v6.1.2 (Dec 31, 2021)

Version 6.1.2, Dec 31, 2021

* performance: revised swap_rule in GrB_mxm, which decides whether
    to compute C=A*B or C=(B'*A')', and variants, resulting in up
    to 3x performance gain over v6.1.1 for GrB_mxm (observed;
    could be higher in other cases).

v6.1.1 (Dec 28, 2021)

Version 6.1.1, Dec 28, 2021

* minor revision to AVX2 and AVX512f selection
* cpu_features/Makefile: remove test of list_cpu_features

v6.1.0 (Dec 26, 2021)

Version 6.1.0, Dec 26, 2021

• added GxB_get options: compiler name and version
• added package: https://github.com/google/cpu_features,
  Nov 30, 2021 version
• performance: faster C+=A*B when C is full, A is bitmap/full, and B is
  sparse/hyper; added saxpy5 kernel. faster C+=A'*B (dot4 kernel).
• bug fix: deserialization of iso and empty matrices/vectors was broken

v6.0.2 (Nov 30, 2021)

bug fix: GrB_Matrix_export; numerical values not properly exported

v5.2.2 (Nov 30, 2021)

bug fix: GrB_Matrix_export; numerical values not properly exported

v6.0.1 (Nov 27, 2021)

Versions 6.0.1 and 5.2.1, Nov 27, 2021

* v6.x and v5.x: differ only in GrB_wait, GrB_Info, GrB_SCMP, and GxB_init
* (37) bug fix:  C+=A'*B when the accum operator is the same as the monoid
    and C is iso-full, and A or B are hypersparse (dot4 method).
* performance: GrB_select with user-defined GrB_IndexUnaryOp ~2x faster.
* performance: faster (MIN,MAX)_(FIRSTJ,SECONDI) semirings

v5.2.1 (Nov 27, 2021)

Versions 6.0.1 and 5.2.1, Nov 27, 2021

* v6.x and v5.x: differ only in GrB_wait, GrB_Info, GrB_SCMP, and GxB_init
* (37) bug fix:  C+=A'*B when the accum operator is the same as the monoid
    and C is iso-full, and A or B are hypersparse (dot4 method).
* performance: GrB_select with user-defined GrB_IndexUnaryOp ~2x faster.
* performance: faster (MIN,MAX)_(FIRSTJ,SECONDI) semirings

v6.0.0 (Nov 15, 2021)

Version 6.0.0, Nov 15, 2021

* v6.0.0: this release contains only a few changes that cause a
    break with backward compatibility.  It is otherwise identical to v5.2.0.
* Fully compliant with the v2.0 C API Specification: three changes from the
    v2.0 C API Spec are not backward compatible: (GrB_*wait, GrB_Info
    enum, and GrB_SCMP).  GxB_init has also changed.
    * GrB_wait (object, mode): was GrB_wait (&object).
    * GrB_Info: changed enum values
    * GrB_SCMP: removed
    * GxB_init (mode, malloc, calloc, realloc, free, is_thread_safe): the
        last parameter, is_thread_safe, is deleted.
        The malloc/calloc/realloc/free functions must be thread-safe.

v5.2.0 (Nov 15, 2021)

Version 5.2.0, Nov 15, 2021

* Added for the v2.0 C API Specification: only features that are
    backward compatible with SuiteSparse:GraphBLAS v5.x have been
    added to v5.2.0.
    * GrB_Scalar: replaces GxB_Scalar, GxB_Scalar_* functions renamed GrB
    * GrB_IndexUnaryOp: new, free, fprint, wait
    * GrB_select: selection via GrB_IndexUnaryOp
    * GrB_apply: with GrB_IndexUnaryOp
    * GrB_reduce: reduce matrix or vector to GrB_Scalar
    * GrB_assign, GrB_subassion: with GrB_Scalar input
    * GrB_*_extractElement_Scalar: get GrB_Scalar from a matrix or vector
    * GrB*build: when dup is NULL, duplicates result in an error.
    * GrB import/export: import/export from/to user-provided arrays
    * GrB_EMPTY_OBJECT, GrB_NOT_IMPLEMENTED: error codes added
    * GrB_*_setElement_Scalar: set an entry in a matrix or vector,
        from a GrB_Scalar
    * GrB_Matrix_diag: same as GxB_Matrix_diag (C, v, k, NULL)
    * GrB_*_serialize/deserialize: with compression
    * GrB_ONEB_T: binary operator, f(x,y)=1, the same as GxB_PAIR_T
* GxB*import* and GxB*export*: now historical; use GxB*pack/unpack*
* GxB_select: is now historical; use GrB_select instead.
* GxB_IGNORE_DUP: special operator for build methods only; if dup
    is this operator, then duplicates are ignored (not an error)
* GxB_IndexUnaryOp_new: create a named index-unary operator
* GxB_BinaryOp_new: create a named binary operator
* GxB_UnaryOp_new: create a named unary operator
* GxB_Type_new: to create a named type
* GxB_Type_name: to query the name of a type
* added GxB_BinaryOp_xtype_name, GxB_BinaryOp_ytype_name,
    GxB_BinaryOp_ztype_name, GxB_deserialize_type_name,
    GxB_Matrix_type_name, GxB_Scalar_type_name, GxB_UnaryOp_xtype_name,
    GxB_UnaryOp_ztype_name, GxB_IndexUnaryOp_xtype_name,
    GxB_IndexUnaryOp_ytype_name, GxB_IndexUnaryOp_ztype_name,
    GxB_Vector_type_name: to query the name of a type as a string
* GxB_BinaryOp_xtype, GxB_BinaryOp_ytype, GxB_BinaryOp_ztype,
    GxB_Matrix_type, GxB_Scalar_type, GxB_SelectOp_ttype,
    GxB_SelectOp_xtype, GxB_UnaryOp_xtype, GxB_UnaryOp_ztype,
    GxB_Vector_type: declared historical; use *type_name methods.
* GxB_*_serialize/deserialize: with compression; optional descriptor
* GxB_Matrix_sort, GxB_Vector_sort: sort a matrix or vector
* GxB_eWiseUnion: like GrB_eWiseAdd except for how entries in A\B
    and B/A are computed.
* added LZ4/LZ4HC: compression library, http://www.lz4.org (BSD 2),
    v1.9.3, Copyright (c) 2011-2016, Yann Collet, All Rights Reserved.
* mis and pagerank demos: removed; mis added to LAGraph/experimental
* disabled free memory pool if OpenMP not available
* (36) bug fix: ewise C=A+B when all matrices are full, GBCOMPACT
    not used, but GB_control.h disabled the operator or type.
    Caught by Roi Lipman, Redis.
* (34) bug fix: C<M>=Z not returning C as iso if Z iso and C initialy
    empty.  Caught by Erik Welch, Anaconda.
* performance improvements: C=A*B: sparse/hyper times bitmap/full,
    and visa versa, including C += A*B when C is full

v5.1.10 (Oct 27, 2021)

Bug fix: GB_selector was not updating ->plen properly.