Changelog.txt

OpenBLAS ChangeLog
====================================================================
Version 0.3.33
23-Apr-2026

general:
 - fixed an incorrect cast in the SBGEMM test case that could lead to spurious test failures
 - fixed an invalid memory access in the converted C version of the CBLAS tests
 - made the BIGNUMA setting automatic when the number of cores exceeds 256
 - Imported recent updates from Reference-LAPACK to realign with its upcoming 3.13.0 release:
   - Implement ?LARF1F and ?ORM2R (Reference-LAPACK PRs 1019,1020,1196,1257)
   - Change loop order in ?GETC2 to improve performance (Reference-LAPACK PR 1023)
   - Change WORK array dimension in ?GELQS/?GEQRS (Reference-LAPACK PR 1094)
   - Add NaN checks for input matrix A in ?GEEV (Reference-LAPACK PR 1136)
   - Fix support for jobu/v in LAPACKE_?GESVDQ_WORK (Reference-LAPACK PRs 1146,1221)
   - Fix display of version number in LAPACK testsuite (Reference-LAPACK PR 1149)
   - Fix DGGES test seed to avoid bad matrix cases (Reference-LAPACK PR 1187)
   - Fix truncation of large WORK array sizes in ZHE (Reference-LAPACK PR 1195)
   - Fix overwriting of LDSWORK parameter in ?TRSYL3 (Reference-LAPACK PR 1206)
   - Fix overwriting of error states in some EIG tests (Reference-LAPACK PR 1207) 
   - Remove unused parameter in DORBDB3/ZUNBDB3 (Reference-LAPACK PR 1209)
   - Re-enable testing of ?BB and ?GG driver functions (Reference-LAPACK PR 1211)
   - Fix workspace size calculation in ?TGSEN (Reference-LAPACK PR 774)
   - Fix typos in the EIG DMD tests and initialized the cutoff variable (PR 1212,1228)
   - Optimized looping in ?LACPY/?LASCL/?LANTR with fat matrix and UPLO=L (PR 1251)

arm64:
 - worked around a serious miscompilation of the DDOT kernel by GCC15, affecting
   most non-SVE targets, and SVE targets in the case of non-unit array stride)
 - fixed an accuracy issue in the GEMV kernel for Neoverse V1 and other SVE targets
 - fixed broken STRMM and SSYMM in DYNAMIC_ARCH builds when running on non-SME hardware
 - added an optimized SHGEMM kernel for Neoverse N2
 - fixed DYNAMIC_ARCH builds under Windows on Arm
 - Added autodetection of Cortex A75/A76 in DYNAMIC_ARCH builds
 - Added autodetection of Neoverse V3, currently supported through V2 kernels
 - Re-added support for the "VORTEX" target in DYNAMIC_ARCH builds with DYNAMIC_LIST
 - Fixed CMake-based builds that use the "Ninja" generator

loongarch64:
 - fixed a build failure due to missing support for the new half-precision float type
 - fixed a long-standing bug in asserting 64bit capability in the c_check helper script

x86_64:
 - added a workaround for miscompilation of the AVX512 GEMM kernels by LLVM on Windows
 - fixed a build failure in the LAED3 code when compiling with MinGW on Windows
 - fixed CMake-based compilation with the NVIDIA HPC compiler
 - Fixed CMake-based builds that use the "Ninja" generator

wasm:
 - added optimized kernels for STRSM and DTRSM

====================================================================
Version 0.3.32
23-Mar-2026

general:
 - Moved the preliminary support for a Web Assembly target to its own WASM
   architecture and WASM128_GENERIC target
 - Fixed a potential performance difference between dedicated compilation for
   a target and its representation in DYNAMIC_ARCH builds by making additional
   cpu-specific parameters available to the DYNAMIC_ARCH configuration
 - Fixed the reimplementation of LAPACK ?GESV to conform to the reference (i.e.
   compute the LU factorization even when NRHS is zero)
 - Improved the error message that is displayed when the compile-time allocation
   of memory buffers is exceeded
 - Fixed a problem with non-serialized accesses to parallelized SYRK by concurrent
   callers
 - Fixed an ABI mismatch in the internal version of CDOT/ZDOT used by the C fallback
   versions of the LAPACK source
 - Improved the f_check script for detecting the Fortran compiler to handle embedded
   dashes in path names
 - Fixed several memory access issues in the utests that were detected by Address
   Sanitizer
 - Fixed Makefile errors in cases where only a subset of precision types was selected
 - Fixed missing function errors in Makefile builds without LAPACK or without threads
 - Fixed a syntax error in the benchmarks Makefile
 - Fixed compiler warnings in the CBLAS testsuite
 - Fixed the OpenMP compiler option used with the Intel Ifx compiler
 - Updated the README sections on supported cpus and operating systems, and added
   notes pertaining to JAVA
 - Updated the documentation page for supported BLAS-like extensions
 - included fixes from the Reference-LAPACK project:
   - Improved step length selection in the fallback path of ?LAED4 
     (Reference-LAPACK PR 1191)
   - Rounding up of LWORK and removal of redundant type conversions in the GVD
     functions (Reference-LAPACK PR 1202)
   - internal errors were getting ignored in calculation of selected eigenvalues
     (Reference-LAPACK PR 1204)

arm64:
 - Fixed a potential miscompilation of the SDOT/DDOT/DSDOT kernels
 - Fixed DYNAMIC_ARCH compilation with CMake and compilers lacking SVE support
 - Improved the performance of BGEMM and SBGEMM kernels for Neoverse V2
 - Added optimized SSUM and DSUM kernels for Neoverse N1
 - Added preliminary support for Neoverse V3 cpus as NEOVERSEV2
 - Added cpu autodetection of Cortex A725 and X925 cpus
 - Fixed a CMake build problem with flang on Mac OS
 - Fixed build problems with gcc versions 12 and earlier that do not support fp16
 - Fixed compilation of GEMM kernels for VORTEXM4/ARMV9SME without multithreading
 - Fixed the optimized CDOT/ZDOT kernel to compile with LLVM under Windows on Arm
 - Renamed the copy of the DllMain function used in static linking on MS Windows to
   OpenBLASDllMain to avoid symbol name conflicts with other libraries

ioongarch64:
 - fixed POTRF returning wrong results on LA464 due to a wrong parameter setting

power:
 - Fixed compilation problems caused by missing support for half-precision floats (FP16)
 - Fixed a potential miscompilation of the POWER10 DGEMV kernel by limiting its optimization
   level
 - Fixed a SCAL issue on PPCG4/PPC970 running Linux
 - Worked around a SCAL issue on PPC970 running FreeBSD by switching to the generic C kernels

riscv64:
 - Optimized the CROT/ZROT kernel for vector length 128 in the non-unit stride path
 - Improved SBGEMM/SHGEMM and related helper functions for type conversion
 - Fixed probing for BFLOAT16 support in DYNAMIC_ARCH cpu detection at runtime

x86_64:
 - Fixed a potential miscompilation (by gcc 15.x) of the AVX512 SGEMM kernel for "small"
   matrix sizes
 - Fixed the SROT and DROT kernels for Haswell to have consistent (FMA) rounding
   in the main loop and tail call
 - Added automatic detection of Intel Arrow Lake H/U, Panther Lake and Jasper Lake
 - Added automatic detection of Intel Emerald Rapids and upcoming cpu models
 - Updated the cache size translation table in the cpu model autodetection code
 - Improved cpu detection fallback to also include Nehalem as a non-AVX option  
 - Fixed a Makefile build issue with clang and the SkylakeX SGEMM kernel 
 - Renamed the copy of the DllMain function used in static linking on MS Windows to
   OpenBLASDllMain to avoid symbol name conflicts with other libraries

wasm:
 - Added optimized intrinsics kernels for SGEMM and DGEMM as well as DOT, ROT and SUM

====================================================================
Version 0.3.31
15-Jan-2026

general:
 - reverted a matrix partitioning optimization from 0.3.30 that could lead to 
   race conditions and subsequent invalid results in GEMM
 - added the bfloat16 extensions BGEMM and BGEMV
 - added a BLAS interface for the ?GEMM_BATCH extensions
 - added the BLAS extensions ?GEMM_BATCH_STRIDED and their CBLAS interface
 - added the basic infrastructure for half-precision float (FP16) format 
   using SH prefix
 - reimplemented the LAPACK SLAED3/DLAED3 function using multithreading, thereby
   improving the performance of the SSYEVD/DSYEVD eigensolver for symmetric matrices 
   on all platforms
 - limited the number of retries for initial memory allocation to avoid infinite
   hanging on low-memory systems
 - fixed a thread lockup situation encountered with python 3.9 or older and numpy
 - introduced a problem size threshold for multithreading in STRMV/DTRMV
 - introduced a problem size threshold for multithreading in CHER/CHER2/CHPR/CHPR2
   and ZHER/ZHER2/ZHPR/ZHPR2
 - improved the problem size thresholds for multithreading in SGER/DGER
 - improved autodetection of the Fortran compiler
 - fixed passing of the INTERFACE64=1 option to the flang-new compiler
 - fixed a potential deadlock in multithreaded code after calling fork()
 - fixed builds using CMake on FreeBSD
 - fixed builds using CMake from within Cygwin on Windows
 - fixed builds using CMake and the NVHPC compiler on ARM64
 - fixed CMake build error from misdetecting compiler or OpenMP versions
 - improved contents of the CMake-generated OpenBLASConfig.cmake file
 - added support for cross-compilation to RISCV targets via CMake
 - fixed cross-compilation to x86 targets from non-x86 architectures
 - fixed failure to install cblas.h if NO_CBLAS=0 was specified
 - fixed missing user-defined pre- and postfixes on functions in lapack.h,lapacke.h 
 - included fixes from the Reference-LAPACK project:
   - fix ordering bug in ?LAED/?LASD (Reference-LAPACK PR 1140)
   - revert changes in ?GEEV from PR 1129 (Reference-LAPACK PR 1142)
   - fix workspace allocation in LAPACKE_?TRSEN (Reference-LAPACK PR 1144)

riscv:
 - added optimized SBGEMM kernels for ZVL128B and ZVL256B targets
 - added optimized SHGEMM kernels for ZVL128B and ZVL256B targets
 - added optimized SBGEMV and SHGEMV kernels for ZVL128B/ZVL256B
 - improved performance of the GEMV kernel for ZVL256B
 - improved the performance of the CROT and ZROT kernels for ZVL128B and x280
 - improved the detection of RVV1.0 capability 
 - improved performance of the matrix packing helper functions for ZVL128B and ZVL256B
 - improved performance of OMATCOPY for ZVL128B and ZVL256B

arm:
 - fixed spurious executable stack in the getarch utility

arm64:
 - fixed spurious executable stack in the getarch utility
 - fixed compiler warnings arising from the timer macro RPCC
 - fixed cache size detection for Qualcomm Oryon under Windows on Arm
 - fixed argument handling in the default SVE kernel for SDOT/DDOT
 - building the BFLOAT16 kernels is now enabled by default
 - improved the overall performance of GEMM,SYMM and HEMM on A64FX
 - improved the performance of SDOT/DDOT on A64FX
 - improved the multithreading performance of SDOT/DDOT on A64FX by
   introduction of a throttling table matching thread count to problem size
 - improved the performance of SGER/DGER on A64FX and NEOVERSEV1
 - improved the multithreading performance of GEMM on A64FX and NEOVERSEV1
 - improved the performance of the GEMV kernel for SVE-capable targets
 - improved the multithreading performance of SGEMM on NEOVERSEV1 and V2
 - added optimized SAXPY/DAXPY SVE kernels for A64FX and NEOVERSEV1
 - added optimized BGEMM and BGEMV kernels for NEOVERSEV1
 - added an optimized BGEMM kernel for NEOVERSEN2
 - added support for the NEOVERSEV2 cpu
 - added dedicated support for the Apple M4 cpu as VORTEXM4
 - added optimized SGEMM/SSYMM/STRMM/SSYRK/SSYR2K for SME-capable targets
   (ARMV9SME and VORTEXM4)
 - improved the precision of the SNRM2 kernel
 - added cpu autodetection and compiler settings for Ampere One processors
 - fixed cpu autodetection for Apple M systems running Linux
 - fixed building on MacOS with AppleClang,gfortran and xcode v16 or newer
 - fixed several errors in the C code replacements for the complex and double
   precision complex LAPACK functions that get used (only) when compiling with
   Microsoft C and NOFORTRAN=1 under MS Windows

power:
 - added initial support for the POWER11 architecture
 - improved performance of DGEMM and DGEMV on POWER10
 - fixed the default compiler flags to use "-O3" instead of the possibly unsafe 
   "-Ofast"
 - fixed building under MacOS (for old G4 Macs) with CMake
 - fixed potential miscompilation of DGEMV and other assembly kernels by gcc15.1
 - fixed compilation with recent versions of flang

loongarch64:
 - fixed warnings and potential inaccuracies arising from incorrect saving of registers
 - fixed enumeration of logical cores on big NUMA servers
 - fixed building with LLVM and the INTERFACE64=1 option

