//===- SparseTensorRuntime.cpp - SparseTensor runtime support lib ---------===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https://llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // // This file implements a light-weight runtime support library for // manipulating sparse tensors from MLIR. More specifically, it provides // C-API wrappers so that MLIR-generated code can call into the C++ runtime // support library. The functionality provided in this library is meant // to simplify benchmarking, testing, and debugging of MLIR code operating // on sparse tensors. However, the provided functionality is **not** // part of core MLIR itself. // // The following memory-resident sparse storage schemes are supported: // // (a) A coordinate scheme for temporarily storing and lexicographically // sorting a sparse tensor by coordinate (SparseTensorCOO). // // (b) A "one-size-fits-all" sparse tensor storage scheme defined by // per-dimension sparse/dense annnotations together with a dimension // ordering used by MLIR compiler-generated code (SparseTensorStorage). // // The following external formats are supported: // // (1) Matrix Market Exchange (MME): *.mtx // https://math.nist.gov/MatrixMarket/formats.html // // (2) Formidable Repository of Open Sparse Tensors and Tools (FROSTT): *.tns // http://frostt.io/tensors/file-formats.html // // Two public APIs are supported: // // (I) Methods operating on MLIR buffers (memrefs) to interact with sparse // tensors. These methods should be used exclusively by MLIR // compiler-generated code. // // (II) Methods that accept C-style data structures to interact with sparse // tensors. These methods can be used by any external runtime that wants // to interact with MLIR compiler-generated code. // // In both cases (I) and (II), the SparseTensorStorage format is externally // only visible as an opaque pointer. // //===----------------------------------------------------------------------===// #include "mlir/ExecutionEngine/SparseTensorRuntime.h" #ifdef MLIR_CRUNNERUTILS_DEFINE_FUNCTIONS #include "mlir/ExecutionEngine/SparseTensor/ArithmeticUtils.h" #include "mlir/ExecutionEngine/SparseTensor/COO.h" #include "mlir/ExecutionEngine/SparseTensor/File.h" #include "mlir/ExecutionEngine/SparseTensor/Storage.h" #include #include using namespace mlir::sparse_tensor; //===----------------------------------------------------------------------===// // // Utilities for manipulating `StridedMemRefType`. // //===----------------------------------------------------------------------===// namespace { #define ASSERT_NO_STRIDE(MEMREF) \ do { \ assert((MEMREF) && "Memref is nullptr"); \ assert(((MEMREF)->strides[0] == 1) && "Memref has non-trivial stride"); \ } while (false) #define MEMREF_GET_USIZE(MEMREF) \ detail::checkOverflowCast((MEMREF)->sizes[0]) #define ASSERT_USIZE_EQ(MEMREF, SZ) \ assert(detail::safelyEQ(MEMREF_GET_USIZE(MEMREF), (SZ)) && \ "Memref size mismatch") #define MEMREF_GET_PAYLOAD(MEMREF) ((MEMREF)->data + (MEMREF)->offset) /// Initializes the memref with the provided size and data pointer. This /// is designed for functions which