of xxlor when copying floating point scalar registers for P9.

XSCPSGNDP has longer latency (6 cycles) than XXLOR (2 cycles) on POWER8 while it has higher throughput with the same latency on POWER9. So XXLOR is preferable for pre-P9.

Also, the two instructions have different behavior for a denormal input value in my understanding; XSCPSGNDP does normalization but XXLOR does not. Does this difference matter?

Aug 1 2018, 11:17 PM

Jul 18 2018

jedilyn added a reviewer for D47332: [PowerPC] Exploit the vector min/max instructions: jedilyn.
Jul 18 2018, 6:13 AM · Restricted Project
jedilyn added inline comments to D47332: [PowerPC] Exploit the vector min/max instructions.
Jul 18 2018, 6:12 AM · Restricted Project

Jul 1 2018

jedilyn updated subscribers of D48813: [PowerPC] Don't make it as pre-inc candidate if displacement isn't 4's multiple for i64 pre-inc load/store.

LGTM

Jul 1 2018, 10:23 PM
jedilyn created D48813: [PowerPC] Don't make it as pre-inc candidate if displacement isn't 4's multiple for i64 pre-inc load/store.
Jul 1 2018, 7:56 PM