x86:
 - fixed building the GEMM3M kernels for the GENERIC target
 - fixed several errors in the C code replacements for the complex and double
   precision complex LAPACK functions that get used (only) when compiling with
   Microsoft C and NOFORTRAN=1 under MS Windows

x86_64:
 - added cpu autodetection for Intel Lunar Lake (Core Ultra 200V)
 - changed all ?MIN and ?MAX assembly kernels to use unaligned operations 
 - fixed several errors in the C code replacements for the complex and double
   precision complex LAPACK functions that get used (only) when compiling with
   Microsoft C and NOFORTRAN=1 under MS Windows
 - fixed potential crashes in builds for Cooper Lake, Sapphire Rapids or Zen5 cpus
   under MS Windows

zarch:
 - added support for building with CMake

sparc:
 - fixed a potential crash in the DNRM2 kernel

====================================================================
Version 0.3.30
19-Jun-2025

general:
 - fixed an installation problem with the thread safety test in gmake builds
 - fixed spurious overwriting of an input array in complex GEMMT/GEMMTR
 - fixed naming of GEMMTR in error messages from XERBLA
 - fixed compilation of SBGEMMT/SBGEMMTR in CMake builds
 - fixed the implementation of ?NRM2 to handle INCX=0 correctly
 - removed tests for CSROT and ZDROT that relied on unspecified behavior
 - fixed a performance regression in multithreaded GEMM that was particularly
   serious on POWER targets
 - fixed linking issues when using LLVM's flang-new with gmake
 - fixed a potential thread safety problem with C11 atomic operations
 - further improved the workload partitioning in parallel GEMM
 - fixed omission of LAPACKE interfaces for CGESVDQ,CTRSYL3 and ?GEQPF in 
   CMake builds
 - fixed mishandling of setting NO_LAPACK to FALSE, and incorrect dependencies
   for LAPACK function SPMV in CMake builds
 - added explicit CMake options for building LAPACKE and shared libraries
 - simplified and improved handling of OpenMP options in CMake builds
 - reworked Windows DLL generation in CMake builds to ensure correct symbol
   renaming (pre/postfixing) and optional generation of PDB files for debugging
 - updated the Perl script version of the gensymbol utility for use with 
   Windows-on-Arm
 - Fixed building with (Mingw) gmake on Windows to ensure completeness of the
   LAPACK included in the static library (potential race condition due to the
   Windows version of the "ln" utility creating snapshot copies rather than links)
 - fixed unwanted deletion of the lapacke_mangling.h file by "make clean"
 - fixed potential duplication of a _64 suffix on library names in CMake builds
 - fixed compilation of the C fallback copies of the LAPACK code with GCC 15
 - included fixed from the Reference-LAPACK project:
   - fixed a truncated error message in the EIG part of the testsuite
     (Reference-LAPACK PR 1119)
   - fixed too strict check in LAPACKE_?gesdd_work (PR #1126)
   - fixed memory corruption when calling ?GEEV with non-finite data (PR #1128)
   - fixed missing initialization of a variable in C/GEQP3RK (PR #1131)
   - fixed 2nd dimension chosen in C/ZUNMLQ transposition operation (PR #1135)

x86_64:
 - fixed an error in the SBGEMV kernel for Cooper Lake/Sapphire Rapids
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - improved the compiler identification code for flang-new
 - fixed a potential build issue in the ZSUM kernel
 - fixed "argument list too long" errors when building on MacOS
 - added cpu autodetection support for several new Arrow Lake models
 - fixed conditional inclusion of the fast path SGEMM kernel in DYNAMIC_ARCH
 - fixed compilation with the MinGW build of GCC 15

arm64:
 - fixed cpu type detection of A64FX and some ThunderX models (broken in 0.3.29)
 - added support for the AmpereOne/1A cpus in DYNAMIC_ ARCH builds
 - added an optimized SBGEMM kernel for NEOVERSEV1
 - improved 1xN SBGEMM performance by forwarding to SBGEMV 
 - introduced a stepwise increase of the thread count used for
   SGEMM and SGEMV on NEOVERSEV1/V2 in relation to problem size
 - introduced a stepwise increase of the thread count used for
   DGEMV on NEOVERSEV1 in relation to problem size
 - introduced a stepwise increase of the thread count used for
   SDOT and DDOT on NEOVERSEV1 in relation to problem size
 - worked around assembler limitations in LLVM for Windows-on-Arm
 - enabled cpu type autodetection from the registry on Windows-on-Arm
 - improved multithreading threshold for GEMV and GESV on Windows-on-Arm
 - fixed overoptimization issues with LLVM's flang in Windows-on-Arm
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - added a fast path SGEMM kernel for small workloads on SME capable targets
 - improved performance of SGEMM and DGEMM kernels for small workloads
 - improved performance of SGEMV and DGEMV on SVE-capable targets
 - improved performance of SGEMV on NEOVERSEN1 and Apple M
 - added optimized SSYMV and DSYMV kernels for NEOVERSEN1, Apple M and all
   SVE capable targets
 - added optimized SBGEMV kernels for NEOVERSEV1/V2/N2
 - improved performance of SGEMM through faster NCOPY kernels
 - added compiler options for the NVIDIA HPC Compiler Suite
 - fixed compilation on OSX with XCode 16.3 and later
 - fixed cpu core type and cache size detection on Apple M4
 - updated GEMM parameter settings for Neoverse cpus in cross-builds with CMake
 - fixed default compiler options for NEOVERSEN1 and CORTEXX2 in CMake builds
 - fixed conditional inclusion of the fast path SGEMM kernel in DYNAMIC_ARCH
 - fixed potential miscompilation of the non-SVE SDOT kernel

riscv64:
 - added optimized SROTM and DROTM kernels for x280
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - improved performance of GEMM_TCOPY on RVV1.0 targets with 
   VLEN of 128 or 256
 - improved performance of OMATCOPY on targets with VLEN 256
 - greatly improved performance of SGEMV/DGEMV
 - improved performance of CGEMV and ZGEMV on C910V and all RVV targets 
   with VLEN 256
 - improved performance of SAXPBY and DAXPBY on C910V and all RVV targets 
   with VLEN 256
 - improved performance of AXPY and DOT on C910V and ZVL256B targets by
   falling back to non-vectorized code for very small N. (Thereby fixing
   poor performance of CHBMV/ZHBMV for very small K)
 - fixed CMake build failures of the TRMM kernels 

loongarch64:
 - improved performance of the LSX versions of SSYMV/DSYMV
 - made the LASX versions of the DSYMV and SSYMV kernels 
   compatible with hardware changes in LA664 and future targets
 - fixed inaccuracies in several LASX kernels
 - improved compatibility of LSX kernels with LA264 targets
 - fixed handling of deprecated target names in CMake builds
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL

power:
 - fixed building for PPCG4 with CMake
 - fixed SSCAL/DSCAL on PPC970 running FreeBSD
 - fixed a potential alignment issue in the POWER8 SGEMV kernel
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL

zarch:
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - fixed unwanted generation of object files with a writable stack

x86:
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - worked around potential miscompilation of CDOT with very old binutils

arm:
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL
 - fixed unwanted generation of object files with a writable stack

sparc:
 - fixed corner cases of NAN and INF input handling in CSCAL and ZSCAL

alpha:
 - fixed build failure caused by spurious Windows-only typecasts

cell:
 - fixed probable build issue caused by spurious Windows-only typecasts
 
====================================================================
Version 0.3.29
12-Jan-2025

general:
 - fixed a potential NULL pointer dereference in multithreaded builds
 - added function aliases for GEMMT using its new name GEMMTR adopted by Reference-BLAS
 - fixed a build failure when building without LAPACK_DEPRECATED functions
 - the minimum required CMake version for CMake-based builds was raised to 3.16.0 in order 
   to remove many compatibility and deprecation warnings
 - added more detailed CMake rules for OpenMP builds (mainly to support recent LLVM)
 - fixed the behavior of the recently added CBLAS_?GEMMT functions with row-major data
 - improved thread scaling of multithreaded SBGEMV 
 - improved thread scaling of multithreaded TRTRI 
 - fixed compilation of the CBLAS testsuite with gcc14 (and no Fortran compiler)
 - added support for option handling changes in flang-new from LLVM18 onwards 
 - added support for recent calling conventions changes in Cray and NVIDIA compilers
 - added support for compilation with the NAG Fortran compiler
 - fixed placement of the -fopenmp flag and libsuffix in the generated pkgconfig file
 - improved the CMakeConfig file generated by the Makefile build
 - fixed const-correctness of cblas_?geadd in cblas.h
 - fixed a potential inaccuracy in multithreaded BLAS3 calls
 - fixed empty implementations of get/set_affinity that print a warning in OpenMP builds
 - fixed function signatures for TRTRS in the converted C version of LAPACK
 - fixed omission of several single-precision LAPACK symbols in the shared library 
 - improved build instructions for the provided "pybench" benchmarks
 - improved documentation, including added build instructions for WoA and HarmonyOS
   as well as descriptions of environment variables that affect build and runtime behavior
 - added a separate "make install_tests" target for use with cross-compilations
 - integrated improvements and corrections from Reference-LAPACK:
   - removed a comparison in LAPACKE ?tpmqrt that is always false (LAPACK PR 1062)
   - fixed the leading dimension for B in tests for GGEV (LAPACK PR 1064)
   - replaced the ?LARFT functions with a recursive implementation (LAPACK PR 1080)

arm:
 - fixed build with recent versions of the NDK (missing .type declaration of symbols)

arm64:
 - fixed a long-standing bug in the (generic) c/zgemm_beta kernel that could lead to
   reads and writes outside the array bounds in some circumstances
 - rewrote cpu autodetection to scan all cores and return the highest performing type
 - improved the DGEMM performance for SVE targets and small matrix sizes
 - improved dimension criteria for forwarding from GEMM to GEMV kernels
 - added SVE kernels for ROT and SWAP
 - improved SVE kernels for SGEMV and DGEMV on A64FX and NEOVERSEV1
 - added support for using the "small matrix" kernels with CMake as well
 - fixed compilation on Windows on Arm
 - improved compile-time detection of SVE capability
 - added cpu autodetection and initial support for Apple M4
 - added support for compilation on systems running IOS
 - added support for compilation on NetBSD ("evbarm" architecture)
 - fixed NRM2 implementations for generic SVE targets and the Neoverse N2
 - fixed compilation for SVE-capable targets with the NVIDIA compiler

x86_64:
 - fixed a wrong storage size in the SBGEMV kernel for Cooper Lake
 - added cpu autodetection for Intel Granite Rapids
 - added cpu autodetection for AMD Ryzen 5 series
 - added optimized SOMATCOPY_CT for AVX-capable targets
 - fixed the fallback implementation of GEMM3M in GENERIC builds
 - tentatively re-enabled builds with the EXPRECISION option
 - worked around a miscompilation of tests with mingw32-gfortran14
 - added support for compilation with the Intel oneAPI 2025.0 compiler on Windows

power:
 - fixed multithreaded SBGEMM
 - fixed a CMake build problem on POWER10
 - improved the performance of SGEMV
 - added vectorized implementations of SBGEMV and support for forwarding 1xN SBGEMM to them
 - fixed illegal instructions and potential memory overflow in SGEMM on PPCG4
 - fixed handling of NaN and Inf arguments in SSCAL and DSCAL on PPC440,G4 and 970
 - added improved CGEMM and ZGEMM kernels for POWER10
 - added Makefile logic to remove all optimization flags in DEBUG builds

mips64:
 - fixed compilation with gcc14
 - fixed GEMM parameter selection for the MIPS64_GENERIC target
 - fixed a potential build failure when compiling with OpenMP

loongarch64:
 - fixed compilation for Loongson3 with recent versions of gmake
 - fixed a potential loss of precision in Loongson3A GEMM
 - fixed a potential build failure when compiling with OpenMP
 - added optimized SOMATCOPY for LASX-capable targets
 - introduced a new cpu naming scheme while retaining compatibility
 - added support for cross-compiling Loongarch64 targets with CMake
 - added support for compilation with LLVM

riscv64:
 - removed thread yielding overhead caused by sched_yield
 - replaced some non-standard intrinsics with their official names
 - fixed and sped up the implementations of CGEMM/ZGEMM TCOPY for vector lenghts 128 and 256
 - improved the performance of SNRM2/DNRM2 for RVV1.0 targets
 - added optimized ?OMATCOPY_CN kernels for RVV1.0 targets

