Page MenuHomePhabricator
Feed Advanced Search

Mar 20 2021

LuoYuanke updated the summary of D93788: [X86] Transform amx pointer..
Mar 20 2021, 5:44 AM · Restricted Project
LuoYuanke updated the diff for D93788: [X86] Transform amx pointer..

Support PHI operation.

Mar 20 2021, 5:41 AM · Restricted Project

Mar 19 2021

LuoYuanke updated the diff for D93788: [X86] Transform amx pointer..

Rebase.

Mar 19 2021, 6:29 PM · Restricted Project
LuoYuanke accepted D98845: [X86] Fix a bug when calculating the ldtilecfg insertion points..

LGTM, thanks.

Mar 19 2021, 7:20 AM · Restricted Project
LuoYuanke added inline comments to D98845: [X86] Fix a bug when calculating the ldtilecfg insertion points..
Mar 19 2021, 7:19 AM · Restricted Project

Mar 18 2021

LuoYuanke added a reviewer for D98845: [X86] Fix a bug when calculating the ldtilecfg insertion points.: xiangzhangllvm.
Mar 18 2021, 6:04 AM · Restricted Project
LuoYuanke committed rGe64adc0b88c2: [X86] Fix compile time regression of D93594. (authored by LuoYuanke).
[X86] Fix compile time regression of D93594.
Mar 18 2021, 1:53 AM
LuoYuanke closed D98773: [X86] Fix compile time regression of D93594..
Mar 18 2021, 1:53 AM · Restricted Project
LuoYuanke added inline comments to D87981: [X86] AMX programming model..
Mar 18 2021, 1:39 AM · Restricted Project, Restricted Project

Mar 17 2021

LuoYuanke added inline comments to D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 10:15 PM · Restricted Project
LuoYuanke updated the diff for D98773: [X86] Fix compile time regression of D93594..

Address Nikita's comments.

Mar 17 2021, 10:13 PM · Restricted Project
LuoYuanke added inline comments to D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 5:23 AM · Restricted Project
LuoYuanke updated the diff for D98773: [X86] Fix compile time regression of D93594..

Address Nikita's comments.

Mar 17 2021, 5:17 AM · Restricted Project
LuoYuanke added inline comments to D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 4:40 AM · Restricted Project
LuoYuanke added a comment to D98773: [X86] Fix compile time regression of D93594..

With "perf stat CLANG_BINARY -w -Werror=date-time -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_STDINT_H=1 -DHAVE_UNISTD_H=1 -DSQLITE_OMIT_LOAD_EXTENSION=1 -DSQLITE_THREADSAFE=0 -I. -MD -MT MultiSource/Applications/sqlite3/CMakeFiles/sqlite3.dir/sqlite3.c.o -MF MultiSource/Applications/sqlite3/CMakeFiles/sqlite3.dir/sqlite3.c.o.d -o MultiSource/Applications/sqlite3/CMakeFiles/sqlite3.dir/sqlite3.c.o -c ../MultiSource/Applications/sqlite3/sqlite3.c", I can see the sqlite3 is fixed in local test.

Mar 17 2021, 4:38 AM · Restricted Project
LuoYuanke added a comment to D98773: [X86] Fix compile time regression of D93594..

@nikic, do you know how to verify the patch fixed the compile time regression?

Mar 17 2021, 4:37 AM · Restricted Project
LuoYuanke updated the summary of D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 4:35 AM · Restricted Project
LuoYuanke updated the diff for D98773: [X86] Fix compile time regression of D93594..

Fix commit message.

Mar 17 2021, 4:34 AM · Restricted Project
LuoYuanke added a comment to D93594: [X86] Pass to transform amx intrinsics to scalar operation..

The fix is uploaded at https://reviews.llvm.org/D98773.

Mar 17 2021, 4:27 AM · Restricted Project, Restricted Project
LuoYuanke added inline comments to D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 4:27 AM · Restricted Project
LuoYuanke added reviewers for D98773: [X86] Fix compile time regression of D93594.: nikic, craig.topper, pengfei, yubing.
Mar 17 2021, 4:26 AM · Restricted Project
LuoYuanke requested review of D98773: [X86] Fix compile time regression of D93594..
Mar 17 2021, 4:24 AM · Restricted Project

Mar 16 2021

LuoYuanke added a comment to D93594: [X86] Pass to transform amx intrinsics to scalar operation..

I can reproduce the regression. I'll help to fix it.

Mar 16 2021, 10:30 PM · Restricted Project, Restricted Project
LuoYuanke added a comment to D98757: [AMX] Not fold constant bitcast into amx intrisic.

Probably we need a .ll test case to for constant folding.

