)]}' { "log": [ { "commit": "0d4a3bfb0045fa61dc520548be7be677b3e36447", "tree": "be5bc0fcceb44d85461e9bd03de4705d16362a5a", "parents": [ "17a98cb3114d99fb0e664d3028195c0f979f604e" ], "author": { "name": "Nikita Popov", "email": "npopov@redhat.com", "time": "Mon Apr 20 13:27:51 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 13:27:51 2026" }, "message": "[LICM] Generate test checks (NFC) (#192970)" }, { "commit": "17a98cb3114d99fb0e664d3028195c0f979f604e", "tree": "10d9b3344b569e6eeb29d97bcb638935a3ad122a", "parents": [ "578c3aae5889a7e29bcb42844adc278920e7ea0a" ], "author": { "name": "J. Ryan Stinnett", "email": "ryan@convolv.es", "time": "Mon Apr 20 13:26:56 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 13:26:56 2026" }, "message": "[NFC][Clang][DebugInfo] Fix constructor type comment (#192949)\n\nThis corrects a constructor comment to say move instead of copy." }, { "commit": "578c3aae5889a7e29bcb42844adc278920e7ea0a", "tree": "beb5d936b464608927791415dedd401f5ea47282", "parents": [ "c3534cf68e4459e8c3141b63b7c720c7d8bbf7a0" ], "author": { "name": "Matthew Devereau", "email": "matthew.devereau@arm.com", "time": "Mon Apr 20 13:13:30 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 13:13:30 2026" }, "message": "[AArch64][Isel] Add lowering for fixed-width interleave3 (#192677)" }, { "commit": "c3534cf68e4459e8c3141b63b7c720c7d8bbf7a0", "tree": "7947133ca0b2145424fb13101ba0aad3d81afc25", "parents": [ "4265ca3ecbcbe7e1df6a5bc95787956c92d2cde5" ], "author": { "name": "woruyu", "email": "1214539920@qq.com", "time": "Mon Apr 20 13:06:42 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 13:06:42 2026" }, "message": "[NFC][AMDGPU] Move PhiLoweringHelper and related types into AMDGPU namespace (#192586)\n\nMove Incoming, createLaneMaskReg, and PhiLoweringHelper into\nllvm::AMDGPU namespace to avoid symbol collisions and improve\ntarget-specific isolation. No functional change." }, { "commit": "4265ca3ecbcbe7e1df6a5bc95787956c92d2cde5", "tree": "644332ad67d1d0bb98d1d3c669ed7bbdbfbdc3e8", "parents": [ "9705f346cbddb30b0e89c869b181838ef738a584" ], "author": { "name": "A. Jiang", "email": "de34@live.cn", "time": "Mon Apr 20 12:50:31 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 12:50:31 2026" }, "message": "[libc++][NFC] Remove `pointer_traits\u003c__wrap_iter\u003e` partial specialization (#178864)\n\nSince commit b5270ba20dc3d84df9d880be11d57667f04c6c28, it was decided\nthat `__wrap_iter` wasn\u0027t intended to perform runtime check, so it\u0027s\n`operator-\u003e` won\u0027t check whether the iterator is deferenceable. (See\nhttps://llvm.org/PR178521.)\n\nCurrently, `pointer_traits\u003c__wrap_iter\u003e` partial specialization is\nprovided and makes `__to_address`/`to_address` call the member\n`to_address` instead of `operator-\u003e`. But given these operations are\nequivalent and `__wrap_iter::operator-\u003e` is improbable to have\nbehavioral change in the future, perhaps it would be better to remove\nthe partial specialization.\n\nThe changes have no negative impact on compile time, and sometimes have\na negligible positive impact.\n\nDrive-by: Also remove inclusion of `\u003c__memory/addressof.h\u003e` from\n`\u003c__iterator/wrap_iter.h\u003e` as we don\u0027t need to call `addressof` or its\nequivalent internal version." }, { "commit": "9705f346cbddb30b0e89c869b181838ef738a584", "tree": "116acb57e0f2c53366bebd012596c5c7f40198ca", "parents": [ "41e2baf25b03120893d7f4bf5be80171b1ce2b30" ], "author": { "name": "Ralf Jung", "email": "post@ralfj.de", "time": "Mon Apr 20 12:44:26 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 12:44:26 2026" }, "message": "[IR] LangRef: document behavior of double-lifetime.start (#184296)\n\nlifetime.start after lifetime.start is defined to reset the contents of the\nallocation to uninitialized memory." }, { "commit": "41e2baf25b03120893d7f4bf5be80171b1ce2b30", "tree": "55f53e463ce6c8073722304dddfaffd4dae79394", "parents": [ "d242dfc3121d9813d57108219cba7ca167ae66d1" ], "author": { "name": "Max Graey", "email": "maxgraey@gmail.com", "time": "Mon Apr 20 12:42:20 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 12:42:20 2026" }, "message": "[KnownBits][APInt] Optimize isConstant (NFC) (#191919)\n\n`isConstant` used is quite often and should be as lightweight as\npossible. Add `APInt::isInverseOf` and utilize it for `isConstant`. It\nis equivalent to `LHS \u003d\u003d ~RHS`, but avoids intermediate allocations." }, { "commit": "d242dfc3121d9813d57108219cba7ca167ae66d1", "tree": "d3f86ff9589e7618663aa2f657188f7a4c69ce2a", "parents": [ "d964f2bbc69fb2d383b9f76cdeb35f813ee409d6" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Mon Apr 20 12:38:50 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 12:38:50 2026" }, "message": "[Polly][NFC] Convert IslNodeBuilder::preloadUnconditionally to isl++ (#192959)" }, { "commit": "d964f2bbc69fb2d383b9f76cdeb35f813ee409d6", "tree": "b540102b5743bcee902c24809fbee4e3a87ff867", "parents": [ "39fbe8bde7ec6d35b69ac7f3bb6ab9191d7f2d71" ], "author": { "name": "David Spickett", "email": "david.spickett@arm.com", "time": "Mon Apr 20 12:12:39 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 12:12:39 2026" }, "message": "[lldb] Remove commented out code in TypeList and TypeMap (#192955)\n\nFixes #192774\n\nThey are obviously unused and not part of any public API." }, { "commit": "39fbe8bde7ec6d35b69ac7f3bb6ab9191d7f2d71", "tree": "ff040b0b3611ad0f86cfd0dfbdef72c8b6bd10bd", "parents": [ "a677acfaae1d67786db266bf3f358951ac23fa1a" ], "author": { "name": "NeKon69", "email": "nobodqwe@gmail.com", "time": "Mon Apr 20 11:54:31 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:54:31 2026" }, "message": "[Sema] Decay functions/array to pointers in `IncompatiblePointerDiscardsQualifiers` and `IncompatibleOBTKinds` cases (#187038)\n\nThis PR adds checks that ensure the compiler has the correct type so it\ndoes not crash when trying to get qualifiers from an incorrect type in\nthe `IncompatiblePointerDiscardsQualifiers` and `IncompatibleOBTKinds`\ncases. It achieves this by performing a decay before accessing pointee\nqualifiers.\n\nIt also adds a test to ensure that such a bug won\u0027t arise again.\n\nFixes #182534\n\n(same as #183616, unfortunately i forgot to create a new branch and\nwiped all changes)" }, { "commit": "a677acfaae1d67786db266bf3f358951ac23fa1a", "tree": "8d5813ad011e42989e0d1b882d578f1246ee82b5", "parents": [ "ea43c3f7e280133997d98de3eda83fb5c5eb92ae" ], "author": { "name": "Aidan Shilling", "email": "aidanshilling91@gmail.com", "time": "Mon Apr 20 11:44:44 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:44:44 2026" }, "message": "AMDGPU/GlobalISel: RegBankLegalize rules for amdgcn_s_alloc_vgpr (#192937)\n\nFrom issue #192497" }, { "commit": "ea43c3f7e280133997d98de3eda83fb5c5eb92ae", "tree": "5cd45e9fe5fc7437b587a1208514b27d83fd6380", "parents": [ "43ccda2703732cab7dd054849174ee03fddfaa22" ], "author": { "name": "zackc6", "email": "112835067+zackc6@users.noreply.github.com", "time": "Mon Apr 20 11:40:04 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:40:04 2026" }, "message": "[mlir][linalg][shard] Fix andi reduction kind in sharding partition (#192381)\n\nlinalg sharding now maps arith.andi combiners to bitwise_and (instead of\nsum) when creating shard.all_reduce. Adds a shard-partition regression\ntest that checks the emitted all-reduce uses reduction \u003d bitwise_and for\nan andi reduction." }, { "commit": "43ccda2703732cab7dd054849174ee03fddfaa22", "tree": "b44a82924b896eb655d1dc89e2f5a6f77331a896", "parents": [ "2c1624b41f61c83b14507dc1c565eeb14317d618" ], "author": { "name": "Michael Kruse", "email": "llvm-project@meinersbur.de", "time": "Mon Apr 20 11:33:58 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:33:58 2026" }, "message": "[Polly][NFC] Convert IslNodeBuilder::preloadInvariantLoad to isl++ (#192942)\n\nAlso fix the wrong lifetime annotation of\n`IslNodeBuilder::materializeParameters`" }, { "commit": "2c1624b41f61c83b14507dc1c565eeb14317d618", "tree": "fefcdb305f0facdbf2cc5d9ae3aa809c598358b4", "parents": [ "d2ffdc35f8124ec54eb7348008cc0a8eb29f6608" ], "author": { "name": "Fraser Cormack", "email": "fcormack@modular.com", "time": "Mon Apr 20 11:30:58 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:30:58 2026" }, "message": "[lldb][PlatformDarwin] Fix build failure (#192934)\n\nPR #191454 introduced a constructor call that breaks in certain build\nconfigs, e.g., with C++20 - https://godbolt.org/z/hTb8166qG.\n\nThis fix mimics what\u0027s done to construct an instance of `XcodeSDK::Info`\nfurther up in the same file." }, { "commit": "d2ffdc35f8124ec54eb7348008cc0a8eb29f6608", "tree": "f754cf6ac8560ccebd2e34f9b17190f071c077ca", "parents": [ "1e6f8aafc75825ca80d60a17cda500a94640a1ba" ], "author": { "name": "forking-google-bazel-bot[bot]", "email": "265904573+forking-google-bazel-bot[bot]@users.noreply.github.com", "time": "Mon Apr 20 11:03:54 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:03:54 2026" }, "message": "[Bazel] Fixes 920b203 (#192922)\n\nThis fixes 920b203fe2a2ad9f4403e8923b2b06518a46129e.