====================================================================
Version 0.3.28
 8-Aug-2024

general:
- Reworked the unfinished implementation of HUGETLB from GotoBLAS
  for allocating huge memory pages as buffers on suitable systems
- Changed the unfinished implementation of GEMM3M for the generic
  target on all architectures to at least forward to regular GEMM
- Improved multithreaded GEMM performance for large non-skinny matrices
- Improved BLAS3 performance on larger multicore systems through improved
  parallelism
- Improved performance of the initial memory allocation by reducing
  locking overhead
- Improved performance of GBMV at small problem sizes by introducing
  a size barrier for the switch to multithreading
- Added an implementation of the CBLAS_GEMM_BATCH extension
- Fixed miscompilation of CAXPYC and ZAXPYC on all architectures in 
  CMAKE builds (error introduced in 0.3.27)
- Fixed corner cases involving the handling of NAN and INFINITY
  arguments in ?SCAL on all architectures
- Added support for cross-compiling to WEBM with CMAKE (in addition
  to the already present makefile support)
- Fixed NAN handling and potential accuracy issues in compilations with
  Intel ICX by supplying a suitable fp-model option by default
- The contents of the github project wiki have been converted into
  a new set of documentation included with the source code.
- It is now possible to register a callback function that replaces
  the built-in support for multithreading with an external backend
  like TBB (openblas_set_threads_callback_function)
- Fixed potential duplication of suffixes in shared library naming
- Improved C compiler detection by the build system to tolerate more
  naming variants for gcc builds
- Fixed an unnecessary dependency of the utest on CBLAS
- Fixed spurious error reports from the BLAS extensions utest
- Fixed unwanted invocation of the GEMM3M tests in cross-compilation
- Fixed a flaw in the makefile build that could lead to the pkgconfig
  file containing an entry of UNKNOWN for the target cpu after installing
- Integrated fixes from the Reference-LAPACK project:
  - Fixed uninitialized variables in the LAPACK tests for ?QP3RK (PR 961)
  - Fixed potential bounds error in ?UNHR_COL/?ORHR_COL (PR 1018)
  - Fixed potential infinite loop in the LAPACK testsuite (PR 1024)
  - Make the variable type used for hidden length arguments configurable (PR 1025)  
  - Fixed SYTRD workspace computation and various typos (PR 1030)
  - Prevent compiler use of FMA that could increase numerical error in ?GEEVX (PR 1033)

x86-64:
- reverted thread management under Windows to its state before 0.3.26
  due to signs of race conditions in some circumstances now under study
- fixed accidental selection of the unoptimized generic SBGEMM kernel
  in CMAKE builds for CooperLake and SapphireRapids targets
- fixed a potential thread buffer overrun in SBSTOBF16 on small systems
- fixed an accuracy issue in ZSCAL introduced in 0.3.26
- fixed compilation with CMAKE and recent releases of LLVM
- added support for Intel Emerald Rapids and Meteor Lake cpus
- added autodetection support for the Zhaoxin KX-7000 cpu
- fixed autodetection of Intel Prescott (probably broken since 0.3.19)
- fixed compilation for older targets with the Yocto SDK
- fixed compilation of the converter-generated C versions
  of the LAPACK sources with gcc-14
- improved compiler options when building with CMAKE and LLVM for
  AVX512-capable targets
- added support for supplying the L2 cache size via an environment
  variable (OPENBLAS_L2_SIZE) in case it is not correctly reported
  (as in some VM configurations)
- improved the error message shown when thread creation fails on startup
- fixed setting the rpath entry of the dylib in CMAKE builds on MacOS

arm:
- fixed building for baremetal targets with make

arm64:
- Added a fast path forwarding SGEMM and DGEMM calls with a 1xN or Mx1
  matrix to the corresponding GEMV kernel 
- added optimized SGEMV and DGEMV kernels for A64FX
- added optimized SVE kernels for small-matrix GEMM
- added A64FX to the cpu list for DYNAMIC_ARCH
- fixed building with support for cpu affinity
- worked around accuracy problems with C/ZNRM2 on NeoverseN1 and
  Apple M targets
- improved GEMM performance on Neoverse V1
- fixed compilation for NEOVERSEN2 with older compilers
- fixed potential miscompilation of the SVE SDOT and DDOT kernels
- fixed potential miscompilation of the non-SVE CDOT and ZDOT kernels
- fixed a potential overflow when using very large user-defined BUFFERSIZE
- fixed setting the rpath entry of the dylib in CMAKE builds on MacOS

power:
- Added a fast path forwarding SGEMM and DGEMM calls with a 1xN or Mx1
  matrix to the corresponding GEMV kernel 
- significantly improved performance of SBGEMM on POWER10
- fixed compilation with OpenMP and the XLF compiler
- fixed building of the BLAS extension utests under AIX
- fixed building of parts of the LAPACK testsuite with XLF
- fixed CSWAP/ZSWAP on big-endian POWER10 targets
- fixed a performance regression in SAXPY on POWER10 with OpenXL
- fixed accuracy issues in CSCAL/ZSCAL when compiled with LLVM
- fixed building for POWER9 under FreeBSD
- fixed a potential overflow when using very large user-defined BUFFERSIZE
- fixed an accuracy issue in the POWER6 kernels for GEMM and GEMV

riscv64:
- Added a fast path forwarding SGEMM and DGEMM calls with a 1xN or Mx1
  matrix to the corresponding GEMV kernel 
- fixed building for RISCV64_GENERIC with OpenMP enabled
- added DYNAMIC_ARCH support (comprising GENERIC_RISCV64 and the two
  RVV 1.0 targets with vector length of 128 and 256)
- worked around the ZVL128B kernels for AXPBY mishandling the special
  case of zero Y increment

loongarch64:
- improved GEMM performance on servers of the 3C5000 generation
- improved performance and stability of DGEMM
- improved GEMV and TRSM kernels for LSX and LASX vector ABIs
- fixed CMAKE compilation with the INTERFACE64 option set
- fixed compilation with CMAKE
- worked around spurious errors flagged by the BLAS3 tests
- worked around a miscompilation of the POTRS utest by gcc 14.1

mips64:
- fixed ASUM and SUM kernels to accept negative step sizes in X
- fixed complex GEMV kernels for MSA

====================================================================
Version 0.3.27
 4-Apr-2024

general:
- added initial (generic) support for the CSKY architecture
- capped the maximum number of threads used in GEMM, GETRF and POTRF to avoid creating
  underutilized or idle threads
- sped up multithreaded POTRF on all platforms
- added extension openblas_set_num_threads_local() that returns the previous thread count
- re-evaluated the SGEMV and DGEMV load thresholds to avoid activating multithreading 
  for too small workloads
- improved the fallback code used when the precompiled number of threads is exceeded,
  and made it callable multiple times during the lifetime of an instance
- added CBLAS interfaces for the BLAS extensions ?AMIN,?AMAX, CAXPYC and ZAXPYC
- fixed a potential buffer overflow in the interface to the GEMMT kernels
- fixed use of incompatible pointer types in GEMMT and C/ZAXPBY as flagged by GCC-14
- fixed unwanted case sensitivity of the character parameters in ?TRTRS
- sped up the OpenMP thread management code
- fixed sizing of logical variables in INTERFACE64 builds of the C version of LAPACK
- fixed inclusion of new LAPACK and LAPACKE functions from LAPACK 3.11 in the shared library
- added a testsuite for the BLAS extensions
- modified the error thresholds for SGS/DGS functions in the LAPACK testsuite to suppress
  spurious errors
- added support for building the benchmark collection with CMAKE
- added rewriting of linker options to avoid linking both libgomp and libomp in CMAKE builds
  with OpenMP enabled that use clang with gfortran
- fixed building on systems with ucLibc
- added support for calling ?NRM2 with a negative increment value on all architectures
- added support for the LLVM18 version of the flang-new compiler
- fixed handling of the OPENBLAS_LOOPS variable in several benchmarks
- Integrated fixes from the Reference-LAPACK project:
  - Increased accuracy in C/ZLARFGP (Reference-LAPACK PR 981)
  
x86:
- fixed handling of NaN and Inf arguments in ZSCAL
- fixed GEMM3M functions failing in CMAKE builds

x86-64:
- removed all instances of sched_yield() on Linux and BSD
- fixed a potential deadlock in the thread server on MSWindows (introduced in 0.3.26)
- fixed GEMM3M functions failing in CMAKE builds
- fixed handling of NaN and Inf arguments in ZSCAL
- added compiler checks for AVX512BF16 compatibility
- fixed LLVM compiler options for Sapphire Rapids 
- fixed cpu handling fallbacks for Sapphire Rapids with
  disabled AVX2 in DYNAMIC_ARCH mode
- fixed extensions SCSUM and DZSUM
- improved GEMM performance for ZEN targets

arm:
- fixed handling of NaN and Inf arguments in ZSCAL

arm64:
- added initial support for the Cortex-A76 cpu
- fixed handling of NaN and Inf arguments in ZSCAL
- fixed default compiler options for gcc (-march and -mtune)
- added support for ArmCompilerForLinux
- added support for the NeoverseV2 cpu in DYNAMIC_ARCH builds
- fixed mishandling of the INTERFACE64 option in CMAKE builds
- corrected SCSUM kernels (erroneously duplicating SCASUM behaviour)  
- added SVE-enabled kernels for CSUM/ZSUM
- worked around an inaccuracy in the NRM2 kernels for NeoverseN1 and Apple M

power:
- improved performance of SGEMM on POWER8/9/10
- improved performance of DGEMM on POWER10
- added support for OpenMP builds with xlc/xlf on AIX
- improved cpu autodetection for DYNAMIC_ARCH builds on older AIX
- fixed cpu core counting on AIX
- added support for building a shared library on AIX

riscv64:
- added support for the X280 cpu
- added support for semi-generic RISCV models with vector length 128 or 256
- added support for compiling with either RVV 0.7.1 or RVV 1.0 standard compilers
- fixed handling of NaN and Inf arguments in ZSCAL
- improved cpu model autodetection
- fixed corner cases in ?AXPBY for C910V
- fixed handling of zero increments in ?AXPY kernels for C910V

loongarch64:
- added optimized kernels for ?AMIN and ?AMAX
- fixed handling of NaN and Inf arguments in ZSCAL
- fixed handling of corner cases in ?AXPBY
- fixed computation of SAMIN and DAMIN in LSX mode
- fixed computation of ?ROT
- added optimized SSYMV and DSYMV kernels for LSX and LASX mode
- added optimized CGEMM and ZGEMM kernels for LSX and LASX mode
- added optimized CGEMV and ZGEMV kernels

mips:
- fixed utilizing MSA on P5600 and related cpus (broken in 0.3.22)
- fixed handling of NaN and Inf arguments in ZSCAL
- fixed mishandling of the INTERFACE64 option in CMAKE builds

zarch:
- fixed handling of NaN and Inf arguments in ZSCAL
- fixed calculation of ?SUM on Z13

====================================================================
Version 0.3.26
 2-Jan-2024

general:
- improved the version of openblas.pc that is created by the CMAKE build
- fixed a CMAKE-specific build problem on older versions of MacOS
- worked around linking problems on old versions of MacOS
- corrected installation location of the lapacke_mangling header in CMAKE builds
- added type declarations for complex variables to the MSVC-specific parts of the LAPACK header
- significantly sped up ?GESV for small problem sizes by introducing a lower bound for multithreading
- imported additions and corrections from the Reference-LAPACK project:
  - added new LAPACK functions for truncated QR with pivoting (Reference-LAPACK PRs 891&941)
  - handle miscalculation of minimum work array size in corner cases (Reference-LAPACK PR 942)
  - fixed use of uninitialized variables in ?GEDMD and improved inline documentation (PR 959)
  - fixed use of uninitialized variables (and consequential failures) in ?BBCSD (PR 967)
  - added tests for the recently introduced Dynamic Mode Decomposition functions (PR 736)
  - fixed several memory leaks in the LAPACK testsuite (PR 953)
  - fixed counting of testsuite results by the Python script (PR 954)
   
x86-64:
- fixed computation of CASUM on SkylakeX and newer targets in the special
  case that AVX512 is not supported by the compiler or operating environment
- fixed potential undefined behaviour in the CASUM/ZASUM kernels for AVX512 targets
- worked around a problem in the pre-AVX kernels for GEMV
- sped up the thread management code on MS Windows

arm64:
- fixed building of the LAPACK testsuite with Xcode 15 on Apple M1 and newer
- sped up the thread management code on MS Windows
- sped up SGEMM and DGEMM on Neoverse V1 and N1
- sped up ?DOT on SVE-capable targets
- reduced the number of targets in DYNAMIC_ARCH builds by eliminating functionally equivalent ones
- included support for Apple M1 and newer targets in DYNAMIC_ARCH builds

power:
- improved the SGEMM kernel for POWER10
- fixed compilation with (very) old versions of gcc
- fixed detection of old 32bit PPC targets in CMAKE-based builds
- added autodetection of the POWERPC 7400 subtype
- fixed CMAKE-based compilation for PPCG4 and PPC970 targets

loongarch64:
- added and improved optimized kernels for almost all BLAS functions

