Started 1 mo 20 days ago
Took 4 hr 40 min on green-dragon-02

Failed Build #15312 (Dec 29, 2019 7:11:55 AM)


Git (git http://labmaster3.local/git/llvm-project.git)

  1. Refactor getUsedValuesDefinedAbove to expose a variant taking a callback (detail)
  2. Use "final" instead of marking method virtual in override (NFC) (detail)
  3. Refactor PassTiming to support nested pipelines. (detail)
  4. Add `parseGenericOperation()` to the OpAsmParser (detail)
  5. Addressing some late review comments on kernel inlining. (detail)
  6. Add support for coalescing adjacent nested pass pipelines. (detail)
  7. Add warpsize and laneid intrinsics to NVVM dialect. (detail)
  8. Overload LLVM::TerminatorOp::build() for empty operands list. (detail)
  9. Explicitly declare the OpPassManager move constructor to avoid undefined (detail)
  10. Convert per channel fake quant attributes to type (detail)
  11. [NFC] Rename ExpressedToUniformQuantizedType to ExpressedToQuantizedType (detail)
  12. Add quant.const_fake_quant_per_axis op (detail)
  13. Add pass generate per block in a function a GraphViz Dot graph with ops (detail)
  14. Avoid sign-compare warning (detail)
  15. Remove redundant qualification (detail)
  16. Remove unused variable (detail)
  17. Convert ConstFakeQuantPerAxis to qcast and dcast pair (detail)
  18. Remove the constraint that min / max should stride zero (detail)
  19. Add folding rule for spv.CompositeExtract (detail)
  20. Don't leak TargetMachine in ExecutionEngine::setupTargetTriple (detail)
  21. Fix a typo in comments. The Inequality and Equality explanations were (detail)
  22. Add logical groups to NVVM op definitions. (detail)
  23. Fix typos in SDBMTest.cpp (detail)
  24. Rename SDBMPositiveExpr to SDBMTermExpr (detail)
  25. [spirv] Add support for spv.loop (de)serialization (detail)
  26. NFC: Clean up constant fold tests (detail)
  27. Cmpf constant folding for nan and inf (detail)
  28. Log name of the generated illegal operation name in DialectConversion (detail)
  29. Fixing typo in documentation. (detail)
  30. NFC: Update comments about rank constraints (detail)
  31. Refactor pass pipeline command line parsing to support explicit pipeline (detail)
  32. Improve verifier error reporting on type mismatch (NFC) (detail)
  33. Forward diagnostics from untracked threads in ParallelDiagnosticHandler. (detail)
  34. Add tablegen class for memrefs with rank constraints (detail)
  35. NFC: Finish replacing FunctionPassBase/ModulePassBase with OpPassBase. (detail)
  36. Update SPIR-V symbols and use GLSL450 instead of VulkanKHR (detail)
  37. Add type constraints for shaped types with same rank and element count (detail)
  38. Publicly expose the functionality to parse a textual pass pipeline. (detail)
  39. add missing memref cast fold pattern for dim op (detail)
  40. Verify that ModuleOps only contain dialect specific attributes. (detail)
  41. Add pattern to canonicalize for loop bounds (detail)
  42. Update regarding op definition and rewrites (detail)
  43. NFC: Fix stray character in error message: 1 -> ' PiperOrigin-RevId: (detail)
  44. Clean up build trip count analysis method - avoid mutating IR (detail)
  45. update normalizeMemRef utility; handle missing failure check + add more (detail)
  46. NFC - Move explicit copy/dma generation utility out of pass and into (detail)
  47. Add convenience methods to create i8 and i16 attributes in Builder. (detail)
  48. NFC: Merge OpPass with OperationPass into just OperationPass. (detail)
  49. NFC: Pass PassInstrumentations by unique_ptr instead of raw pointer. (detail)
  50. Update the IRPrinter instrumentation to work on non function/module (detail)
  51. NFC: Update the expected outputs of pass-timing. (detail)
  52. NFC: Update the PassInstrumentation section. (detail)
  53. Fix typo in test/AffineOps/ops.mlir (detail)
  54. Update the pass registration section and add a sub-section on the (detail)
  55. Add mechanism to specify extended instruction sets in SPIR-V. (detail)
  56. Drop makePositionAttr and the like in favor of Builder::getI64ArrayAttr (detail)
  57. Error out when kernel function is not found while translating GPU calls. (detail)
  58. Unify how errors are emitted in LaunchFuncOp verification. (detail)
  59. Introduce SDBMDirect expression into the SDBM expression hierarchy (detail)
  60. Overhaul the SDBM expression kind hierarchy (detail)
  61. [spirv] Add support for BitEnumAttr (detail)
  62. Add support for multi-level value mapping to DialectConversion. (detail)
  63. [spirv] Add support for function calls. (detail)
  64. Autogenerate (de)serialization for Extended Instruction Sets (detail)
  65. Add missing CMake dependency from libAnalysis to the Vector dialect (detail)
  66. Add rewrite pattern to compose maps into affine load/stores (detail)
  67. Change MLIR translation functions signature (detail)
  68. Support file-to-file translation in mlir-translate (detail)
  69. Add a preprocess pass to remove sequences that are problematic with (detail)
  70. Register a -test-spirv-roundtrip hook to mlir-translate (detail)
  71. Add (de)serialization support for OpRuntimeArray. (detail)
  72. Add support to OpAsmParser for parsing unknown keywords. (detail)
  73. Simplify SDBM expressions more aggressively in operators and conversions (detail)
  74. SDBM: support sum expressions on the LHS of stripe expressions (detail)
  75. Unify error messages to start with lower-case. (detail)
  76. Support symbolic operands for memref replacement; fix memrefNormalize (detail)
  77. Fix nested dominance relationship between parent results and child (detail)
  78. NFC: Remove stray logging from ~Block(). PiperOrigin-RevId: 269941815 (detail)
  79. Outline GPU kernel function into a nested module. (detail)
  80. Add address space attribute to LLVMIR's GlobalOp. (detail)
  81. NFC: Fix return indentation in generated op definitions. (detail)
  82. Quantize attribute values by per axis quantization parameters (detail)
  83. Automated rollback of commit 5684a12434f923d03b6870f2aa16226bfb0b38b6 (detail)
  84. Allow specification of decorators on SPIR-V StructType members. (detail)
  85. Remove unused import and two import forms (detail)
  86. Add utility to extract strides from layout map in MemRefType. (detail)
  87. Fix public build (detail)
  88. Use SmallVectorImpl in getStrides (detail)
  89. NFC: Pass OpAsmParser by reference instead of by pointer. (detail)
  90. Avoid iterator invalidation when recursively computing pattern depth. (detail)
  91. NFC: Pass OperationState by reference instead of by pointer. (detail)
  92. NFC: Pass OpAsmPrinter by reference instead of by pointer. (detail)
  93. Make GlobalOp's value attribute optional. (detail)
  94. [ODS] Add support for FloatElementsAttr (detail)
  95. Add a note about the use of recursion in the developer guide (detail)
  96. Upgrade/fix/simplify store to load forwarding (detail)
  97. [spirv] Add OpControlBarrier and OpMemoryBarrier. (detail)
  98. minor spelling tweaks (detail)
  99. Add integer sign- and zero-extension and truncation to standard. (detail)
  100. Fix undefined reference to mlir::getElementTypeOrSelf(mlir::Type) (detail)
  101. update layout map examples in doc, drop stale ones (detail)
  102. Update TruncateIOp builder to pass OperationState by reference. (detail)
  103. Specalize f32->i8/u8 Quanitization with C++ native arithmetic to (detail)
  104. [NFC] Replace std::clamp with inline implementation. (detail)
  105. Fix a number of Clang-Tidy warnings. (detail)
  106. Outline GPU kernel function into a nested module. (detail)
  107. Add variants of interleave that take separator (detail)
  108. NFC: Fix warning for uninitialized field. PiperOrigin-RevId: 270704572 (detail)
  109. Update the Pass, Analysis, and PassManager sections of (detail)
  110. Refactor DiagnosticEngine to support multiple registered diagnostic (detail)
  111. Add interfaces for call-like/callable operations. (detail)
  112. Add initial callgraph support. (detail)
  113. Add convenience methods to set an OpBuilder insertion point after an (detail)
  114. Simplify the way spirv::StructTypes are parsed. (detail)
  115. Use constant's location for reporting errors in parsing of hex constant (detail)
  116. Handle OpMemberName instruction in SPIR-V deserializer. (detail)
  117. Let mlir-translate support -split-input-file (detail)
  118. Make spirv::RuntimeArrayType part of spirv::CompositeType. (detail)
  119. [spirv] NFC: clean up (de)serialization tests (detail)
  120. Allow null Attribute for value when building GlobalOp. (detail)
  121. Clone called functions into nested GPU module. (detail)
  122. Normalize lowering of MemRef types (detail)
  123. Introduce splat op + provide its LLVM lowering (detail)
  124. Allow attaching descriptions to OpInterfaces and InterfaceMethods. (detail)
  125. NFC: Add a description to the ODS interface example. PiperOrigin-RevId: (detail)
  126. [spirv] Replace bitwiseCast with llvm::bit_cast (detail)