\n\nCo-authored-by: Google Bazel Bot \u003cgoogle-bazel-bot@google.com\u003e" }, { "commit": "1e6f8aafc75825ca80d60a17cda500a94640a1ba", "tree": "142a73a79809a8918c7df4a7dcedb6f356912708", "parents": [ "de5ac001c5f45608a600ef661363a8b32b252248" ], "author": { "name": "Oleksandr \"Alex\" Zinenko", "email": "git@ozinenko.com", "time": "Mon Apr 20 11:02:35 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:02:35 2026" }, "message": "[mlir] MlirOptMain: avoid double verification (#192661)\n\nMlirOptMain would run verification twice at the end of the processing:\n 1. after the last pass in the pipeline;\n2. prior to printing. Since there is no logic that could mutate, and\nthus potentially invalidate, the IR between the two, the second\nverification is redundant. Skip it when possible." }, { "commit": "de5ac001c5f45608a600ef661363a8b32b252248", "tree": "b16cc7ec6c41e8624efa67ce2674c3a3b26113a7", "parents": [ "0b0cb9d9cb557d6ba377642c236bfdd444986705" ], "author": { "name": "Xinlong Chen", "email": "xinlongchen@tencent.com", "time": "Mon Apr 20 11:02:22 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 11:02:22 2026" }, "message": "[SelectionDAG] Return poison instead of undef for out-of-bounds EXTRACT_VECTOR_ELT (#192844)\n\nOut-of-bounds EXTRACT_VECTOR_ELT on fixed-length vectors is undefined\nbehavior.\n\nReturn poison instead of undef to be consistent with LangRef semantics.\n\nPrep work to help with https://github.com/llvm/llvm-project/pull/190307" }, { "commit": "0b0cb9d9cb557d6ba377642c236bfdd444986705", "tree": "01b419697f44ef8bd5808e22c07c517b345f9ad4", "parents": [ "18c3c86cd30fef80e30382c1eda54e05f530a3ca" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Mon Apr 20 10:49:30 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 10:49:30 2026" }, "message": "ValueTracking: Use SimplifyQuery for computeConstantRange (#191726)\n\nDoes introduce new context passing in a few of the updated contexts." }, { "commit": "18c3c86cd30fef80e30382c1eda54e05f530a3ca", "tree": "7823a106b7809deea419711acb3f69065e8350dd", "parents": [ "8b20914a38205f56ae026a7cef8835b8a3cc3b06" ], "author": { "name": "Nathan Gauër", "email": "brioche@google.com", "time": "Mon Apr 20 10:09:36 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 10:09:36 2026" }, "message": "[HLSL][Clang] Start emitting @llvm.structured.alloca (#190157)\n\nAllowing some pattterns in the FE to emit this new instruction to emit\nlogical pointers. Renamed the experimental-emit-sgep flag to reflect the\nbroader logic it gates.\nThis also updates the few frontend tests to reflect the newly emitted\nalloca.\n\nNext step is to handle the Mem2Reg/Reg2Mem." }, { "commit": "8b20914a38205f56ae026a7cef8835b8a3cc3b06", "tree": "fd6206807d2cd0a47d32993142cc7b917cca708a", "parents": [ "318a1ea9275eb69ae2ec2bcbdeb22a885dee86ca" ], "author": { "name": "Kerry McLaughlin", "email": "kerry.mclaughlin@arm.com", "time": "Mon Apr 20 10:06:34 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 10:06:34 2026" }, "message": "[AArch64] Lower masked.expandload intrinsic using SVE2p2/SME2p2 EXPAND (#190999)\n\nThe masked.expandload intrinsic can be lowered using the EXPAND instruction\nwhen available, where the source vector is the result of a contiguous load\nof the number of active elements in the predicate. EXPAND is available with\neither feature in non-streaming mode. It is available in streaming-mode\nwith SME2p2, or with SVE2p2 when SME_FA64 is also enabled.\n\nIntrinsics which return a fixed-width result can also be lowered using SVE\ninstructions when preferred, otherwise they will be scalarised by falling\nback on scalarizeMaskedExpandLoad.\n\nScalable vectors are not supported when EXPAND is not available. In this\ncase, the cost model will return an invalid cost for the intrinsic." }, { "commit": "318a1ea9275eb69ae2ec2bcbdeb22a885dee86ca", "tree": "a446114b1fca5036647382095d56c51c8fd0af92", "parents": [ "4a41e9cbd78a8ebbd9253b2d14b1df87c10a4f93" ], "author": { "name": "A. Jiang", "email": "de34@live.cn", "time": "Mon Apr 20 10:03:30 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 10:03:30 2026" }, "message": "[libc++][test] Unblock cases for `ranges::sort` with proxy ranges (#188490)\n\nlibc++ switched to use `iter_move`/`iter_swap` long time ago, so we\nshould unblock these cases." }, { "commit": "4a41e9cbd78a8ebbd9253b2d14b1df87c10a4f93", "tree": "304647b5391b1baf59f44a0a2b651d0cf48138d5", "parents": [ "24ec6d485d584dae819ca1bbe61d288eb1d50d10" ], "author": { "name": "Diana Picus", "email": "Diana-Magda.Picus@amd.com", "time": "Mon Apr 20 10:00:14 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 10:00:14 2026" }, "message": "AMDGPU: Don\u0027t save FP/BP for noreturn functions (#187668)\n\nAs suggested here:\nhttps://github.com/llvm/llvm-project/pull/184616#discussion_r2889401998\n\nWe could probably skip saving other regs too, but that can be a future\npatch.\n\nAssisted-by: Sline with Claude Opus." }, { "commit": "24ec6d485d584dae819ca1bbe61d288eb1d50d10", "tree": "36321cf943355dc773f5780f51403d93c03cbf0a", "parents": [ "90ecec0f1a607dedc7bd329086119d55967eda00" ], "author": { "name": "Arseniy Obolenskiy", "email": "arseniy.obolenskiy@amd.com", "time": "Mon Apr 20 09:57:57 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:57:57 2026" }, "message": "[SPIR-V] Fix image query and sampler type (#190767)\n\n- Use OpImageQuerySize instead of OpImageQuerySizeLod for multisampled\nSPIR-V spec requires MS\u003d0 for OpImageQuerySizeLod\n- Use `target(\"spirv.Sampler\")` instead of i32 for non-constant sampler\nkernel parameters so they produce OpTypeSampler as required by\nOpSampledImage\n\nrelated to https://github.com/llvm/llvm-project/issues/190736" }, { "commit": "90ecec0f1a607dedc7bd329086119d55967eda00", "tree": "257a39ed6e88392a0317e73d27ebac5f517cbe75", "parents": [ "050da78ab7898d49c11dbf6b968df10249537412" ], "author": { "name": "🍌Shawn", "email": "m18824909883@163.com", "time": "Mon Apr 20 09:43:43 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:43:43 2026" }, "message": "[clang][NFC] Simplify boolean return in `SemaARM::checkTargetClonesAttr` (#192832)" }, { "commit": "050da78ab7898d49c11dbf6b968df10249537412", "tree": "52ae93fb45dd9aea610e1fe24279e61b6e2a7519", "parents": [ "d3141b465382e4b10d5ed9cba6c8c480856e1092" ], "author": { "name": "Luke Lau", "email": "luke@igalia.com", "time": "Mon Apr 20 09:43:28 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:43:28 2026" }, "message": "[RISCV] Remove codegen for vp_fsub (#191832)\n\nPart of the work to remove trivial VP intrinsics from the RISC-V\nbackend, see\nhttps://discourse.llvm.org/t/rfc-remove-codegen-support-for-trivial-vp-intrinsics-in-the-risc-v-backend/87999\n\nThis splits off the vp_fsub intrinsic from #179622." }, { "commit": "d3141b465382e4b10d5ed9cba6c8c480856e1092", "tree": "67a4edbeb1c4ec58084059ff6b1376a2d47f2619", "parents": [ "69c566bf7ecda471ecc8625c50c6e4d261508ef3" ], "author": { "name": "Madhur Amilkanthwar", "email": "madhura@nvidia.com", "time": "Mon Apr 20 09:36:36 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:36:36 2026" }, "message": "[LoopFusion] Validate loop structure before creating LoopCandidates (#192280)\n\nThis patch deletes the assert which required the loop to have \npreheader. It is not guaranteed to have preheader when loops\nare structured using `indirectbr`. Instead, we now rely on header.\n\nFixes #156670." }, { "commit": "69c566bf7ecda471ecc8625c50c6e4d261508ef3", "tree": "10811bf44b59e56368edb2b5d744b22b228a0a09", "parents": [ "332390385017b1d8ab3ed4c8d4a47545175e3b14" ], "author": { "name": "Stefan Gränitz", "email": "stefan.graenitz@gmail.com", "time": "Mon Apr 20 09:23:11 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:23:11 2026" }, "message": "[lldb] Add caching and _NT_SYMBOL_PATH parsing in SymbolLocatorSymStore (#191782)\n\nThe _NT_SYMBOL_PATH environment variable is the idiomatic way to set a\nsystem-wide lookup order of symbol servers and a local cache for\nSymStore. It holds a semicolon-separated list of entries in the\nfollowing notations:\n* srv*[\u003ccache\u003e*]\u003csource\u003e sets a source and an optional explicit cache\n* cache*\u003ccache\u003e sets an implicit cache for all subsequent entries\n* all other entries are bare local directories\n\nSince symbol paths are closely intertwined with the caching of symbol\nfiles, this patch proposes support in LLDB for both features at once.\nParseEnvSymbolPaths() implements the parsing logic, which processes\nentries of the symbol path string from left to right to create a series\nof LookupEntry objects that each store a source and a cache location.\nThe source of a LookupEntry can be a local directory or an HTTP server\naddress. The cache is a local directory or empty. This representation\nunifies the implicit vs. explicit caching options from the SymStore\nprotocol.\n\nThe lookup remains in LocateSymStoreEntry() which we now invoke for\neach LookupEntry. Here we distinguish sources between HTTP servers and\nlocal directories. The latter doesn\u0027t change. We just moved the logging\nto clearly express the new cases. For HTTP downloads we now check the\ncache first and add files to it after download. The download itself\nkeeps targeting a temporary file to avoid corrupt cache entries in case\nof interruptions. After all, the SymStore protocol has no check-sums!