====================================================================
Version 0.3.25
 12-Nov-2023

general:
- improved the error message shown on exceeding the maximum thread count
- improved the code to add supplementary thread buffers in case of overflow
- fixed a potential division by zero in ?ROTG
- improved the ?MATCOPY functions to accept zero-sized rows or columns
- corrected empty prototypes in function declarations
- cleaned up unused declarations in the f2c-converted versions of the LAPACK sources
- fixed compilation with the Cray CCE Compiler suite
- improved link line rewriting to avoid mixed libgomp/libomp builds with clang&gfortran
- worked around OPENMP builds with LLVM14's libomp hanging on FreeBSD
- improved the Makefiles to require less option duplication on "make install"
- imported the following changes from the upcoming release 3.12 of Reference-LAPACK
  - deprecate utility functions ?GELQS and ?GEQRS (LAPACK PR 900)
  - apply rounding up to workspace calculations done in floating point (LAPACK PR 904)
  - avoid overflow in STGEX2/DTGEX2 (LAPACK PR 907)
  - fix accumulation in ?LASSQ (LAPACK PR 909)
  - fix handling of NaN values in ?GECON (LAPACK PR 926)
  - avoid overflow in CBDSQR/ZBDSQR (LAPACK PR 927)
  - fix poor vector orthogonalizations in ?ORBDB5/?UNBDB5 (LAPACK PR 928 & 930)

x86-64:
- fixed compile-time autodetection of AMD Ryzen3 and Ryzen4 cpus
- fixed capability-based fallback selection for unknown cpus in DYNAMIC_ARCH
- added AVX512 optimizations for ?ASUM on Sapphire Rapids and Cooper Lake

ARM64:
- fixed building on Apple with homebrew gcc
- fixed building with XCODE 15
- fixed building on A64FX and Cortex A710/X1/X2
- increased the default buffer size for recent ARM server cpus 

POWER:
- fixed building with the IBM xlf 16.1.1 compiler
- fixed building with IBM XL C
- added support for DYNAMIC_ARCH builds with clang
- fixed union declaration in the BFLOAT16 test case
- enable optimizations for the AIX assembler on POWER10

LOONGARCH64:
- added an optimized SGEMV kernel
- added an optimized DTRSM kernel

====================================================================
Version 0.3.24
 03-Sep-2023

general:
   - declared the arguments of cblas_xerbla as const (in accordance with the reference implementation 
     and others, the previous discrepancy appears to have dated back to GotoBLAS)
   - fixed the implementation of ?GEMMT that was added in 0.3.23
   - made cpu-specific SWITCH_RATIO parameters for GEMM available to DYNAMIC_ARCH builds
   - fixed application of SYMBOLSUFFIX in CMAKE builds
   - fixed missing SSYCONVF function in the shared library
   - fixed parallel build logic used with gmake
   - added support for compilation with LLVM17, in particular its new Fortran compiler
   - added support for CMAKE builds using the NVIDIA HPC compiler
   - fixed INTERFACE64 builds with CMAKE and the f95 Fortran compiler
   - fixed cross-build detection and management in c_check
   - disabled building of the tests with CMAKE when ONLY_CBLAS is defined
   - fixed several issues with the handling of runtime limits on the number of OPENMP threads
   - corrected the error code returned by SGEADD/DGEADD when LDA is too small
   - corrected the error code returned by IMATCOPY when LDB is too small
   - updated ?NRM2 to support negative increment values (as introduced in release 3.10 
     of the reference BLAS)
   - fixed OpenMP builds with CLANG for the case where libomp is not in a standard location
   - fixed a potential overwrite of unrelated memory during thread initialisation on startup
   - fixed a potential integer overflow in the multithreading threshold for ?SYMM/?SYRK
   - fixed build of the LAPACKE interfaces for the LAPACK 3.11.0 ?TRSYL functions added in 0.3.22
   - fixed installation of .cmake files in concurrent 32 and 64bit builds with CMAKE
   - applied additions and corrections from the development branch of Reference-LAPACK:
   - fixed actual arguments passed to a number of LAPACK functions (from Reference-LAPACK PR 885)
   - fixed workspace query results in LAPACK ?SYTRF/?TRECV3 (from Reference-LAPACK PR 883)
   - fixed derivation of the UPLO parameter in LAPACKE_?larfb (from Reference-LAPACK PR 878)
   - fixed a crash in LAPACK ?GELSDD on NRHS=0 (from Reference-LAPACK PR 876)
   - added new LAPACK utility functions CRSCL and ZRSCL (from Reference-LAPACK PR 839)
	- corrected the order of eigenvalues for 2x2 matrices in ?STEMR (Reference-LAPACK PR 867)
	- removed spurious reference to OpenMP variables outside OpenMP contexts (Reference-LAPACK PR 860)
	- updated file comments on use of LAMBDA variable in LAPACK (Reference-LAPACK PR 852)
	- fixed documentation of LAPACK SLASD0/DLASD0 (Reference-LAPACK PR 855)
	- fixed confusing use of "minor" in LAPACK documentation (Reference-LAPACK PR 849)
	- added new LAPACK functions ?GEDMD for dynamic mode decomposition (Reference-LAPACK PR 736)
	- fixed potential stack overflows in the EIG part of the LAPACK testsuite (Reference-LAPACK PR 854)
	- applied small improvements to the variants of Cholesky and QR functions (Reference-LAPACK PR 847)
	- removed unused variables from LAPACK ?BDSQR (Reference-LAPACK PR 832)
	- fixed a potential crash on allocation failure in LAPACKE SGEESX/DGEESX (Reference-LAPACK PR 836)
	- added a quick return from SLARUV/DLARUV for N < 1 (Reference-LAPACK PR 837)
	- updated function descriptions in LAPACK ?GEGS/?GEGV (Reference-LAPACK PR 831)
	- improved algorithm description in ?GELSY (Reference-LAPACK PR 833)
	- fixed scaling in LAPACK STGSNA/DTGSNA (Reference-LAPACK PR 830)
	- fixed crash in LAPACKE_?geqrt with row-major data (Reference-LAPACK PR 768)
	- added LAPACKE interfaces for C/ZUNHR_COL and S/DORHR_COL (Reference-LAPACK PR 827)
	- added error exit tests for SYSV/SYTD2/GEHD2 to the testsuite (Reference-LAPACK PR 795)
	- fixed typos in LAPACK source and comments (Reference-LAPACK PRs 809,811,812,814,820)
	- adopt refactored ?GEBAL implementation (Reference-LAPACK PR 808)

x86_64:
   - added cpu model autodetection for Intel Alder Lake N
   - added activation of the AMX tile to the Sapphire Rapids SBGEMM kernel
   - worked around miscompilations of GEMV/SYMV kernels by gcc's tree-vectorizer
   - fixed compilation of Cooperlake and Sapphire Rapids kernels with CLANG
   - fixed runtime detection of Cooperlake and Sapphire Rapids in DYNAMIC_ARCH
   - fixed feature-based cputype fallback in DYNAMIC_ARCH
   - added support for building the AVX512 kernels with the NVIDIA HPC compiler
   - corrected ZAXPY result on old pre-AVX hardware for the INCX=0 case
   - fixed a potential use of uninitialized variables in ZTRSM

ARM64:
   - added cpu model autodetection for Apple M2
   - fixed wrong results of CGEMM/CTRMM/DNRM2 under OSX (use of reserved register)
   - added support for building the SVE kernels with the NVIDIA HPC compiler
   - added support for building the SVE kernels with the Apple Clang compiler
   - fixed compiler option handling for building the SVE kernels with LLVM
   - implemented SWITCH_RATIO parameter for improved GEMM performance on Neoverse
   - activated SVE SGEMM and DGEMM kernels for Neoverse V1
   - improved performance of the SVE CGEMM and ZGEMM kernels on Neoverse V1
   - improved kernel selection for the ARMV8SVE target and added it to DYNAMIC_ARCH
   - fixed runtime check for SVE availability in DYNAMIC_ARCH builds to take OS or
     container restrictions into account
   - fixed a potential use of uninitialized variables in ZTRSM
   - fix a potential misdetection of ARMV8 hardware as 32bit in CMAKE builds

LOONGARCH64:
   - added ABI detection
   - added support for cpu affinity handling
   - fixed compilation with early versions of the Loongson toolchain
   - added an optimized SGEMM kernel for 3A5000
   - added optimized DGEMV kernels for 3A5000
   - improved the performance of the DGEMM kernel for 3A5000

MIPS64:
   - fixed miscompilation of TRMM kernels for the MIPS64_GENERIC target

POWER:
   - fixed compiler warnings in the POWER10 SBGEMM kernel

RISCV:
   - fixed application of the INTERFACE64 option when building with CMAKE
   - fix a potential misdetection of RISCV hardware as 32bit in CMAKE builds
   - fixed IDAMAX and DOT kernels for C910V
   - fixed corner cases in the ROT and SWAP kernels for C910V
   - fixed compilation of the C910V target with recent vendor compilers

====================================================================
Version 0.3.23
 01-Apr-2023
 
 general:
   - fixed a serious regression in GETRF/GETF2 and ZGETRF/ZGETF2 where
     subnormal but nonzero data elements triggered the singularity flag
   - fixed a long-standing bug in CSPR/ZSPR in single-threaded operation
     for cases where elements of the X vector are real numbers (or 
     complex with only the real part zero)
   - fixed gmake builds with the option NO_LAPACK
   - fixed a few instances in the gmake Makefiles where expressly
     setting NO_LAPACK=0 or NO_LAPACKE=0 would have the opposite effect
   
x86_64:
   - added further CPUID values for Intel Raptor Lake

====================================================================
Version 0.3.22
 26-Mar-2023

general:
 - Updated the included LAPACK to Reference-LAPACK release 3.11.0
   plus post-release corrections and improvements
 - Added initial support for processing with the EMSCRIPTEN javascript
   converter (yielding a single-threaded build only)
 - Added a threshold for multithreading in SYMM, SYMV and SYR2K
 - Increased the threshold for multithreading in SYRK
 - OpenBLAS no longer decreases the global OMP_NUM_THREADS when it
   exceeds the maximum thread count the library was compiled for.
 - fixed ?GETF2 potentially returning NaN with tiny matrix elements
 - fixed openblas_set_num_threads to work in USE_OPENMP builds
 - fixed cpu core counting in USE_OPENMP builds returning the number
   of OMP "places" rather than cores
 - fixed interpretation of USE_PERL=0 in build scripts
 - fixed linking of the library with libm in CMAKE builds
 - fixed startup delays resulting from a wrong default setting of 
   NO_WARMUP in CMAKE builds
 - fixed inconsistent defaults for overriding of LAPACK SPMV, SPR, 
   SYMV, SYR functions in gmake and CMAKE builds
 - fixed stride calculation in the optimized small-matrix path of 
   complex SYR
 - fixed compilation of ReLAPACK with CMAKE
 - fixed pkgconfig file contents for INTERFACE64 builds
 - fixed building of Reference-LAPACK with recent gfortran
 - fixed building with only a subset of precision types on Windows
 - added new environment variable OPENBLAS_DEFAULT_NUM_THREADS
 - added a GEMV-based implementation of GEMMT 
 - added support for building under QNX
 - updated support for (cross-)building for ALPHA targets

x86_64:
 - added autodetection of Intel Raptor Lake cpu models
 - added SSCAL microkernels for Haswell and newer targets
 - improved the performance of the Haswell DSCAL microkernel
 - added CSCAL and ZSCAL microkernels for SkylakeX targets
 - fixed detection of gfortran and Cray CCE compilers
 - fixed detection of recent versions of the Intel Fortran compiler
 - fixed compilation with LLVM to no longer run out of AVX512 registers
 - fix cpu type option setting with recent NVIDIA HPC compiler versions
 - fixed compilation for/on AMD Ryzen 4 cpus
 - fixed compilation of AVX2-capable targets with Apple Clang
 - fixed runtime selection of COOPERLAKE in DYNAMIC_ARCH builds
 - worked around gcc/llvm using risky FMA operations in CSCAL/ZSCAL
 - worked around miscompilations of GEMV, SYMV and ZDOT kernels
   by gcc12's tree-vectorizer on OSX and Windows

ARM:
 - fixed cross-compilation to ARMV5 and ARMV6 targets with CMAKE

ARMV8:
 - fixed cross-compilation to CortexA53 with CMAKE
 - fixed compilation with CMAKE and "Arm Compiler for Linux 22.1"
 - added cpu autodetection for Cortex X3 and A715
 - fixed conditional compilation of SVE-capable targets in DYNAMIC_ARCH
 - sped up SVE kernels by removing unnecessary prefetches
 - improved the GEMM performance of Neoverse V1
 - added SVE kernels for SDOT and DDOT
 - added an SBGEMM kernel for Neoverse N2
 - improved cpu-specific compiler option selection for Neoverse cpus
 - added support for setting CONSISTENT_FPCSR 

MIPS64:
 - improved MSA capability detection and handling
 - added a MIPS64_GENERIC build target
 - fixed corner cases in DNRM2

