Home
last modified time | relevance | path

Searched refs:sparse (Results 1 – 25 of 113) sorted by relevance

12345

/llvm-project-15.0.7/mlir/test/Integration/Dialect/SparseTensor/CPU/
H A Dsparse_scale.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
9 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=4" | \
26 // Integration test that lowers a kernel annotated as sparse to actual sparse
27 // code, initializes a matching sparse storage scheme from a dense tensor,
32 // A kernel that scales a sparse matrix A by a factor of 2.0.
46 // Main driver that converts a dense tensor into a sparse tensor
47 // and then calls the sparse scaling kernel with the sparse tensor
66 // Convert dense tensor to sparse tensor and call sparse kernel.
H A Dsparse_out_simple.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
10 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=4" | \
33 // Integration test that lowers a kernel annotated as sparse to
34 // actual sparse code, initializes a matching sparse storage scheme
39 // A kernel that multiplies a sparse matrix A with itself
41 // a sparse tensor as output, but although the values of the
42 // sparse tensor change, its nonzero structure remains the same.
58 // Main driver that reads matrix from file and calls the sparse kernel.
64 // Read the sparse matrix from file, construct sparse storage.
H A Dsparse_sum.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
10 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=2" | \
33 // Integration test that lowers a kernel annotated as sparse to
34 // actual sparse code, initializes a matching sparse storage scheme
56 // Main driver that reads matrix from file and calls the sparse kernel.
66 // Read the sparse matrix from file, construct sparse storage.
H A Dsparse_tanh.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
18 // Performs zero-preserving math to sparse vector.
30 // Dumps a sparse vector of type f64.
32 // Dump the values array to verify only sparse contents are stored.
49 // Setup sparse vector.
50 %v1 = arith.constant sparse<
57 // Call sparse vector kernel.
H A Dsparse_vector_ops.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
48 // Scales a sparse vector into a new sparse vector.
64 // Scales a sparse vector in place.
76 // Adds two sparse vectors into a new sparse vector.
92 // Multiplies two sparse vectors into a new sparse vector.
124 // Sum reduces dot product of two sparse vectors.
139 // Dumps a sparse vector.
159 // Setup sparse vectors.
160 %v1 = arith.constant sparse<
164 %v2 = arith.constant sparse<
[all …]
H A Dsparse_matrix_ops.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
38 // Scales a sparse matrix into a new sparse matrix.
56 // Scales a sparse matrix in place.
68 // Adds two sparse matrices element-wise into a new sparse matrix.
86 // Multiplies two sparse matrices element-wise into a new sparse matrix.
104 // Dump a sparse matrix.
119 // Setup sparse matrices.
120 %m1 = arith.constant sparse<
124 %m2 = arith.constant sparse<
129 // TODO: Use %sm1 when we support sparse tensor copies.
[all …]
H A Dsparse_matvec.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
11 // RUN: --sparse-compiler="vectorization-strategy=2 vl=16 enable-simd-index32" | \
37 // Integration test that lowers a kernel annotated as sparse to
38 // actual sparse code, initializes a matching sparse storage scheme
43 // A kernel that multiplies a sparse matrix A with a dense vector b
64 // Main driver that reads matrix from file and calls the sparse kernel.
73 // Read the sparse matrix from file, construct sparse storage.
H A Dsparse_flatten.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
10 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=4" | \
23 // since, even though it impacts the sparse storage scheme layout,
39 // Integration test that lowers a kernel annotated as sparse to
40 // actual sparse code, initializes a matching sparse storage scheme
63 // Main driver that reads tensor from file and calls the sparse kernel.
75 // Read the sparse tensor from file, construct sparse storage.
H A Dsparse_sum_c32.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
24 // Integration test that lowers a kernel annotated as sparse to
25 // actual sparse code, initializes a matching sparse storage scheme
47 // Main driver that reads matrix from file and calls the sparse kernel.
60 // Read the sparse matrix from file, construct sparse storage.
H A Dsparse_spmm.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
10 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=2" | \
34 // Integration test that lowers a kernel annotated as sparse to
35 // actual sparse code, initializes a matching sparse storage scheme
40 // A kernel that multiplies a sparse matrix A with a dense matrix B
60 // Main driver that reads matrix from file and calls the sparse kernel.
69 // Read the sparse matrix from file, construct sparse storage.
H A Ddense_output_f16.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
20 // Creates a dense vector using the minimum values from two input sparse vectors.
47 // Dump the values array to verify only sparse contents are stored.
61 // Setup sparse vectors.
62 %v1 = arith.constant sparse<
66 %v2 = arith.constant sparse<
73 // Call the sparse vector kernel.
H A Ddense_output_bf16.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
20 // Creates a dense vector using the minimum values from two input sparse vectors.
47 // Dump the values array to verify only sparse contents are stored.
61 // Setup sparse vectors.
62 %v1 = arith.constant sparse<
66 %v2 = arith.constant sparse<
73 // Call the sparse vector kernel.
H A Dsparse_constant_to_sparse_tensor.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