Mar 16 2021, 10:13 PM · Restricted Project, Restricted Project
LuoYuanke added a comment to D98757: [AMX] Not fold constant bitcast into amx intrisic.

Would you add a test case for it?

Mar 16 2021, 8:39 PM · Restricted Project, Restricted Project
LuoYuanke added a comment to D98595: [NFC] Refactor the code to avoid check target type..

I'll look into this further to see if there is a better solution.

Mar 16 2021, 12:06 AM · Restricted Project

Mar 15 2021

LuoYuanke added a comment to D98595: [NFC] Refactor the code to avoid check target type..

Is this really NFC? If it is, then as @lebedev.ri commented - it feels like you're hiding an AMX issue. If it isn't NFC we need more test coverage.

Mar 15 2021, 5:12 PM · Restricted Project

Mar 13 2021

LuoYuanke updated the diff for D98595: [NFC] Refactor the code to avoid check target type..

Fix the comments.

Mar 13 2021, 7:33 PM · Restricted Project
LuoYuanke updated subscribers of D98595: [NFC] Refactor the code to avoid check target type..
Mar 13 2021, 7:31 PM · Restricted Project
LuoYuanke added reviewers for D98595: [NFC] Refactor the code to avoid check target type.: craig.topper, pengfei, spatel, lebedev.ri.
Mar 13 2021, 7:31 PM · Restricted Project
LuoYuanke requested review of D98595: [NFC] Refactor the code to avoid check target type..
Mar 13 2021, 7:30 PM · Restricted Project
LuoYuanke committed rG66fbf5fafb16: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*. (authored by LuoYuanke).
[X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*.
Mar 13 2021, 5:25 PM
LuoYuanke closed D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..
Mar 13 2021, 5:25 PM · Restricted Project

Mar 12 2021

LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

@lebedev.ri, after applying Sanjay and Pengfei's suggestion for the patch, current code check if the the bitcast is lossless before transform. I think it conform the previous code infrastructure and looks neat. It is much better that previous patch. May I commit the patch now?

Mar 12 2021, 6:03 PM · Restricted Project

Mar 11 2021

LuoYuanke updated the diff for D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Fix bug. We need check bitcast is lossless.

Mar 11 2021, 11:43 PM · Restricted Project
LuoYuanke updated the diff for D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Address Pengfei's comments.

Mar 11 2021, 9:56 PM · Restricted Project
LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

I don't know anything about the AMX type / functionality, so I'm probably not the best judge.
I agree that we want to avoid type-based hacks (but what does it mean that we even have target-specific types in IR?)...
OTOH there is already precedent for AMX exceptions in instcombine (and MMX before that). I think we managed to make some of the MMX hacks less obviously bad by excluding all target-specific types from a given transform. Is that a possibility here? That is, could we limit the transform using isIntOrIntVectorTy() or similar?

Mar 11 2021, 5:04 PM · Restricted Project
LuoYuanke updated the diff for D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Address Sanjay's comments.

Mar 11 2021, 5:01 PM · Restricted Project

Mar 10 2021

LuoYuanke updated the diff for D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Rebase.

Mar 10 2021, 7:09 PM · Restricted Project
LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

@lebedev.ri, may I commit the patch? I think the risk of this patch is pretty low. We can revise our solution if we have any better ideas in the future.

Mar 10 2021, 6:27 PM · Restricted Project

Mar 9 2021

LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

I do not understand how load x86_amx can cause trouble while load <256 x i32> works.
More generally, if this proceeds, what will happen if i directly feed the current (bad) IR to llc?
It will still have issues, will it not?
If they are interchangeable like that, why the lowering can not be extended to be agnostic of the actual type?

Mar 9 2021, 10:30 PM · Restricted Project
LuoYuanke updated the diff for D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Address Pengfei's comments.

Mar 9 2021, 9:46 PM · Restricted Project
LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

See e.g. https://lists.llvm.org/pipermail/llvm-dev/2019-December/137684.html

There are already several pointer cast transform in LLVM. Does opaque pointer require change of all pointer cast in LLVM? Nevertheless, it seems opaque pointer is not related with this patch.

Mar 9 2021, 5:43 AM · Restricted Project
LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

Hm. Two things: pointer types are going away, and we can't make any optimizations based on the pointer type regardless.

Mar 9 2021, 5:08 AM · Restricted Project
LuoYuanke added a comment to D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..

I'm not really convinced this is moving in the right direction.
other side of the problem should be fixed.