LOONGARCH64:
 - fixed handling of the INTERFACE64 option

RISCV:
 - fixed handling of the INTERFACE64 option

====================================================================
Version 0.3.21
 07-Aug-2022

general:
 - Updated the included LAPACK to Reference-LAPACK release 3.10.1
 - when no Fortran compiler is available, OpenBLAS builds will now automatically
   build LAPACK from an f2c-converted copy of LAPACK 3.9.0 unless the NO_LAPACK option
   is specified
 - similarly added C versions of the BLAS and CBLAS tests
 - enabled building of the ReLAPACK GEMMT kernels when ReLAPACK is built
 - function LAPACKE_lsame is now annotated with the GCC attribute "const" to aid static analyzers
 - added USE_TLS to the list of options reported by the openblas_get_config() function
 - CMAKE builds now support the BUILD_TESTING keyword (to disable the LAPACK testsuite) of Reference-LAPACK
 - fixed CMAKE builds of the laswp_ncopy and neg_tcopy kernels
 - removed the build system requirements for PERL (while keeping the original perl scripts as backup)
 - handle building and running OpenBLAS on systems that report zero available cpu cores
 - added SYMBOLPREFIX/SYMBOLSUFFIX handling for LAPACK 3.10.0 functions added in 0.3.20
 - fixed linking of the utests on QNX
 - Added support for compilation with the Intel ifx compiler
 - Added support for compilation with the Fujitsu FCC compiler for Fugaku
 - Added support for compilation with the Cray C and Fortran compilers
 - reverted OpenMP threadpool behaviour in the exec_blas call to its state before 0.3.11, that is
   the threadpool will no longer grow or shrink on demand as the overhead for this is too big at least with
   GNU OpenMP. The adaptive behaviour introduced in 0.3.11 can still be requested at runtime by setting
   the environment variable OMP_ADAPTIVE
 - worked around spurious STFSM/CTFSM errors reported by the LAPACK testsuite

x86_64:
 - fixed determination of compiler support for AVX512 and removed the 0.3.19
   workaround for building SKYLAKEX kernels on Sandybridge hardware
 - fixed compilation for the SKYLAKEX target with gcc 6
 - fixed compilation of the CooperLake SBGEMM kernel with LLVM
 - fixed compilation of the SkyLakeX small matrix GEMM kernels with LLVM or ICC
 - fixed compilation of some BFLOAT16 kernels with CMAKE
 - added support for the Zhaoxin/Centaur KH40000 cpu
 - fixed a potential crash in the ZSYMV kernel used for all targets except generic
 - fixed gmake compilation for DYNAMIC_ARCH with a DYNAMIC_LIST including ATOM
 - fixed compilation of LAPACKE with the INTEGER64 option on Windows
 - added support for cross-compiling to individual Intel or AMD targets using CMAKE
   (previously only CORE2 supported, added targets are ATOM, PRESCOTT, NEHALEM, SANDYBRIDGE,
   HASWELL,SKYLAKEX, COOPERLAKE, SAPPHIRERAPIDS, OPTERON, BARCELONA, BULLDOZER, PILEDRIVER,
   STEAMROLLER,EXCAVATOR, ZEN)

SPARC:
 - worked around an overflow error in the DNRM2 kernel

POWER:
 - worked around an overflow error in the POWER6 DNRM2 kernel
 - fixed compilation on PPC440
 - fixed a performance regression in the level1 BLAS on POWER10
 - fixed the POWER10 ZGEMM kernel
 - fixed singlethreaded builds for POWER10
 - fixed compilation of the POWER10 DGEMV kernel with older gcc versions
 - enabled compilation of the BFLOAT16 kernels by default
 - enabled the small matrix kernels by default for DYNAMIC_ARCH builds
 - added a workaround for a miscompilation of the CDOT and ZDOT kernels by GCC 12

- RISCV:
 - fixed cpu autodetection logic

ARMV8:
 - added an SBGEMM kernel for Neoverse N2
 - worked around an overflow error in the DNRM2 kernel used on M1, NeoverseN1, ThunderX2T99
 - added support for ARM64 systems running MS Windows
 - added support for cross-compiling to the GENERIC ARMV8 target under CMAKE (Windows/MSVC)
 - fixed a performance regression in the generic ARMV8 DGEMM kernel introduced in 0.3.19
 - added initial support for the Apple M1 cpu under Linux
 - added initial support for the Phytium FT2000 cpu
 - added initial support for the Cortex A510, A710, X1 and X2 cpu
 - fixed an accidental mixup of cpu identifiers in the autodetection code introduced in 0.3.20
 - fixed linking of Apple M1 builds on macOS 12 and later with recent XCode
 - made Neoverse N2 available in DYNAMIC_ARCH builds

MIPS,MIPS64:
 - worked around an overflow error in the DNRM2 kernel

LOONGARCH64:
 - worked around an overflow error in the DNRM2 kernel
 - added preliminary support for the LOONGSON2K1000 cpu
 - added DYNAMIC_ARCH support

====================================================================
Version 0.3.20
 20-Feb-2022

general:
 - some code cleanup, with added casts etc.
 - fixed obtaining the cpu count with OpenMP and OMP_PROC_BIND unset
 - fixed pivot index calculation by ?LASWP for negative increments other than one
 - fixed input argument check in LAPACK ? GEQRT2
 - improved the check for a Fortran compiler in CMAKE builds
 - disabled building OpenBLAS' optimized versions of LAPACK complex SPMV,SPR,SYMV,SYR with NO_LAPACK=1
 - fixed building of LAPACK on certain distributed filesystems with parallel gmake
 - fixed building the shared library on MacOS with classic flang

x86_64:
 - fixed cross-compilation with CMAKE for CORE2 target
 - fixed miscompilation of AVX512 code in DYNAMIC_ARCH builds
 - added support for the "incidental" AVX512 hardware in Alder Lake when enabled in BIOS

E2K:
 - add new architecture (Russian Elbrus E2000 family)

SPARC:
 - fix IMIN/IMAX

ARMV8:
 - added SVE-enabled CGEMM and ZGEMM kernels for ARMV8SVE and A64FX
 - added support for Neoverse N2 and V1 cpus

MIPS,MIPS64:
 - fixed autodetection of MSA capability

LOONGARCH64:
 - added an optimized DGEMM kernel

====================================================================
Version 0.3.19
 19-Dec-2021
 
 general:
 - reverted unsafe TRSV/ZRSV optimizations introduced in 0.3.16
 - fixed a potential thread race in the thread buffer reallocation routines
   that were introduced in 0.3.18
 - fixed miscounting of thread pool size on Linux with OMP_PROC_BIND=TRUE
 - fixed CBLAS interfaces for CSROT/ZSROT and CROTG/ZROTG
 - made automatic library suffix for CMAKE builds with INTERFACE64 available
   to CBLAS-only builds

x86_64:
 - DYNAMIC_ARCH builds now fall back to the cpu with most similar capabilities
   when an unknown CPUID is encountered, instead of defaulting to Prescott
 - added cpu detection for Intel Alder Lake
 - added cpu detection for Intel Sapphire Rapids
 - added an optimized SBGEMM kernel for Sapphire Rapids
 - fixed DYNAMIC_ARCH builds on OSX with CMAKE
 - worked around DYNAMIC_ARCH builds made on Sandybridge failing on SkylakeX
 - fixed missing thread initialization for static builds on Windows/MSVC
 - fixed an excessive read in ZSYMV

POWER:
 - added support for POWER10 in big-endian mode
 - added support for building with CMAKE
 - added optimized SGEMM and DGEMM kernels for small matrix sizes

ARMV8:
 - added basic support and cputype detection for Fujitsu A64FX
 - added a generic ARMV8SVE target  
 - added SVE-enabled SGEMM and DGEMM kernels for ARMV8SVE and A64FX
 - added optimized CGEMM and ZGEMM kernels for Cortex A53 and A55 cpus
 - fixed cpuid detection for Apple M1 and improved performance
 - improved compiler flag setting in CMAKE builds

RISCV64:
 - fixed improper initialization in CSCAL/ZSCAL for strided access patterns

MIPS:
 - added a GENERIC target for MIPS32
 - added support for cross-compiling to MIPS32 on x86_64 using CMAKE

MIPS64:
 - fixed misdetection of MSA capability

====================================================================
Version 0.3.18
 02-Oct-2021

general:
 - when the build-time number of preconfigured threads is exceeded
   at runtime (typically by an external program calling BLAS functions
   from a larger number of threads in parallel), OpenBLAS will now 
   allocate an auxiliary control structure for up to 512 additional
   threads instead of aborting
 - added support for Loongson's LoongArch64 cpu architecture
 - fixed building OpenBLAS with CMAKE and -DBUILD_BFLOAT16=ON
 - added support for building OpenBLAS as a CMAKE subproject
 - added support for building for Windows/ARM64 targets with clang
 - improved support for building with the IBM xlf compiler
 - imported Reference-LAPACK PR 625 (out-of-bounds reads in ?LARRV)
 - imported Reference-LAPACK PR 597 for testsuite compatibility with
   LLVM's libomp

x86_64:
 - added SkylakeX S/DGEMM kernels for small problem sizes (M*N*K<=1000000)
 - added optimized SBGEMM for Intel Cooper Lake
 - reinstated the performance patch for AVX512 SGEMV_T with a proper fix
 - added a workaround for a gcc11 tree-vectorizer bug that caused spurious
   failures in the test programs for complex BLAS3 when compiling at -O3
   (the default for cmake "release" builds)
 - added support for runtime cpu count detection under Haiku OS
 - worked around a long-standing miscompilation issue of the Haswell DGEMV_T
   kernel with gcc that could produce NaN output in some corner cases

POWER:
 - improved performance of DASUM on POWER10

ARMV8:
 - fixed crashes (use of reserved register x18) on Apple M1 under OSX
 - fixed building with gcc releases earlier than 5.1

MIPS:
 - fixed building under BSD

MIPS64:
 - fixed building under BSD

====================================================================
Version 0.3.17
 15-Jul-2021

common:
 - reverted the optimization of SGEMV_N/DGEMV_N for small input sizes
   and consecutive arguments as it led to stack overflows on x86_64
   with some operating systems (notably OSX and Windows)
 
 x86_64:
  - reverted the performance patch for SGEMV_T on AVX512 as it caused
    wrong results in some applications
    
 SPARC:
  - fixed compilation with compilers other than gcc
====================================================================
Version 0.3.16
 11-Jul-2021

common:
 - drastically reduced the stack size requirements for running the LAPACK
   testsuite (Reference-LAPACK PR 553)
 - fixed spurious test failures in the LAPACK testsuite (Reference-LAPACK
   PR 564)
 - expressly setting DYNAMIC_ARCH=0 no longer enables dynamic_arch mode 
 - improved performance of xGER, xSPR, xSPR2, xSYR, xSYR2, xTRSV, SGEMV_N 
   and DGEMV_N, for small input sizes and consecutive arguments
 - improved performance of xGETRF, xPORTF and xPOTRI for small input sizes
   by disabling multithreading
 - fixed installing with BSD versions of the "install" utility

RISCV:
 - fixed the implementation of xIMIN
 - improved the performance of DSDOT
 - fixed linking of the tests on C910V with current vendor gcc

POWER:
- fixed SBGEMM computation for some odd value inputs
- fixed compilation for PPCG4, PPC970, POWER3, POWER4 and POWER5

x86_64:
 - improved performance of SGEMV_N and SGEMV_T for small N on AVX512-capable cpus
 - worked around a miscompilation of ZGEMM/ZTRMM on Sandybridge with old gcc
   versions
 - fixed compilation with MS Visual Studio versions older than 2017
 - fixed macro name collision with winnt.h from the latest Win10 SDK
 - added cpu type autodetection for Intel Ice Lake SP
 - fixed cpu type autodetection for Intel Tiger Lake
 - added cpu type autodetection for recent Centaur/Zhaoxin models
 - fixed compilation with musl libc

ARM64:
- fixed compilation with gcc/gfortran on the Apple M1
- fixed linking of the tests on FreeBSD
- fixed missing restore of a register in the recently rewritten DNRM2 kernel
  for ThunderX2 and Neoverse N1 that could cause spurious failures in e.g.
  DGEEV
- added compiler optimization flags for the EMAG8180
- added initial support for Cortex A55

ARM:
- fixed linking of the tests on FreeBSD

====================================================================
Version 0.3.15
  2-May-2021

common:
 - imported improvements and bugfixes from Reference-LAPACK 3.9.1
 - imported LAPACKE interface fixes from Reference-LAPACK PRs 534 + 537
 - fixed a problem in the cpu detection of 0.3.14 that prevented cross-compilation
 - fixed a sequence problem in the generation of softlinks to the library in GMAKE

RISC V:
 - fixed compilation on RISCV (missing entry in getarch)
 - fixed a potential division by zero in CROTG and ZROTG