want to "return" a memref that aliases /// into memory owned by some other object (e.g., `SparseTensorStorage`), /// without doing any actual copying. (The "return" is in scarequotes /// because the `_mlir_ciface_` calling convention migrates any returned /// memrefs into an out-parameter passed before all the other function /// parameters.) template static inline void aliasIntoMemref(DataSizeT size, T *data, StridedMemRefType &ref) { ref.basePtr = ref.data = data; ref.offset = 0; using MemrefSizeT = std::remove_reference_t; ref.sizes[0] = detail::checkOverflowCast(size); ref.strides[0] = 1; } } // anonymous namespace extern "C" { //===----------------------------------------------------------------------===// // // Public functions which operate on MLIR buffers (memrefs) to interact // with sparse tensors (which are only visible as opaque pointers externally). // //===----------------------------------------------------------------------===// #define CASE(p, c, v, P, C, V) \ if (posTp == (p) && crdTp == (c) && valTp == (v)) { \ switch (action) { \ case Action::kEmpty: { \ return SparseTensorStorage::newEmpty( \ dimRank, dimSizes, lvlRank, lvlSizes, lvlTypes, dim2lvl, lvl2dim); \ } \ case Action::kFromReader: { \ assert(ptr && "Received nullptr for SparseTensorReader object"); \ SparseTensorReader &reader = *static_cast(ptr); \ return static_cast(reader.readSparseTensor( \ lvlRank, lvlSizes, lvlTypes, dim2lvl, lvl2dim)); \ } \ case Action::kPack: { \ assert(ptr && "Received nullptr for SparseTensorStorage object"); \ intptr_t *buffers = static_cast(ptr); \ return SparseTensorStorage::newFromBuffers( \ dimRank, dimSizes, lvlRank, lvlSizes, lvlTypes, dim2lvl, lvl2dim, \ dimRank, buffers); \ } \ case Action::kSortCOOInPlace: { \ assert(ptr && "Received nullptr for SparseTensorStorage object"); \ auto &tensor = *static_cast *>(ptr); \ tensor.sortInPlace(); \ return ptr; \ } \ } \ fprintf(stderr, "unknown action %d\n", static_cast(action)); \ exit(1); \ } #define CASE_SECSAME(p, v, P, V) CASE(p, p, v, P, P, V) // Assume index_type is in fact uint64_t, so that _mlir_ciface_newSparseTensor // can safely rewrite kIndex to kU64. We make this assertion to guarantee // that this file cannot get out of sync with its header. static_assert(std::is_same::value, "Expected index_type == uint64_t"); // The Swiss-army-knife for sparse tensor creation. void *_mlir_ciface_newSparseTensor( // NOLINT StridedMemRefType *dimSizesRef, StridedMemRefType *lvlSizesRef, StridedMemRefType *lvlTypesRef, StridedMemRefType *dim2lvlRef, StridedMemRefType *lvl2dimRef, OverheadType posTp, OverheadType crdTp, PrimaryType valTp, Action action, void *ptr) { ASSERT_NO_STRIDE(dimSizesRef); ASSERT_NO_STRIDE(lvlSizesRef); ASSERT_NO_STRIDE(lvlTypesRef); ASSERT_NO_STRIDE(dim2lvlRef); ASSERT_NO_STRIDE(lvl2dimRef); const uint64_t dimRank = MEMREF_GET_USIZE(dimSizesRef); const uint64_t lvlRank = MEMREF_GET_USIZE(lvlSizesRef); ASSERT_USIZE_EQ(lvlTypesRef, lvlRank); ASSERT_USIZE_EQ(dim2lvlRef, lvlRank); ASSERT_USIZE_EQ(lvl2dimRef, dimRank); const index_type *dimSizes = MEMREF_GET_PAYLOAD(dimSizesRef); const index_type *lvlSizes = MEMREF_GET_PAYLOAD(lvlSizesRef); const LevelType *lvlTypes = MEMREF_GET_PAYLOAD(lvlTypesRef); const index_type *dim2lvl = MEMREF_GET_PAYLOAD(dim2lvlRef); const index_type *lvl2dim = MEMREF_GET_PAYLOAD(lvl2dimRef); // Rewrite kIndex to kU64, to avoid introducing a bunch of new cases. // This is safe because of the static_assert above. if (posTp == OverheadType::kIndex) posTp = OverheadType::kU64; if (crdTp == OverheadType::kIndex) crdTp = OverheadType::kU64; // Double matrices with all combinations of overhead storage. CASE(OverheadType::kU64, OverheadType::kU64, PrimaryType::kF64, uint64_t, uint64_t, double); CASE(OverheadType::kU64, OverheadType::kU32, PrimaryType::kF64, uint64_t, uint32_t, double); CASE(OverheadType::kU64, OverheadType::kU16, PrimaryType::kF64, uint64_t, uint16_t, double); CASE(OverheadType::kU64, OverheadType::kU8, PrimaryType::kF64, uint64_t, uint8_t, double); CASE(OverheadType::kU32, OverheadType::kU64, PrimaryType::kF64, uint32_t, uint64_t, double); CASE(OverheadType::kU32, OverheadType::kU32, PrimaryType::kF64, uint32_t, uint32_t, double); CASE(OverheadType::kU32, OverheadType::kU16, PrimaryType::kF64, uint32_t, uint16_t, double); CASE(OverheadType::kU32, OverheadType::kU8, PrimaryType::kF64, uint32_t, uint8_t, double); CASE(OverheadType::kU16, OverheadType::kU64, PrimaryType::kF64, uint16_t, uint64_t, double); CASE(OverheadType::kU16, OverheadType::kU32, PrimaryType::kF64, uint16_t, uint32_t, double); CASE(OverheadType::kU16, OverheadType::kU16, PrimaryType::kF64, uint16_t, uint16_t, double); CASE(OverheadType::kU16, OverheadType::kU8, PrimaryType::kF64, uint16_t, uint8_t, double); CASE(OverheadType::kU8, OverheadType::kU64, PrimaryType::kF64, uint8_t, uint64_t, double); CASE(OverheadType::kU8, OverheadType::kU32, PrimaryType::kF64, uint8_t, uint32_t, double); CASE(OverheadType::kU8, OverheadType::kU16, PrimaryType::kF64, uint8_t, uint16_t, double); CASE(OverheadType::kU8, OverheadType::kU8, PrimaryType::kF64, uint8_t, uint8_t, double); // Float matrices with all combinations of overhead storage. CASE(OverheadType::kU64, OverheadType::kU64, PrimaryType::kF32, uint64_t, uint64_t, float); CASE(OverheadType::kU64, OverheadType::kU32, PrimaryType::kF32, uint64_t, uint32_t, float); CASE(OverheadType::kU64, OverheadType::kU16, PrimaryType::kF32, uint64_t, uint16_t, float); CASE(OverheadType::kU64, OverheadType::kU8, PrimaryType::kF32, uint64_t, uint8_t, float); CASE(OverheadType::kU32, OverheadType::kU64, PrimaryType::kF32, uint32_t, uint64_t, float); CASE(OverheadType::kU32, OverheadType::kU32, PrimaryType::kF32, uint32_t, uint32_t, float); CASE(OverheadType::kU32, OverheadType::kU16, PrimaryType::kF32, uint32_t, uint16_t, float); CASE(OverheadType::kU32, OverheadType::kU8, PrimaryType::kF32, uint32_t, uint8_t, float); CASE(OverheadType::kU16, OverheadType::kU64, PrimaryType::kF32, uint16_t, uint64_t, float); CASE(OverheadType::kU16, OverheadType::kU32, PrimaryType::kF32, uint16_t, uint32_t, float); CASE(OverheadType::kU16, OverheadType::kU16, PrimaryType::kF32, uint16_t, uint16_t, float); CASE(OverheadType::kU16, OverheadType::kU8, PrimaryType::kF32, uint16_t, uint8_t, float); CASE(OverheadType::kU8, OverheadType::kU64, PrimaryType::kF32, uint8_t, uint64_t, float); CASE(OverheadType::kU8, OverheadType::kU32, PrimaryType::kF32, uint8_t, uint32_t, float); CASE(OverheadType::kU8, OverheadType::kU16, PrimaryType::kF32, uint8_t, uint16_t, float); CASE(OverheadType::kU8, OverheadType::kU8, PrimaryType::kF32, uint8_t, uint8_t, float); // Two-byte