  127. Add support for GLSL Binary ops, and use it to implement GLSL FMax. (detail)
  128. NFC: Add 'const' dereference method to OperandTypeIterator. (detail)
  129. Miscellaneous fixes to SPIR-V Deserializer (details below). (detail)
  130. Forward declare LogicalResult as a struct to be consistent (detail)
  131. Fix memref-stride-calculation on Windows (detail)
  132. Add tablegen verification traits for comparing different properties (detail)
  133. Emit function name being tested in TestMemRefStrideCalculation (detail)
  134. Minor spelling tweaks (detail)
  135. [spirv] Add SPV_UnaryOp and spv.FNegate (detail)
  136. Change the return type of createPrintCFGGraphPass to match other passes. (detail)
  137. Add spv.Bitcast operation to SPIR-V dialect (detail)
  138. Remove unused variables and methods to address compiler warnings (detail)
  139. Add AllReduceOp to GPU dialect with lowering to NVVM. (detail)
  140. Drop support for memrefs from JitRunner (detail)
  141. Decouple tiling from fusion in Linalg. (detail)
  142. Make result ops generated output deterministic (detail)
  143. Disable failing tests (detail)
  144. [ROCm] Adding ROCDL Dialect. (detail)
  145. Fix missing links in the documentation (detail)
  146. Fix Documentation (detail)
  147. Fix JitRunner.cpp Error creation pattern and reactivate tests. (detail)
  148. Promote MemRefDescriptor to a pointer to struct when passing function (detail)
  149. NFC - clean up op accessor usage, std.load/store op verify, other stale (detail)
  150. Add TODO to revisit coupling of CallOp to MemRefType lowering (detail)
  151. Remove spurious debug spew in tests (detail)
  152. Append a newline when dumping a Value. (detail)
  153. Tablegen helpers for accessing properties of shaped types (detail)
  154. Switch explicit create methods to match generated build's order (detail)
  155. Fix syntax of 'call' and 'splat' ops (detail)
  156. Fix a typo in Toy Chapter 2 tutorial documentation (detail)
  157. update - remove outdated info (detail)
  158. Add InferTypeOpTrait & enable generating its member function definition (detail)
  159. Switch comments from GPU dialect terms to CUDA terms (NFC). (detail)
  160. Fix MemRefType::getStrides corner case (detail)
  161. Use MaybeAlign when setting alignment (detail)
  162. Add support for Logical Ops in SPIR-V dialect (detail)
  163. [DRR] Explain result type deduction in doc (detail)
  164. Normalize MemRefType lowering to LLVM as strided MemRef descriptor (detail)
  165. Adding some missing SPIR-V core and GLSL extended ops. (detail)
  166. NFC: Change `classof` on registered operations to use pointer (detail)
  167. Enable autogenerating OpInterface method declarations (detail)
  168. Add missing file from cmakelist (detail)
  169. Adding signed integer ops for abs, sign, min, and max in the GLSL (detail)
  170. [spirv] Add array length check. (detail)
  171. Format markdown list. (detail)
  172. Pass the pointer of the parent pipeline collection pass to (detail)
  173. Fold away reduction over 0 dimensions. (detail)
  174. Add integer shift ops to LLVM dialect. (detail)
  175. Add verification error message for ops that require at least one operand (detail)
  176. Change all_reduce lowering to support 2D and 3D blocks. (detail)
  177. [spirv] NFC: rename SPV_ArithmeticOp to SPV_ArithmeticBinaryOp (detail)
  178. Unify Linalg types by using strided memrefs (detail)
  179. Fix and simplify CallOp/CallIndirectOp to LLVM::CallOp conversion (detail)
  180. Add a pair of hooks to DominanceInfo. (detail)
  181. [spirv] Change enum case uniquing in (detail)
  182. Show type even if elementsattr is elided in graph (detail)
  183. [ROCm] Adding pass to lower GPU Dialect to ROCDL Dialect. (detail)
  184. Replace spurious `long` stride type by int64_t - NFC (detail)
  185. Fix example in OpInterfaces documentation (detail)
  186. [spirv] Add support for spv.selection (detail)
  187. Extract MemRefType::getStridesAndOffset as a free function and fix (detail)
  188. Disallow index types in memrefs. (detail)
  189. NFC: rename Conversion/ControlFlowToCFG to Conversion/LoopToStandard (detail)
  190. Add parentheses around boolean operators in assert (detail)
  191. Give modules a name (detail)
  192. Make Module::getName return Optional<StringRef> (detail)
  193. Add syntactic sugar for strided memref parsing. This CL implements the (detail)
  194. Generalize parse/printBinaryOp to parse/printOneResultOp. (detail)
  195. Add fpext and fptrunc to the Standard dialect and includes conversion to (detail)
  196. Add `axis` attribute to the quant.stats op (detail)
  197. Update the Inliner pass to work on SCCs of the CallGraph. (detail)
  198. Add support for inlining calls with different arg/result types from the (detail)
  199. Fix typos, NFC. (detail)
  200. Moving the GPUIndexIntrinsicOpLowering template to a common location (detail)
  201. Add missing Linalg lowerings to allow roundtrip.mlir to lower to LLVM (detail)
  202. Replace constexpr MemRefType::kDynamicStrideOrOffset by a (detail)
  203. Add some utility builder functions for SPIR-V operations. (detail)
  204. Add spv.Undef op to support OpUndef instruction in SPIR-V. (detail)
  205. Add missing dependency on the TypeInferOpInterface from the Test dialect (detail)
  206. [spirv] Allow return ops to be in control flow ops (detail)
  207. NFC: Cleanup test ops and traits tests (detail)
  208. Allow element type traits to operate on scalars (detail)
  209. Enable emitting dialect summary & description during op generation (detail)
  210. Support reduction of partial warps. (detail)
  211. Add OpaqueLoc to MLIR locations. (detail)
  212. Fix CMake build after adding TestOpaqueLoc.cpp (detail)
  213. Add DialectType and generate docs for dialect types (detail)
  214. Support AllocOp terminal in Linalg::AliasAnalysis. (detail)
  215. Start a minimal mlir_utils runtime library for testing debugging (detail)
  216. Fix a comment in the OperationInterface example. (detail)
  217. Change Block::getParent() to be a const function. This is only necessary (detail)
  218. fix simplify-affine-structures bug (detail)
  219. [spirv] Fix function entry block erase after moving to spv.selection (detail)
  220. Update UndefOp (de)serialization to generate OpUndef at module level. (detail)
  221. Add a new class, OpPrintingFlags, to enable programmatic control of (detail)
  222. [spirv] Disable a crashing spv.loop test (detail)
  223. Do not add spirv::BitcastOp for cast from signed to unsigned type. (detail)
  224. Expose `fuseProducerOf` in Linalg/Utils/Utils.h. (detail)
  225. Print result types when dumping graphviz. (detail)
  226. Add a flag to the AsmPrinter for eliding large ElementsAttrs. (detail)
  227. Update upgrade some uses of mlir::interleave API to take container (detail)
  228. Use named modules for gpu.launch_func (detail)
  229. Fuse GenerateCubinAccessors pass into LaunchFunctToCuda (detail)
  230. GPUToCUDA: emit addressof directly instead of wrapping it into a getter (detail)
  231. GPUToCUDA: attach CUBIN to the nested module rather than to the function (detail)
  232. Linalg to LLVM lowering: decrease the reliance on symbol lookup in a (detail)
  233. NFC: Remove unused default cl::opt value. (detail)
  234. Add support for walking the uses of a symbol. (detail)
  235. unroll and jam: fix order of jammed bodies (detail)
  236. Add a PatternRewriter hook for cloning a region into another. (detail)
  237. [spirv] Add a pass to decorate the composite types with layout info. (detail)
  238. [ROCm] Fix the return type for the device function calls from i32 to (detail)
  239. Add support for parsing/printing non bare-identifier SymbolRefs. (detail)
  240. Add Instance Specific Pass Options. (detail)
  241. Update the symbol utility methods to handle the case of unknown (detail)
  242. Allow dynamic but ranked types in ops with SameOperandsAndResultShape (detail)
  243. NFC: Fully qualify use of std::string. PiperOrigin-RevId: 273668957 (detail)
  244. Assert that region is not cloned into itself. (detail)
  245. Change to doxygen comments. NFC. (detail)
  246. Add exp operation to (detail)
  247. Update the usage and comments in (detail)
  248. Add support for some multi-store cases in affine fusion (detail)
  249. Fix Windows linkage error (detail)
  250. Make SPIR-V lowering infrastructure follow Vulkan SPIR-V validation. (detail)
  251. Guard rewriter insertion point during signature conversion. (detail)
  252. Add ::printAsTextualPipeline to Pass and OpPassManager. (detail)
  253. minor spelling tweaks (detail)
  254. Pre-allocate space for results from a regex match that uses 3 match (detail)
  255. Fix typo in QuantizedType method names (detail)
  256. Add test for fix to tablegen for custom folders for ops that return a (detail)
  257. Use llvm.func to define functions with wrapped LLVM IR function type (detail)
  258. NFC: Cleanup of type checking tests (detail)
  259. Mark GPU dialect as illegal when lowering to NVVM. (detail)