POWER:
 - fixed LAPACK testsuite failures seen with the NVIDIA HPC compiler
 - improved CGEMM, DGEMM and ZGEMM performance on POWER10
 - added an optimized ZGEMV kernel for POWER10
 - fixed a potential division by zero in CROTG and ZROTG

x86_64:
 - added support for Intel Control-flow Enforcement Technology (CET)
 - reverted the DOMATCOPY_RT code to the generic C version
 - fixed a bug in the AVX512 SGEMM kernel introduced in 0.3.14
 - fixed misapplication of -msse flag to non-SSE cpus in DYNAMIC_ARCH
 - added support for compilation of the benchmarks on older OSX versions
 - fix propagation of the NO_AVX512 option in CMAKE builds
 - fix compilation of the AVX512 SGEMM kernel with clang-cl on Windows
 - fixed compilation of the CTESTs with INTERFACE64=1 (random faults on OSX)
 - corrected the Haswell DROT kernel to require AVX2/FMA3 rather than AVX512

ARM:
 - fixed a potential division by zero in CROTG and ZROTG
 - fixed a potential overflow in IMATCOPY/ZIMATCOPY and the CTESTs

ARM64:
 - fixed spurious reads outside the array in the SGEMM tcopy macro
 - fixed a potential division by zero in CROTG and ZROTG
 - fixed a segmentation fault in DYNAMIC_ARCH builds (reappeared in 0.3.14)

MIPS
 - fixed a potential division by zero in CROTG and ZROTG
 - fixed a potential overflow in IMATCOPY/ZIMATCOPY and the CTESTs

MIPS64:
 - fixed a potential division by zero in CROTG and ZROTG

SPARC:
 - fixed a potential division by zero in CROTG and ZROTG

====================================================================
Version 0.3.14
 17-Mar-2021
 
 common:
	* Fixed a race condition on thread shutdown in non-OpenMP builds
	* Fixed custom BUFFERSIZE option getting ignored in gmake builds
	* Fixed CMAKE compilation of the TRMM kernels for GENERIC platforms
	* Added CBLAS interfaces for CROTG, ZROTG, CSROT and ZDROT
	* Improved performance of OMATCOPY_RT across all platforms
	* Changed perl scripts to use env instead of a hardcoded /usr/bin/perl
	* Fixed potential misreading of the GCC compiler version in the build scripts
	* Fixed convergence problems in LAPACK complex GGEV/GGES (Reference-LAPACK #477)
	* Reduced the stacksize requirements for running the LAPACK testsuite (Reference-LAPACK #335)

 RISCV:
 	* Fixed compilation on RISCV (missing entry in getarch)
	
 POWER:
 	* Fixed compilation for DYNAMIC_ARCH with clang and with old gcc versions
	* Added support for compilation on FreeBSD/ppc64le
 	* Added optimized POWER10 kernels for SSCAL, DSCAL, CSCAL, ZSCAL
	* Added optimized POWER10 kernels for SROT, DROT, CDOT, SASUM, DASUM
	* Improved SSWAP, DSWAP, CSWAP, ZSWAP performance on POWER10
	* Improved SCOPY and CCOPY performance on POWER10
	* Improved SGEMM and DGEMM performance on POWER10
	* Added support for compilation with the NVIDIA HPC compiler

 x86_64:
	* Added an optimized bfloat16 GEMM kernel for Cooperlake
	* Added CPUID autodetection for Intel Rocket Lake and Tiger Lake cpus
	* Improved the performance of SASUM,DASUM,SROT,DROT on AMD Ryzen cpus
	* Added support for compilation with the NAG Fortran compiler
	* Fixed recognition of the AMD AOCC compiler
	* Fixed compilation for DYNAMIC_ARCH with clang on Windows
	* Added support for running the BLAS/CBLAS tests on Windows
	* Fixed signatures of the tls callback functions for Windows x64
	* Fixed various issues with fma intrinsics support handling 

 ARM:
	* Added support for embedded Cortex M targets via a new option EMBEDDED

 ARMV8:
 	* Fixed the THUNDERX2T99 and NEOVERSEN1 DNRM2/ZNRM2 kernels for inputs with Inf
	* Added support for the DYNAMIC_LIST option
	* Added support for compilation with the NVIDIA HPC compiler
 	* Added support for compiling with the NAG Fortran compiler

====================================================================
Version 0.3.13
 12-Dec-2020
 
 common:
	* Added a generic bfloat16 SBGEMV kernel
	* Fixed a potentially severe memory leak after fork in OpenMP builds
	  that was introduced in 0.3.12
	* Added detection of the Fujitsu Fortran compiler
	* Added detection of the (e)gfortran compiler on OpenBSD
	* Added support for overriding the default name of the library independently
	  from symbol suffixing in the gmake builds (already supported in cmake)

RISCV:
	* Added a RISC V port optimized for C910V

POWER:
	* Added optimized POWER10 kernels for SAXPY, CAXPY, SDOT, DDOT and DGEMV_N
	* Improved DGEMM performance on POWER10
	* Improved STRSM and DTRSM performance on POWER9 and POWER10
	* Fixed segmemtation faults in DYNAMIC_ARCH builds
 	* Fixed compilation with the PGI compiler

x86:
	* Fixed compilation of kernels that require SSE2 intrinsics since 0.3.12
	
x86_64:
	* Added an optimized bfloat16 SBGEMV kernel for SkylakeX and Cooperlake
	* Improved the performance of SASUM and DASUM kernels through parallelization
	* Improved the performance of SROT and DROT kernels
	* Improved the performance of multithreaded xSYRK
	* Fixed OpenMP builds that use the LLVM Clang compiler together with GNU gfortran
	  (where linking of both the LLVM libomp and GNU libgomp could lead to lockups or
	  wrong results)
	* Fixed miscompilations by old gcc 4.6
	* Fixed misdetection of AVX2 capability in some Sandybridge cpus
	* Fixed lockups in builds combining DYNAMIC_ARCH with TARGET=GENERIC on OpenBSD

ARM64:
	* Fixed segmemtation faults in DYNAMIC_ARCH builds

MIPS:
	* Improved kernels for Loongson 3R3 ("3A") and 3R4 ("3B") models, including MSA
	* Fixed bugs in the MSA kernels for CGEMM, CTRMM, CGEMV and ZGEMV
	* Added handling of zero increments in the MSA kernels for SSWAP and DSWAP
	* Added DYNAMIC_ARCH support for MIPS64 (currently Loongson3R3/3R4 only)

SPARC:
	* Fixed building 32 and 64 bit SPARC kernels with the SolarisStudio compilers

====================================================================
Version 0.3.12
 24-Oct-2020

common:
	* Fixed missing BLAS/LAPACK functions (inadvertently dropped during
	  the build system restructuring)
	* Fixed argument conversion macro in LAPACKE_zgesvdq (LAPACK #458)

POWER:
	* Added optimized SCOPY/CCOPY kernels for POWER10
	* Increased and unified the default size of the GEMM BUFFER
	* Fixed building for POWER10 in DYNAMIC_ARCH mode
	* POWER10 compatibility test now checks binutils version as well
	* Cleaned up compiler warnings

x86_64:
	* corrected compiler version checks for AVX2 compatibility
	* added compiler option -mavx2 for building with flang
	* fixed direct SGEMM pathway for small matrix sizes (broken by
	  the code refactoring in 0.3.11)
	* fixed unhandled partial register clobbers in several kernels
	  for AXPY,DOT,GEMV_N and GEMV_T flagged by gcc10 tree-vectorizer

ARMV8:
	* improved Apple Vortex support to include cross-compiling

====================================================================
Version 0.3.11
 17-Oct-2020

common:
 	* API change:
	  the newly added BFLOAT16 functions were renamed to use the
	  letter "B" instead of "H" to avoid potential confusion with
	  the IEEE "half precision float" type, i.e. the 0.3.10
	  SHGEMM is now SBGEMM and the corresponding build option
	  was changed from "BUILD_HALF" to "BUILD_BFLOAT16".
	* Reduced the default BLAS3_MEM_ALLOC_THRESHOLD (used as an upper
	  limit for placing temporary arrays on the stack) to be compatible
	  with a stack size of 1mb (as imposed by the JAVA runtime library) 
	* Added mixed-precision dot function SBDOT and utility functions
	  shstobf16, shdtobf16, sbf16tos and dbf16tod to convert between
	  single or double precision float arrays and bfloat16 arrays
	* Fixed prototypes of LAPACK_?ggsvp and LAPACK_?ggsvd functions
	  in lapack.h
	* Fixed underflow and rounding errors in LAPACK SLANV2 and DLANV2
	  (causing miscalculations in e.g. SHSEQR/DHSEQR, LAPACK issue #263)
	* Fixed workspace calculation in LAPACK ?GELQ (LAPACK issue #415)
	* Fixed several bugs in the LAPACK testsuite
	* Improved performance of TRMM and TRSM for certain problem sizes
	* Fixed infinite recursions and workspace miscalculations in ReLAPACK
	* CMAKE builds no longer require pkg-config for creating the .pc file
	* Makefile builds no longer misread NO_CBLAS=0 or NO_LAPACK=0 as 
	  enabling these options
	* Fixed detection of gfortran when invoked through an mpi wrapper
	* Improve thread reinitialization performance with OpenMP after a fork 
	* Added support for building only the subset of the library required
	  for a particular precision by specifying BUILD_SINGLE, BUILD_DOUBLE
	* Optional function name prefixes and suffixes are now correctly
	  reflected in the generated cblas.h
	* Added CMAKE build support for the LAPACK and multithreading tests

POWER:
	* Added optimized support for POWER10
	* Added support for compiling for POWER8 in 32bit mode
	* Added support for compilation with LLVM/clang
	* Added support for compilation with NVIDIA/PGI compilers
	* Fixed building on big-endian POWER8
	* Fixed miscompilation of ZDOTC by gcc10
	* Fixed alignment errors in the POWER8 SAXPY kernel
	* Improved CPU detection on AIX
	* Supported building with older compilers on POWER9

x86_64:
	* Added support for Intel Cooperlake
	* Added autodetection of AMD Renoir/Matisse/Zen3 cpus
	* Added autodetection of Intel Comet Lake cpus
	* Reimplemented ?sum, ?dot and daxpy using universal intrinsics
	* Reset the fpu state before using the fpu on Windows as a workaround
	  for a problem introduced in Windows 10 build 19041 (a.k.a. SDK 2004)
	* Fixed potentially undefined behaviour in the dot and gemv_t kernels
	* Fixed a potential segmentation fault in DYNAMIC_ARCH builds
	* Fixed building for ZEN with PGI/NVIDIA and AMD AOCC compilers
	
ARMV7:
	* Fixed cpu detection on BSD-like systems

ARMV8:
	* Added preliminary support for Apple Vortex cpus
	* Added support for the Cavium ThunderX3T110 cpu
	* Fixed cpu detection on BSD-like systems
	* Fixed compilation in -std=C18 mode

IBM Z:
	* Added support for compiling with the clang compiler
	* Improved GEMM performance on Z14

====================================================================
Version 0.3.10
 14-Jun-2020

common:
	* Improved thread locking behaviour in blas_server and parallel getrf                         
	* Imported bugfix 394 from LAPACK (spurious reference to "XERBL"
	  due to overlong lines)
	* Imported bugfix 403 from LAPACK (compile option "recursive" required
	  for correctness with Intel and PGI)
	* Imported bugfix 408 from LAPACK (wrong scaling in ZHEEQUB)
	* Imported bugfix 411 from LAPACK (infinite loop in LARGV/LARTG/LARTGP)
	* Fixed mismatches between BUFFERSIZE and GEMM_UNROLL parameters that
	  could lead to crashes at large matrix sizes  
	* Restored internal soname in dynamic libraries on FreeBSD and Dragonfly 
	* Added API (openblas_setaffinity) to set the thread affinity on Linux
	* Added initial infrastructure for half-precision floating point 
	  (bfloat16) support with a generic implementation of SHGEMM     
	* Added CMAKE build system support for building the cblas_Xgemm3m
	  functions
	* Fixed CMAKE support for building in a path with embedded spaces
	* Fixed CMAKE (non)handling of NO_EXPRECISION and MAX_STACK_ALLOC
	* Fixed GCC version detection in the Makefiles
	* Allowed overriding the names of AR, AS and LD in Makefile builds

POWER:
	* Fixed big-endian POWER8 ELFv2 builds on FreeBSD
	* Fixed GCC version checks and DYNAMIC_ARCH builds on POWER9
	* Fixed CMAKE build support for POWER9
	* fixed a potential race condition in the thread buffer allocation
	* Worked around LAPACK test failures on PPC G4

MIPS:
	* Fixed a potential race condition in the thread buffer allocation
	* Added support for MIPS 24K/24KE family based on P5600 kernels

MIPS64:
	* fixed a potential race condition in the thread buffer allocation
	* Added TARGET=GENERIC

ARMV7:
	* Fixed a race condition in the thread buffer allocation

ARMV8:
	* Fixed a race condition in the thread buffer allocation
	* Fixed zero initialisation in the assembly for SGEMM and DGEMM BETA
	* Improved performance of the ThunderX2 DAXPY kernel
	* Added an optimized SGEMM kernel for Cortex A53    
	* Fixed Makefile support for INTERFACE64 (8-byte integer)

x86_64:
	* Fixed a syntax error in the CMAKE setup for SkylakeX
	* Improved performance of STRSM on Haswell, SkylakeX and Ryzen
	* Improved SGEMM performance on SGEMM for workloads with ldc a
	  multiple of 1024
	* Improved DGEMM performance on Skylake X
	* Fixed unwanted AVX512-dependency of SGEMM in DYNAMIC_ARCH
	  builds created on SkylakeX
	* Removed data alignment requirement in the SSE2 copy kernels
	  that could cause spurious crashes
	* Added a workaround for an optimizer bug in AppleClang 11.0.3
	* Fixed LAPACK test failures due to wrong options for Intel Fortran
	* Fixed compilation and LAPACK test results with recent Flang
	  and AMD AOCC
	* Fixed DYNAMIC_ARCH builds with CMAKE on OS X
	* Fixed missing exports of cblas_i?amin, cblas_i?min, cblas_i?max,
	  cblas_?sum, cblas_?gemm3m in the shared library on OS 
	* Fixed reporting of cpu name in DYNAMIC_ARCH builds (would sometimes
	  show the name of an older generation chip supported by the same kernels)

IBM Z:
	* Improved performance of SGEMM/STRMM and DGEMM/DTRMM on Z14

====================================================================
Version 0.3.9
 1-Mar-2020
 
 common:
 	* Fixed a miscompilation of the GETRF functions with CMAKE
	* Imported bugfix 390 from LAPACK (missing NaN propagation in xCOMBSSQ)
	* The size of the memory buffer used for splitting GEMM tasks across
          multiple threads can now be configured in the build system.

