README and SpGEMM example update

pawelszczerbuk · Feb 1, 2022 · 1adfd9b · 1adfd9b
1 parent 73635a6
commit 1adfd9b
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 1 deletion.
diff --git a/cuSPARSE/README.md b/cuSPARSE/README.md
@@ -14,6 +14,7 @@ This folder demonstrates cuSPARSE Generic APIs usage.
 - [Conversion](#conversion)
 - [Legacy APIs](#legacy-apis)
 - [Optimizations](#optimizations)
+- [Sparse Iterative Methods](#sparse-iterative-methods)
 
 ---
 
@@ -137,3 +138,8 @@ This folder demonstrates cuSPARSE Generic APIs usage.
 
     The sample demonstrates how to optimize *sparse vector - dense vector scaling and sum* (`cusparseAxpby`) by exploiting NVIDIA Ampere architecture *Hardware Memory Compression*
 
+##### Sparse Iterative Methods
+
+* [Preconditioned BiCGStab](bicgstab/)
+
+    The sample describes how to use the cuSPARSE and cuBLAS libraries to implement the Incomplete-LU preconditioned iterative *Biconjugate Gradient Stabilized Method (BiCGStab)*
diff --git a/cuSPARSE/spgemm/spgemm_example.c b/cuSPARSE/spgemm/spgemm_example.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1993-2021 NVIDIA Corporation.  All rights reserved.
+ * Copyright 1993-2022 NVIDIA Corporation.  All rights reserved.
  *
  * NOTICE TO LICENSEE:
  *
@@ -195,6 +195,10 @@ int main(void) {
     // allocate matrix C
     CHECK_CUDA( cudaMalloc((void**) &dC_columns, C_nnz1 * sizeof(int))   )
     CHECK_CUDA( cudaMalloc((void**) &dC_values,  C_nnz1 * sizeof(float)) )
+
+    // NOTE: if 'beta' != 0, the values of C must be update after the allocation
+    //       of dC_values, and before the call of cusparseSpGEMM_copy
+
     // update matC with the new pointers
     CHECK_CUSPARSE(
         cusparseCsrSetPointers(matC, dC_csrOffsets, dC_columns, dC_values) )