목록LLVM-STUDY (67)

Why should I know this?

ARM Instruction] UBFM, UBFX, UBFIZ, LSR, LSL

UBFM Syntax UBFM Wd, Wn, #, # ; 32-bit general registers UBFM Xd, Xn, #, # ; 64-bit general registers Usage Unsigned Bitfield Move copies any number of low-order bits from a source register into the same number of adjacent bits at any position in the destination register, with zeros in the upper and lower bits. Important! UBFM is an instruction that copies by specifying a memory range and can be..
LLVM-STUDY/BACKEND 2024. 4. 20. 19:40
[AArch64] LDR (LDRroX, LDRXui)

Load 의 IR 원형 ex) %0 = load i32, ptr %arrayidx = load [volatile] , ptr [, align ][, !nontemporal !][, !invariant.load !][, !invariant.group !][, !nonnull !][, !dereferenceable !][, !dereferenceable_or_null !][, !align !][, !noundef !]  LDR(immediate)3130292827262524232221201918171615141312111098765432101x111000010imm901RnRtsize   opc   LDRXui (immediate) MIR 예시renameable $x9 = LDRXui renamable $x..
LLVM-STUDY/MIR 2024. 4. 16. 15:02
[AArch64] Generate stp for complex repeating constansts#514

https://github.com/llvm/llvm-project/issues/51483 [AArch64] Generate stp for complex repeating constansts · Issue #51483 · llvm/llvm-project Bugzilla Link 52141 Version trunk OS Windows NT Reporter LLVM Bugzilla Contributor CC @Arnaud-de-Grandmaison-ARM,@DMG862,@smithp35 Extended Description For complex repeating constants like: void fo... github.com 유사한 최적화 제의와 성능 리포트. https://github.com/dotnet..
LLVM-STUDY/BACKEND 2024. 3. 21. 14:05
Backend 주요 로직 모음

SmallVector WorkList; WorkList.push_back(RetagReg); while (!WorkList.empty()) { Register UseReg = WorkList.pop_back_val(); for (auto &UseI : MRI->use_instructions(UseReg)) { unsigned Opcode = UseI.getOpcode(); if (Opcode == AArch64::STGi || Opcode == AArch64::ST2Gi || Opcode == AArch64::STZGi || Opcode == AArch64::STZ2Gi || Opcode == AArch64::STGPi || Opcode == AArch64::STGloop || Opcode == AArc..
LLVM-STUDY/BACKEND 2024. 3. 19. 01:39
[AArch64] STPX

원문 출처 : https://developer.arm.com/documentation/ddi0602/2023-12/Base-Instructions/STP--Store-Pair-of-Registers-?lang=en#iclass_post_index Documentation – Arm Developer developer.arm.com // FIXME: Use dedicated range-checked addressing mode operand here.defm STPW : StorePairOffset;defm STPX : StorePairOffset; multiclass StorePairOffset opc, bit V, RegisterOperand regtype,                         ..
LLVM-STUDY/MIR 2024. 3. 18. 21:26
LLVM 원라인 패치 목록

* 유관 패치를 작성하게 되면 반영하기 위해 남겨놓는 목적 * * 다른 패치 작성자 분들이 혹시 보게 되면 활용해주세요. * AArch64MIPeepholeOpt.cpp::555 // replaceRegWith changes MIs definition register. Keep it for SSA form until // deleting MI. Only if we made a new destination register. if (DstReg != NewDstReg) { MRI->replaceRegWith(DstReg, NewDstReg); MI.getOperand(0).setReg(DstReg); //
LLVM-STUDY 2024. 3. 17. 19:28
LLVM 최적화 패치 제출까지의 순서 정리

https://die4taoam.tistory.com/122 LLVM 에 패치 보내기 튜토리얼 https://youtu.be/C5Y977rLqpw?si=JXvJ7RsQ-26XSW4i 패치 만들고 LLVM 에 보내기 코드를 수정한 뒤 LLVM 에 기여하는 방식은 code-review 과정을 별도의 플랫폼에서 거친다는 차이가 있습니다. 이에 대한 자세한 과정 die4taoam.tistory.com LLVM에 패치 보내는 튜터리얼은, Nikita Popov 가 작성한 글을 기반으로 작성됐습니다. 현재는 이 글에 포함되어 있는 리뷰과정은 모두 완전 github로 옮겨간 점 등이 달라졌으며, github 환경에 익숙한 사람들에게는 오히려 더 쉽게 LLVM에 기여할 수 있는 기회가 될 수 있을 것 같습니다. 이런저..
LLVM-STUDY 2024. 3. 3. 03:57
[TODO] Alive: Provably Correct InstCombine Optimizations

https://youtu.be/iV5jBtE35r4?si=DrrIq7UaO-bJSlZb 이걸 왜 못봤어나 싶다.
LLVM-STUDY/ALIVE2 2024. 1. 14. 20:55
LLVM debug tips

https://youtu.be/y4b-sgp6VYA?si=FBwXN-abkikLJwx0 cmake .. -GNinja \ -DCMAKE_BUILD_TYPE=Release \ -DLLVM_ENABLE_LLD=ON \ -DLLVM_ENABLE_PROJECTS="clang;lld;compiler-rt" \ -DLLVM_TARGETS_TO_BUILD="AArch64;ARM;X86" \ -DLLVM_ENABLE_ASSERTION=OFF -E : Stop before compiling -S : Stop before assembling -c : Stop before linking Dumping LEXER Tokens clang -Xclang -dump-tokens -E foo.cpp clang -Xclang -hel..
LLVM-STUDY 2024. 1. 14. 20:53
improve bitfield arithmetic #33784

https://github.com/llvm/llvm-project/issues/33874 Improve bitfield arithmetic · Issue #33874 · llvm/llvm-project Bugzilla Link 34526 Version trunk OS Windows NT CC @alexey-bataev,@legrosbuffle,@rotateright Extended Description We could improve math ops on irregular bitfields with the relevant SWAR style patte... github.com 패치목적 // 기존에 생성되던 IR define dso_local i8 @AddCounters(Counters, Counters)(..
LLVM-STUDY/PATCH 2023. 12. 20. 04:30
Prev 1 2 3 4 5 ··· 7 Next