graphblas: handling gcc compiler bug on power and s390

.github/workflows/cross-compile.yaml

@@ -51,9 +51,11 @@ jobs:
             ccache-max: 64M
           - arch: s390x
             target-triple: s390x-linux-gnu
-            flags: --target=s390x-linux-gnu
-            cc: clang
-            cxx: clang++
+            # flags: --target=s390x-linux-gnu
+            # cc: clang
+            # cxx: clang++
+            cc: s390x-linux-gnu-gcc
+            cxx: s390x-linux-gnu-g++
             ccache-max: 64M
           - arch: riscv64
             target-triple: riscv64-linux-gnu

ChangeLog

@@ -1,10 +1,9 @@
-June 1, 2025: version 7.11.0
+June 2, 2025: version 7.11.0
 
     * LAGraph v1.2: many new experimental algorithms, LAGraph_Random in src/
-    * GraphBLAS v10.1.0: risc-V vectorization, GRAPHBLAS_VANILLA (optional;
-        to disable GxB extensions), and cmake revisions to ensure gcc is not
-        used to compile GraphBLAS on the Power or s390 processors (as a
-        workaround around for a bug in the gcc compiler).
+    * GraphBLAS v10.1.1: risc-V vectorization, GRAPHBLAS_VANILLA (optional;
+        to disable GxB extensions), and workaround for gcc compiler bug on
+        Power and s390x processors.
     * nearly all packages: cmake updates, to move static targets into separate
         exports (needed for Fedora packaging)
     * Package versions in this release: (* denotes a new version)
@@ -18,7 +17,7 @@ June 1, 2025: version 7.11.0
         CSparse 4.3.2
         CXSparse 4.4.2              *
         Example 1.8.9               *
-        GraphBLAS 10.1.0            *
+        GraphBLAS 10.1.1            *
         KLU 2.3.6                   *
         LDL 3.3.3                   *
         LAGraph 1.2.0               *

Example/CMakeLists.txt

@@ -96,7 +96,7 @@ find_package ( CCOLAMD 3.3.5 REQUIRED )
 find_package ( CHOLMOD 5.3.4 REQUIRED )
 find_package ( COLAMD 3.3.5 REQUIRED )
 find_package ( CXSparse 4.4.2 REQUIRED )
-find_package ( GraphBLAS 10.1.0 )
+find_package ( GraphBLAS 10.1.1 )
 find_package ( KLU 2.3.6 REQUIRED )
 find_package ( KLU_CHOLMOD 2.3.6 REQUIRED )
 find_package ( LDL 3.3.3 REQUIRED )

Example/Demo/mydemo.out

@@ -1,25 +1,25 @@
 My demo
-Date from #include 'my.h':  May 5, 2025
-Date from compiled library: May 5, 2025
-version from #include 'my.h.': 1.8.8
-version from compiled library: 1.8.8
+Date from #include 'my.h':  June 1, 2025
+Date from compiled library: June 1, 2025
+version from #include 'my.h.': 1.8.9
+version from compiled library: 1.8.9
 
 ------------------------------------------------------------
-MY: v1.8.8 (May 5, 2025)
-MY: v1.8.8 (in library)
-MY version code: 1008008
-MY date: May 5, 2025
+MY: v1.8.9 (June 1, 2025)
+MY: v1.8.9 (in library)
+MY version code: 1008009
+MY date: June 1, 2025
 
 ------------------------------------------------------------
-SuiteSparse_config: v7.10.3 (May 5, 2025)
-SuiteSparse_config: v7.10.3 (in library)
-SuiteSparse_config version code: 7010003
+SuiteSparse_config: v7.11.0 (June 1, 2025)
+SuiteSparse_config: v7.11.0 (in library)
+SuiteSparse_config version code: 7011000
 
