Accept symmetric sparse matrix in Matrix Market Exchange Format.

[bixia]

Removed a comment.

[aartbik]

That was fast! Thanks Bixia.

mlir/lib/ExecutionEngine/SparseTensorUtils.cpp

514

update comment general/symmetric

518

I found this very hard to read, because strcmp returns 0 on failure here. Do you see chance to make this more readable?

622

Add a note that for now we chose to deal with symmetric matrices by fully constructing them.
In the future we of course may want to make the symmetry implicit for storage reasons.

mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum.mlir

7

it is a bit strange to just change the input matrix of an existing test for this
(and also to change it here, but not in the "vectorized" line L21).

I assume you picked the values such that the outcome is the same?

If you don't want to add a whole new file just make sure you change both L7 and L21.
Also, I would pick a new matrix, and just fix the expected outcome in L88

[aartbik]

mlir/test/Integration/data/test_symmetric.mtx

14

it does not matter of course, but Matrix Market uses a lower triangular or upper triangular matrix consistently for symmetric matrices. This is a bit bad example in the sense that we have a_14 in the upper triangular and a_52 in the lower trianguar.

Please pick a consistent region for the values

[bixia]

mlir/lib/ExecutionEngine/SparseTensorUtils.cpp

518

Would introduce a lambda like this help?

not_equal = [](const char* src, const* expected) {return strcmp(toLower(src), expected);}
if (not_equal(header, "%%matrixmarket") || not_equal(object, "matrix") ||

not_euqal(format, "coordinate") || not_equal(field, "real") ||
(not_equal(symmetry, "general") && !(*is_symmetric))) {

[bixia]

Addressed review comments.

[bixia]

mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum.mlir

7

Fixed L7 and L21, used a new matrix.

mlir/test/Integration/data/test_symmetric.mtx

14

Fixed this and used a new matrix instead of modifying test.mtx.

[aartbik]

mlir/lib/ExecutionEngine/SparseTensorUtils.cpp

518

Let's leave as is.

[bixia]

Modified test/CMakeLists.txt to copy the new tensor data to the test directory.

mlir/lib/ExecutionEngine/SparseTensorUtils.cpp

 /// Helper to convert string to lower case.
 static char *toLower(char *token) {
   for (char *c = token; *c; c++)
     *c = tolower(*c);
   return token;
 }
 
 /// Read the MME header of a general sparse matrix of type real.
-static void readMMEHeader(FILE *file, char *name, uint64_t *idata) {
+static void readMMEHeader(FILE *file, char *name, uint64_t *idata,
+                          bool *is_symmetric) {
   char line[1025];
   char header[64];
   char object[64];
   char format[64];
   char field[64];
   char symmetry[64];
   // Read header line.
   if (fscanf(file, "%63s %63s %63s %63s %63s\n", header, object, format, field,
              symmetry) != 5) {
     fprintf(stderr, "Corrupt header in %s\n", name);
     exit(1);
   }
+  *is_symmetric = (strcmp(toLower(symmetry), "symmetric") == 0);
   // Make sure this is a general sparse matrix.
   if (strcmp(toLower(header), "%%matrixmarket") ||
       strcmp(toLower(object), "matrix") ||
       strcmp(toLower(format), "coordinate") || strcmp(toLower(field), "real") ||
-      strcmp(toLower(symmetry), "general")) {
+      (strcmp(toLower(symmetry), "general") && !(*is_symmetric))) {
     fprintf(stderr,
             "Cannot find a general sparse matrix with type real in %s\n", name);
     exit(1);
   }
   // Skip comments.
   while (1) {
     if (!fgets(line, 1025, file)) {
       fprintf(stderr, "Cannot find data in %s\n", name);
static SparseTensorCOO<V> *openSparseTensorCOO(char *filename, uint64_t rank,
   // Open the file.
   FILE *file = fopen(filename, "r");
   if (!file) {
     fprintf(stderr, "Cannot find %s\n", filename);
     exit(1);
   }
   // Perform some file format dependent set up.
   uint64_t idata[512];
+  bool is_symmetric = false;
   if (strstr(filename, ".mtx")) {
-    readMMEHeader(file, filename, idata);
+    readMMEHeader(file, filename, idata, &is_symmetric);
   } else if (strstr(filename, ".tns")) {
     readExtFROSTTHeader(file, filename, idata);
   } else {
     fprintf(stderr, "Unknown format %s\n", filename);
     exit(1);
   }
   // Prepare sparse tensor object with per-dimension sizes
   // and the number of nonzeros as initial capacity.
for (uint64_t k = 0; k < nnz; k++) {
     // The external formats always store the numerical values with the type
     // double, but we cast these values to the sparse tensor object type.
     double value;
     if (fscanf(file, "%lg\n", &value) != 1) {
       fprintf(stderr, "Cannot find next value in %s\n", filename);
       exit(1);
     }
     tensor->add(indices, value);
+    // We currently chose to deal with symmetric matrices by fully constructing
+    // them. In the future, we may want to make symmetry implicit for storage
+    // reasons.
+    if (is_symmetric && indices[0] != indices[1])
+      tensor->add({indices[1], indices[0]}, value);
   }
   // Close the file and return tensor.
   fclose(file);
   return tensor;
 }
 