floats with both overheads of the same type. CASE_SECSAME(OverheadType::kU64, PrimaryType::kF16, uint64_t, f16); CASE_SECSAME(OverheadType::kU64, PrimaryType::kBF16, uint64_t, bf16); CASE_SECSAME(OverheadType::kU32, PrimaryType::kF16, uint32_t, f16); CASE_SECSAME(OverheadType::kU32, PrimaryType::kBF16, uint32_t, bf16); CASE_SECSAME(OverheadType::kU16, PrimaryType::kF16, uint16_t, f16); CASE_SECSAME(OverheadType::kU16, PrimaryType::kBF16, uint16_t, bf16); CASE_SECSAME(OverheadType::kU8, PrimaryType::kF16, uint8_t, f16); CASE_SECSAME(OverheadType::kU8, PrimaryType::kBF16, uint8_t, bf16); // Integral matrices with both overheads of the same type. CASE_SECSAME(OverheadType::kU64, PrimaryType::kI64, uint64_t, int64_t); CASE_SECSAME(OverheadType::kU64, PrimaryType::kI32, uint64_t, int32_t); CASE_SECSAME(OverheadType::kU64, PrimaryType::kI16, uint64_t, int16_t); CASE_SECSAME(OverheadType::kU64, PrimaryType::kI8, uint64_t, int8_t); CASE_SECSAME(OverheadType::kU32, PrimaryType::kI64, uint32_t, int64_t); CASE_SECSAME(OverheadType::kU32, PrimaryType::kI32, uint32_t, int32_t); CASE_SECSAME(OverheadType::kU32, PrimaryType::kI16, uint32_t, int16_t); CASE_SECSAME(OverheadType::kU32, PrimaryType::kI8, uint32_t, int8_t); CASE_SECSAME(OverheadType::kU16, PrimaryType::kI64, uint16_t, int64_t); CASE_SECSAME(OverheadType::kU16, PrimaryType::kI32, uint16_t, int32_t); CASE_SECSAME(OverheadType::kU16, PrimaryType::kI16, uint16_t, int16_t); CASE_SECSAME(OverheadType::kU16, PrimaryType::kI8, uint16_t, int8_t); CASE_SECSAME(OverheadType::kU8, PrimaryType::kI64, uint8_t, int64_t); CASE_SECSAME(OverheadType::kU8, PrimaryType::kI32, uint8_t, int32_t); CASE_SECSAME(OverheadType::kU8, PrimaryType::kI16, uint8_t, int16_t); CASE_SECSAME(OverheadType::kU8, PrimaryType::kI8, uint8_t, int8_t); // Complex matrices with wide overhead. CASE_SECSAME(OverheadType::kU64, PrimaryType::kC64, uint64_t, complex64); CASE_SECSAME(OverheadType::kU64, PrimaryType::kC32, uint64_t, complex32); // Unsupported case (add above if needed). fprintf(stderr, "unsupported combination of types: \n", static_cast(posTp), static_cast(crdTp), static_cast(valTp)); exit(1); } #undef CASE #undef CASE_SECSAME #define IMPL_SPARSEVALUES(VNAME, V) \ void _mlir_ciface_sparseValues##VNAME(StridedMemRefType *ref, \ void *tensor) { \ assert(ref &&tensor); \ std::vector *v; \ static_cast(tensor)->getValues(&v); \ assert(v); \ aliasIntoMemref(v->size(), v->data(), *ref); \ } MLIR_SPARSETENSOR_FOREVERY_V(IMPL_SPARSEVALUES) #undef IMPL_SPARSEVALUES #define IMPL_GETOVERHEAD(NAME, TYPE, LIB) \ void _mlir_ciface_##NAME(StridedMemRefType *ref, void *tensor, \ index_type lvl) { \ assert(ref &&tensor); \ std::vector *v; \ static_cast(tensor)->LIB(&v, lvl); \ assert(v); \ aliasIntoMemref(v->size(), v->data(), *ref); \ } #define IMPL_SPARSEPOSITIONS(PNAME, P) \ IMPL_GETOVERHEAD(sparsePositions##PNAME, P, getPositions) MLIR_SPARSETENSOR_FOREVERY_O(IMPL_SPARSEPOSITIONS) #undef IMPL_SPARSEPOSITIONS #define IMPL_SPARSECOORDINATES(CNAME, C) \ IMPL_GETOVERHEAD(sparseCoordinates##CNAME, C, getCoordinates) MLIR_SPARSETENSOR_FOREVERY_O(IMPL_SPARSECOORDINATES) #undef IMPL_SPARSECOORDINATES #define IMPL_SPARSECOORDINATESBUFFER(CNAME, C) \ IMPL_GETOVERHEAD(sparseCoordinatesBuffer##CNAME, C, getCoordinatesBuffer) MLIR_SPARSETENSOR_FOREVERY_O(IMPL_SPARSECOORDINATESBUFFER) #undef IMPL_SPARSECOORDINATESBUFFER #undef IMPL_GETOVERHEAD #define IMPL_LEXINSERT(VNAME, V) \ void _mlir_ciface_lexInsert##VNAME( \ void *t, StridedMemRefType *lvlCoordsRef, \ StridedMemRefType *vref) { \ assert(t &&vref); \ auto &tensor = *static_cast(t); \ ASSERT_NO_STRIDE(lvlCoordsRef); \ index_type *lvlCoords = MEMREF_GET_PAYLOAD(lvlCoordsRef); \ assert(lvlCoords); \ V *value = MEMREF_GET_PAYLOAD(vref); \ tensor.lexInsert(lvlCoords, *value); \ } MLIR_SPARSETENSOR_FOREVERY_V(IMPL_LEXINSERT) #undef IMPL_LEXINSERT #define IMPL_EXPINSERT(VNAME, V) \ void _mlir_ciface_expInsert##VNAME( \ void *t, StridedMemRefType *lvlCoordsRef, \ StridedMemRefType *vref, StridedMemRefType *fref, \ StridedMemRefType *aref, index_type count) { \ assert(t); \ auto &tensor = *static_cast(t); \ ASSERT_NO_STRIDE(lvlCoordsRef); \ ASSERT_NO_STRIDE(vref); \ ASSERT_NO_STRIDE(fref); \ ASSERT_NO_STRIDE(aref); \ ASSERT_USIZE_EQ(vref, MEMREF_GET_USIZE(fref)); \ index_type *lvlCoords = MEMREF_GET_PAYLOAD(lvlCoordsRef); \ V *values = MEMREF_GET_PAYLOAD(vref); \ bool *filled = MEMREF_GET_PAYLOAD(fref); \ index_type *added = MEMREF_GET_PAYLOAD(aref); \ uint64_t expsz = vref->sizes[0]; \ tensor.expInsert(lvlCoords, values, filled, added, count, expsz); \ } MLIR_SPARSETENSOR_FOREVERY_V(IMPL_EXPINSERT) #undef IMPL_EXPINSERT void *_mlir_ciface_createCheckedSparseTensorReader( char *filename, StridedMemRefType *dimShapeRef, PrimaryType valTp) { ASSERT_NO_STRIDE(dimShapeRef); const uint64_t dimRank = MEMREF_GET_USIZE(dimShapeRef); const index_type *dimShape = MEMREF_GET_PAYLOAD(dimShapeRef); auto *reader = SparseTensorReader::create(filename, dimRank, dimShape, valTp); return static_cast(reader); } void _mlir_ciface_getSparseTensorReaderDimSizes( StridedMemRefType *out, void *p) { assert(out && p); SparseTensorReader &reader = *static_cast(p); auto *dimSizes = const_cast(reader.getDimSizes()); aliasIntoMemref(reader.getRank(), dimSizes, *out); } #define IMPL_GETNEXT(VNAME, V, CNAME, C) \ bool _mlir_ciface_getSparseTensorReaderReadToBuffers##CNAME##VNAME( \ void *p, StridedMemRefType *dim2lvlRef, \ StridedMemRefType *lvl2dimRef, \ StridedMemRefType *cref, StridedMemRefType *vref) { \ assert(p); \ auto &reader = *static_cast(p); \ ASSERT_NO_STRIDE(dim2lvlRef); \ ASSERT_NO_STRIDE(lvl2dimRef); \ ASSERT_NO_STRIDE(cref); \ ASSERT_NO_STRIDE(vref); \ const uint64_t dimRank = reader.getRank(); \ const uint64_t lvlRank = MEMREF_GET_USIZE(dim2lvlRef); \ const uint64_t cSize = MEMREF_GET_USIZE(cref); \ const uint64_t