  260. Python bindings: export index_cast (detail)
  261. Remove the need to convert operations in regions of operations that have (detail)
  262. Add support for canonicalizing callable regions during inlining. (detail)
  263. Add trait for specified shapes matching (detail)
  264. Add lowering of constant ops to SPIR-V. (detail)
  265. Translation to LLVM: check the validity of module-level Ops (detail)
  266. Standard-to-LLVM conversion: check that operands have LLVM types (detail)
  267. NFC: Initialize pass manager option fields inline instead of the class (detail)
  268. Add support for generating reproducers on pass crash and failure. (detail)
  269. NFC: Print the generic op form after pass failure. (detail)
  270. LLVM conversion: harden a test to check for LLVM funcs rather than any (detail)
  271. Add unary ops and ExpOp to Standard Dialect. (detail)
  272. Rename LLVM::exp and LLVM::fmuladd to LLVM::ExpOP and LLVM::FMulAddOp. (detail)
  273. Drop obsolete code from std to llvm memref lowering (detail)
  274. LLVM Dialect: introduce llvm.mlir.null operation (detail)
  275. Emit LLVM IR equivalent of sizeof when lowering alloc operations (detail)
  276. Add LLVM IR dialect hooks for FP128 and X86_FP80 types (detail)
  277. Add lowering of VectorOps dialect to LLVM to the Linalg LLVM lowering (detail)
  278. Fix linalg.subview behavior in (partially) static cases. (detail)
  279. Use single quotes to wrap '@HOST_LDFLAGS@' in LIT config file (detail)
  280. Allowing replacing non-root operations in DialectConversion. (detail)
  281. NFC: Replace usages of Value::getKind with explicit isa/casts. (detail)
  282. Merge Ch3 of the Toy tutorial into chapter 2. (detail)
  283. AsmPrinter: avoid unused-variable warning (detail)
  284. Expose mlir::parseType to bindings (detail)
  285. Add conversion for splat of vectors of 2+D (detail)
  286. Consistent use of int in mlir_runner_utils.cpp (detail)
  287. Fix typos in LangRef and OpDefinitions (detail)
  288. Update Chapter 3 to demonstrate pattern match and rewrite optimizations (detail)
  289. Fix typos in InferTypeOpInterface. (detail)
  290. Fix minor typos (detail)
  291. [spirv] Add support for SpecId decoration on spv.specConstant (detail)
  292. Fix typo (detail)
  293. NFC: Split out ToyOpsIncGen into a separate CMakeLists.txt. (detail)
  294. Fix typos (detail)
  295. Fix typo (detail)
  296. Add support for OpBitwiseOr, OpBitwiseXor, and OpBitwiseAnd in SPIR-V (detail)
  297. NFC: Various code cleanups for Ch3. (detail)
  298. NFC: Remove NoSideEffect traits from all ops except for ConstantOp. (detail)
  299. Implement simple loop-invariant-code-motion based on dialect interfaces. (detail)
  300. Fix typo in (detail)
  301. Fix CMake configuration after introduction of LICM and LoopLikeInterface (detail)
  302. Add support for PatternRewriter::eraseOp. (detail)
  303. Allow for remapping argument to a Value in SignatureConversion. (detail)
  304. Support custom accumulator provided as region to gpu.all_reduce. (detail)
  305. Makes spv.module generated by GPU->SPIRV conversion spec compliant (detail)
  306. NFC: Update VectorOrTensor -> Shaped (detail)
  307. Update comments in ast.toy (detail)
  308. Update Chapter 4 of the Toy tutorial (detail)
  309. Code cleanups on Ch.4 (detail)
  310. Add ComplexType to TableGen with Tensor support (detail)
  311. Fix typo in tutorial. (detail)
  312. Add support for inlining toy call operations. (detail)
  313. Add Ch.5 of the toy tutorial. (detail)
  314. [spirv] Implement inliner interface (detail)
  315. Fix invalid transpose in example and add proper verification. (detail)
  316. Add LLVM_DEBUG in RewritersGen.cpp and Pattern.cpp (detail)
  317. Fix RewriterGen to support using NativeCodeCall as auxiliary pattern (detail)
  318. [DRR] Allow capturing and referencing no-result ops (detail)
  319. Use a SmallVector instead of an ArrayRef to materialize a temporary (detail)
  320. Rename LoopNestBuilder to AffineLoopNestBuilder - NFC (detail)
  321. Allow '_' when pretty printing dialect symbols (detail)
  322. [spirv] Add a canonicalization pattern for spv.selection. (detail)
  323. Decouple Linalg promotion from Linalg tiling - NFC (detail)
  324. Add Ch.6 of the Toy tutorial. (detail)
  325. NFC: Delete the Linalg tutorial. (detail)
  326. Add EDSC support for loop.for operations (detail)
  327. Introduce a wrapper around ConversionPattern that operates on the (detail)
  328. Automated rollback of commit b65c8bb5d6ab418bb3fcd0302aee19d3615d90f1 (detail)
  329. NFC: Remove trivial builder get methods. (detail)
  330. Add gpu.barrier op to synchronize invocations of a local workgroup. (detail)
  331. Add documentation on restrictions to dialect conversion rewrites (detail)
  332. Use StrEnumAttr for gpu.allreduce op instead of StringAttr to better (detail)
  333. Fix OSS target name GPUtoNVVMTransforms -> MLIRGPUtoNVVMTransforms (detail)
  334. Automated rollback of commit 575405f4d6762830c1c4520569de4e4ed3c8eed5 (detail)
  335. [DRR] Use eraseOp() to replace no-result ops (detail)
  336. Slightly rephrase a difficult-to-parse sentence. (detail)
  337. Implement lowering of VectorTypeCastOp to LLVM (detail)
  338. Lower vector transfer ops to loop.for operations. (detail)
  339. Add support for function result attributes. (detail)
  340. NFC: Rename SPIR-V serializer find*ID() to get*ID() to be consistent (detail)
  341. NFC: Add missing include for StringMap. PiperOrigin-RevId: 275588019 (detail)
  342. NFC: Cleanup the implementation of walkSymbolUses. (detail)
  343. Get active source lane predicate from shuffle instruction. (detail)
  344. Use new eraseOp instead of replaceOp with empty values (detail)
  345. Slight rewording in toy ch2 to make persistence of name clearer (detail)
  346. Add missing include to llvm Allocator.h (detail)
  347. Add missing include to StringMap in Verifier and DialectConversion. (detail)
  348. Add SourceMgrDiagnosticHandler to toy (detail)
  349. Fix minor spelling tweaks (NFC) (detail)
  350. Fix minor spelling tweaks (NFC) (detail)
  351. [DRR] Address GCC warning by wrapping for statement body with {} (detail)
  352. Update Ch1 to reflect new changes in the tutorial. (detail)
  353. NFC: Fix typo : Retur -> Return PiperOrigin-RevId: 275745931 (detail)
  354. Add a Symbol trait to simplify defining operations that represent (detail)
  355. NFC: Elide the value of a UnitAttr within nested attribute dictionaries. (detail)
  356. Unify GPU op definition names with other dialects. (detail)
  357. NFC: Fix remaining usages of MulOp as matrix multiplication. (detail)
  358. Cleanup and rewrite (detail)
  359. Use LLVM_Type instead of AnyType in the definition of LLVM_CallOp (detail)
  360. [spirv] Allow block arguments on spv.Branch(Conditional) (detail)
  361. [DRR] Allow interleaved operands and attributes (detail)
  362. NFC: Add support for parsing attributes programmatically via (detail)
  363. NFC: Remove a right parenthesis from comment. PiperOrigin-RevId: (detail)
  364. Expose optimizations flags in Python bindings (detail)
  365. Update loop.for verifier message (detail)
  366. Update chapter 3 code snippet to match the actual output of the code (detail)
  367. Drop MemRefUtils from the ExecutionEngine (detail)
  368. NFC: Remove references to the toy.generic attribute. (detail)
  369. Fix "set-but-unused" warning in DialectConversion (detail)
  370. Add @below and @above directives to verify-diagnostics. (detail)
  371. Wrap ODS to 80 lines and remove const qualifier for local `int` variable (detail)
  372. Add custom lowering of ExpOp for NVVM and ROCM. (detail)
  373. Add missing dependency on MLIRIR on MLIREDSCInterface (detail)
  374. Add support for replacing all uses of a symbol. (detail)
  375. Convert the Canonicalize and CSE passes to generic Operation Passes. (detail)
  376. [spirv] AccessChainOp canonicalization. (detail)
  377. Add support for parsing multiple result name groups. (detail)
  378. Define AnyRankedTensor Type in TableGen (detail)
  379. Fix include guards and add tests for OpToFuncCallLowering. (detail)
  380. Lookup function declaration in SymbolTable not ModuleOp. (detail)
  381. Print reason why dynamic library could not be loaded during execution. (detail)
  382. Add support for marking an operation as recursively legal. (detail)
  383. Standardize Linalg transformations to take an OpBuilder and an (detail)
  384. Parse locations in parseGenericOperation (detail)
  385. [spirv] Support OpPhi using block arguments (detail)
  386. Add a convenient operation build method for spirv::SelectOp (detail)
  387. [spirv] Use LLVM graph traversal utility for PrettyBlockOrderVisitor (detail)
  388. Bugfix: Keep worklistMap in sync with worklist in GreedyPatternRewriter (detail)
  389. Add "[TOC]" to generated documentation (detail)
  390. [spirv] Mark control flow ops as InFunctionScope (detail)
  391. [spirv] Fix and add spv.Unreachable (detail)