\n\nWe define a default cache path that is used as a fallback, if the symbol\npath doesn\u0027t specify any. It can be overridden through the\nplugin.symbol-locator.symstore.cache property. This is analog to\nDebuginfod and very handy since we want to move files out from temp\nafter download. The default cache path is what we use if there are no\nother options.\n\nThe symbol path notation in the SymStore protocol carries quite some\nlegacy. The SymStoreTest unittest checks that we can parse all the\nobscure combinations of possible server and cache entries.\n\n---------\n\nCo-authored-by: Nerixyz \u003cnero.9@hotmail.de\u003e" }, { "commit": "332390385017b1d8ab3ed4c8d4a47545175e3b14", "tree": "97c527f37809fdf0c7d1028620532045966f8291", "parents": [ "7a6422567a915e50826f6811a22fbc7ba5687042" ], "author": { "name": "David Green", "email": "david.green@arm.com", "time": "Mon Apr 20 09:16:23 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:16:23 2026" }, "message": "[AArch64][GlobalISel] Move KnownBitsVectorTest to mir. NFC (#192536)\n\nThis ports some of the older C++ GlobalISel known-bits tests to use\nprint\u003cgisel-value-tracking\u003e in a mir file. This is mostly autogenerated,\nbut attempts to keep the existing comments. Some tests have not been\nported as they are entirely in C++ or tested isKnownToBeAPowerOfTwo,\nwhich is not tested in the print output." }, { "commit": "7a6422567a915e50826f6811a22fbc7ba5687042", "tree": "cea33489452697284f9351418498c542f735d15b", "parents": [ "57e35e93b95841521e89d1e102e1cc602060ce43" ], "author": { "name": "Elio", "email": "xiongzile@bytedance.com", "time": "Mon Apr 20 09:12:28 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:12:28 2026" }, "message": "[CIR] add vsqrt and vsqrtq support (#192282)\n\nPart of https://github.com/llvm/llvm-project/issues/185382\n\nco-authored by: @Kouunnn \u003cxerw1314@gmail.com\u003e\n\n---------\n\nCo-authored-by: Zile Xiong \u003cxiongzile99@gmail.com\u003e\nCo-authored-by: ZCkouun \u003c1765074320@qq.com\u003e" }, { "commit": "57e35e93b95841521e89d1e102e1cc602060ce43", "tree": "6953774fd6af6ecd00d8e1aa7bf57884e258c9fd", "parents": [ "b37a60793347046bb85739da3e50e2408eb5b7a9" ], "author": { "name": "David Sherwood", "email": "david.sherwood@arm.com", "time": "Mon Apr 20 09:02:48 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:02:48 2026" }, "message": "[LV][NFC] Regen CHECK lines in LoopVectorize/X86/replicating-load-store-costs.ll (#192682)" }, { "commit": "b37a60793347046bb85739da3e50e2408eb5b7a9", "tree": "8d99d9393303ed9a0abad6c66b895226a1ed476c", "parents": [ "6fd5b522c529393535212c14e07b0c2c011b3cd1" ], "author": { "name": "Benjamin Maxwell", "email": "benjamin.maxwell@arm.com", "time": "Mon Apr 20 09:00:18 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 09:00:18 2026" }, "message": "[compiler-rt] Don\u0027t provide `__arm_sme_state` for baremetal targets (#191434)\n\nPreviously, we required baremetal runtimes to implement an undocumented\n`__aarch64_sme_accessible` hook to check if SME is available (as\nchecking CPU features may vary across targets).\n\nThis allowed us to provide a generic `__arm_sme_state` implementation\nbut caused some friction for toolchains that depend on compiler-rt.\n\nThis patch instead removes the implementation of `__arm_sme_state` for\nbaremetal. This makes it the responsibility of the runtime (e.g. libc)\nto provide this function for baremetal targets.\n\nThe requirements of this function are documented in the AAPCS64:\nhttps://github.com/ARM-software/abi-aa/blob/main/aapcs64/aapcs64.rst#811__arm_sme_state\n\nAll other SME ABI rountines are still provided by compiler-rt." }, { "commit": "6fd5b522c529393535212c14e07b0c2c011b3cd1", "tree": "7017d19985178e1192834d0baba16c071f3ab98e", "parents": [ "75d63af41e85eea0e6b13714f1be32fc08efb16e" ], "author": { "name": "Oleksandr \"Alex\" Zinenko", "email": "git@ozinenko.com", "time": "Mon Apr 20 08:52:46 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:52:46 2026" }, "message": "[mlir] reduce excessive verification in transform (#192653)\n\n`mergeSymbolsInto` called by the transform interpreter for named\nsequence management was calling a full verifier after renaming symbols.\nThe renaming could have potentially broken symbol table-related\ninvariants, but not really anything else. Only verify the symbol\ntable-related invariants instead." }, { "commit": "75d63af41e85eea0e6b13714f1be32fc08efb16e", "tree": "362204ba7b276e01360037de980f48f68ee68424", "parents": [ "f2efeabe314bb0a9b1ef46c07b11f605ed351b9c" ], "author": { "name": "David Green", "email": "david.green@arm.com", "time": "Mon Apr 20 08:51:04 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:51:04 2026" }, "message": "[AArch64] Fix lowering of non-power2 uitofp (#190921)\n\nThe code in DAGTypeLegalizer::SplitVecOp_TruncateHelper attempts to use\ngetFloatingPointVT(InElementSize/2), which is invalid for non-power2\ntype sizes. Fall back to the existing SplitVecOp_UnaryOp in this case." }, { "commit": "f2efeabe314bb0a9b1ef46c07b11f605ed351b9c", "tree": "8cf86970b6a25ee8ae59086a04d1e863958624cd", "parents": [ "920b203fe2a2ad9f4403e8923b2b06518a46129e" ], "author": { "name": "Cullen Rhodes", "email": "cullen.rhodes@arm.com", "time": "Mon Apr 20 08:48:21 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:48:21 2026" }, "message": "[CodeGen] Parse frame-pointer attribute once when creating MachineFunction (#191974)\n\nTargetOptions::DisableFramePointerElim is hot and showing up in\ncompile-time profiles via AArch64FrameLowering::hasFPImpl on\naarch64-O0-g builds. Repeatedly looking up the function attribute is\nexpensive. Parsing it once at MachineFunction initialisation and storing\nas FramePointerKind on MachineFrameInfo is a -0.21% geomean improvement\non CTMark stage1-aarch64-O0-g. Also helps debug builds on other targets.\n\nhttps://llvm-compile-time-tracker.com/compare.php?from\u003d215f35eb8f1c313ac135ad47db1cc0b99b3ae694\u0026to\u003d51f6617517177bea1cc49baeab3acaf62d5e9df9\u0026stat\u003dinstructions%3Au" }, { "commit": "920b203fe2a2ad9f4403e8923b2b06518a46129e", "tree": "3b83948e1194278d153a6634ccf084cb465d72e0", "parents": [ "936643678fa80eceea717ad304300d8a90340df0" ], "author": { "name": "Georgiy Samoylov", "email": "Ignitor21838@gmail.com", "time": "Mon Apr 20 08:46:46 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:46:46 2026" }, "message": "[lldb][RISCV] Implement access to TLS variables on RISC-V (#191410)\n\nOn RISC-V Linux, LLDB computes TLS variable addresses incorrectly:\n`GetThreadLocalData` returns a correct tls_block, but then\nunconditionally adds tls_file_addr from `DW_OP_GNU_push_tls_address`,\nwhich on RISC-V/glibc is a VMA inside PT_TLS, not a pure offset. This\nresults in an over-shifted address.\n\nThis patch:\n\n* Adds a small helper that, for an ELF module, finds the PT_TLS program\nheader and reads its p_vaddr.\n\n* In `DynamicLoaderPOSIXDYLD::GetThreadLocalData`, normalizes\ntls_file_addr to an offset: if `PT_TLS` is found and tls_file_addr \u003e\u003d\np_vaddr, it uses tpoff \u003d tls_file_addr - p_vaddr, otherwise keeps the\nold value.\n\n* Returns tls_block + tpoff instead of always tls_block + tls_file_addr.\n\nThis fixes reading TLS variables on RISC-V/glibc while leaving behavior\nunchanged on targets where DWARF already provides a plain offset (e.g.\nx86_64).\n\nAlso this patch fixes `TestTlsGlobals.py`\n\n---------\n\nCo-authored-by: Jonas Devlieghere \u003cjonas@devlieghere.com\u003e" }, { "commit": "936643678fa80eceea717ad304300d8a90340df0", "tree": "136914353985da42189e5eed25924b04345064b5", "parents": [ "c95a333de71081dc7c23412bd4456d6241cba9fd" ], "author": { "name": "Hocky Yudhiono", "email": "hocky.yudhiono@gmail.com", "time": "Mon Apr 20 08:44:25 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:44:25 2026" }, "message": "[mlir][tensor] Preserve tensor encodings when materializing tensor.empty in some passes (#192411)\n\nThis PR fixes tensor encoding propagation bugs in some `tensor.empty`\nmaterialization paths that could produce type-invalid IR (encoded result\nexpected, unencoded value produced).\n\nAssisted-by: Cursor (Codex 5.3)" }, { "commit": "c95a333de71081dc7c23412bd4456d6241cba9fd", "tree": "f3080947c8da696d14dfe3bf7b7eba7b0adc2260", "parents": [ "bafbd54a904ffd69f36e6d9ffcd28e6c1431db97" ], "author": { "name": "Ryan Cowan", "email": "ryan.cowan@arm.com", "time": "Mon Apr 20 08:38:16 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:38:16 2026" }, "message": "[AArch64][GlobalISel] FP Info implementation for AArch64. (#177158)\n\nThis work sits on top of #155107. The aim is to implement support for\nextended types in the AArch64 backend.\n\nMuch of the implementation just builds upon #155107 but features changes to\nthe MatchTableExecutor to allow for matching multiple patterns to reduce\nthe need for duplicated patterns. This patch also features a new match\ntable opcode to match a pattern based on the shape of a type.