vSize = MEMREF_GET_USIZE(vref); \ ASSERT_USIZE_EQ(lvl2dimRef, dimRank); \ assert(cSize >= lvlRank * reader.getNSE()); \ assert(vSize >= reader.getNSE()); \ (void)dimRank; \ (void)cSize; \ (void)vSize; \ index_type *dim2lvl = MEMREF_GET_PAYLOAD(dim2lvlRef); \ index_type *lvl2dim = MEMREF_GET_PAYLOAD(lvl2dimRef); \ C *lvlCoordinates = MEMREF_GET_PAYLOAD(cref); \ V *values = MEMREF_GET_PAYLOAD(vref); \ return reader.readToBuffers(lvlRank, dim2lvl, lvl2dim, \ lvlCoordinates, values); \ } MLIR_SPARSETENSOR_FOREVERY_V_O(IMPL_GETNEXT) #undef IMPL_GETNEXT void _mlir_ciface_outSparseTensorWriterMetaData( void *p, index_type dimRank, index_type nse, StridedMemRefType *dimSizesRef) { assert(p); ASSERT_NO_STRIDE(dimSizesRef); assert(dimRank != 0); index_type *dimSizes = MEMREF_GET_PAYLOAD(dimSizesRef); std::ostream &file = *static_cast(p); file << dimRank << " " << nse << '\n'; for (index_type d = 0; d < dimRank - 1; d++) file << dimSizes[d] << " "; file << dimSizes[dimRank - 1] << '\n'; } #define IMPL_OUTNEXT(VNAME, V) \ void _mlir_ciface_outSparseTensorWriterNext##VNAME( \ void *p, index_type dimRank, \ StridedMemRefType *dimCoordsRef, \ StridedMemRefType *vref) { \ assert(p &&vref); \ ASSERT_NO_STRIDE(dimCoordsRef); \ const index_type *dimCoords = MEMREF_GET_PAYLOAD(dimCoordsRef); \ std::ostream &file = *static_cast(p); \ for (index_type d = 0; d < dimRank; d++) \ file << (dimCoords[d] + 1) << " "; \ V *value = MEMREF_GET_PAYLOAD(vref); \ file << *value << '\n'; \ } MLIR_SPARSETENSOR_FOREVERY_V(IMPL_OUTNEXT) #undef IMPL_OUTNEXT //===----------------------------------------------------------------------===// // // Public functions which accept only C-style data structures to interact // with sparse tensors (which are only visible as opaque pointers externally). // //===----------------------------------------------------------------------===// index_type sparseLvlSize(void *tensor, index_type l) { return static_cast(tensor)->getLvlSize(l); } index_type sparseDimSize(void *tensor, index_type d) { return static_cast(tensor)->getDimSize(d); } void endLexInsert(void *tensor) { return static_cast(tensor)->endLexInsert(); } void delSparseTensor(void *tensor) { delete static_cast(tensor); } char *getTensorFilename(index_type id) { constexpr size_t bufSize = 80; char var[bufSize]; snprintf(var, bufSize, "TENSOR%" PRIu64, id); char *env = getenv(var); if (!env) { fprintf(stderr, "Environment variable %s is not set\n", var); exit(1); } return env; } index_type getSparseTensorReaderNSE(void *p) { return static_cast(p)->getNSE(); } void delSparseTensorReader(void *p) { delete static_cast(p); } void *createSparseTensorWriter(char *filename) { std::ostream *file = (filename[0] == 0) ? &std::cout : new std::ofstream(filename); *file << "# extended FROSTT format\n"; return static_cast(file); } void delSparseTensorWriter(void *p) { std::ostream *file = static_cast(p); file->flush(); assert(file->good()); if (file != &std::cout) delete file; } } // extern "C" #undef MEMREF_GET_PAYLOAD #undef ASSERT_USIZE_EQ #undef MEMREF_GET_USIZE #undef ASSERT_NO_STRIDE #endif // MLIR_CRUNNERUTILS_DEFINE_FUNCTIONS