 } // anonymous namespace
 

mlir/test/CMakeLists.txt

if (MLIR_INCLUDE_INTEGRATION_TESTS)
   option(MLIR_RUN_X86VECTOR_TESTS "Run X86Vector tests.")
   option(MLIR_RUN_CUDA_TENSOR_CORE_TESTS "Run CUDA Tensor core WMMA tests.")
   # Passed to lit.site.cfg.py.in to set up the path where to find the libraries.
   set(MLIR_INTEGRATION_TEST_DIR ${CMAKE_LIBRARY_OUTPUT_DIRECTORY})
 
   # Copy test data over.
   file(COPY ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/mttkrp_b.tns
             ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/test.mtx
+            ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/test_symmetric.mtx
             ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/test.tns
             ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/wide.mtx
             ${CMAKE_CURRENT_SOURCE_DIR}/Integration/data/zero.mtx
           DESTINATION ${MLIR_INTEGRATION_TEST_DIR}/data/)
 endif()
 
 llvm_canonicalize_cmake_booleans(
   LLVM_BUILD_EXAMPLES

mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum.mlir

 // RUN: mlir-opt %s \
 // RUN:   --sparsification --sparse-tensor-conversion \
 // RUN:   --convert-vector-to-scf --convert-scf-to-std \
 // RUN:   --func-bufferize --tensor-constant-bufferize --tensor-bufferize \
 // RUN:   --std-bufferize --finalizing-bufferize --lower-affine \
 // RUN:   --convert-vector-to-llvm --convert-memref-to-llvm --convert-std-to-llvm --reconcile-unrealized-casts | \
-// RUN: TENSOR0="%mlir_integration_test_dir/data/test.mtx" \
+// RUN: TENSOR0="%mlir_integration_test_dir/data/test_symmetric.mtx" \
 // RUN: mlir-cpu-runner \
 // RUN:  -e entry -entry-point-result=void  \
 // RUN:  -shared-libs=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
 // RUN: FileCheck %s
 //
 // Do the same run, but now with SIMDization as well. This should not change the outcome.
 //
 // RUN: mlir-opt %s \
 // RUN:   --sparsification="vectorization-strategy=2 vl=2" --sparse-tensor-conversion \
 // RUN:   --convert-vector-to-scf --convert-scf-to-std \
 // RUN:   --func-bufferize --tensor-constant-bufferize --tensor-bufferize \
 // RUN:   --std-bufferize --finalizing-bufferize --lower-affine \
 // RUN:   --convert-vector-to-llvm --convert-memref-to-llvm --convert-std-to-llvm --reconcile-unrealized-casts | \
-// RUN: TENSOR0="%mlir_integration_test_dir/data/test.mtx" \
+// RUN: TENSOR0="%mlir_integration_test_dir/data/test_symmetric.mtx" \
 // RUN: mlir-cpu-runner \
 // RUN:  -e entry -entry-point-result=void  \
 // RUN:  -shared-libs=%mlir_integration_test_dir/libmlir_c_runner_utils%shlibext | \
 // RUN: FileCheck %s
 
 !Filename = type !llvm.ptr<i8>
 
 #SparseMatrix = #sparse_tensor.encoding<{
func @entry() {
     %a = sparse_tensor.new %fileName : !Filename to tensor<?x?xf64, #SparseMatrix>
 
     // Call the kernel.
     %0 = call @kernel_sum_reduce(%a, %x)
       : (tensor<?x?xf64, #SparseMatrix>, tensor<f64>) -> tensor<f64>
 
     // Print the result for verification.
     //
-    // CHECK: 28.2
+    // CHECK: 30.2
     //
     %m = memref.buffer_cast %0 : memref<f64>
     %v = memref.load %m[] : memref<f64>
     vector.print %v : f64
 
     // Release the resources.
     memref.dealloc %xdata : memref<f64>
     sparse_tensor.release %a : tensor<?x?xf64, #SparseMatrix>
 
     return
   }
 }

mlir/test/Integration/data/test_symmetric.mtx

+%%MatrixMarket matrix coordinate real symmetric
+%
+% This is a test sparse matrix in Matrix Market Exchange Format.
+% see https://math.nist.gov/MatrixMarket
+%
+5 5 7
+1 1 5.0
+1 3 4.1
+2 2 3.0
+2 4 2.0
+3 3 1.0
+4 4 4.0
+5 5 5.0