\n\n---------\n\nCo-authored-by: David Green \u003cdavid.green@arm.com\u003e" }, { "commit": "bafbd54a904ffd69f36e6d9ffcd28e6c1431db97", "tree": "584d1f00ba87fb01557eb325cef2a2a85e49770a", "parents": [ "76fdcce7f3ef0e24f9e9f214a21c437356dc3d7b" ], "author": { "name": "Pavel Labath", "email": "pavel@labath.sk", "time": "Mon Apr 20 08:36:40 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:36:40 2026" }, "message": "[libc] Add \"struct linger\" (#192606)\n\nAdd a simple test to get/set the socket option. I didn\u0027t try to test the\nactual lingering behavior. That sounds complicated and I\u0027m not sure if\nit\u0027s even doable on a loopback connection." }, { "commit": "76fdcce7f3ef0e24f9e9f214a21c437356dc3d7b", "tree": "5431430e18cb921319e5194c82f20c9a165c4f0a", "parents": [ "f1fe791ab552990185688fa096173149403eb59f" ], "author": { "name": "David Spickett", "email": "david.spickett@arm.com", "time": "Mon Apr 20 08:32:34 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:32:34 2026" }, "message": "[lldb][docs] Update standalone build instructions (#192613)\n\n* LLVM requires CMake 3.20\n(https://llvm.org/docs/GettingStarted.html#software) so we do not need\nto mention 3.14 anymore.\n* CMAKE_BUILD_TYPE was listed twice in one command.\n* \"ninja\" only works when in the build directory or given `-C \u003cdir\u003e`, so\nI have changed that to \"cmake --build\" which works with ninja and other\nbuild tools." }, { "commit": "f1fe791ab552990185688fa096173149403eb59f", "tree": "238825c40b4023109b3b002e6ad31e8199cda8d0", "parents": [ "255190d92f94fcedc54ef794666eadcb1e16f51b" ], "author": { "name": "Pavel Labath", "email": "pavel@labath.sk", "time": "Mon Apr 20 08:32:13 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:32:13 2026" }, "message": "[DWARFYAML] Add support for v5 debug_line file/dir entries (#192226)\n\nThis lets us specify all fields in the v5 header. Since v5 entries are\nform-based, I\u0027ve extracted the relevant parts of the debug_info DIE\nwriting code so it could be reused here as well.\n\nThe v5 file and directory entries are more expressive than \u003c\u003dv4 ones, so\none could in theory store everything in the v5 format, while still\nreading (YAML) and writing (raw DWARF) in the old format. However, that\nwould create more corner cases (what if the data cannot be represented\nin the older format), and it didn\u0027t seem like it was particularly\nworthwhile to handle those." }, { "commit": "255190d92f94fcedc54ef794666eadcb1e16f51b", "tree": "97fad27d8dc63f3233025e42ca14530e03c329b0", "parents": [ "9ca929f53508f93e282f37cb3e1295b53d7e58fb" ], "author": { "name": "David Spickett", "email": "david.spickett@arm.com", "time": "Mon Apr 20 08:30:07 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:30:07 2026" }, "message": "[libcxx][ci] Set CMAKE_C_COMPILER_TARGET for all Arm/AArch64 builds (#192645)\n\nAs requested on #192493.\n\nThis is not strictly needed for native builds, but setting only\nCMAKE_CXX_COMPILER_TARGET does look suspicious. Especially as we often\nset both CXX_FLAGS and C_FLAGS in the same builds.\n\nSet both C_COMPILER_TARGET and CXX_COMPILER_TARGET so on one has to\nwonder if it\u0027s the cause of a problem.\n\n(note that picolibc builds are already setting both)" }, { "commit": "9ca929f53508f93e282f37cb3e1295b53d7e58fb", "tree": "f5b159a97041cec9a6bfdcd805b15817ceece2e4", "parents": [ "e5064d7c6c7a1b9dcb9da1106980495f6f131bc2" ], "author": { "name": "Donát Nagy", "email": "donat.nagy@ericsson.com", "time": "Mon Apr 20 08:25:22 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:25:22 2026" }, "message": "[NFC][analyzer] Eliminate BranchNodeBuilder (#192744)\n\nThis commit removes the class `BranchNodeBuilder` because it didn\u0027t\nprovide enough advantages to justify its existence. (This class wasn\u0027t\nas bad as `SwitchNodeBuilder` and `IndirectGotoNodeBuilder`, but its\nsingle method was very simple so I still think that it is better to\neliminate it.)\n\nTo be able to unify the use of `LocationContext`s in `processBranch`,\nthis commit also replaces the overcomplicated helper function\n`getInlinedLocationContext` with use of `LocationContext::inTopFrame()`.\nThe code of this helper function was written in March 2012, before the\nintroduction of `inTopFrame` (November 2012). It was originally in the\nbody of the method `ExprEngine::processCFGBlockEntrance` and when I\nextracted it in 2025 by commit 9600a12f0de233324b559f60997b9c2db153fede\n(because I also needed to call it from `processBranch`) I didn\u0027t notice\nthat I can achieve the same goal with simpler logic.\n\nAdditionally, a last stray reference to `friend class SwitchNodeBuilder`\nis removed (it was overlooked I removed that class in\nc80443cd37b2e2788cba67ffa180a6331e5f0791)." }, { "commit": "e5064d7c6c7a1b9dcb9da1106980495f6f131bc2", "tree": "2cec4f1e34468775a327eb9687878dd73e1e6034", "parents": [ "069616af55692f47372c01a0f20f1ca615d034f5" ], "author": { "name": "jeanPerier", "email": "jperier@nvidia.com", "time": "Mon Apr 20 08:19:48 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:19:48 2026" }, "message": "[flang] preserve representation of logical constants from TRANSFER (#192417)\n\nThis patch fixes the reproducer from\nhttps://github.com/llvm/llvm-project/issues/192234.\n\nLogical constant have canonical representation in flang (0 and 1, which\nis compiler specific), but other values can be obtained via TRANSFER\nfrom integer (in which case they are true if and only if different from\nzero, which is a compiler specific interpretation).\n\nWhile using TRANSFER to obtain logical is not recommended because it is\nnot portable, the standard at guarantee that round trips should preserve\nthe value.\n\nThis was not the case with constants because lowering was always\nlowering logical constants to either 0 or 1 even when they are obtained\nvia TRANSFER. This patch fixes this by lowering non canonical logical\nconstants to an integer constant + bitcast. A folder is added to fold\nconverts to i1 of such bitcast so that usage of such TRANSFER in\nbranches for instance can still be optimized at the MLIR level." }, { "commit": "069616af55692f47372c01a0f20f1ca615d034f5", "tree": "653f60356bb2b29db905be34b3d8d32f16de46cb", "parents": [ "832d6dba668d8ea559dd21d52b264d38f97f0ae6" ], "author": { "name": "Sander de Smalen", "email": "sander.desmalen@arm.com", "time": "Mon Apr 20 08:05:38 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 08:05:38 2026" }, "message": "[LV] NFC: Address post-merge comments on splice-cost.ll (#192624)\n\nAddresses comments left on #192473 after merging." }, { "commit": "832d6dba668d8ea559dd21d52b264d38f97f0ae6", "tree": "53f96999d22b83aaface05856fccc4362e089238", "parents": [ "e3db5e0cb17199bcb30fba51ff6cf58094fa1f7c" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Mon Apr 20 04:41:30 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 04:41:30 2026" }, "message": "Revert \"[JTS][Passes] Enable JTS By Default\" (#192894)\n\nReverts llvm/llvm-project#192884\n\nStill causes failures on the BOLT builder." }, { "commit": "e3db5e0cb17199bcb30fba51ff6cf58094fa1f7c", "tree": "aba62013b4e61803b0ac7c8c485f96f8a461e536", "parents": [ "36f8b96c88e44a0ef4735f6befecd1a2f04f3eb6" ], "author": { "name": "lrzlin", "email": "linrunze@loongson.cn", "time": "Mon Apr 20 02:33:09 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 02:33:09 2026" }, "message": "[clang][LoongArch] Check target features in CheckLoongArchBuiltinFunctionCall (#191811)\n\nAdd target features check in `CheckLoongArchBuiltinFunctionCall`, thus\nwe could through an error\nwhen pass the `-mno-lsx` to clang while using the builtin LSX intrinsics\nfor global variables instead of\ntrigger an ICE.\n\nMinimal Example:\n```\n// clang-20 --march\u003dloongarch64 -mno-lsx -S -o - \"x.cc\"\n__attribute__((__vector_size__(16))) long foo \u003d __builtin_lsx_vinsgr2vr_w(foo, 0, 0);\n```\nand the compiler will output\n```\nx.cc:1:49: error: builtin needs target feature lsx\n 1 | __attribute__((__vector_size__(16))) long foo \u003d __builtin_lsx_vinsgr2vr_w(foo, 0, 0);\n | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n1 error generated.\n```\n\nFix https://github.com/llvm/llvm-project/issues/131771" }, { "commit": "36f8b96c88e44a0ef4735f6befecd1a2f04f3eb6", "tree": "f764118abdfdc13877b267835725b2c952d785d2", "parents": [ "d8e9862e71b6bb7f5b8c06159429f31a1516299c" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Mon Apr 20 02:28:37 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 02:28:37 2026" }, "message": "Reapply \"[JTS][Passes] Enable JTS By Default\" (#192884)\n\nReverts llvm/llvm-project#192737\n\nReapplies https://github.com/llvm/llvm-project/pull/190674\n\nThe one remaining issue should have been fixed by\nhttps://github.com/llvm/llvm-project/pull/192877." }, { "commit": "d8e9862e71b6bb7f5b8c06159429f31a1516299c", "tree": "cb125cd3dcbeca5c5264767f89a9f630da6e5876", "parents": [ "cbd802b07289c210d0f9f299d25176cd797be1b7" ], "author": { "name": "Fangrui Song", "email": "i@maskray.me", "time": "Mon Apr 20 02:22:24 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 02:22:24 2026" }, "message": "[Driver] Gnu: Move -s/-t/-u emission to match GCC order (#192883)\n\nGCC places -s, -t, and -u sym in one contiguous group just before\nScrt1.o / crt1.o, with -L paths after the CRT files.