 ------------------------------------------------------------
-CXSparse: v4.4.1 (June 20, 2024)
-CXSparse: v4.4.1 (in library)
-CXSparse version code: 4004001
-CXSparse Version 4.4.1, June 20, 2024.  Copyright (c) Timothy A. Davis, 2006-2024
+CXSparse: v4.4.2 (June 1, 2025)
+CXSparse: v4.4.2 (in library)
+CXSparse version code: 4004002
+CXSparse Version 4.4.2, June 1, 2025.  Copyright (c) Timothy A. Davis, 2006-2024
 2-by-2, nzmax: 4 nnz: 4, 1-norm: 34
     col 0 : locations 0 to 1
       0 : 11
@@ -29,78 +29,78 @@ CXSparse Version 4.4.1, June 20, 2024.  Copyright (c) Timothy A. Davis, 2006-202
       1 : 22
 
 ------------------------------------------------------------
-AMD: v3.3.3 (June 20, 2024)
-AMD: v3.3.3 (in library)
-AMD version code: 3003003
+AMD: v3.3.4 (June 1, 2025)
+AMD: v3.3.4 (in library)
+AMD version code: 3003004
 P [0] = 0
 P [1] = 1
 
 ------------------------------------------------------------
-BTF: v2.3.2 (Mar 22, 2024)
-BTF: v2.3.2 (in library)
-BTF version code: 2003002
+BTF: v2.3.3 (June 1, 2025)
+BTF: v2.3.3 (in library)
+BTF version code: 2003003
 P [0] = 0
 P [1] = 1
 Q [0] = 0
 Q [1] = 1
 nblocks 1
 
 ------------------------------------------------------------
-CAMD: v3.3.4 (May 5, 2025)
-CAMD: v3.3.4 (in library)
-CAMD version code: 3003004
+CAMD: v3.3.5 (June 1, 2025)
+CAMD: v3.3.5 (in library)
+CAMD version code: 3003005
 P [0] = 0
 P [1] = 1
 
 ------------------------------------------------------------
-CCOLAMD: v3.3.4 (June 20, 2024)
-CCOLAMD: v3.3.4 (in library)
-CCOLAMD version code: 3003004
+CCOLAMD: v3.3.5 (June 1, 2025)
+CCOLAMD: v3.3.5 (in library)
+CCOLAMD version code: 3003005
 P [0] = 0
 P [1] = 1
 
 ------------------------------------------------------------
-COLAMD: v3.3.4 (June 20, 2024)
-COLAMD: v3.3.4 (in library)
-COLAMD version code: 3003004
+COLAMD: v3.3.5 (June 1, 2025)
+COLAMD: v3.3.5 (in library)
+COLAMD version code: 3003005
 P [0] = 0
 P [1] = 1
 
 ------------------------------------------------------------
-CHOLMOD: v5.3.3 (May 5, 2025)
-CHOLMOD: v5.3.3 (in library)
-CHOLMOD version code: 5003003
+CHOLMOD: v5.3.4 (June 1, 2025)
+CHOLMOD: v5.3.4 (in library)
+CHOLMOD version code: 5003004
 
 ------------------------------------------------------------
-GraphBLAS: v10.0.5 (May 5, 2025)
-GraphBLAS: v10.0.5 (in library)
-GraphBLAS version code: 10000005
+GraphBLAS: v10.1.1 (June 2, 2025)
+GraphBLAS: v10.1.1 (in library)
+GraphBLAS version code: 10001001
 
 ------------------------------------------------------------
-LAGraph: v1.1.5 (Feb 20, 2025)
-LAGraph: v1.1.5 (in library)
-LAGraph version code: 1001005
+LAGraph: v1.2.0 (June 1, 2025)
+LAGraph: v1.2.0 (in library)
+LAGraph version code: 1002000
 
 ------------------------------------------------------------
-KLU: v2.3.5 (Oct 10, 2024)
-KLU: v2.3.5 (in library)
-KLU version code: 2003005
+KLU: v2.3.6 (June 1, 2025)
+KLU: v2.3.6 (in library)
+KLU version code: 2003006
 x [0] = 36.4
 x [1] = -32.7
 error: 7.10543e-14
 
 ------------------------------------------------------------
-LDL: v3.3.2 (Mar 22, 2024)
-LDL: v3.3.2 (in library)
-LDL version code: 3003002
+LDL: v3.3.3 (June 1, 2025)
+LDL: v3.3.3 (in library)
+LDL version code: 3003003
 x2 [0] = 36.4
 x2 [1] = -32.7
 error: 0
 