  392. Add basic support for declarative Linalg transformations (detail)
  393. Fix segfault when no symbol is given to an constraint operand (detail)
  394. Add a utility accessor 'has_single_element' for ranges. (detail)
  395. Add support to GreedyPatternRewriter for erasing unreachable blocks. (detail)
  396. Dump op location in createPrintOpGraphPass for easier debugging. (detail)
  397. Use `not` to invert return code in expected to fail tests (detail)
  398. [spirv] Add cast operations (detail)
  399. NFC: Simplify UseRange::iterator to just be a (detail)
  400. LinalgDependenceGraph: add const modifiers to accessors (detail)
  401. Add a test.symbol_scope operation that has the SymbolTable Traits to the (detail)
  402. Add a test for lowering GPU ops that cover cases where the symbol table (detail)
  403. NFC: Use #ifndef in various .td files instead of #ifdef and #else (detail)
  404. [mlir][llvm] Add missing cast ops (detail)
  405. Add Linalg pattern for producer-consumer fusion (detail)
  406. Support lowering of imperfectly nested loops into GPU dialect. (detail)
  407. Move BitEnumAttr from to (detail)
  408. Delete spurious file (detail)
  409. Add ReferToOp attribute constraint for SymbolRefAttr (detail)
  410. Add DialectAsmParser/Printer classes to simplify dialect attribute and (detail)
  411. Remove the need for passing a location to parseAttribute/parseType. (detail)
  412. Refactor QuantOps TypeParser to use the DialectAsmParser methods (detail)
  413. Refactor LinalgDialect::parseType to use the DialectAsmParser methods (detail)
  414. Drop spurious debug spew. (detail)
  415. Update the SPV dialect type parser to use the methods on (detail)
  416. [NVVM] Add mma.sync operation. (detail)
  417. Rename Region::RegionType to Region::BlockListType (detail)
  418. Emit empty lines after headers when generating op docs (detail)
  419. Add a PatternRewriter hook to merge blocks, and use it to support for (detail)
  420. NFC: Rename parseOptionalAttributeDict -> parseOptionalAttrDict to match (detail)
  421. [llvm] Add initial import of LLVM modules to mlir-translate (detail)
  422. [llvm] Allow GlobalOp to take a region for complex initializers (detail)
  423. Fix typos in the Standard Dialect documentation (detail)
  424. Add (parse|print)OptionalAttrDictWithKeyword hooks to simplify parsing (detail)
  425. NFC: Remove an extra space when printing the 'attributes' prefix before (detail)
  426. Add support for the LLVM FNeg instruction (detail)
  427. Add ViewOp to the StandardOps dialect, which casts a 1D/i8 element type (detail)
  428. [DRR] List some limitations clearly in the doc (detail)
  429. Add ViewOp verification for dynamic strides, and address some comments (detail)
  430. Adding an m_NonZero constant integer matcher. (detail)
  431. Add affine load/store/dma_start/dma_wait to dialect doc. (detail)
  432. Add lowering of std.view to LLVM (detail)
  433. Drop spurious test file (detail)
  434. Replace some remnant uses of "inst" with "op". (detail)
  435. NFC: Uniformize parser naming scheme in Toy tutorial to camelCase and (detail)
  436. NFC: Tidy up the implementation of operations in the Toy tutorial (detail)
  437. Add IndexedGenericOp to Linalg. (detail)
  438. Update Linalg to use std.view (detail)
  439. Add compatible query method to infer type interface (detail)
  440. Fix parameter name and document option in linalg::promoteSubViews (detail)
  441. Add canonicalizer for ViewOp which folds constants into the ViewOp (detail)
  442. Add Ch-7 of the toy tutorial detailing how to define new types. (detail)
  443. Swap operand order in std.view operation so that offset appears before (detail)
  444. Fix asm printer for affine expr (detail)
  445. mlir-translate: support -verify-diagnostics (detail)
  446. Move description from GenericOpBase to linalg.(indexed_)generic. (detail)
  447. [spirv] Add bit ops (detail)
  448. Explicitly initialize isRecursivelyLegal (detail)
  449. Look for SymbolRefAttr in KernelOutlining instead of hard-coding CallOp (detail)
  450. Add a short TableGen HowTo to tutorial chapter 2. (detail)
  451. Fix segfault (nullptr dereference) when passing a non-existent file to (detail)
  452. Also consider index constants when folding integer arithmetics with (detail)
  453. Adds std.subview operation which takes dynamic offsets, sizes and (detail)
  454. Add support for nested symbol references. (detail)
  455. Restructure comment lexing to not recurse. (detail)
  456. Add support for alignment attribute in std.alloc. (detail)
  457. Adds affine.min operation which returns the minimum value from a (detail)
  458. Add LLVM lowering of std.subview (detail)
  459. Update outdated comment for NativeCodeCall (detail)
  460. Update textmate syntax file (detail)
  461. Add a printer flag to use local scope when printing IR. (detail)
  462. [spirv] Properly return when finding error in serialization (detail)
  463. Add Conversion to lower loop::ForOp to spirv::LoopOp. (detail)
  464. Expose an isSubclassOf() method on AttrConstraint (detail)
  465. Add an option to print an operation if a diagnostic is emitted on it (detail)
  466. Add support for OpPhi in loop header block (detail)
  467. Make legality check in GPU->SPIR-V lowering of FuncOp kernel specific. (detail)
  468. NFC: Change DictionaryAttr::get(StringRef) to use binary search instead (detail)
  469. Add operations needed to support lowering of AffineExpr to SPIR-V. (detail)
  470. Make VariableOp instructions be in the first block in the function. (detail)
  471. Rename the current parseSymbolName to parseOptionalSymbolName (detail)
  472. Create and begin writing glossary. (detail)
  473. NFC: Refactor block signature conversion to not erase the original (detail)
  474. Add FuncOp::eraseArgument (detail)
  475. Fix glossary formatting. (detail)
  476. Deprecate linalg.subview in favor of std.subview (detail)
  477. Replace explicit concatenation by llvm::concat (detail)
  478. Do not emit aliases when printing local form (detail)
  479. Concentrate memref descriptor manipulation logic in one place (detail)
  480. [doc] Add debugging tips in ODS and DRR doc regarding mlir-tblgen (detail)
  481. Use MemRefDescriptor in Linalg-to-LLVM conversion (detail)
  482. Move VectorOps to Tablegen - (almost) NFC (detail)
  483. Adapt code to LLVM API updates. (detail)
  484. Use MemRefDescriptor in Vector-to-LLVM convresion (detail)
  485. Make positions of elements in MemRef descriptor private (detail)
  486. Move Affine to Standard conversion to lib/Conversion (detail)
  487. [ODS] Fix operation argument population to avoid crash (detail)
  488. Add CMakeLists.txt for AffineToStandard conversion (detail)
  489. Adds canonicalizer to SubViewOp which folds constants from base memref (detail)
  490. NFC: Refactor Dialect Conversion targeting SPIR-V. (detail)
  491. Refactor the LowerVectorTransfers pass to use the RewritePattern infra - (detail)
  492. Expand on operation definition to clarify the difference between (detail)
  493. Add more navigation to the MLIR toy tutorial. (detail)
  494. Mark std.subview as no-sideeffect. (detail)
  495. Mark std.view as no-sideeffect. (detail)
  496. Templatize linalg::LowerToLoops - NFC (detail)
  497. Use aggregate-parameter builder for ops having autogen type-deduction (detail)
  498. Fix build warnings (detail)
  499. Use simpler highlighting textmate syntax (detail)
  500. Modify tutorial and other documentation for consistency, clarity, and (detail)
  501. NFC: Convert CmpIPredicate in StandardOps to use EnumAttr (detail)
  502. Clarify that identity maps are discarded from the MemRef type (detail)
  503. [spirv] Add bit ops (detail)
  504. Fix attribute dict syntax in the docs (detail)
  505. Fix mismatched-tags warning (detail)
  506. Rename CLI flags -lower-gpu-ops-to-*-ops to -convert-gpu-to-* (detail)
  507. Implement folding of pattern dim(subview(_)[...][s1, ..., sn][...], i) (detail)
  508. Standardize all VectorOps class names to be prefixed by Vector - NFC (detail)
  509. Fix Affine Loop Fusion test case reported on github. This CL utilizies (detail)
  510. ConvertStandardToLLVM: replace assertion with graceful failure (detail)
  511. Also elide large array attribute in OpGraph Dump (detail)
  512. [spirv] Add a canonicalizer for BitcastOp. (detail)
  513. Fix regarding 1D/2D/3D Dim symbol name (detail)
  514. Add a parseAttribute<AttrType> overload for the non-type case. (detail)
  515. Fix SubViewOp stride calculation in constant folding. Adds unit tests (detail)
  516. Drop unnecessary dependences from mlir-translate (detail)
  517. Lower linalg.indexed_generic to loops. (detail)
  518. Use SmallVectorImpl instead of SmallVector for function parameters (NFC) (detail)
  519. Support SPIR-V constant op to take DenseElementsAttr as input. (detail)
  520. Fix the shape of the outcome in the example code. (detail)
  521. Add dialect-attribute-entry requirement to docs (detail)
  522. Change conversion CLI flag from -lower-to-llvm to -convert-std-to-llvm (detail)