\n\nMatch that ordering by dropping the early `push_back(\"-s\")` and rolling\n-s, -t, and -u into one addAllArgs call placed immediately after -o\noutput. This keeps -Wl,... after -s/-t/-u so that user overrides like\n-Wl,--strip-debug still take precedence. Update linux-ld-args.c to\n\n-T remains at the end so earlier -L paths take precedence." }, { "commit": "cbd802b07289c210d0f9f299d25176cd797be1b7", "tree": "df8ded3fd4af1a460cf660b67c944bf8419b42b8", "parents": [ "9542cfa1ad91c8510904f04fd780e4fcad319694" ], "author": { "name": "Akimasa Watanuki", "email": "mencotton0410@gmail.com", "time": "Mon Apr 20 01:54:13 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 01:54:13 2026" }, "message": "[MLIR][Vector] Fix `scf.for` block-argument yields in warp distribution (#192247)\n\nTeach WarpOpScfForOp to remap yielded `scf.for` body block arguments\nthrough `argMapping` before creating the replacement `gpu.yield`.\n\nHandle yielded loop-carried values and other `scf.for` body block\narguments when moving the loop body into the new inner warp op, instead\nof reusing the pre-merge values.\n\nAdd a regression test for yielding a loop-carried block argument during\nwarp distribution.\n\nFix #186573" }, { "commit": "9542cfa1ad91c8510904f04fd780e4fcad319694", "tree": "051f191b10c170c40747a478614933b8cd09e1c0", "parents": [ "837f7e958dea6e86a680f4016e3fcafa00dcd4af" ], "author": { "name": "Alexander Shaposhnikov", "email": "ashaposhnikov@google.com", "time": "Mon Apr 20 01:46:48 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 01:46:48 2026" }, "message": "[JumpTableToSwitch] Fix wrong function used for GUID computation (#192877)\n\nThe FuncToGuid lambda\u0027s fallback path (when target functions lack !guid\nmetadata) was using \u0027F\u0027 (the caller) instead of \u0027Fct\u0027 (the callee) in\ngetIRPGOFuncName, causing all GUID lookups to resolve to the caller\u0027s\nGUID." }, { "commit": "837f7e958dea6e86a680f4016e3fcafa00dcd4af", "tree": "9c2148defa58b28feaaefb0aab89952a1d2f3598", "parents": [ "257706f21227ba596c81b5c97a3d4f0d33aa15d5" ], "author": { "name": "Mend Renovate", "email": "bot@renovateapp.com", "time": "Mon Apr 20 01:43:23 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 01:43:23 2026" }, "message": "Update [Github] Update GHA Dependencies (major) (#192881)\n\nThis PR contains the following updates:\n\n| Package | Type | Update | Change |\n|---|---|---|---|\n|\n[actions/create-github-app-token](https://redirect.github.com/actions/create-github-app-token)\n| action | major | `v2.2.1` → `v3.1.1` |\n|\n[actions/github-script](https://redirect.github.com/actions/github-script)\n| action | major | `v8.0.0` → `v9.0.0` |" }, { "commit": "257706f21227ba596c81b5c97a3d4f0d33aa15d5", "tree": "6f68c6c26ee66de4d12837431418422b17ab1798", "parents": [ "1b2ccf9c6c79d6e57ba6f0a4cf8a3bd1075edebb" ], "author": { "name": "HighW4y2H3ll", "email": "zhenghaohuu@gmail.com", "time": "Mon Apr 20 01:30:31 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 01:30:31 2026" }, "message": "[llvm-profgen] Add branch/target validation (#188620)\n\nAdd extra branch source and target validation checks for LBR samples.\nThis is to check whether there are branch source samples that do not\nmatch a call/branch/ret instruction in the binary, and branch target\nsamples that do not match a resolved Imm target address, or a function\nstart address (in case of an indirect call).\n\nExample output:\n```\n# X86\nwarning: 0.01%(27/376876) of sampled target addresses do not match the binary.\n# AArch64\nwarning: 0.01%(63782/795824826) of branch samples do not match the binary.\nwarning: 0.01%(70468/795824826) of branch targets do not match the binary.\n```\nRun time overhead:\n```\nBefore:\nreal 61m38.081s\nuser 60m50.768s\nsys 0m42.931s\n\nAfter:\nreal 62m33.176s\nuser 61m47.650s\nsys 0m42.204s\n```" }, { "commit": "1b2ccf9c6c79d6e57ba6f0a4cf8a3bd1075edebb", "tree": "3a8b3dc4f11b2959c6910c5c39442fea5e67ce63", "parents": [ "89f185654a2c59537c38c0ba2447c29b13586ee1" ], "author": { "name": "Shoreshen", "email": "372660931@qq.com", "time": "Mon Apr 20 00:43:22 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 00:43:22 2026" }, "message": "[AMDGPU] Generate waterfall for calls with SGPR(inreg) argument (#146997)\n\nFixing issue https://github.com/llvm/llvm-project/issues/140780\n\nGenerate waterfall loop for call using SGPR(inreg) argument but result\nfrom divergent source (e.g. VGPR)." }, { "commit": "89f185654a2c59537c38c0ba2447c29b13586ee1", "tree": "7106a246a4782d56b5c70cf01e33f714039a1ab4", "parents": [ "cea56f648ef4c0972c04c5230393f1e2a5d590b9" ], "author": { "name": "Victor Chernyakin", "email": "chernyakin.victor.j@outlook.com", "time": "Mon Apr 20 00:33:45 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Apr 20 00:33:45 2026" }, "message": "[clang-tidy][NFC] Add test case confirming #190944 is fixed (#192707)\n\nCloses #190944.\n\nThis issue is already fixed, and this change just adds a test case to\nconfirm that." }, { "commit": "cea56f648ef4c0972c04c5230393f1e2a5d590b9", "tree": "905c08790f8abb6c65c37712ddc527d99c0ac964", "parents": [ "e08ec62ddc4b1ffa75bf8dd9486620a9f1c9e78c" ], "author": { "name": "Florian Hahn", "email": "flo@fhahn.com", "time": "Sun Apr 19 21:33:57 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 21:33:57 2026" }, "message": "[LV] Add additional test cases with predicated inductions. (NFC) (#192875)\n\nExtend test coverage with predicated IVs both with and without\nadditional predicates from LoopAccessInfo." }, { "commit": "e08ec62ddc4b1ffa75bf8dd9486620a9f1c9e78c", "tree": "7e59cf3debe8ba7731d25a8bd08d17b31f461d32", "parents": [ "cac7fe50e0fbedfb14028c170d83386efeb1265b" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Sun Apr 19 21:26:14 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 21:26:14 2026" }, "message": "[SelectOptimize] Emit Fatal Error instead of Asserting on null PSI (#192871)\n\nSelectOptimize expects to have PSI available which will normally be\navailable if the pipeline is set up correctly to require\nProfileSummaryInfo at the beginning. However, we do not want to assert\nif someone sets up the pipeline incorrectly, instead reporting a fatal\nusage error.\n\nFixes #192759." }, { "commit": "cac7fe50e0fbedfb14028c170d83386efeb1265b", "tree": "62d45066fc889f77706a9b4797ff6c41a9ebb86a", "parents": [ "abc37e6aef740b0af087331e6db807b916f559f5" ], "author": { "name": "Florian Hahn", "email": "flo@fhahn.com", "time": "Sun Apr 19 21:21:04 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 21:21:04 2026" }, "message": "[VPlan] Make canonical IV part of the region (#156262)\n\nThe canonical IV is directly tied to a loop region. To directly ensure\nthere\u0027s a single, unique canonical IV, directly define it by the region.\n\nDepends on https://github.com/llvm/llvm-project/pull/161589.\n\nPR: https://github.com/llvm/llvm-project/pull/156262" }, { "commit": "abc37e6aef740b0af087331e6db807b916f559f5", "tree": "e7ba7e921d9e36c3911880e0b46fafce44fa3321", "parents": [ "4bd6175f718f032cc391de9263a3d4c14ae5109d" ], "author": { "name": "Nico Weber", "email": "thakis@chromium.org", "time": "Sun Apr 19 20:16:07 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 20:16:07 2026" }, "message": "[gn build] Port 4b0dd87d3a29 (#192869)" }, { "commit": "4bd6175f718f032cc391de9263a3d4c14ae5109d", "tree": "b63a70a06eef8c6fa8c0dfcc2121b3e8eb386d23", "parents": [ "e7cdce484bd273b3d6a26d4edc54ae4c37cea3b0" ], "author": { "name": "Benjamin Maxwell", "email": "benjamin.maxwell@arm.com", "time": "Sun Apr 19 19:48:55 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 19:48:55 2026" }, "message": "[AArch64][SME] Remove a couple of unused functions (NFC) (#192649)\n\nFollow up to https://github.com/llvm/llvm-project/pull/190950" }, { "commit": "e7cdce484bd273b3d6a26d4edc54ae4c37cea3b0", "tree": "ebcfcd4bc0572df3e7c1747227e9760b44bce2a4", "parents": [ "c12a3bb65edf117cfd209e1bb8029beadb0c1c01" ], "author": { "name": "Alex MacLean", "email": "amaclean@nvidia.com", "time": "Sun Apr 19 19:45:39 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 19:45:39 2026" }, "message": "[BitcodeReader] Simplify CST_CODE_DATA constant parsing (NFC) (#190846)\n\nCleanup boilerplate in the CST_CODE_DATA case for bitcode parser. This\nis generally an aesthetic improvement and can significantly reduce stack\nusage on some compilers (such as MSVC) where each SmallVector previously\nwas allocated it\u0027s own individual space." }, { "commit": "c12a3bb65edf117cfd209e1bb8029beadb0c1c01", "tree": "7e582cd8294700dd197dc394b70ece5c8a73ad5b", "parents": [ "d34661b89c7978b28005811c557ca8aca1a23d0d" ], "author": { "name": "Fangrui Song", "email": "i@maskray.me", "time": "Sun Apr 19 19:02:45 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 19:02:45 2026" }, "message": "[test] Improve driver option coverage (#192861)\n\nSplit the Linux GNU-linker assertions into a dedicated file and extend\nthem with previously-untested passthrough flags handled by\ngnutools::Linker::ConstructJob: -s, -u sym, -rdynamic (-export-dynamic),\n-Wl,-z keyword pairs.\n\nLoongArch on Linux previously had no test for -X / --no-relax, which are\nemitted by Gnu.cpp for LoongArch and RISC-V." }, { "commit": "d34661b89c7978b28005811c557ca8aca1a23d0d", "tree": "9d7c1b335c4960f4237814bc3be19256167e5b3f", "parents": [ "5492150b2078f03c9c1942c13f6d0428a530e165" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Sun Apr 19 19:02:24 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 19:02:24 2026" }, "message": "[LPM][LegacyPM] Reenable LCSSA Verification\n\nThis was disabled about a decade ago due to issues with LoopSink.