Mar 9 2021, 4:53 AM · Restricted Project
LuoYuanke updated the summary of D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..
Mar 9 2021, 4:01 AM · Restricted Project
LuoYuanke requested review of D98247: [X86][AMX] Prevent transforming load pointer from <256 x i32>* to x86_amx*..
Mar 9 2021, 3:57 AM · Restricted Project
LuoYuanke committed rG0875c2f7f6d6: [X86][AMX] Add test case for combining AMX bitcast. (authored by LuoYuanke).
[X86][AMX] Add test case for combining AMX bitcast.
Mar 9 2021, 3:48 AM

Mar 7 2021

LuoYuanke added inline comments to D98011: [X86][NFC] Adding one flag to imply whether the instruction should check the predicate when compress EVEX instructions to VEX encoding..
Mar 7 2021, 6:47 PM · Restricted Project

Mar 5 2021

LuoYuanke added inline comments to D98011: [X86][NFC] Adding one flag to imply whether the instruction should check the predicate when compress EVEX instructions to VEX encoding..
Mar 5 2021, 4:32 PM · Restricted Project

Mar 4 2021

LuoYuanke added a comment to D93594: [X86] Pass to transform amx intrinsics to scalar operation..

LGTM too.

Mar 4 2021, 11:52 PM · Restricted Project, Restricted Project
LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Mar 4 2021, 5:09 AM · Restricted Project, Restricted Project

Mar 2 2021

LuoYuanke accepted D97444: [NFC] Add x86_amx and some missed half, bfloat keywords to llvm plugin syntaxes.

LGTM.

Mar 2 2021, 3:46 AM · Restricted Project
LuoYuanke added a comment to D97444: [NFC] Add x86_amx and some missed half, bfloat keywords to llvm plugin syntaxes.

It seems half is already support in the vim plugin. So may remove half from the commit message.

Mar 2 2021, 2:10 AM · Restricted Project
LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Mar 2 2021, 1:24 AM · Restricted Project, Restricted Project

Feb 27 2021

LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Feb 27 2021, 5:40 AM · Restricted Project, Restricted Project
LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Feb 27 2021, 4:49 AM · Restricted Project, Restricted Project
LuoYuanke added inline comments to D97453: [X86] Allow PTILEZEROV and PTILELOADDV to be rematerializable.
Feb 27 2021, 4:29 AM · Restricted Project
LuoYuanke added inline comments to D97453: [X86] Allow PTILEZEROV and PTILELOADDV to be rematerializable.
Feb 27 2021, 3:04 AM · Restricted Project
LuoYuanke added inline comments to D97453: [X86] Allow PTILEZEROV and PTILELOADDV to be rematerializable.
Feb 27 2021, 12:21 AM · Restricted Project

Feb 26 2021

LuoYuanke accepted D97453: [X86] Allow PTILEZEROV and PTILELOADDV to be rematerializable.

LGTM. Thanks.

Feb 26 2021, 4:57 AM · Restricted Project

Feb 23 2021

LuoYuanke added inline comments to D97358: [X86] Support amx-bf16 intrinsic..
Feb 23 2021, 11:32 PM · Restricted Project, Restricted Project

Feb 22 2021

LuoYuanke committed rG8f48ddd19358: [X86][AMX] Lower tile copy instruction. (authored by LuoYuanke).
[X86][AMX] Lower tile copy instruction.
Feb 22 2021, 3:50 PM
LuoYuanke closed D97112: [X86][AMX] Lower tile copy instruction..
Feb 22 2021, 3:50 PM · Restricted Project

Feb 20 2021

LuoYuanke added reviewers for D97112: [X86][AMX] Lower tile copy instruction.: craig.topper, LiuChen3.
Feb 20 2021, 4:20 AM · Restricted Project
LuoYuanke added inline comments to D97112: [X86][AMX] Lower tile copy instruction..
Feb 20 2021, 2:16 AM · Restricted Project
LuoYuanke updated the diff for D97112: [X86][AMX] Lower tile copy instruction..

Address Pengfei's comments.

Feb 20 2021, 2:15 AM · Restricted Project
LuoYuanke added inline comments to D97112: [X86][AMX] Lower tile copy instruction..
Feb 20 2021, 2:11 AM · Restricted Project
LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Feb 20 2021, 12:57 AM · Restricted Project, Restricted Project

Feb 19 2021

LuoYuanke updated the diff for D97112: [X86][AMX] Lower tile copy instruction..

Remove useless code.

Feb 19 2021, 11:12 PM · Restricted Project
LuoYuanke added reviewers for D97112: [X86][AMX] Lower tile copy instruction.: pengfei, xiangzhangllvm, yubing.
Feb 19 2021, 11:07 PM · Restricted Project
LuoYuanke requested review of D97112: [X86][AMX] Lower tile copy instruction..
Feb 19 2021, 11:05 PM · Restricted Project

Feb 17 2021

LuoYuanke accepted D96927: [X86] Zero AMX config buffer for non AVX512 cases..