  523. Add '*' and '?' and optional brace parse calls to the Parser (detail)
  524. Add getRemappedValue to ConversionPatternRewriter (detail)
  525. Add mention to avoid cl::opt for MLIR passes in the developer guide (detail)
  526. Add a note on commit messages to our developer guide (detail)
  527. Fix pretty printer corner case in mlir_runner_utils.cpp. (detail)
  528. Add VectorOps.StridedSliceOp (detail)
  529. Make type and rank explicit in mcuMemHostRegister function. (detail)
  530. Add some CMake rules for installing headers, mlir-tblgen, and mlir-opt (detail)
  531. Extend kernel outlining to also consider dim worth inlining. (detail)
  532. Fix 'the the' typo. (detail)
  533. Fix the comment to Region block iterators. (detail)
  534. Add StridedMemRef<>::operator[] - NFC (detail)
  535. Add a new OpAsmOpInterface to allow for ops to directly hook into the (detail)
  536. Implement unrolling of vector ops to finer-grained vector ops as a (detail)
  537. Changes to SubViewOp to make it more amenable to canonicalization. (detail)
  538. Add multi-level DCE pass. (detail)
  539. missing outer index %i in search_body (detail)
  540. Verify subview op result has dynamic shape, when sizes are specified. (detail)
  541. Add VectorContractionOp to the VectorOps dialect. (detail)
  542. Merge DCE and unreachable block elimination into a new utility (detail)
  543. Correctly parse empty affine maps. (detail)
  544. Add a document detailing operation traits, how to define them, and the (detail)
  545. NFC: Use Region::getBlocks to fix build failure with drop_begin. (detail)
  546. Fixed typo in 2-d tiled layout (detail)
  547. Fix OSS build (detail)
  548. Don't force newline before function attributes (detail)
  549. Add missing include after LLVM 049043b598ef5b12a5894c0c22db8608be70f517 (detail)
  550. Split Linalg declarative patterns from specific test patterns - NFC (detail)
  551. Drop unused function - NFC (detail)
  552. Fix OSS builds - NFC (detail)
  553. NFC: Add wrappers around DenseIntElementsAttr/DenseFPElementsAttr::get (detail)
  554. Change CUDA tests to use print_memref. (detail)
  555. Add support for using the ODS result names as the Asm result names for (detail)
  556. Move Linalg Transforms that are actually Conversions - NFC (detail)
  557. Add more detail about locations in Chapter 2 of tutorial. (detail)
  558. Unify vector op names with other dialects. (detail)
  559. Small formatting fix in Tutorial Ch2. (detail)
  560. Add more canonicalizations for SubViewOp. (detail)
  561. [spirv] Add a canonicalizer for `spirv::LogicalNotOp`. (detail)
  562. Fix Windows Build (detail)
  563. NFC: Remove unnecessarily guarded tablegen includes. (detail)
  564. Make isValidSymbol more powerful (detail)
  565. [spirv] NFC: rename test files and sort tests inside (detail)
  566. NFC: Wire up DRR settings for SPIR-V canonicalization patterns (detail)
  567. Support folding of StandardOps with DenseElementsAttr. (detail)
  568. Introduce gpu.func (detail)
  569. Add vector.insertelement op (detail)
  570. Allow memref_cast from static strides to dynamic strides. (detail)
  571. Introduce attributes that specify the final ABI for a spirv::ModuleOp. (detail)
  572. De-duplicate EnumAttr overrides by defining defaults (detail)
  573. Changing directory shortcut for CPU/GPU runner utils. (detail)
  574. Update VectorContractionOp to take iterator types and index mapping (detail)
  575. Add missing rule to generate SPIR-V ABI Attribute using tblgen to CMake. (detail)
  576. NFC: Actually expose the implementation of createGPUToSPIRVLoweringPass. (detail)
  577. Make std.divis and std.diviu support ElementsAttr folding. (detail)
  578. Allow LLVM::ExtractElementOp to have non-i32 indices. (detail)
  579. Add a vector.InsertStridedSliceOp (detail)
  580. Use vector.InsertStridedSlice in Vector -> Vector unrolling (detail)
  581. Add support for AttrSizedOperandSegments/AttrSizedResultSegments (detail)
  582. Some minor corrections and improvements to LangRef (detail)
  583. Relax restriction on affine_apply dim and symbol operands (detail)
  584. Automated rollback of commit d60133f89bb08341718bb3132b19bc891f7d4f4d (detail)
  585. Misc changes to lowering to SPIR-V. (detail)
  586. Add OpenMP dialect to the dialect registry (detail)
  587. [VectorOps] Add a BroadcastOp to the VectorOps dialect (detail)
  588. Add create method that takes equivalent of OperationState with (detail)
  589. [VectorOps] Refine BroadcastOp in VectorOps dialect (detail)
  590. Implement Linalg to loops lowering as a pattern (detail)
  591. Fixed typo in Toy tutorial (second var e -> var f) (detail)
  592. [spirv] Add folders for spv.IAdd and spv.IMul (detail)
  593. [spirv] NFC: Add getZero() and getOne() static method to ConstantOp (detail)
  594. NFC: A few cleanups for SPIRVLowering (detail)
  595. [Linalg] Change attribute n_loop_types to iterator (detail)
  596. Split out FunctionLike printing/parsing into (detail)
  597. [spirv] Check that operand of `spirv::CompositeExtractOp` is constant (detail)
  598. Fixed typo in Ch-1 of Toy tutorial (detail)
  599. Fix redundant convert and use NamedAttributeList as value (detail)
  600. mlir-tblgen: Dump input records when no generator is set (detail)
  601. Introduce Linkage attribute to the LLVM dialect (detail)
  602. Lower linalg.indexed_generic with libcall to LLVM. (detail)
  603. add missing '>' in Ch-2 (detail)
  604. NFC: Update std.subview op to use AttrSizedOperandSegments (detail)
  605. [DRR] Introduce `$_` to ignore op argument match (detail)
  606. Add missing `>` to the description of std.view. (detail)
  607. docs: minor spelling tweaks (detail)
  608. Generate dialect documentations in the doc folder for every dialect (detail)
  609. [ODS] Generate builders taking unwrapped value and defaults for (detail)
  610. [VectorOps] Add legality rules to broadcast (detail)
  611. NFC: use `&&` instead of `and` PiperOrigin-RevId: 283392575 (detail)
  612. [spirv] NFC: reorder sections in (detail)
  613. Add linkage support to LLVMFuncOp (detail)
  614. [Linalg] Update/fix documentation for linalg.indexed_generic. (detail)
  615. [spirv] Add spv.SubgroupBallotKHROp (detail)
  616. Extend conversion of SubViewOp to llvm to also support cases where size (detail)
  617. AffineLoopFusion: Prevent fusion of multi-out-edge producer loops (detail)
  618. Fix ViewOp to have at most one offset operand (detail)
  619. Add python bindings for ArrayAttr, AffineMapAttr. (detail)
  620. Add Python bindings for affine expressions with binary operators. (detail)
  621. Convert MemRefType to a linearized array in SPIR-V lowering. (detail)
  622. Allow analyses to provide a hook 'isInvalidated' to determine if they (detail)
  623. Verifier: Better error message in case of successor operand mismatch. (detail)
  624. Add CreateMaskOp to the VectorOps dialect. (detail)
  625. Make diagnostic a bit clearer. (detail)
  626. Add a pass to legalize operations before lowering to SPIR-V. (detail)
  627. [spirv] Add spv.GroupNonUniformBallot (detail)
  628. Refactor dependencies to expose Vector transformations as patterns - NFC (detail)
  629. Avoid variable name conflict in MLIR tutorial code snippet (detail)
  630. minor spelling tweaks (detail)
  631. Adds support for unrolling single-result vector operations with iterator (detail)
  632. Added new FAbs, FCeil, Cos, Neg, Sign, Tanh operations. (detail)
  633. Loop coalescing: fix pointer chainsing in use-chain traversal (detail)
  634. [spirv] Adding sqrt op in the GLSL extension. (detail)
  635. NFC - fix name / comments - isAccessInvariant (detail)
  636. Print out large elementsattr's such that they are parseable. (detail)
  637. [spirv] Define a few more extensions in (detail)
  638. NFC: Fix mismatches between and actual parser implementation. (detail)
  639. Drop MaterializeVectorTransfers in favor of simpler declarative (detail)
  640. [CSE] NFC: Hash the attribute dictionary pointer instead of the list of (detail)
  641. Add canonicalization patterns for vector CreateMaskOp and StridedSliceOp (detail)
  642. Add a CL option to Standard to LLVM lowering to use alloca instead of (detail)
  643. Add emitOptional(Error|Warning|Remark) functions to simplify emission (detail)
  644. Optimize operation ordering to support non-congruent indices. (detail)
  645. Add MLIRIR as a dependency to LLVM and related dialects (detail)
  646. Move ModuleManager functionality into mlir::SymbolTable. (detail)
  647. Fix MLIR Build after LLVM upstream JIT changes (getMainJITDylib removed) (detail)
  648. [spirv] Fix nested loop (de)serialization (detail)
  649. Add a flag to dump the current stack trace when emitting a diagnostic. (detail)
  650. Add spv.AtomicCompareExchangeWeak (detail)
  651. Allow specification of the workgroup size for GPUToSPIRV lowering. (detail)
  652. Add support for instance specific pass statistics. (detail)
  653. [spirv] Add CompositeInsertOp operation (detail)
