[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added a comment. The fix is uploaded at https://reviews.llvm.org/D98773. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commits mailing list cf

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added a comment. I can reproduce the regression. I'll help to fix it. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commits mailing list cfe-c

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

nikic added a comment. @yubing In this case I would recommend building `sqlite3.c` from test-suite under `perf stat` and look at the `instructions` metric. For me the command looks like this: perf stat CLANG_BINARY -w -Werror=date-time -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added a comment. In D93594#2628497 , @nikic wrote: > It looks like this has caused a compile-time regression at `O0`: > https://llvm-compile-time-tracker.com/compare.php?from=9341bcbdc93a251b632ffaa51a84452a7a4a5e4e&to=4f198b0c27b04e830a3069aaf4b39

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

nikic added a comment. It looks like this has caused a compile-time regression at `O0`: https://llvm-compile-time-tracker.com/compare.php?from=9341bcbdc93a251b632ffaa51a84452a7a4a5e4e&to=4f198b0c27b04e830a3069aaf4b39cf203eaae4a&stat=instructions The cause is probably the computation of DomTree a

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG4f198b0c27b0: [X86] Pass to transform amx intrinsics to scalar operation. (authored by yubing). Repository: rG LLVM Github Monorepo CHANGES SINCE

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 329204. yubing added a comment. Fix buildfail when it is -DBUILD_SHARED_LIBS=ON Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added a comment. In D93594#2606157 , @RKSimon wrote: > @yubing I've reverted this as it was failing on a lot of buildbots: > http://lab.llvm.org:8011/#/builders/109/builds/9867 Hi, @RKSimon @nicolasvasilache , it seems we haven't told libLLVMX86C

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added a comment. Thanks all for reporting and reverting this. I will do bugfix asap. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commits mailin

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added a comment. Thanks all for reporting and reverting this. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commits mailing list cfe-commits@lis

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

davezarzycki added a comment. In D93594#2606157 , @RKSimon wrote: > @yubing I've reverted this as it was failing on a lot of buildbots: > http://lab.llvm.org:8011/#/builders/109/builds/9867 Thanks – I was just about to point out this broke downstream tes

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

RKSimon added a comment. @yubing I've reverted this as it was failing on a lot of buildbots: http://lab.llvm.org:8011/#/builders/109/builds/9867 Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

nicolasvasilache added a comment. This seems to break the build https://buildkite.com/mlir/mlir-core/builds/12026#91ec4dfe-542f-4312-92db-7d555f05ce06. I could repro locally, reverting locally fixes the build. Please address, thanks! Repository: rG LLVM Github Monorepo CHANGES SINCE LAST A

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

This revision was landed with ongoing or failed builds. This revision was automatically updated to reflect the committed changes. Closed by commit rG8198d83965ba: [X86] Pass to transform amx intrinsics to scalar operation. (authored by LuoYuanke, committed by yubing). Changed prior to commit: h

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added a comment. LGTM too. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.or

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 328408. yubing added a comment. Address pengfei's comments Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes.h llvm/lib/Target/

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:82 + DTU.applyUpdatesPermissive({ + {DominatorTree::Delete, Preheader, Tmp}, + {DominatorTree::Insert, Header, Body}, pengfei wrote: > Do we need to remove the s

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei accepted this revision. pengfei added a comment. This revision is now accepted and ready to land. LGTM with some nitpicks 😊 Comment at: llvm/include/llvm/CodeGen/Passes.h:496 + + /// The pass transform amx intrinsics to scalar operation if the function has + /// optno

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:311 + Value *ResElt = B.CreateAdd(EltC, SubVecR); + Value *NewVecC = B.CreateInsertElement(VecCPhi, ResElt, IdxC); + Value *NewVecD = B.CreateInsertElement(VecDPhi, ResElt, IdxC); -

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:311 + Value *ResElt = B.CreateAdd(EltC, SubVecR); + Value *NewVecC = B.CreateInsertElement(VecCPhi, ResElt, IdxC); + Value *NewVecD = B.CreateInsertElement(VecDPhi, ResElt, IdxC); --

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing marked 15 inline comments as done. yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:311 + Value *ResElt = B.CreateAdd(EltC, SubVecR); + Value *NewVecC = B.CreateInsertElement(VecCPhi, ResElt, IdxC); + Value *NewVecD = B.CreateInser

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 327362. yubing edited the summary of this revision. yubing added a comment. address comments above Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:99 + Loop *RowLoop = LI.AllocateLoop(); + Loop *ColLoop = LI.AllocateLoop(); + RowLoop->addChildLoop(ColLoop); pengfei wrote: > Not sure how about the arithmetic intrinsics