 ------------------------------------------------------------
-RBio: v4.3.4 (Oct 10, 2024)
-RBio: v4.3.4 (in library)
-RBio version code: 4003004
+RBio: v4.3.5 (June 1, 2025)
+RBio: v4.3.5 (in library)
+RBio version code: 4003005
 njumbled 0, nzeros 0
 result 0
 mtype: iua
@@ -113,14 +113,14 @@ iua                        2             2             4             0
  11 21 12 22
 
 ------------------------------------------------------------
-SPEX: v3.2.3 (Feb 20, 2025)
-SPEX: v3.2.3 (in library)
-SPEX version code: 3002003
+SPEX: v3.2.4 (June 1, 2025)
+SPEX: v3.2.4 (in library)
+SPEX version code: 3002004
 
 ------------------------------------------------------------
-SuiteSparseQR: v4.3.4 (June 20, 2024)
-SuiteSparseQR: v4.3.4 (in library)
-SuiteSparseQR version code: 4003004
+SuiteSparseQR: v4.3.5 (June 1, 2025)
+SuiteSparseQR: v4.3.5 (in library)
+SuiteSparseQR version code: 4003005
 
 CHOLMOD dense:   X from QR:  2-by-1, 
   leading dimension 2, nzmax 2, real, double
@@ -131,10 +131,10 @@ CHOLMOD dense:   X from QR:  2-by-1,
 
 
 ------------------------------------------------------------
-UMFPACK: v6.3.5 (Sept 23, 2024)
-UMFPACK: v6.3.5 (in library)
-UMFPACK version code: 6003005
-UMFPACK V6.3.5 (Sept 23, 2024)
+UMFPACK: v6.3.6 (June 1, 2025)
+UMFPACK: v6.3.6 (in library)
+UMFPACK version code: 6003006
+UMFPACK V6.3.6 (June 1, 2025)
 UMFPACK:  Copyright (c) 2005-2024 by Timothy A. Davis.  All Rights Reserved.
 
 UMFPACK License: SPDX-License-Identifier: GPL-2.0+
@@ -155,9 +155,9 @@ UMFPACK License: SPDX-License-Identifier: GPL-2.0+
 
 
 Availability: http://www.suitesparse.com
-UMFPACK V6.3.5 (Sept 23, 2024): OK
+UMFPACK V6.3.6 (June 1, 2025): OK
 
-UMFPACK V6.3.5 (Sept 23, 2024), Info:
+UMFPACK V6.3.6 (June 1, 2025), Info:
     matrix entry defined as:          double
     Int (generic integer) defined as: int64_t
     BLAS library used: Intel MKL 64lp BLAS (32-bit integers).  size of BLAS integer: 4

Example/Include/my_internal.h

@@ -56,8 +56,8 @@
 #if ! defined (NO_GRAPHBLAS)
     #include "GraphBLAS.h"
     #if !defined ( GxB_SUITESPARSE_GRAPHBLAS ) || \
-        GxB_IMPLEMENTATION < GxB_VERSION (10,1,0)
-    #error "This library requires SuiteSparse:GraphBLAS 10.1.0 or later"
+        GxB_IMPLEMENTATION < GxB_VERSION (10,1,1)
+    #error "This library requires SuiteSparse:GraphBLAS 10.1.1 or later"
     #endif
 #endif
 

GraphBLAS/CUDA/GB_cuda_AxB_dot3.cpp

@@ -98,8 +98,9 @@ GrB_Info GB_cuda_AxB_dot3           // C<M> = A'*B using dot product method
     int device = -1;
 
     // FIXME: control the GPU to use via the descriptor
-    CUDA_OK (cudaSetDevice ( 0 )) ;
-    CUDA_OK (cudaGetDevice (&device)) ;
+//  CUDA_OK (cudaSetDevice ( 0 )) ;
+    CUDA_OK (cudaGetDevice (&device)) ;     // FIXME
+    printf ("dot3 using cuda device %d\n", device) ;
     int number_of_sms = GB_Global_gpu_sm_get (0) ;
 