POWER:
	* Fixed several compilation problems related to endianness
	  and ELF version on POWER8 and POWER9
	* Fixed use of the absolute value IAMIN/IAMAX instead of IMIN/IMAX
	* Fixed a race condition in the level3 blas code
	
MIPS64:
	* Fixed use of the absoltute value IAMIN/IAMAX instead of IMIN/IMAX
	
ARMV7:
	* Fixed a race condition in the level3 blas code
	* Fixed compilation on Android
ARMV8:
	* Added support for Ampere EMAG8180
	* Added support for Neoverse N1
	* Improved performance of the blas_lock function
	* Fixed a race condition in the level3 blas code
	* Fixed a performance regression on TSV110-based servers

x86_64:
	* Fixed a long-standing error with undeclared register overwrites
	  in the DSCAL microkernel for HASWELL,SKYLAKEX and ZEN
	* Fixed a long-standing bug in the SSE implementation of IAMAX
	* Fixed a CMAKE build failure with DYNAMIC_ARCH
	* Fixed cpu autodetection of Goldmont+, Cannon Lake and Ice Lake
	* Fixed a compilation failure on OSX with compiler name containing dash
	* Fixed compilation with MinGW on SkylakeX
	* Improved speed of the AVX512 GEMM3M kernel on SkylakeX
	* Added an AVX512 STRMM kernel for SkylakeX
	* Improved GEMM performance on Haswell and Zen

zarch:
	* fixed compilation of the DYNAMIC_ARCH code
	
====================================================================
Version 0.3.8
 9-Feb-2020

common: 
`	* LAPACK has been updated to 3.9.0 (plus patches up to 
	  January 2nd, 2020)
	* CMAKE support has been improved in several areas including
	  cross-compilation
	* a thread race condition in the GEMM3M kernels was resolved
	* the "generic" (plain C) gemm beta kernel used by many targets
	  has been sped up
	* an optimized version of the LAPACK trtrs functions has been added
	* an incompatibilty between the LAPACK tests and the OpenBLAS
	  implementation of XERBLA was resolved, removing the numerous
	  warnings about wrong error exits in the former 
	* support for NetBSD has been added
	* support for compilation with g95 and non-GNU versions of ld
	  has been improved
	* support for compilation with (upcoming) gcc 10 has been added

POWER:
	* worked around miscompilation of several POWER8 and POWER9 
	  kernels by older versions of gcc
	* added support for big-endian POWER8 and for compilation on AIX
	* corrected bugs in the big-endian support for PPC440 and PPC970
	* DYNAMIC_ARCH support is now available in CMAKE builds as well

ARMV8:
	* performance of DGEMM_BETA and SGEMM_NCOPY has been improved
	* compilation for 32bit works again 
	* performance of the RPCC function has been improved
	* improved performance on small systems
	* DYNAMIC_ARCH support is now available in CMAKE builds as well
	* cross-compilation from OSX to IOS was simplified
 
x86_64:
	* a new AVX512 DGEMM kernel was added and the AVX512 SGEMM kernel
	  was significantly improved
	* optimized AVX512 kernels for CGEMM and ZGEMM have been added
	* AVX2 kernels for STRMM, SGEMM, and CGEMM have been significantly
	  sped up and optimized CGEMM3M and ZGEMM3M kernels have been added 
	* added support for QEMU virtual cpus
	* a compilation problem with PGI and SUN compilers was fixed
	* Intel "Goldmont plus" is now autodetected
	* a potential crash on program exit on MS Windows has been fixed

x86:
	* an unwanted case sensitivity in the implementation of LSAME
  	  on older 32bit AMD cpus was fixed

zarch:
	* Z15 is now supported as Z14
	* DYNAMIC_ARCH is now available on ZARCH as well

====================================================================
Version 0.3.7
11-Aug 2019

common:
  * having the gmake special variables TARGET_ARCH or TARGET_MACH
    defined no longer causes build failures in ctest or utest
  * defining NO_AFFINITY or USE_TLS to 0 in gmake builds no longer
    has the same effect as setting them to 1
  * a new test program was added to allow checking the library for
    thread safety
  * a new option USE_LOCKING was added to ensure thread safety when
    OpenBLAS itself is built without multithreading but will be 
    called from multiple threads.
  * a build failure on Linux with glibc versions earlier than 2.5
    was fixed
  * a runtime error with CPU enumeration (and NO_AFFINITY not set)
    on glibc 2.6 was fixed
  * NO_AFFINITY was added to the CMAKE options (and defaults to being
    active on Linux, as in the gmake builds)

x86_64:
  * the build-time logic for detection of AVX512 availability in
    the processor and compiler was fixed
  * gmake builds on OSX now set the internal name of the library to 
    libopenblas.0.dylib (consistent with CMAKE)
  * the Haswell DGEMM kernel received a significant speedup through
    improved prefetch and load instructions
  * performance of DGEMM, DTRMM, DTRSM and ZDOT on Zen/Zen2 was markedly
    increased by avoiding vpermpd instructions
  * the SKYLAKEX (AVX512) DGEMM helper functions have now been disabled
    to fix remaining errors in DGEMM, DSYMM and DTRMM

POWER:
  * added support for building on FreeBSD/powerpc64 and FreeBSD/ppc970
  * added optimized kernels for POWER9 SGEMM and STRMM

ARMV7:
  * fixed the softfp implementations of xAMAX and IxAMAX
  * removed the predefined -march= flags on both ARMV5 and ARMV6 as
    they were appropriate for only a subset of platforms

====================================================================
Version 0.3.6
29-Apr-2019

common:
	* the build tools now check that a given cpu TARGET is actually valid
	* the build-time check of system features (c_check) has been made
  	  less dependent on particular perl features (this should mainly
  	  benefit building on Windows)
	* several problem with the ReLAPACK integration were fixed,
	  including INTERFACE64 support and building a shared library
	* building with CMAKE on BSD systems was improved
	* a non-absolute SUM function was added based on the
  	  existing optimized code for ASUM
	* CBLAS interfaces to the IxMIN and IxMAX functions were added
	* a name clash between LAPACKE and BOOST headers was resolved
	* CMAKE builds with OpenMP failed to include the appropriate getrf_parallel
	  kernels
	* a crash on thread (key) deletion with the USE_TLS=1 memory management
	  option was fixed
	* restored several earlier fixes, in particular for OpenMP performance,
  	  building on BSD, and calling fork on CYGWIN, which had inadvertently
  	  been dropped in the 0.3.3 rewrite of the memory management code.

x86_64:
	* the AVX512 DGEMM kernel has been disabled again due to unsolved problems
	* building with old versions of MSVC was fixed
	* it is now possible to build a static library on Windows with CMAKE
	* accessing environment variables on CYGWIN at run time was fixed
	* the CMAKE build system now recognizes 32bit userspace on 64bit hardware
	* Intel "Denverton" atom and Hygon "Dhyana" zen CPUs are now autodetected
	* building for DYNAMIC_ARCH with a DYNAMIC_LIST of targets is now supported
  	  with CMAKE as well
	* building for DYNAMIC_ARCH with GENERIC as the default target is now supported
	* a buffer overflow in the SSE GEMM kernel for Intel Nano targets was fixed
	* assembly bugs involving undeclared modification of input operands were fixed
  	  in the AXPY, DOT, GEMV, GER, SCAL, SYMV and TRSM microkernels for Nehalem,
	  Sandybridge, Haswell, Bulldozer and Piledriver. These would typically cause
	  test failures or segfaults when compiled with recent versions of gcc from 8 onward.
	* a similar bug was fixed in the blas_quickdivide code used to split workloads
	  in most functions
	* a bug in the IxMIN implementation for the GENERIC target made it return the result of IxMAX
	* fixed building on SkylakeX systems when either the compiler or the (emulated) operating
	  environment does not support AVX512
	* improved GEMM performance on ZEN targets

x86:
	* build failures caused by the recently added checks for AVX512 were fixed
	* an inline assembly bug involving undeclared modification of an input argument was
  	  fixed in the blas_quickdivide code used to split workloads in most functions
	* a bug in the IMIN implementation for the GENERIC target made it return the result of IMAX

MIPS32:
	* a bug in the IMIN implementation made it return the result of IMAX

POWER:
	* single precision BLAS1/2 functions have received optimized POWER8 kernels
	* POWER9 is now a separate target, with an optimized DGEMM/DTRMM kernel
	* building on PPC970 systems under OSX Leopard or Tiger is now supported
	* out-of-bounds memory accesses in the gemm_beta microkernels were fixed
	* building a shared library on AIX is now supported for POWER6
	* DYNAMIC_ARCH support has been added for POWER6 and newer

ARMv7:
	* corrected xDOT behaviour with zero INC_X or INC_Y
	* a bug in the IMIN implementation made it return the result of IMAX

ARMv8:
	* added support for HiSilicon TSV110 cpus
	* the CMAKE build system now recognizes 32bit userspace on 64bit hardware
	* cross-compilation with CMAKE now works again
	* a bug in the IMIN implementation made it return the result of IMAX
	* ARMV8 builds with the BINARY=32 option are now automatically handled as ARMV7

IBM Z:
	* optimized microkernels for single precicion BLAS1/2 functions have been added
	  for both Z13 and Z14

====================================================================
Version 0.3.5
31-Dec-2018

common:
	* loop unrolling in TRMV has been enabled again.
	* A domain error in the thread workload distribution for SYRK
	  has been fixed.
	* gmake builds will now automatically add -fPIC to the build
	  options if the platform requires it.
	* a pthreads key leakage (and associate crash on dlclose) in
	  the USE_TLS codepath was fixed.
	* building of the utest cases on systems that do not provide
	  an implementation of complex.h was fixed.
	  
x86_64:
	* the SkylakeX code was changed to compile on OSX.
	* unwanted application of the -march=skylake-avx512 option
	  to the common code parts of a DYNAMIC_ARCH build was fixed.
	* improved performance of SGEMM for small workloads on Skylake X.
	* performance of SGEMM and DGEMM was improved on Haswell.

ARMV8:
	* a configuration error that broke the CNRM2 kernel was corrected.
	* compilation of the GEMM kernels with CMAKE was fixed.
	* DYNAMIC_ARCH builds are now available with CMAKE as well.
	* using CMAKE for cross-compilation to the new cpu TARGETs
	  introduced in 0.3.4 now works.
	  
POWER:
	* a problem in cpu autodetection for AIX has been corrected.
	
