cuda testing

Author: DrTimothyAldenDavis

CMakeLists.txt

@@ -39,10 +39,10 @@
 cmake_minimum_required ( VERSION 3.20 ) # LAGraph can be built stand-alone
 
 # version of LAGraph
-set ( LAGraph_DATE "Sept 8, 2025" )
+set ( LAGraph_DATE "FIXME, 2025" )
 set ( LAGraph_VERSION_MAJOR 1 CACHE STRING "" FORCE )
-set ( LAGraph_VERSION_MINOR 2 CACHE STRING "" FORCE )
-set ( LAGraph_VERSION_SUB   1 CACHE STRING "" FORCE )
+set ( LAGraph_VERSION_MINOR 3 CACHE STRING "" FORCE )
+set ( LAGraph_VERSION_SUB   0 CACHE STRING "" FORCE )
 
 message ( STATUS "Building LAGraph version: v"
     ${LAGraph_VERSION_MAJOR}.

experimental/test/test_MaxFlow.c

@@ -27,8 +27,11 @@ char msg[LAGRAPH_MSG_LEN];
 LAGraph_Graph G = NULL;
 GrB_Matrix A = NULL;
 #define LEN 512
+#ifdef GRAPHBLAS_HAS_CUDA
 #define NTESTS 4
-// #define NTESTS 7
+#else
+#define NTESTS 7
+#endif
 char filename[LEN + 1];
 
 typedef struct{
@@ -44,10 +47,13 @@ test_info tests[] = {
   {"matrix_random_flow.mtx", 0,9, 22, LAGraph_ADJACENCY_DIRECTED},
   {"rand.mtx", 0, 19, 37, LAGraph_ADJACENCY_DIRECTED},
   {"mcl.mtx", 0, 9, 0, LAGraph_ADJACENCY_DIRECTED},
-// FIXME: re-enable these matrices:
-//{"cycle_flow.mtx", 0, 89, 1, LAGraph_ADJACENCY_DIRECTED},
-//{"random_weighted_general2.mtx", 0, 299, 11098623877, LAGraph_ADJACENCY_UNDIRECTED},
-//{"random_weighted_general1.mtx", 0, 499, 6264009335, LAGraph_ADJACENCY_UNDIRECTED}
+#ifndef GRAPHBLAS_HAS_CUDA
+// FIXME: the CUDA cases are currently very slow for these matrices,
+// when the GPU is hacked to always be used regardless of problem size:
+  {"cycle_flow.mtx", 0, 89, 1, LAGraph_ADJACENCY_DIRECTED},
+  {"random_weighted_general2.mtx", 0, 299, 11098623877, LAGraph_ADJACENCY_UNDIRECTED},
+  {"random_weighted_general1.mtx", 0, 499, 6264009335, LAGraph_ADJACENCY_UNDIRECTED}
+#endif
 };
 
 //399 11098623877 alt sink and src for test 6