  654. Add UnrankedMemRef Type (detail)
  655. Refactor the IRPrinting instrumentation to take a derivable config. (detail)
  656. [Linalg] Add permutation information to tiling (detail)
  657. Add include path to the TestDialect to fix broken build. (detail)
  658. NFC: Add documentation for `-mlir-print-op-on-diagnostic` and (detail)
  659. LLVM::AddressOfOp: properly take into account the address space (detail)
  660. minor spelling tweaks (detail)
  661. DimOp folding for alloc/view dynamic dimensions (detail)
  662. [spirv] Reorder `erase` and `emplace` to avoid "invalid iterator (detail)
  663. Unroll vector masks along with their associated vector arguments. (detail)
  664. Use regex to fix failure when stats are disabled. (detail)
  665. fix examples in comments (detail)
  666. Add conversions of GPU func with memory attributions to LLVM/NVVM (detail)
  667. Generate builder for ops that use InferTypeOpInterface trait in ODS (detail)
  668. [VectorOps] Add lowering of vector.broadcast to LLVM IR (detail)
  669. Provide a way to get the type of a ValueHandle. (detail)
  670. Move GPU::FuncOp definition to ODS - NFC (detail)
  671. LLVM::GlobalOp: take address space as builder argument (detail)
  672. [VecOps] Rename vector.[insert|extract]element to just (detail)
  673. Use named traits in the ODS definition of LLVMFuncOp (detail)
  674. Move GPU::LaunchOp to ODS. NFC. (detail)
  675. During serialization do a walk of ops in module to find spv.module. (detail)
  676. Replace custom getBody method with an ODS-generated in gpu::LaunchOp (detail)
  677. Change inferReturnTypes to return LogicalResult and values (detail)
  678. NFC: Separate implementation and definition in (detail)
  679. Fix langref code snippet - NFC (detail)
  680. NFC - update doc, comments, vim syntax file (detail)
  681. Add a flag to the IRPrinter instrumentation to only print after a pass (detail)
  682. Improve Linalg documentation following the Structured Ops presentation. (detail)
  683. Add a new ValueRange class. (detail)
  684. Update the builder API to take ValueRange instead of ArrayRef<Value *> (detail)
  685. NFC: Expose constFoldBinaryOp via a header (detail)
  686. Add a layer of recursive matchers that compose. (detail)
  687. Replace spurious SmallVector constructions with ValueRange (detail)
  688. Post-submit cleanups in RecursiveMatchers (detail)
  689. Add RegionRange for when need to abstract over different region (detail)
  690. [StructuredOps][Linalg] Add a primitive pattern to rewrite the (detail)
  691. Minor spelling tweaks (detail)
  692. Unify vector op unrolling transformation. (detail)
  693. Add lowering for module with gpu.kernel_module attribute. (detail)
  694. ODS: Generate named accessors for raw attributes (detail)
  695. Factor out commonly reusable names across structured ops dialects (detail)
  696. [spirv] Add spv.IAdd, spv.ISub, and spv.IMul folders (detail)
  697. [spirv] Add CompositeConstruct operation. (detail)
  698. Fix minor spelling tweaks. (detail)
  699. Add new indexed_accessor_range_base and indexed_accessor_range classes (detail)
  700. Refactor the Block support classes. (detail)
  701. [VectorOps] Fix off-by-one error in insert/extract validation (detail)
  702. [VectorOps] Add a ShuffleOp to the VectorOps dialect (detail)
  703. Fix build breakage on gcc-5 (detail)
  704. Drop Markdown style annotations (detail)
  705. Add a doc on guidelines for contributing a new dialect to the MLIR core (detail)
  706. More convenience build methods for SPIR-V ops. (detail)
  707. Add Py API for composing an affine expression with a map. Also allows (detail)
  708. Uniformize Vector transforms as patterns on the model of Linalg - NFC (detail)
  709. [Linalg] Add a Linalg iterator permutation transformation (detail)
  710. Refactor the various operand/result/type iterators to use (detail)
  711. Fold TestLinalgTilePermutePatterns into TestLinalgTransformPatterns - (detail)
  712. Move gpu.launch_func to ODS. NFC (detail)
  713. More affine expr simplifications for floordiv and mod (detail)
  714. Add VectorOp transform pattern which splits vector TransferReadOps to (detail)
  715. [VectorOps] Add lowering of vector.insert to LLVM IR (detail)
  716. Add initial liveness analysis including test cases. (detail)
  717. [Linalg] Add tiling for IndexedGenericOp with a region. (detail)
  718. Add a function to get lowering patterns from GPU to NVVM. (detail)
  719. Automated rollback of commit 98fbf41044d3364dbaf18db81b9e8d9520d14761 (detail)
  720. Roll-forward initial liveness analysis including test cases. (detail)
  721. NFC: Fix naming inconsistency: FuncOpLowering -> GPUFuncOpLowering. (detail)
  722. Continue refactoring StructuredOps utilities (detail)
  723. [spirv] Add lowering for std.fdiv, std.frem, std.fsub (detail)
  724. Expose a convenience function to add interface attributes to a function. (detail)
  725. Fix OSS build (detail)
  726. Add missing CMake dependency for MLIRTestIR. (detail)
  727. Add std.log* and llvm.intr.log* that correspond to the LLVMIR intrinsics (detail)
  728. Make OpBuilder::insert virtual instead of OpBuilder::createOperation. (detail)
  729. Switch from shfl.bfly to shfl.down. (detail)
  730. Automated rollback of commit f68ac464d818629e0fe10c23b44ac782d64a12d2 (detail)
  731. Move cpu runner utils templates to .h (detail)
  732. Added lowering of `std.tanh` to llvm function call to `tanh` and (detail)
  733. [Linalg] Add test for fusion of GenericOp with IndexedGenericOp. (detail)
  734. Retire !linalg.buffer type - NFC (detail)
  735. Add type inference variant for separate params builder generated (detail)
  736. [VectorOps] Add lowering of vector.shuffle to LLVM IR (detail)
  737. Fix logic on when to emit collective type but separate arg builder (detail)
  738. NFC: Cleanup the various Op::print methods. (detail)
  739. Fix maskAndClamp in gpu.all_reduce. (detail)
  740. Add a type range for the XLA HLO dialect. (detail)
  741. Try to fold operations in DialectConversion when trying to legalize. (detail)
  742. Add a layer of EDSC for linalg.GenericOp (detail)
  743. Skip generating C++ for "DeclareOpInterfaceMethods" in op interface gen. (detail)
  744. Refactor various canonicalization patterns as in-place folds. (detail)
  745. Apply a level of sugaring to the linalg.generic EDSC - NFC (detail)
  746. Reconcile struct and class for NestedPatternMatchers - NFC (detail)
  747. Add verifyCompatibleShape function overload with shapes (detail)
  748. Splat op doc - fix misformat / update tablegen op desc. comment (detail)
  749. Remove LLVM dependency on mlir::Module and instead check Traits. (detail)
  750. Make memref promotion during std->LLVM lowering the default calling (detail)
  751. Adds ExtractSlicesOp to the VectorOps dialect. (detail)
  752. [VectorOps] Add [insert/extract]element definition together with (detail)
  753. Remove unused variable (fix warning) NFC (detail)
  754. [Linalg] Expose subview promotion as a declarative pattern (detail)
  755. Make "LowerToCFG" an operation pass (detail)
  756. Insert signature-converted blocks into a region with a parent operation. (detail)
  757. Plug gpu.func into the GPU lowering pipelines (detail)
  758. Add InsertSlicesOp to the VectorOps dialect. (detail)
  759. Add edsc::ops for pointwise, conv and dilated_conv (detail)
  760. [spirv] Re-enable nested loop (de)serialization test (detail)
  761. Fix (de)serialization generation for SPV_ScopeAttr, (detail)
  762. Add atomic operations to SPIR-V dialect. (detail)
  763. Update vector op unrolling transformation to generate ExtractSlicesOp (detail)
  764. Replace code with equivalent satisfiesLLVMModule() function call. (detail)
  765. Add pattern rewrite which splits a vector TransferReadOp into slices (detail)
  766. Integrate from upstream at revision e4fce659a759. (detail)
  767. Add a new utility class TypeSwitch to ADT. (detail)
  768. fix a typo in OpDefinitions doc (detail)
  769. Add pattern rewrite to forward vector tuple elements to their users. (detail)
  770. Homogenize the description of the MemRef conversion to the LLVM dialect (detail)
  771. StdToLLVM conversion: drop getMemRefElementType utility function (detail)
  772. ConversionToLLVMDialect doc: update the syntax for LLVM types (detail)
  773. Make it possible to override the lowering of MemRef to the LLVM dialect. (detail)
  774. Add missing virtual inliner interface method in SPIR-V dialect. (detail)
  775. Add pattern rewrite which splits a vector TransferWriteOp into slices (detail)
  776. NFC: Use TypeSwitch to simplify existing code. PiperOrigin-RevId: (detail)
  777. NFC: Use this-> to appease GCC bug related to template lambda. (detail)
  778. Add support for float and string attributes to the C API and python (detail)
  779. Harden the requirements to memory attribution types in gpu.func (detail)
  780. Fix line break in LangRef (detail)
  781. Move function template definition to the header file. NFC (detail)