LGTM. Thanks.

Feb 17 2021, 9:24 PM · Restricted Project
LuoYuanke added inline comments to D96927: [X86] Zero AMX config buffer for non AVX512 cases..
Feb 17 2021, 7:53 PM · Restricted Project
LuoYuanke added inline comments to D96927: [X86] Zero AMX config buffer for non AVX512 cases..
Feb 17 2021, 7:53 PM · Restricted Project

Feb 9 2021

LuoYuanke added inline comments to D93594: [X86] Pass to transform amx intrinsics to scalar operation..
Feb 9 2021, 4:24 AM · Restricted Project, Restricted Project

Feb 8 2021

LuoYuanke added inline comments to D96315: [X86][AMX] Assemble new AMX interface to an experiment header file..
Feb 8 2021, 10:19 PM · Restricted Project
LuoYuanke updated the diff for D96315: [X86][AMX] Assemble new AMX interface to an experiment header file..

Address Pengfei's comments.

Feb 8 2021, 10:18 PM · Restricted Project
LuoYuanke added reviewers for D96315: [X86][AMX] Assemble new AMX interface to an experiment header file.: craig.topper, pengfei, xiangzhangllvm, annita.zhang.
Feb 8 2021, 9:37 PM · Restricted Project
LuoYuanke requested review of D96315: [X86][AMX] Assemble new AMX interface to an experiment header file..
Feb 8 2021, 9:36 PM · Restricted Project

Feb 5 2021

LuoYuanke added a comment to D93594: [X86] Pass to transform amx intrinsics to scalar operation..

Would you rebase to see if the lit test failure is related to this patch?

Feb 5 2021, 12:58 AM · Restricted Project, Restricted Project
LuoYuanke added a reviewer for D93594: [X86] Pass to transform amx intrinsics to scalar operation.: craig.topper.
Feb 5 2021, 12:02 AM · Restricted Project, Restricted Project
LuoYuanke added a reviewer for D93594: [X86] Pass to transform amx intrinsics to scalar operation.: xiangzhangllvm.
Feb 5 2021, 12:00 AM · Restricted Project, Restricted Project

Jan 29 2021

LuoYuanke accepted D95136: [X86] Fix tile config register spill issue..

LGTM. Thank you!

Jan 29 2021, 4:42 PM · Restricted Project
LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 29 2021, 3:50 AM · Restricted Project

Jan 28 2021

LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 28 2021, 5:42 AM · Restricted Project
LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 28 2021, 5:03 AM · Restricted Project
LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 28 2021, 5:00 AM · Restricted Project
LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 28 2021, 4:56 AM · Restricted Project
LuoYuanke committed rGbf649181504f: [X86][AMX] Prevent shape def being scheduled across ldtilecfg. (authored by LuoYuanke).
[X86][AMX] Prevent shape def being scheduled across ldtilecfg.
Jan 28 2021, 12:21 AM
LuoYuanke closed D95582: [X86][AMX] Prevent shape def being scheduled across ldtilecfg..
Jan 28 2021, 12:20 AM · Restricted Project

Jan 27 2021

LuoYuanke updated the diff for D95582: [X86][AMX] Prevent shape def being scheduled across ldtilecfg..

Address Pengfei's comments.

Jan 27 2021, 8:58 PM · Restricted Project
LuoYuanke added inline comments to D95582: [X86][AMX] Prevent shape def being scheduled across ldtilecfg..
Jan 27 2021, 8:56 PM · Restricted Project
LuoYuanke added reviewers for D95582: [X86][AMX] Prevent shape def being scheduled across ldtilecfg.: craig.topper, pengfei, xiangzhangllvm.
Jan 27 2021, 6:50 PM · Restricted Project
LuoYuanke requested review of D95582: [X86][AMX] Prevent shape def being scheduled across ldtilecfg..
Jan 27 2021, 6:48 PM · Restricted Project
LuoYuanke added inline comments to D91927: [X86] Add x86_amx type for intel AMX..
Jan 27 2021, 5:20 PM · Restricted Project, Restricted Project

Jan 24 2021

LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 24 2021, 8:21 PM · Restricted Project
LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 24 2021, 4:02 AM · Restricted Project

Jan 23 2021

LuoYuanke added inline comments to D95136: [X86] Fix tile config register spill issue..
Jan 23 2021, 8:55 PM · Restricted Project
LuoYuanke added reviewers for D95136: [X86] Fix tile config register spill issue.: craig.topper, qcolombet.
Jan 23 2021, 4:56 PM · Restricted Project