\nLoopSink has since had its LegacyPM version removed and is now a\nfunction pass due to not needing too much loop infrastructure. So we can\ntry enabling this again to prevent backsliding on important cases while\nwe work on switching to the NewPM which does enforce these things.\n\nEventually we will want to add assertions here for LoopStrengthReduce,\nbut given it does not correctly preserve LCSSA, postpone that for now.\n\nReviewers: arsenm, Meinersbur, nikic, fhahn\n\nPull Request: https://github.com/llvm/llvm-project/pull/191667" }, { "commit": "5492150b2078f03c9c1942c13f6d0428a530e165", "tree": "2629d42772a7d78ef11789759e5a5b756b9ff835", "parents": [ "e93f7e595cbcbc995c97942ae343a1382f921f18" ], "author": { "name": "Pedro Lobo", "email": "pedro.lobo@tecnico.ulisboa.pt", "time": "Sun Apr 19 18:58:19 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 18:58:19 2026" }, "message": "[ConstantFold] Support byte values in `bitcast` constant folding (#188030)\n\nAdd support for constant folding `bitcast` instructions including\n`ConstantByte` values. This patch handles bitcasts between byte types\nand integer, FP, and other byte types in both directions.\n\n`poison` source bytes are preserved, rather than letting the generic\ninteger fold refine them to `undef` or zero. This is because some\nthreading optimizations just compare the result of constant folding\n(e.g., https://github.com/llvm/llvm-project/pull/114280).\n\nFolds are skipped for byte `bitcast`s where element counts don\u0027t divide\nevenly and the source contains `poison` values. Some of these casts can\nbe folded. However, this is left for a future PR." }, { "commit": "e93f7e595cbcbc995c97942ae343a1382f921f18", "tree": "2baa50893d4924426713dbbb4594dbc508a5d456", "parents": [ "f5e2b8b1df079a90f0756d75dcafff8e8efb8eb1" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Sun Apr 19 18:56:30 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 18:56:30 2026" }, "message": "[NFC][Loop] Remove unused verbose loop debug output\n\nThis has been defined out using the preprocessor for ~14 years. Given it\ndoesn\u0027t look like it has gotten any use since then, just remove it to\nclean the code up a bit.\n\nReviewers: nikic\n\nPull Request: https://github.com/llvm/llvm-project/pull/192830" }, { "commit": "f5e2b8b1df079a90f0756d75dcafff8e8efb8eb1", "tree": "18273e14cf6ab5077a81f7c4d810729d3d24ca68", "parents": [ "b6a4fdb8035d78934f98e823f67a5adb8c333828" ], "author": { "name": "Jonas Devlieghere", "email": "jonas@devlieghere.com", "time": "Sun Apr 19 18:54:21 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 18:54:21 2026" }, "message": "[lldb] Rally around triple rather than arch in the API tests (#192818)\n\nThis PR removes as much uses of arch as possible, in favor of using\ntriple directly. Most of the changes are in the builder, which no longer\npasses ARCH to Make, and of course in Makefile.rules.\n\nThis significantly simplifies the remote Darwin test suite, as it\npreviously had to try and piece together the triple from the platform\nand the arch. As an added benefit, we now go through the same code path\nfor host and remote test runs.\n\nI have tested this on Darwin and Linux and made the changes with the\nremote test suites in mind, but it\u0027s possible I missed something not\ncaught by my local testing." }, { "commit": "b6a4fdb8035d78934f98e823f67a5adb8c333828", "tree": "ac0063989af20741efae6240b654b653149df86b", "parents": [ "fcc61927ad4c9d455079728daa1e6b99106ed355" ], "author": { "name": "Aiden Grossman", "email": "aidengrossman@google.com", "time": "Sun Apr 19 18:49:03 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 18:49:03 2026" }, "message": "[LSR] Autogenerate some tests\n\npr25541.ll - Was a regression test for a crash. Make a note and\nautogenerate the tests.\nbin_power.ll - Was essentially doing what UTC would do already, minus\none test where the assembly output is somewhat large, but the large\nassembly output shouldn\u0027t be an issue with an autoupdate script and\nisn\u0027t big enough to justify not including in my opinion.\n\nThis makes updating some tests easier for planned changes to LCSSA\npreservation.\n\nReviewers: fhahn, nikic\n\nPull Request: https://github.com/llvm/llvm-project/pull/191664" }, { "commit": "fcc61927ad4c9d455079728daa1e6b99106ed355", "tree": "b04bdfb516d3485f451e65998ece8e9d8a423203", "parents": [ "a3f244e2d5550e110daee562b88bee247b75d66c" ], "author": { "name": "Henrich Lauko", "email": "xlauko@mail.muni.cz", "time": "Sun Apr 19 16:35:35 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 16:35:35 2026" }, "message": "[CIR] Emit inbounds nuw flags on GetMemberOp GEP lowering (#186738)\n\nStruct member accesses via GetMemberOp are always inbounds and cannot\nunsigned-wrap, matching LLVM\u0027s IRBuilder::CreateStructGEP behavior." }, { "commit": "a3f244e2d5550e110daee562b88bee247b75d66c", "tree": "b1aafb22bdec8f53c5f759044133043bf7edd675", "parents": [ "17e1a3b875faddbbfd4f5329d490d164943260d0" ], "author": { "name": "Jakub Mazurkiewicz", "email": "mazkuba3@gmail.com", "time": "Sun Apr 19 15:59:28 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 15:59:28 2026" }, "message": "[libc++][functional] Implement `std::bind_front\u003cNTTP\u003e` (#165096)\n\nAdd `std::bind_front\u003cNTTP\u003e` function from\n[P2714R1](https://wg21.link/p2714r1).\n\nTowards https://github.com/llvm/llvm-project/issues/105388." }, { "commit": "17e1a3b875faddbbfd4f5329d490d164943260d0", "tree": "741a01840e36d0fe9de83f2a5f31246263178db2", "parents": [ "26b7fd51095221e03ce2a2170b6d168652ba3c7d" ], "author": { "name": "eiytoq", "email": "eiytoq@outlook.com", "time": "Sun Apr 19 13:13:57 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 13:13:57 2026" }, "message": "[libc++] Enable `mdspan::operator[]` constraint tests for AppleClang (#192475)\n\nAlthough I couldn\u0027t find a corresponding AppleClang release note, this\nissue appears to be fixed in AppleClang 16." }, { "commit": "26b7fd51095221e03ce2a2170b6d168652ba3c7d", "tree": "03f8bd33383afaf12e6f20fb5f669243bbc06550", "parents": [ "d4974dfe67c9a322f3b9a913dbbaf4e4f4a58a08" ], "author": { "name": "Florian Hahn", "email": "flo@fhahn.com", "time": "Sun Apr 19 09:56:21 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 09:56:21 2026" }, "message": "[VPlan] Support plain CFG plans in the verifier. (#181817)\n\nUpdate the verifier to support verifying plain CFG plans in the verifier\nand add missing support for switch opcodes in type analysis.\n\nThis allows the verifier to handle early plain-CFG plans, allowing using\nRUN_PASS wrapper for early transforms.\n\nPR: https://github.com/llvm/llvm-project/pull/181817" }, { "commit": "d4974dfe67c9a322f3b9a913dbbaf4e4f4a58a08", "tree": "749d99bc84bcd23c1f968243a4a8603f3a9c1e89", "parents": [ "bd5b29314660232e74a43b684cbecf6f0c1e3983" ], "author": { "name": "Matthias Springer", "email": "me@m-sp.org", "time": "Sun Apr 19 09:51:52 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 09:51:52 2026" }, "message": "[mlir][docs] Add poison example to canonicalization section (#192782)\n\nMention that folding to `ub.poison` is a common canonicalization." }, { "commit": "bd5b29314660232e74a43b684cbecf6f0c1e3983", "tree": "8bc029ba897ebc04a156bd9770b6af657da9f8ba", "parents": [ "5c723d819668aaa13b6dd57f669e9fad27129e60" ], "author": { "name": "jumerckx", "email": "31353884+jumerckx@users.noreply.github.com", "time": "Sun Apr 19 09:47:09 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 09:47:09 2026" }, "message": "[mlir] Support full commutative operation equality (#192652)\n\nPrevious, commutative equality only works if the operand lists are\nexact permutations of one another.\nBy treating the `equivalentValues` map as a map onto a common set of\nvalues, we can achieve full commutative equality." }, { "commit": "5c723d819668aaa13b6dd57f669e9fad27129e60", "tree": "82fb158aee32fd5f2d55126b11ee99dca2e4052d", "parents": [ "d8c14330dcc17610f634b45d1730b7a6e46d91cd" ], "author": { "name": "Jason Molenda", "email": "jmolenda@apple.com", "time": "Sun Apr 19 08:56:43 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 08:56:43 2026" }, "message": "[lldb][macOS] add additional logging print to TestRosetta (#192837)\n\nto clearly show that we\u0027re executing x86_64 binaries in this debug\nsession." }, { "commit": "d8c14330dcc17610f634b45d1730b7a6e46d91cd", "tree": "4cae6ae3561eb01e1d5eefcc84e5e6f7719d6622", "parents": [ "2a3667bbb54081dee0eb7eb28de81631e393199a" ], "author": { "name": "Alisson Azzolini", "email": "37222419+aazzolini@users.noreply.github.com", "time": "Sun Apr 19 08:07:56 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 08:07:56 2026" }, "message": "[mlir][python] Add location source composition to loc_tracebacks() (#192310)\n\n## Summary\n\nAdd two new kwargs to `loc_tracebacks()` controlling how the three\nlocation sources (explicit `loc\u003d`, traceback, `Location.current`)\ncompose:\n\n- **`on_explicit`**: `OnExplicitAction.USE_EXPLICIT` (default) |\n`OnExplicitAction.USE_TRACEBACK` — what to do when explicit `loc\u003d` is\npassed\n- **`current_loc`**: `CurrentLocAction.FALLBACK` (default) |\n`CurrentLocAction.NAMELOC_WRAP` — how to compose `Location.current`\nNameLoc scopes on top\n\nBoth enums are exposed to Python via `nb::enum_\u003c\u003e` bindings, following\nthe `PassDisplayMode` pattern in `Pass.cpp`.