12 // Integration tests for conversions from sparse constants to sparse tensors.
22 %ti = arith.constant sparse<[[0, 0], [0, 7], [1, 2], [4, 2], [5, 3], [6, 4], [6, 6], [9, 7]],
25 // Convert the tensor in COO format to a sparse tensor with annotation #Tensor1.
H A Dsparse_sampled_matmul.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
11 // RUN: --sparse-compiler="vectorization-strategy=2 vl=4 enable-simd-index32" | \
39 // Integration test that lowers a kernel annotated as sparse to
40 // actual sparse code, initializes a matching sparse storage scheme
66 // Main driver that reads matrix from file and calls the sparse kernel.
97 // Read the sparse matrix from file, construct sparse storage.
H A Dsparse_conversion_sparse2sparse.mlir2 // RUN: mlir-opt %s --sparse-compiler="s2s-strategy=2" | \
55 // Convert dense tensor directly to various sparse tensors.
62 // Convert sparse tensor directly to another sparse format.
70 // Convert sparse tensor back to dense.
88 // Release sparse tensors.
H A Dsparse_quantized_matmul.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
8 // RUN: mlir-opt %s --sparse-compiler="vectorization-strategy=2 vl=2" | \
15 // An example of a quantized sparse matmul. With the zero offset for the
16 // sparse input, the sparse compiler generates very efficient code for the
H A Dsparse_unary.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
29 // Invert the structure of a sparse vector. Present values become missing.
159 // Dumps a sparse vector of type f64.
174 // Dumps a sparse vector of type i32.
189 // Dump a sparse matrix.
206 // Setup sparse vectors.
207 %v1 = arith.constant sparse<
213 // Setup sparse matrices.
214 %m1 = arith.constant sparse<
220 // Call sparse vector kernels.
[all …]
H A Dsparse_dot.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
26 // Setup two sparse vectors.
27 %d1 = arith.constant sparse<
30 %d2 = arith.constant sparse<
H A Dsparse_tensor_mul.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
23 // Multiplies two 3-d sparse tensors element-wise into a new sparse tensor.
48 // Setup sparse tensor A
60 // Setup sparse tensor B
76 // Call sparse tensor multiplication kernel.
H A Dsparse_transpose.mlir1 // RUN: mlir-opt %s --sparse-compiler | \
27 // Transposing a sparse row-wise matrix into another sparse row-wise
30 // matrix into a sparse column-wise matrix first.
51 // However, even better, the sparse compiler is able to insert such a
75 // Setup input sparse matrix from compressed constant.
/llvm-project-15.0.7/mlir/include/mlir/Dialect/SparseTensor/Transforms/
H A DPasses.td15 let summary = "Automatically generate sparse tensor code from sparse tensor types";
17 A pass that implements the core functionality of a **sparse compiler**.
19 sparse tensor types is converted into code in which the sparsity is
21 selected sparse storage schemes.
38 // Multiply a sparse matrix A with a dense vector b into a dense vector x.
79 def SparseTensorConversion : Pass<"sparse-tensor-conversion", "ModuleOp"> {
82 A pass that converts sparse tensor primitives to calls into a runtime
83 support library. All sparse tensor types are converted into opaque
84 pointers to the underlying sparse storage schemes.
89 on the selected sparse tensor storage schemes.
[all …]
/llvm-project-15.0.7/mlir/include/mlir/Dialect/SparseTensor/IR/
H A DSparseTensorAttrDefs.td28 of tensors. The encoding is eventually used by a **sparse compiler**
29 pass to generate sparse code fully automatically for all tensor
30 expressions that involve tensors with a sparse encoding. Compiler
31 passes that run before this sparse compiler pass need to be
49 // Data in sparse tensor encoding.
61 // Unlike dense storage, most sparse storage schemes do not provide
63 // dimensions that should be support by the sparse storage scheme
86 // Dimension level types that define sparse tensors:
88 // Compressed - dimension is sparse, only nonzeros are stored
103 : TensorOf<allowedTypes, [IsSparseTensorPred], "sparse tensor">;
[all …]
H A DSparseTensorOps.td40 typed sparse tensor with inital contents into a computation.
123 Returns the indices array of the sparse storage format at the
171 // may be refined over time as our sparse abstractions evolve.
189 may be refined over time as our sparse abstractions evolve.
230 may be refined over time as our sparse abstractions evolve.
261 may be refined over time as our sparse abstractions evolve.
298 may be refined over time as our sparse abstractions evolve.
344 - overlap (elements present in both sparse tensors)
345 - left (elements only present in the left sparse tensor)
456 - present (elements present in the sparse tensor)
[all …]
H A DSparseTensorBase.td19 operations, and passes that are required to make sparse tensor
21 The dialect forms a bridge between high-level operations on sparse
22 tensors types and lower-level operations on the actual sparse storage
25 means of a small sparse runtime support library.
28 implementation detail**, by letting a **sparse compiler** generate
29 sparse code automatically was pioneered for linear algebra by [Bik96]
30 in MT1 (see https://www.aartbik.com/sparse.php) and formalized
34 The MLIR implementation closely follows the "sparse iteration theory"
49 lattices drive actual sparse code generation, which consists of a
63 sparse systems of linear equations. In Sparse Matrices and Their
/llvm-project-15.0.7/lldb/test/API/functionalities/postmortem/FreeBSDKernel/tools/
H A DREADME.rst34 7. Use the ``copy-sparse.py`` tool to create a sparse version of the vmcore::
37 grep '^% RD' | python copy-sparse.py /path/to/core vmcore.sparse
39 8. Compress the sparse vmcore file using ``bzip2``::
41 bzip2 -9 vmcore.sparse

12345