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:88 + +template LuoYuanke wrote: > > pengfei wrote: > > > Can we just use `template `? I think it also can reduce the > > > branch. > > Why do we need a template instead of passing a param

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:88 + +template pengfei wrote: > > Can we just use `template `? I think it also can reduce the > > branch. > Why do we need a template instead of passing a parameter `bool IsLoad`? Bing thou

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:88 + +template Can we just use `template `? I think it also can reduce the > branch. Why do we need a template instead of passing a parameter `bool IsLoad`? Repository: rG LLVM Github

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:311 + Value *ResElt = B.CreateAdd(EltC, SubVecR); + Value *NewVecC = B.CreateInsertElement(VecCPhi, ResElt, IdxC); + Value *NewVecD = B.CreateInsertElement(VecDPhi, ResElt, IdxC);

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:311 + Value *ResElt = B.CreateAdd(EltC, SubVecR); + Value *NewVecC = B.CreateInsertElement(VecCPhi, ResElt, IdxC); + Value *NewVecD = B.CreateInsertElement(VecDPhi, ResElt, IdxC); -

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added inline comments. Comment at: llvm/include/llvm/CodeGen/Passes.h:496 + + /// The pass transform amx intrinsics to scalar operation if the function has + /// optnone attribute or it is O0. transforms Comment at: llvm/lib/Target/X8

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 325964. yubing added a comment. Fix some comments and commit message Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes.h llvm/l

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing marked 13 inline comments as done. yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:487 + SmallVector WorkList; + for (BasicBlock *BB : depth_first(&Func)) { +for (BasicBlock::iterator II = BB->begin(), IE = BB->end(); II != IE;

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 325670. yubing added a comment. Address comments above Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes.h llvm/lib/Target/X86/

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:490 + Instruction &Inst = *II++; + if (match(&Inst, m_Intrinsic()) || + match(&Inst, m_Intrinsic()) || Should be better to use ``` if (auto *Inst = dyn_ca

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/include/llvm/CodeGen/Passes.h:496 + + FunctionPass *createX86LowerAMXIntrinsicsPass(); } // End llvm namespace Add comments to describe what the pass does? Comment at: llvm/lib/Target/X86/X86L

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 325155. yubing marked 5 inline comments as done. yubing added a comment. Small fix for some code Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/ll

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 325152. yubing added a comment. Address the commments above. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes.h llvm/lib/Targe

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:211-212 +IRBuilderBase &B, DomTreeUpdater &DTU, +LoopInfo &LI, Value *Row, Value *Col, +

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

xiangzhangllvm added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:211-212 +IRBuilderBase &B, DomTreeUpdater &DTU, +LoopInfo &LI, Value *Row, Value *Col, +

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added inline comments. Comment at: llvm/lib/Target/X86/X86TargetMachine.cpp:416 + if (TM->getOptLevel() == CodeGenOpt::None) +addPass(createX86LowerAMXIntrinsicsPass()); + else { We may add both pass anyway and skip the pass based on the option le

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

pengfei added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:433 +bool X86LowerAMXIntrinsics::visit() { + bool C; + SmallVector TileDPBSSDs; `bool C = false` Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cp

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

xiangzhangllvm added inline comments. Comment at: llvm/lib/Target/X86/X86LowerAMXIntrinsics.cpp:356 + I->eraseFromParent(); +} + } I see you need force match bitcast then replace, add assert for no bitcast case Comment at: llvm/lib/Ta

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added inline comments. Comment at: llvm/test/CodeGen/X86/AMX/amx-low-intrinsics.ll:77 +; CHECK-NEXT:[[TILEDPBSSD_UNROLL_ROWS_IV:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ], [ [[TILEDPBSSD_UNROLL_ROWS_STEP:%.*]], [[TILEDPBSSD_UNROLL_ROWS_LATCH:%.*]] ] +; CHECK-NEXT:[[VEC

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing added a comment. Strange. llvm/test/CodeGen/X86/AMX/amx-low-intrinsics.ll can pass in my local machine. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 __

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 321673. yubing added a comment. Rebase and fix the bug in amx_api.c Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 Files: llvm/include/llvm/CodeGen/Passes.h llvm/li

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

LuoYuanke added a comment. Would you rebase to see if the lit test failure is related to this patch? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D93594/new/ https://reviews.llvm.org/D93594 ___ cfe-commi

[PATCH] D93594: [X86] Pass to transform amx intrinsics to scalar operation.

yubing updated this revision to Diff 319797. yubing added a comment. Herald added a project: clang. Herald added a subscriber: cfe-commits. Fix some bugs in lowerTileDPBSSD, lowerTileStore, lowerTileLoad Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/