\n\nThe two flags are orthogonal and only take effect when\n`loc_tracebacks()` is active.\n\n## Use cases\n\n1. **DSL profiling attribution**:\n`loc_tracebacks(current_loc\u003dCurrentLocAction.NAMELOC_WRAP,\non_explicit\u003dOnExplicitAction.USE_TRACEBACK)` — push NameLoc scopes for\nprofiling and have them survive traceback generation\n2. **Override explicit locs with tracebacks**:\n`loc_tracebacks(on_explicit\u003dOnExplicitAction.USE_TRACEBACK)` — get full\nPython stack traces without removing every `loc\u003d` argument\n\n## Test plan\n\n- [x] All 102 existing Python binding tests pass (5 unsupported\nunchanged)\n- [x] 6 new tests covering: nameloc_wrap, on_explicit default,\non_explicit\u003duse_traceback, DSL profiling combo, orthogonality\n(use_explicit + nameloc_wrap), tracebacks-disabled fallback\n\nCo-authored by Claude Code ;)\n\n---------\n\nCo-authored-by: Claude Opus 4.6 (1M context) \u003cnoreply@anthropic.com\u003e" }, { "commit": "2a3667bbb54081dee0eb7eb28de81631e393199a", "tree": "fd751a8c1935e4c4ec46137534fc616e0dc3db82", "parents": [ "8de4c61386b3f94ad77060c1047d7cc4c979415c" ], "author": { "name": "Jason Molenda", "email": "jmolenda@apple.com", "time": "Sun Apr 19 08:05:11 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 08:05:11 2026" }, "message": "Revert \"[lldb][CI] force TestRosetta.py to fail for a minute, so I (#… (#192836)\n\n…192835)\"\n\nThis reverts commit 8de4c61386b3f94ad77060c1047d7cc4c979415c." }, { "commit": "8de4c61386b3f94ad77060c1047d7cc4c979415c", "tree": "5a0b9c014dfb7bdd0f8fa5e3ed701207235d464c", "parents": [ "ce27dcc79de9a1effbe62d9fbce737d26634ae88" ], "author": { "name": "Jason Molenda", "email": "jmolenda@apple.com", "time": "Sun Apr 19 07:36:09 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 07:36:09 2026" }, "message": "[lldb][CI] force TestRosetta.py to fail for a minute, so I (#192835)\n\ncan collect the trace logging on the CI bot and see what its environment\nlooks like today." }, { "commit": "ce27dcc79de9a1effbe62d9fbce737d26634ae88", "tree": "fd751a8c1935e4c4ec46137534fc616e0dc3db82", "parents": [ "4cf7796dfb540215f06ba9ad5fe92534f8ab034b" ], "author": { "name": "Vitaly Buka", "email": "vitalybuka@google.com", "time": "Sun Apr 19 07:10:38 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 07:10:38 2026" }, "message": "[NFC][CodeGen] Make a few MachineBlockHashInfo functions static (#192825)" }, { "commit": "4cf7796dfb540215f06ba9ad5fe92534f8ab034b", "tree": "144920783760c2926871d6e8df9b741d955e2c03", "parents": [ "f7abe260a548e5b98ca193cb3b42c1e6896b5a48" ], "author": { "name": "Ramkumar Ramachandra", "email": "artagnon@tenstorrent.com", "time": "Sun Apr 19 06:57:27 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 06:57:27 2026" }, "message": "[VPlan] Mark DerivedIV unconditionally single-scalar (NFC) (#189706)\n\nThe result must be single-scalar, independently of operands. The patch\nis an NFC because all operands happen to be single-scalar currently." }, { "commit": "f7abe260a548e5b98ca193cb3b42c1e6896b5a48", "tree": "f0b557ca0349fe549d3ec4b86daffc83cbad4350", "parents": [ "8225bcc141e749bcdb8b900d1e1a9f9a36fdac00" ], "author": { "name": "Vitaly Buka", "email": "vitalybuka@google.com", "time": "Sun Apr 19 06:20:54 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 06:20:54 2026" }, "message": "[NFC][CodeGen] Use DenseMap instead of unordered_map in BasicBlockMatchingAndInference (#192824)\n\nWe don\u0027t need reference stability of unordered_map here." }, { "commit": "8225bcc141e749bcdb8b900d1e1a9f9a36fdac00", "tree": "f0c8af354f5aaaeddc8adb4f6571a8eff21dde86", "parents": [ "404b75651f7220e9de1801fa987ccf92c1fc95bb" ], "author": { "name": "Michael Buch", "email": "michaelbuch12@gmail.com", "time": "Sun Apr 19 06:03:33 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 06:03:33 2026" }, "message": "[lldb][PlatformDarwin] Make PlatformDarwin define a safe-path for auto-loading scripting resources (#191454)\n\nThis patch adds a new API (`Platform::GetSafeAutoLoadPaths`) which gives\nplatforms a chance to advertise their safe-paths. We have a\n`LLDB_SAFE_AUTO_LOAD_PATHS` CMake variable for this that vendors can\nset, but for sensible defaults we wanted to bake them into LLDB for\nconvenience. We could set the defaults of the CMake variable\nper-platform, but for Apple platforms that\u0027s trickier because the path\nisn\u0027t statically known (it\u0027s the SDK path derived from the target\u0027s\ntriple).\n\nDepends on:\n* https://github.com/llvm/llvm-project/pull/191446\n* https://github.com/llvm/llvm-project/pull/192703\n\nAssisted-by: Claude\n- Used Claude to write the skeleton of the test before manually cleaning\nit up." }, { "commit": "404b75651f7220e9de1801fa987ccf92c1fc95bb", "tree": "2eb751e40b29d66c41f0382904f7862412bd1749", "parents": [ "4113d79a20253cd34820d8ff2031c7e0307b9e66" ], "author": { "name": "Michael Buch", "email": "michaelbuch12@gmail.com", "time": "Sun Apr 19 05:54:45 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 05:54:45 2026" }, "message": "[lldb][Target] Move Debugger::GetSafeAutoLoadPaths into Target (#192703)\n\nRequired for https://github.com/llvm/llvm-project/pull/191454\n\nWe want to make the `Target`/`Platform` define commonly used default\npaths (without configuring them in CMake). This is easiest if this logic\nlived in `Target` (since then we have access to the associated\n`Platform`)." }, { "commit": "4113d79a20253cd34820d8ff2031c7e0307b9e66", "tree": "adf9d62f7932bce14cdd50f57f32329d6a2fdaec", "parents": [ "59785514f9f981c69b933ffb5739201dede521c2" ], "author": { "name": "Vitaly Buka", "email": "vitalybuka@google.com", "time": "Sun Apr 19 05:30:27 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 05:30:27 2026" }, "message": "[NFC][CodeGen] Optimize hash table lookups in MachineBlockHashInfo (#192823)" }, { "commit": "59785514f9f981c69b933ffb5739201dede521c2", "tree": "5942a85a4bc5200f9778eead209cc6433fbc6b25", "parents": [ "5b381d5c00009c73eb85fa8b5341590a1adf8521" ], "author": { "name": "Timm Baeder", "email": "tbaeder@redhat.com", "time": "Sun Apr 19 05:30:14 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 05:30:14 2026" }, "message": "[clang][bytecode][NFC] Avoid classify calls in visitZeroRecordInitializer (#192676)\n\nWe already have the PrimType in the Descriptor." }, { "commit": "5b381d5c00009c73eb85fa8b5341590a1adf8521", "tree": "da4f9f838c6392cd388d3487bd8554a27b8f2ee0", "parents": [ "80a11185b670796a8c4b4cfcf8b1e877e85e486c" ], "author": { "name": "Jason Molenda", "email": "jmolenda@apple.com", "time": "Sun Apr 19 05:01:14 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 05:01:14 2026" }, "message": "[lldb][AArch64] Fix typeo in AArch64 DoFixAddr highmem (#191952)\n\nCode and Data masks are the same on AArch64, but someone could adopt a\nCode mask that cleared the low 2 bits, so it\u0027s good to correct the\nmistake.\n\nrdar://174463000" }, { "commit": "80a11185b670796a8c4b4cfcf8b1e877e85e486c", "tree": "a5da46d1f32b175bb5840c60f02a8569a1088c0b", "parents": [ "2c6ce15199c9c97900f24b308afef38731db8adb" ], "author": { "name": "aokblast", "email": "aokblast@FreeBSD.org", "time": "Sun Apr 19 04:17:26 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 04:17:26 2026" }, "message": "[llvm-ar] Enable failed testacase on FreeBSD (#192612)\n\nFreeBSD used to allow the read system call on directory file descriptors\n(dirfd). In those cases, it returned the internal file data structure\ninstead of failing. This caused the llvm-ar test— which expects a\nfailure when attempting to read a directory— to pass incorrectly.\n\nThis behavior existed for compatibility with older Unix implementations,\nbut it was removed in commit dcef4f65ae3978c50eab745f67364db4660a9f43.\nThe change applies to FreeBSD 13 and later. Since FreeBSD 12 has also\nreached end-of-life, the outdated behavior is no longer relevant.\n\nAs a result, the test now behaves as intended and can be safely\nre-enabled." }, { "commit": "2c6ce15199c9c97900f24b308afef38731db8adb", "tree": "861dc9c35d583b4ea3bfec802f1aeaba3e8fbc03", "parents": [ "26ba308be0eeb652ecbc0ad3c660ef69404c0185" ], "author": { "name": "Andrei (Andrey) Khropov", "email": "andrey.khropov@gmail.com", "time": "Sun Apr 19 00:39:00 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sun Apr 19 00:39:00 2026" }, "message": "[llvm] Fix misspelling in DWARFLinkerTypeUnit (#184335)" }, { "commit": "26ba308be0eeb652ecbc0ad3c660ef69404c0185", "tree": "b7dd13cd02d1644e2eceb75ce7c04f4c17881e13", "parents": [ "0053d6628ba22263a0d0dd12854daa06846b4bf8" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@outlook.com", "time": "Sat Apr 18 22:02:53 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 22:02:53 2026" }, "message": "[SLP][NFC] Simplify range/type accessors\n\nUse the range-friendly llvm::accumulate() overload instead of the\ntwo-iterator std::accumulate() form, drop redundant cast\u003c\u003e/dyn_cast\u003c\u003e\ncalls where the source already provides a typed accessor.