     //--------------------------------------------------------------------------

GraphBLAS/Doc/ChangeLog

@@ -1,3 +1,9 @@
+June 2, 2025: version 10.1.1
+
+    * gcc can now be used the Power or s390:  workaround added;
+        see Source/mxm/factory/GB_AxB_saxpy3_symbolic_fine_template.c.
+	This resolves the issue found in v10.1.0.
+
 June 1, 2025: version 10.1.0
 
     * add support for RISC-V vectorization: contributed by Rodion Suvorov

GraphBLAS/Doc/GraphBLAS_version.tex

@@ -1,5 +1,5 @@
 % version of SuiteSparse:GraphBLAS
 \date{VERSION
-10.1.0,
-June 1, 2025}
+10.1.1,
+June 3, 2025}
 

GraphBLAS/Doc/UserGuide/GrB_installing.tex

@@ -199,18 +199,6 @@ \subsubsection{On the Intel-based Mac}
     \begin{verbatim}
     make CC=icc CXX=icpc \end{verbatim} }
 
-%----------------------------------------
-\subsubsection{On IBM Power or IBM System/390}
-%----------------------------------------
-
-You cannot use use \verb'gcc' to compile GraphBLAS on the Power or s390
-architectures.  At least gcc versions up to 14.2.0 have a bug in the atomic
-capture on the IBM Power8 (see the comments in the \verb'Source/mxm/factory'
-folder).  The bug also occurs on the s390 architecture.  The cmake build script
-will report a fatal error if gcc is attempted to be used on these systems.  Use
-\verb'clang', \verb'ibm-clang', or \verb'xlc' instead.  Those compilers are
-able to correctly compile GraphBLAS on those systems.
-
 %----------------------------------------
 \subsubsection{On Microsoft Windows}
 \label{sec:windows}

GraphBLAS/Doc/UserGuide/GrB_release.tex

@@ -5,6 +5,14 @@ \section{Release Notes}
 
 \begin{itemize}
 
+\item June 2, 2025: version 10.1.1
+
+    \begin{itemize}
+    \item gcc can now be used the Power or s390:  workaround added;
+        see \verb'Source/mxm/factory/GB_AxB_saxpy3_symbolic_fine_template.c.'
+	This resolves the issue found in v10.1.0.
+    \end{itemize}
+
 \item June 1, 2025: version 10.1.0
 
     \begin{itemize}

GraphBLAS/Include/GraphBLAS.h

@@ -1,4 +1,4 @@
-// SuiteSparse:GraphBLAS 10.1.0
+// SuiteSparse:GraphBLAS 10.1.1
 //------------------------------------------------------------------------------
 // GraphBLAS.h: definitions for the GraphBLAS package
 //------------------------------------------------------------------------------
@@ -286,10 +286,10 @@
 
 // The version of this implementation, and the GraphBLAS API version:
 #define GxB_IMPLEMENTATION_NAME "SuiteSparse:GraphBLAS"
-#define GxB_IMPLEMENTATION_DATE "June 1, 2025"
+#define GxB_IMPLEMENTATION_DATE "June 3, 2025"
 #define GxB_IMPLEMENTATION_MAJOR 10
 #define GxB_IMPLEMENTATION_MINOR 1
-#define GxB_IMPLEMENTATION_SUB   0
+#define GxB_IMPLEMENTATION_SUB   1
 #define GxB_SPEC_DATE "Dec 22, 2023"
 #define GxB_SPEC_MAJOR 2
 #define GxB_SPEC_MINOR 1

GraphBLAS/README.md

@@ -4,7 +4,7 @@ SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2025, All Rights Reserved.
 
 SPDX-License-Identifier: Apache-2.0
 
-VERSION 10.1.0, June 1, 2025
+VERSION 10.1.1, June 3, 2025
 
 SuiteSparse:GraphBLAS is a complete implementation of the GraphBLAS standard,
 which defines a set of sparse matrix operations on an extended algebra of