  782. NFC: Remove unnecessary 'llvm::' prefix from uses of llvm symbols (detail)
  783. Introduce prefetch op: affine -> std -> llvm intrinsic (detail)
  784. NFC: Cleanup non-conforming usages of namespaces. (detail)
  785. Update code block designations (detail)
  786. Add support for providing a default implementation for an interface (detail)
  787. NFC: Remove forbidden include of <iostream> (detail)
  788. [VectorOps] Add vector.print definition, with lowering support (detail)
  789. Added LLVM ops and lowering phases from standard dialect for FAbs, (detail)
  790. Move the specializations of VectorTransferRewriter::matchAndRewrite back (detail)
  791. Add a couple useful LLVM_DEBUG's to the inliner. (detail)
  792. [VectorOps] Replace iostream with stdio in support lib for vector.print (detail)
  793. Add runtime utils support for print_memref_i8 (detail)
  794. Detemplatize ModuleTranslation::lookupValues (detail)
  795. [VectorOps] minor cleanup: vector dialect "subscripts" are i32 (detail)
  796. LLVMFuncOp: implement addEntryBlock (detail)
  797. [VectorOps] Add vector ReshapeOp to the VectorOps dialect. (detail)
  798. Restructure and update Linalg ODS and documentation - NFC (detail)
  799. [VectorOps] Update vector transfer_read/write ops to operatate on (detail)
  800. Unique trait list during ODS Operator trait construction (detail)
  801. Allow dialect to create friendly names for region arguments (detail)
  802. Add gpu.shuffle op. (detail)
  803. fix isValidDim for block arg case (detail)
  804. [VectorOps] remove redundant returns from invalid ops test (detail)
  805. [VectorOps] unify vector dialect "subscripts" (detail)
  806. NFC: Move the classes related to Pass options into a new header file (detail)
  807. Make Type and Attribute classes trivially copyable (detail)
  808. Add integer bit-shift operations to the standard dialect. (detail)
  809. NFC: Introduce new ValuePtr/ValueRef typedefs to simplify the transition (detail)
  810. [InstCombine] add test for copysign; NFC (detail)
  811. Adjust License.txt file to use the LLVM license (detail)
  812. ReImplement the Value classes as value-typed objects wrapping an (detail)
  813. Automated rollback of commit f603a50109107b447b835dac11f0eb541288393e (detail)
  814. Resubmit: ReImplement the Value classes as value-typed objects wrapping (detail)
  815. Update (detail)
  816. Change the `notifyRootUpdated` API to be transaction based. (detail)
  817. [CFG] Fix an assertion failure with static initializers (detail)
  818. NFC: Replace ValuePtr with Value and remove it now that Value is (detail)
  819. Refactor the way that pass options are specified. (detail)
  820. Remove pybind11-based bindings (detail)
  821. Revert "[lldb] Don't process symlinks deep inside DWARFUnit" (detail)
  822. Adjust some MLIR paths and docs (detail)
  823. [DWARF] Check that all fields of a Unit Header are read. (detail)
  824. DebugInfo: Support dumping any exprloc as an expression (detail)
  825. NFC: Rename printOptionValue to printValue to fix MSVC build. (detail)
  826. test: correct flags for Windows (detail)
  827. [SelectionDAG] Don't repeatedly add a node to the worklist in (detail)
  828. [SelectionDAG] Change SelectionDAGISel::{funcInfo,SDB} to use unique_ptr (detail)
  829. [DebugInfo] Fix v4 macinfo for dwo files. (detail)
  830. Add `mlir` to -DLLVM_ALL_PROJECTS CMake option (detail)
  831. [docs] fix typo in Lexicon.rst (detail)
  832. Remove static MLIR doc ; they are already on the website (detail)
  833. [llvm-readobj] - Remove an excessive helper for printing dynamic tags. (detail)
  834. Add the Apache2 with LLVM exceptions license to MLIR (detail)
  835. DebugInfo: Add {} to address -Wdangling-else warning. (detail)
  836. DebugInfo: Correct the form of DW_AT_macro_info in .dwo files (detail)
  837. [TypePromotion] Make TypeSize a class member (detail)
  838. [clang] [ast] CXXRecordDecl::getVisibleConversionFunctions() could be (detail)
  839. Revert "[Support] Extend TimeProfiler to support multiple threads" (detail)
  840. clang-doc remove trailing whitespaces (detail)
  841. doc: Document that extra-arg/extra-arg-before can be used several times (detail)
  842. [lldb][NFC] Remove ClangExternalASTSourceCommon (detail)
  843. mlir Fix the syntax (detail)
  844. VariableName doc: fix the link to the mozilla doc (detail)
  845. [NFC] Remove some dead code from CGBuiltin.cpp. (detail)
  846. AMDGPU/GlobalISel: Lower llvm.amdgcn.else (detail)
  847. GlobalISel: Define equivalent node for G_INTRINSIC_TRUNC (detail)
  848. AMDGPU/GlobalISel: Legalize some 16-bit round instructions (detail)
  849. Fix false positive in magic number checker. (detail)
  850. Fix readability-const-return-type identifying the wrong `const` token (detail)
  851. [lldb][NFC] Move ClangASTContext::m_scratch_ast_source_up to the (detail)
  852. GlobalISel: Fix naming variables "brank" instead of "bank" (detail)
  853. GlobalISel: Define equivalent node for G_INTRINSIC_ROUND (detail)
  854. GlobalISel: Update syntax in debug printing (detail)
  855. [OPENMP50]Basic support for conditional lastprivate. (detail)
  856. [OPENMP][DOCS]Update status of OpenMP 5.0 features, NFC. (detail)
  857. [X86] Add STRICT versions of CVTTP2SI, CVTTP2UI, CMPM, and CMPP. (detail)
  858. test: ensure that we dead-strip in the linker (detail)
  859. [mlir] Fix -Wunneeded-internal-declaration (detail)
  860. [X86] Use 128-bit vector instructions for f32/f64->i64 conversions on (detail)
  861. [compiler-rt] Adapt stop-the-world for ptrace changes in NetBSD-9.99.30 (detail)
  862. [compiler-rt] Adapt for ptrace(2) changes in NetBSD-9.99.30 (detail)
  863. [lldb] Adapt for NetBSD-9.99.30 ptrace(2) API changes (detail)
  864. AMDGPU/GlobalISel: Fix mapping and selection of llvm.amdgcn.div.fixup (detail)
  865. [OPENMP]Add extra checks and initialization for clause modifier. (detail)
  866. Migrate function attribute "no-frame-pointer-elim" to (detail)
  867. Migrate function attribute "no-frame-pointer-elim-non-leaf" to (detail)
  868. Migrate function attribute "no-frame-pointer-elim"="false" to (detail)
  869. [Support] Fix behavior of StringRef::count with overlapping occurrences, (detail)
  870. [Thumb][test] Fix CodeGen/Thumb/PR17309.ll  after (detail)
  871. [WinEH] Delete addFnAttr("no-frame-pointer-elim") which seems no longer (detail)
  872. [MCJIT] Migrate function attribute "no-frame-pointer-elim" to (detail)
  873. [Attributor] Function level undefined behavior attribute (detail)
  874. [Attributor] UB Attribute now handles all instructions that access (detail)
  875. [compiler-rt] [netbsd] Define _RTLD_SOURCE to fix build (detail)
  876. [X86FixupSetCC] Use MachineInstr::readRegister/definesRegister to check (detail)
  877. [Attributor] Reach optimistic fixpoint in AAValueSimplify when the (detail)
  878. [MLIR][NFC] Insert const_cast to avoid warning (detail)
  879. Add missing strict_fp_to_int (detail)
  880. [llvm-readobj/llvm-readelf][test] - Add testing for EI_OSABI and (detail)
  881. [llvm-readobj] - Merge `gnu-symbols.test` to `symbols.test` and cleanup. (detail)
  882. [AMDGPU][MC][DOC] Updated AMD GPU assembler syntax description. (detail)
  883. [compiler-rt] [netbsd] Correct the fallback definition of PT_LWPNEXT (detail)
  884. [llvm-nm] Display STT_GNU_IFUNC as 'i' (detail)
  885. [X86] Merge together some common code in LowerFP_TO_INT now that we have (detail)
  886. [lldb][NFC] Use StringRef in (detail)
  887. [X86FixupSetCC] Remember the preceding eflags defining instruction while (detail)
  888. [X86] Use zero vector to extend to 512-bits for strict_fp_to_uint (detail)
  889. [OpenMP][IR-Builder] Introduce the finalization stack (detail)
  890. Make lazyload_metadata.ll resilient to the addition of new metadata (detail)
  891. [OpenMP][IR-Builder] Introduce "pragma omp parallel" code generation (detail)
  892. test commit (detail)
  893. Revert "test commit" (detail)
  894. [X86] Enable STRICT_SINT_TO_FP/STRICT_UINT_TO_FP on X86 backend (detail)
  895. Revert "Make lazyload_metadata.ll resilient to the addition of new (detail)
  896. [PowerPC] Modify the hasSideEffects of MTLR and MFLR from 1 to 0 (detail)
  897. [NFC][LoopFusion] Fix printing of the guard branch. Reviewer: kbarton, (detail)
  898. [NFC][PowerPC] Add a function tryAndWithMask to handle all the cases (detail)
  899. [Bitstream] Delete skipAbbreviatedField which duplicates (detail)
  900. [PowerPC] stop folding if result rlwinm mask is wrap while original (detail)
  901. Fix the MLIR Vim syntax file: the keyword group was missing (detail)
  902. [clang][test] Minor fixes in testcase absolute-paths-symlinks.c (detail)
  903. [lldb] Remove some calls to GetASTContext (detail)
  904. [compiler-rt] Disable QSORT interception on watchOS and tvOS. (detail)
  905. [lldb][NFC] Use ClangASTContext in AppleObjCRuntime interfaces (detail)
  906. [lldb][tests] Posix function strdup requires macro _POSIX_C_SOURCE (detail)
  907. [lldb][tests] Platform triple can be None (detail)