\n\nReviewers: \n\nPull Request: https://github.com/llvm/llvm-project/pull/192811" }, { "commit": "0053d6628ba22263a0d0dd12854daa06846b4bf8", "tree": "bc45554862770a6e36471bfda4250584fc317d58", "parents": [ "5bcac7bb5c4957138c929a13359d1350994230bb" ], "author": { "name": "Javed Absar", "email": "javed.absar@gmail.com", "time": "Sat Apr 18 21:32:32 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 21:32:32 2026" }, "message": "[mlir][linalg][elementwise] Add tests for tiling elementwise operation. (#192798)" }, { "commit": "5bcac7bb5c4957138c929a13359d1350994230bb", "tree": "5e928112a4b0194c992be4ad4898184fb6f17033", "parents": [ "e70a2e84ce954c58c2da5bfbee7f69b8c2b1109d" ], "author": { "name": "Florian Hahn", "email": "flo@fhahn.com", "time": "Sat Apr 18 21:17:44 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 21:17:44 2026" }, "message": "[VPlan] Fix isLatch to handle BranchOnTwoConds, check verifier.(NFC) (#192808)\n\nUpdate `isLatch` to handle BranchOnTwoConds,\nin preparation to enable running the verifier on earlier VPlans." }, { "commit": "e70a2e84ce954c58c2da5bfbee7f69b8c2b1109d", "tree": "638867c5a7a569234ebfb7d1568f70140f90b879", "parents": [ "b7f54fd7ecbcd25906b19e472718dd6f93159070" ], "author": { "name": "Florian Hahn", "email": "flo@fhahn.com", "time": "Sat Apr 18 20:42:50 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 20:42:50 2026" }, "message": "[VPlan] Allow same predecessor multiple times in the verifier. (NFC) (#192809)\n\nIR instructions like switch lead to initial VPlans with blocks that have\nthe same predecessor multiple times. Allow it in the verifier." }, { "commit": "b7f54fd7ecbcd25906b19e472718dd6f93159070", "tree": "d2fe87c91dd50af97477e73e2ba7647ef5d577c5", "parents": [ "6ad7d557b9f5b969e96acd49b216e87cd3f1efc6" ], "author": { "name": "Zorojuro", "email": "sawantsukumar@gmail.com", "time": "Sat Apr 18 19:41:15 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 19:41:15 2026" }, "message": "[libc][math] Refactor logbbf16 to header-only (#189194)\n\nCo-authored-by: bassiounix \u003cmuhammad.m.bassiouni@gmail.com\u003e" }, { "commit": "6ad7d557b9f5b969e96acd49b216e87cd3f1efc6", "tree": "fc915f3b229e066a28083eeecf955e806de26d2e", "parents": [ "580910073482c0b49be5364f9595ba2dc2bf8f2e" ], "author": { "name": "Anshul Nigham", "email": "nigham@google.com", "time": "Sat Apr 18 18:59:11 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 18:59:11 2026" }, "message": "[NewPM] Adds a port for AArch64ConditionalComparesPass (#192755)\n\nAdds a standard porting for AArch64ConditionalCompares." }, { "commit": "580910073482c0b49be5364f9595ba2dc2bf8f2e", "tree": "b70ebac23fdc03e757845718327198db5c276e4c", "parents": [ "9d193ada7056dcc61ab061d2e35a4c05d7e61e15" ], "author": { "name": "Fangrui Song", "email": "i@maskray.me", "time": "Sat Apr 18 18:33:36 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 18:33:36 2026" }, "message": "Revert \"[FileCheck] Add a diff output option for FileCheck (#187120)\" (#192803)\n\nThis reverts commit 40333cde2bbd62936d67478eab15ed1fa55b0532.\n\nThe implementation is buggy. There are thousands of failures in\n```\nFILECHECK_OPTS\u003d--diff ninja -C /tmp/Rel check-llvm-mc\nFILECHECK_OPTS\u003d--diff ninja -C /tmp/Rel check-lld-elf\n```" }, { "commit": "9d193ada7056dcc61ab061d2e35a4c05d7e61e15", "tree": "2ffb9eaad944cd174f4b75ddff4d9f63b55126e2", "parents": [ "d1dcce6b6580edfd4c501c17ded6b7cb86ae4ccf" ], "author": { "name": "Jonas Devlieghere", "email": "jonas@devlieghere.com", "time": "Sat Apr 18 18:24:48 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 18:24:48 2026" }, "message": "[lldb] Avoid NotRequired for Python \u003c 3.11 compatibility (#192723)\n\nNotRequired was added in Python 3.11, while the minimum supported Python\nfor LLDB is 3.8. Avoid it by using `Dict[str, Any]` instead." }, { "commit": "d1dcce6b6580edfd4c501c17ded6b7cb86ae4ccf", "tree": "5176bdd540f580e400faf1c05c189777235fa0d4", "parents": [ "4fb21a1d226bff588a66b51452a159b41c5373d5" ], "author": { "name": "Simon Pilgrim", "email": "llvm-dev@redking.me.uk", "time": "Sat Apr 18 17:10:49 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 17:10:49 2026" }, "message": "[X86] Add crash test coverage for #169485 (#192797)" }, { "commit": "4fb21a1d226bff588a66b51452a159b41c5373d5", "tree": "68554ba6fd119deb49f0c4415e4bb9039e1fb875", "parents": [ "da6ca203a677f51764cc9dfa1bbc9660a5c199e3" ], "author": { "name": "Andreas Jonson", "email": "andjo403@hotmail.com", "time": "Sat Apr 18 15:33:26 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 15:33:26 2026" }, "message": "[IR] Add icmp like matcher (NFC) (#192746)\n\nmatches icmp and trunc nuw x to i1 (icmp ne x,0)" }, { "commit": "da6ca203a677f51764cc9dfa1bbc9660a5c199e3", "tree": "7d674f347ad6b1aa594558cd612a7a38d5f62ce4", "parents": [ "90b1416c3136244433d9f3917fe3029145687618" ], "author": { "name": "Ramkumar Ramachandra", "email": "artagnon@tenstorrent.com", "time": "Sat Apr 18 14:58:56 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 14:58:56 2026" }, "message": "[VPlan] Strip vp_post_order_{deep,shallow} (NFC) (#192787)\n\nPost 691a130 ([ADT] Refactor post order traversal, #191047),\nPostOrderTraversal\u0027s lifetime needs to exceed the lifetime of the\niterator. The vp_post_order_{deep,shallow} helpers now have the\npotential for being used incorrectly: hence, strip them, and require the\nPostOrderTraversal to be constructed explictly, similar to RPOT." }, { "commit": "90b1416c3136244433d9f3917fe3029145687618", "tree": "8a559fbc9140e0af7d3801c75685052fbfdde084", "parents": [ "b144b9c483335d55c0678d177d0f58b5650e2802" ], "author": { "name": "flovent", "email": "flbven@protonmail.com", "time": "Sat Apr 18 14:24:08 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 14:24:08 2026" }, "message": "[clang-tidy] Fix crash in `cppcoreguidelines-pro-type-member-init` with alias template in constructor initializer (#192786)\n\nCloses #192510." }, { "commit": "b144b9c483335d55c0678d177d0f58b5650e2802", "tree": "1dc964c1422dcdfb65db6e53bb38338ec1974664", "parents": [ "0d69fb3b995151ebbd83b03e94f7a301c7482b30" ], "author": { "name": "Balázs Benics", "email": "benicsbalazs@gmail.com", "time": "Sat Apr 18 13:57:57 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 13:57:57 2026" }, "message": "[clang][ssaf] Rework NestedBuildNamespaces of TU and LU summaries (#191489)\n\nProblem:\n\nASTEntityMapping::getEntityName() creates EntityName with an empty\nNestedBuildNamespace, but the linker expected the TU namespace to\nalready be present in the EntityName. This caused internal linkage\nsymbols to conflict during linking: since their NestedBuildNamespace was\nempty, they all inherited the same LU namespace during resolution,\nmaking internal symbols from different TUs appear identical. For\nexample, two \"static inline\" functions with the same USR in separate TUs\nwould be incorrectly merged into a single LU entity instead of remaining\ndistinct.\n\nThis is now fixed and demonstrated by the\nInternalLinkageWithEmptyNamespaceAcrossTUs test, which creates two\ninternal linkage symbols across two TUs and verifies they remain\nseparate after linking.\n\n---\n\nTU summaries now omit the \"namespace\" field from entity names in JSON,\nsince the linker adds TU namespace qualification at link time. LU (and\nWPA) summaries retain the \"namespace\" field as before, since their\nentity names are fully resolved.\n\nSplit entityNameFromJSON/entityNameToJSON into TU and LU variants:\n- tuEntityNameFromJSON: reads only \"usr\" and \"suffix\"\n- tuEntityNameToJSON: writes only \"usr\" and \"suffix\"\n- luEntityNameFromJSON/luEntityNameToJSON: full behavior with required\n\"namespace\" field\n\nSimilarly split entityIdTableEntry and entityIdTable functions. Shared\nhelpers (entityNameCoreFromJSON, entityIdTableFromJSONImpl) to avoid\ncode duplication.\n\nUpdate all TU JSON test inputs to remove the \"namespace\" field, and\nupdate unit tests accordingly.\n\n---\n\nAlternatively, we could choose to populate the NestedBuildNamespace with\nthe TU namespace when creating an EntityName, thus meet the linker\nexpectation. However, this would bloat the representation, and also make\nit redundant.\n\nAssisted-By: claude" }, { "commit": "0d69fb3b995151ebbd83b03e94f7a301c7482b30", "tree": "15d30977d2dbea1153f0f424403ba23119eb7b03", "parents": [ "28c43a6aa115dfa754edc110064958c55d3668dc" ], "author": { "name": "David Green", "email": "david.green@arm.com", "time": "Sat Apr 18 13:40:15 2026" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Sat Apr 18 13:40:15 2026" }, "message": "[AArch64][GISel] Regbank G_BITCAST using src regbank. (#191522)\n\nThis takes the regbank of a G_BITCAST from the input srcreg that has\nalready been allocated, keeping the bitcast on the same bank. A copy\nthen might be added to the dst if needed.\n\nonlyUsesFP is also improved for G_BITCAST, allowing operations to guess\nthe correct type through a onlyUsesFP too.\n\nOriginally from #177158 by Ryan Cowan with modifications." } ], "next": "28c43a6aa115dfa754edc110064958c55d3668dc" }