  908. [lldb][NFC] Simplify if-return (detail)
  909. [lldb][NFC] Add "lldb-vscode" to all_categories (detail)
  910. [lldb][NFC] Move lock scope where it should begin (detail)
  911. [lldb] Specify unsigned underlying type for an enumeration explicitly (detail)
  912. [lldb][test] Don't include "test_common.h" in the debug macros test (detail)
  913. [libcxx] Fix a typo in (detail)
  914. Partially revert "Add initial tests for (detail)
  915. [X86] Add avx512f and avx512dq+vl command lines to the vector strictfp (detail)
  916. [X86] Add test cases for v2i64->v2f32 (detail)
  917. [X86] Add custom lowering for v2i64->v2f32 (detail)
  918. [X86] Merge the SINT_TO_FP/UINT_TO_FP handlers in ReplaceNodeResults (detail)
  919. [OpenMP][NFCI] Use the libFrontend ProcBindKind in Clang (detail)
  920. [BPF] Enable relocation location for load/store/shifts (detail)
  921. [Attributor] Add helper to change an instruction to `unreachable` inst (detail)
  922. [ELF] Support input section description .gnu.version* in /DISCARD/ (detail)
  923. [DebugInfo][SelectionDAG] Change order while transferring SDDbgValue to (detail)
  924. Make lazyload_metadata.ll resilient to the addition of new metadata (detail)
  925. Allow newlines in AST Matchers in clang-query files (detail)
  926. Revert "Allow newlines in AST Matchers in clang-query files" (detail)
  927. Revert "[msan] Check qsort input." (detail)
  928. Allow newlines in AST Matchers in clang-query files (detail)
  929. [X86] Add custom widening for v2f64->v2i32 strict_fp_to_uint with (detail)
  930. [ELF] Improve the condition to create .interp (detail)
  931. [X86] Add custom widening for v2i32->v2f64 strict_uint_to_fp with (detail)
  932. Add a clang-tidy configuration file for MLIR, it is using camelBack for (detail)
  933. [ELF][test] Fix dynamic-linker.s (detail)
  934. Revert "Make lazyload_metadata.ll resilient to the addition of new (detail)
  935. [X86] Custom widen 128/256-bit vXi32 uint_to_fp on avx512f targets (detail)
  936. Make lazyload_metadata.ll resilient to the addition of new metadata (detail)
  937. Fix -Wunused-lambda-capture warnings. (detail)
  938. Remove SrcVT only used in an assert and propagate query. (detail)
  939. [compiler-rt] [netbsd] Add support for versioned statvfs interceptors (detail)
  940. add custom operation for strict fpextend/fpround (detail)
  941. Revert "Allow newlines in AST Matchers in clang-query files" + 1 (detail)
  942. Remove unused static function. (detail)
  943. Fix a -Wcovered-switch-default warning by moving the unreachable out of (detail)
  944. Remove an unused static function. (detail)
  945. Fix a -Wcovered-switch-default warning by moving the unreachable out of (detail)
  946. Temporarily Revert "[compiler-rt] [netbsd] Add support for versioned (detail)
  947. Add an __attribute__((unused)) to populateWithGenerated since it might (detail)
  948. [X86] Add test cases for v2i32->v2f32 (detail)
  949. [X86] Add custom legalization for strict_uint_to_fp v2i32->v2f32. (detail)
  950. [X86] Add tests for constrained float intrinsics on (detail)
  951. [X86][FPEnv] Promote some float strictfp operations to double on (detail)
  952. [X86] Fix -Wmisleading-indentation after D71892 (detail)
  953. [X86] Custom widen strict v2f32->v2i32 by padding with zeroes. (detail)
  954. [X86] Custom widen 128/256-bit vXi32 fp_to_uint on avx512f targets (detail)
  955. Delete llvm.{sig,}{setjmp,longjmp} remnant after r136821 (detail)
  956. [clangd] Fix crash in hover (detail)
  957. [llvm-readobj][llvm-objdump][test] - Improve dynamic section testing. (detail)
  958. [X86] Custom widen v2i32->v2f32 strict_sint_to_fp to avoid (detail)
  959. [X86] Add v2i32->v2f32 non-strict sint_to_fp/uint_to_fp tests. NFC (detail)
  960. [X86] Allow v2i32->v2f32 strict and non-strict uint_to_fp to be widened (detail)
  961. [clangd] Reformat `HoverTests.cpp` NFC (detail)
  962. [mlir] Floating constants for import-llvm (detail)
  963. [MLIR] [NFC] fix unused var warning (detail)
  964. [mlir] fix typo in a comment (detail)
  965. [lldb] Silent random xpass on aarch64-linux buildbot (detail)
  966. [OpenCL] Pretty print __private addr space (detail)
  967. AMDGPU: Use correct DebugLoc (detail)
  968. AMDGPU/GlobalISel: Select some 128-bit load/stores (detail)
  969. AMDGPU/GlobalISel: Fix extra result register in fdiv64 lowering (detail)
  970. [MLIR][spirv] Fix links in docs after repo migration (detail)
  971. [OpenMP] Implementation of OMPT reduction callbacks (detail)
  972. [NFC][DA] Remove duplicate code in checkSrcSubscript and (detail)
  973. Allow newlines in AST Matchers in clang-query files (detail)
  974. [Matrix] Propagate and use shape info for binary operators. (detail)
  975. [OpenCL] Fix inconsistency between opencl and c11 atomic fetch max/min (detail)
  976. [OpenCL] Fixed printing of __private in AMDGPU test (detail)
  977. [ConstantRange] Respect destination bitwidth for cast results. (detail)
  978. [compiler-rt] [netbsd] Add support for versioned statvfs interceptors (detail)
  979. [VFS] Don't run symlink test on Windows, it may pass or fail (detail)
  980. TailDuplication: Clear NoPHIs property (detail)
  981. Revert "[msan] Intercept qsort, qsort_r." (detail)
  982. Reland "[msan] Intercept qsort, qsort_r." (detail)
  983. [mlir] Convert std.and/std.or ops to spv.LogicalAnd/spv.LogicalOr (detail)
  984. Revert "CWG2352: Allow qualification conversions during reference (detail)
  985. [mlir][spirv] Update docs regarding how to define new ops and types (detail)
  986. [PowerPC][LoopVectorize]Add floating point reg usage test (detail)
  987. Support powerpc and sparc when building without init_array. (detail)
  988. Revert "[ELF] Improve the condition to create .interp" (detail)
  989. [mlir][spirv] Add basic definitions for supporting availability (detail)
  990. AMDGPU: Use Register (detail)
  991. TII: Fix using Register for a subregister index argument (detail)
  992. Hexagon: Fix missing tablegen mode comment (detail)
  993. AMDGPU/GlobalISel: Use SReg_32 for readfirstlane constraining (detail)
  994. [sanitizer] Link Sanitizer-x86_64-Test-Nolibc with -static (detail)
  995. [ELF] Improve the condition to create .interp (detail)
  996. AMDGPU: Adjust test so it will work with GlobalISel (detail)
  997. DebugInfo: Fix rangesBaseAddress DICompileUnit bitcode (detail)
  998. lld: Remove explicit copy ops from AssociatedIterator, relying on (detail)
  999. [Intrinsic] Delete tablegen rules of llvm.{sig,}{setjmp,longjmp} (detail)
  1000. Delete setjmp_undefined_for_msvc workaround after llvm.setjmp was (detail)
  1001. [ADT] Fix FoldingSet documentation typos (detail)
  1002. [mlir] Merge the successor operand count into BlockOperand. (detail)
  1003. [TargetLowering] Update comment to reference the correct compiler-rt (detail)
  1004. [PowerPC] Modify the hasSideEffects of some VSX instructions from 1 to 0 (detail)
  1005. [lldb][NFC] Remove GetASTContext call in ClangDeclVendor (detail)
  1006. Fix crash in getFullyQualifiedName for inline namespace (detail)
  1007. SimplifyDemandedBits - Remove duplicate getOperand() call. NFC. (detail)
  1008. [PowerPC] Change default for unaligned FP access for older subtargets (detail)
  1009. Fix bots after a9ad65a2b34f (detail)
  1010. [X86] Add test cases for v4i64->v4f32 and v8i64->v8f32 (detail)
  1011. [CMake] Fix lld detection after D69685 (detail)
  1012. [COFF] Make the autogenerated .weak.<name>.default symbols static (detail)
  1013. Allow redeclaration of __declspec(uuid) (detail)
  1014. [lldb][NFC] Remove GetASTContext call in ClangPersistentVariables (detail)
  1015. Revert "[COFF] Make the autogenerated .weak.<name>.default symbols (detail)
  1016. [mlir] Fix the wrong computation of dynamic strides for lowering AllocOp (detail)
  1017. [X86] Fix -enable-machine-outliner for x86-32 after D48683 (detail)
  1018. [X86] Remove a redundant (scalar_to_vector (extract_vector_elt X))) in (detail)
  1019. [LV] Use getMask() when printing recipe [NFCI] (detail)
  1020. [X86] Stop accidentally custom type legalizing v4i32->v4f32 on SSE1 only (detail)
  1021. [Attributor] AAUndefinedBehavior: Check for branches on undef value. (detail)
  1022. Fix handling of newlines in clang-query (detail)
  1023. Fix newline handling in clang-query parser (detail)
  1024. Fix use of named values surrounded by newlines in clang-query (detail)

Started by timer (132 times)

This run spent:

  • 5 days 11 hr waiting;
  • 4 hr 40 min build duration;
  • 5 days 15 hr total from scheduled to completion.
Revision: bee973f4df39f3eaf7fd91f4c3df6d0e64e9ee62
  • origin/master
Revision: 3cdabe9ad312fedaf0feb85046f9a8ef6e3189f9
  • refs/remotes/origin/master
Test Result (1 failure / ±0)

Identified problems

Ninja target failed

Below is a link to the first failed ninja target.
Indication 1

Regression test failed

This build failed because a regression test in the test suite FAILed. See the test report for details.
Indication 2