====================================================================
Version 0.3.4
02-Dec-2018

common:
	* the new, experimental thread-local memory allocation had 
	  inadvertently been left enabled for gmake builds in 0.3.3
	  despite the announcement. It is now disabled by default, and
	  single-threaded builds will keep using the old allocator even
	  if the USE_TLS option is turned on.
	* OpenBLAS will now provide enough buffer space for at least 50
	  threads by default.
	* The output of openblas_get_config() now contains the version
	  number.
	* A serious thread safety bug in GEMV operation with small M and
	  large N size has been fixed.
	* The code will now automatically call blas_thread_init after a
	  fork if needed before handling a call to openblas_set_num_threads
	* Accesses to parallelized level3 functions from multiple callers
	  are now serialized to avoid thread races (unless using OpenMP).
	  This should provide better performance than the known-threadsafe
	  (but non-default) USE_SIMPLE_THREADED_LEVEL3 option.
	* When building LAPACK with gfortran, -frecursive is now (again)
	  enabled by default to ensure correct behaviour.
        * The OpenBLAS version cblas.h now supports both CBLAS_ORDER and
	  CBLAS_LAYOUT as the name of the matrix row/column order option.
	* Externally set LDFLAGS are now passed through to the final compile/link
	  steps to facilitate setting platform-specific linker flags.
	* A potential race condition during the build of LAPACK (that would 
	  usually manifest itself as a failure to build TESTING/MATGEN) has been 
	  fixed.
	* xHEMV has been changed to stay single-threaded for small input sizes
	  where the overhead of multithreading exceeds any possible gains
	* CSWAP and ZSWAP have been limited to a single thread except on ARMV8 or
	  ThunderX hardware with sizable input.
	* Linker flags for the PGI compiler have been updated
	* Behaviour of AXPY with zero increments is now handled in the C interface,
	  correcting the result on at least Intel Atom.
	* The result matrix from calling SGELSS with an all-zero input matrix is 
	  now zeroed completely.
	  
x86_64:
	* Autodetection of AMD Ryzen2 has been fixed (again).
        * CMAKE builds now support labeling of an INTERFACE64=1 build of
	  the library with the _64 suffix.
	* AVX512 version of DGEMM has been added and the AVX512 SGEMM kernel
	  has been sped up by rewriting with C intrinsics
	* Fixed compilation on RHEL5/CENTOS5 (issue with typename __WAIT_STATUS)
	
POWER:
	* added support for building on AIX (with gcc and GNU tools from AIX Toolbox).
	* CPU type detection has been implemented for AIX.
	* CPU type detection has been fixed for NETBSD.
	
MIPS64:
	* AXPY on LOONGSON3A has been corrected to pass "zero increment" utest.
	* DSDOT on LOONGSON3A has been fixed.
	* the SGEMM microkernel has been hardened against potential data loss.
	
ARMV8:
	* DYNAMic_ARCH support is now available for 64bit ARM
	* cross-compiling for ARMV8 under iOS now works.
	* cpu-specific code has been rearranged to make better use of both
	  hardware commonalities and model-specific compiler optimizations.
	* XGENE1 has been removed as a TARGET, superseded by the improved generic
	  ARMV8 support.
	
ARMV7:
	* Older assembly mnemonics have been converted to UAL form to allow
	  building with clang 7.0
	* Cross compiling LAPACKE for Android has been fixed again (broken by
	  update to LAPACK 3.7.0 some while ago).  
	  
====================================================================
Version 0.3.3
31-Aug-2018

common:
	* thread memory allocation has been switched back to the method
	  used before version 0.3.1 due to unexpected problems caused by
	  the new code under some circumstances. A new compile-time option
	  USE_TLS has been added to enable the new code, and it is hoped
	  that this can become the default again in the next version.
	* LAPAck PR272 has been integrated, which fixes spurious errors
	  in DSYEVR and related functions caused by missing conversion 
	  from ILAENV to ILAENV_2STAGE in several _2stage routines.
	* the cmake-generated OpenBLASConfig.cmake now uses correct case
	  for the name of the library
	* added support for Haiku OS  

x86_64:
	* added AVX512 implementations of SDOT, DDOT, SAXPY, DAXPY,
	  DSCAL, DGEMVN and DSYMVL 
	* added a workaround for a cygwin issue that prevented compilation
	  of AVX512 code
	  
IBM Z:
	* added autodetection of Z14
	* fixed TRMM errors in the generic target
	
====================================================================
Version 0.3.2
30-Jul-2018

common:
	* fixes for regressions caused by the rewrite of the thread
	  initialization code in 0.3.1

POWER:
	* fixed cpu autodetection for the BSDs

MIPS64:
	* fixed utest errors in AXPY, DSDOT, ROT and SWAP

x86_64:
	* added autodetection of AMD Ryzen 2
	* fixed build with older versions of MSVC

====================================================================
Version 0.3.1
01-Jul-2018

common:
	* rewritten thread initialization code with significantly reduced overhead
	* added CBLAS interfaces to the IxAMIN BLAS extension functions
	* fixed the lapack-test target
	* CMAKE builds now create an OpenBLASConfig.cmake file
	* ZAXPY now uses a single thread for small input sizes
	* the LAPACK code was updated from Reference-LAPACK/lapack#253
          (fixing LAPACKE interfaces to Aasen's functions)

POWER:
	* corrected CROT and ZROT behaviour with zero INC_X

ARMV7:
	* corrected xDOT behaviour with zero INC_X or INC_Y

x86_64:
	* retired some older targets of DYNAMIC_ARCH builds to a new option DYNAMIC_OLDER,
	  this affects PENRYN,DUNNINGTON,OPTERON,OPTERON_SSE3,BOBCAT,ATOM and NANO
	  (which will still be supported via the slower PRESCOTT kernels when this option is not set)
	* added an option DYNAMIC_LIST that (used in conjunction with DYNAMIC_ARCH) allows to 
	  specify the list of x86_64 targets to include. Any target not on the list will be supported 
	  by the Sandybridge or Nehalem kernels if available, or by Prescott.
	* improved SWITCH_RATIO on Haswell for increased GEMM throughput
	* added initial support for Intel Skylake X, including an AVX512 SGEMM kernel
	* added autodetection of Intel Cannon Lake series as Skylake X
	* added a default L2 cache size for hypervisors that return zero here (Chromebook)
	* fixed a name clash with recent Windows10 headers that broke the build with (at least)
	  recent mingw from MSYS2
	* fixed a link error in mixed clang/gfortran builds with OpenMP
	* updated the OSX deployment target to 10.8
	* switched on parallel make for builds on MS Windows by default

x86:
	* fixed SSWAP and DSWAP behaviour with zero INC_X and INC_Y

====================================================================
Version 0.3.0
23-May-2108

common:
	* fixed some more thread race and locking bugs
	* added preliminary support for calling an OpenMP build of the library from multiple threads
	* removed performance impact of thread locks added in 0.2.20 on OpenMP code
	* general code cleanup 
	* optimized DSDOT implementation
	* improved thread distribution for GEMM
	* corrected IMATCOPY/OMATCOPY implementation
	* fixed out-of-bounds accesses in the multithreaded xBMV/xPMV and SYMV implementations
	* cmake build improvements
	* pkgconfig file now contains build options
	* openblas_get_config() now reports USE_OPENMP and NUM_THREADS settings used for the build
	* corrections and improvements for systems with more than 64 cpus
	* LAPACK code updated to 3.8.0 including later fixes
	* added ReLAPACK, a recursive implementation of several LAPACK functions
	* Rewrote ROTMG to handle cases that the netlib code failed to address
	* Disabled (broken) multithreading code for xTRMV
	* corrected prototypes of complex CBLAS functions to make our cblas.h match the generally accepted standard
	* shared memory access failures on startup are now handled more gracefully
	* restored utests from earlier releases (and made them pass on all affected systems)

SPARC:
	* several fixes for cpu autodetection

POWER:
	* corrected vector register overwriting in several Power8 kernels
	* optimized additional BLAS functions

ARM:
	* added support for CortexA53 and A72 
	* added autodetection for ThunderX2T99
	* made most optimized kernels the default for generic ARMv8 targets 

x86_64:
	* parallelized DDOT kernel for Haswell
	* changed alignment directives in assembly kernels to boost performance on OSX
	* fixed register handling in the GEMV microkernels (bug exposed by gcc7)
	* added support for building on OpenBSD and Dragonfly 
	* updated compiler options to work with Intel release 2018
	* support fully optimized build with clang/flang on Microsoft Windows
	* fixed building on AIX

IBM Z:
	* added optimized BLAS 1/2 functions

MIPS:
	* fixed cpu autodetection helper code
	* added mips32 1004K cpu (Mediatek MT7621 and similar SoC)
	* added mips64 I6500 cpu

====================================================================
Version 0.2.20
24-Jul-2017

common:
        * Improved CMake support
        * Fixed several thread race and locking bugs
        * Fixed default LAPACK optimization level
        * Updated LAPACK to 3.7.0
        * Added ReLAPACK (https://github.com/HPAC/ReLAPACK, make BUILD_RELAPACK=1)

POWER:
        * Optimizations for Power9
        * Fixed several Power8 assembly bugs

ARM:
        * New optimized Vulcan and ThunderX2T99 targets
        * Support for ARMV7 SOFT_FP ABI (make ARM_SOFTFP_ABI=1)
        * Detect all cpu cores including offline ones
        * Fix compilation with CLANG
        * Support building a shared library for Android

MIPS:
        * Fixed several threading issues
        * Fix compilation with CLANG

x86_64:
        * Detect Intel Bay Trail and Apollo Lake
        * Detect Intel Sky Lake and Kaby Lake
        * Detect Intel Knights Landing
        * Detect AMD A8, A10, A12 and Ryzen
        * Support 64bit builds with Visual Studio
        * Fix building with Intel and PGI compilers
        * Fix building with MINGW and TDM-GCC
        * Fix cmake builds for Haswell and related cpus
        * Fix building for Sandybridge with CLANG 3.9
	* Add support for the FLANG compiler

IBM Z:
        * New target z13 with BLAS3 optimizations

====================================================================
Version 0.2.19
1-Sep-2016
common:
	* Improved cross compiling.
	* Fix the bug on musl libc.

POWER:
	* Optimize BLAS on Power8
	* Fixed Julia+OpenBLAS bugs on Power8

MIPS:
	* Optimize BLAS on MIPS P5600 and I6400 (Thanks, Shivraj Patil, Kaustubh Raste)

ARM:
	* Improved on ARM Cortex-A57. (Thanks, Ashwin Sekhar T K)


====================================================================
Version 0.2.18
12-Apr-2016
common:
	* If you set MAKE_NB_JOBS flag less or equal than zero,
	  make will be without -j.

x86/x86_64:
	* Support building Visual Studio static library. (#813, Thanks, theoractice)
	* Fix bugs to pass buidbot CI tests (http://build.openblas.net)

ARM:
	* Provide DGEMM 8x4 kernel for Cortex-A57 (Thanks, Ashwin Sekhar T K)

POWER:
	* Optimize S and C BLAS3 on Power8
	* Optimize BLAS2/1 on Power8

====================================================================
Version 0.2.17
20-Mar-2016
common:
	* Enable BUILD_LAPACK_DEPRECATED=1 by default.

====================================================================
Version 0.2.16
15-Mar-2016
common:
	* Avoid potential getenv segfault. (#716)
	* Import LAPACK svn bugfix #142-#147,#150-#155

x86/x86_64:
	* Optimize c/zgemv for AMD Bulldozer, Piledriver, Steamroller
	* Fix bug with scipy linalg test.

ARM:
	* Improve DGEMM for ARM Cortex-A57. (Thanks, Ashwin Sekhar T K)

POWER:
	* Optimize D and Z BLAS3 functions for Power8.

====================================================================
Version 0.2.16.rc1
23-Feb-2016
common:
	* Upgrade LAPACK to 3.6.0 version.
	  Add BUILD_LAPACK_DEPRECATED option in Makefile.rule to build
	  LAPACK deprecated functions.
	* Add MAKE_NB_JOBS option in Makefile.
	  Force number of make jobs.This is particularly
	  useful when using distcc. (#735. Thanks, Jerome Robert.)
	* Redesign unit test. Run unit/regression test at every build (Travis-CI and Appveyor).
	* Disable multi-threading for small size swap and ger. (#744. Thanks, Jerome Robert)
	* Improve small zger, zgemv, ztrmv using stack alloction (#727. Thanks, Jerome Robert)
	* Let openblas_get_num_threads return the number of active threads. 
	  (#760. Thanks, Jerome Robert)
	* Support illumos(OmniOS). (#749. Thanks, Lauri Tirkkonen)
	* Fix LAPACK Dormbr, Dormlq bug. (#711, #713. Thanks, Brendan Tracey)
	* Update scipy benchmark script. (#745. Thanks, John Kirkham)

x86/x86_64:
	* Optimize trsm kernels for AMD Bulldozer, Piledriver, Steamroller.
	* Detect Intel Avoton.
	* Detect AMD Trinity, Richland, E2-3200.
	* Fix gemv performance bug on Mac OSX Intel Haswell.
	* Fix some bugs with CMake and Visual Studio

ARM:
	* Support and optimize Cortex-A57 AArch64.
	  (#686. Thanks, Ashwin Sekhar TK)
	* Fix Android build on ARMV7 (#778. Thanks, Paul Mustiere)
	* Update ARMV6 kernels.

POWER:
	* Fix detection of POWER architecture
	  (#684. Thanks, Sebastien Villemot)

====================================================================
Version 0.2.15
27-Oct-2015
common:
	* Support cmake on x86/x86-64. Natively compiling on MS Visual Studio.
	  (experimental. Thank Hank Anderson for the initial cmake porting work.)
	  
	  On Linux and Mac OSX, OpenBLAS cmake supports assembly kernels.
	  e.g. cmake .
	       make
	       make test (Optional)

	  On Windows MS Visual Studio, OpenBLAS cmake only support C kernels.
	  (OpenBLAS uses AT&T style assembly, which is not supported by MSVC.)
	  e.g. cmake -G "Visual Studio 12 Win64" .
	       Open OpenBLAS.sln and build.
	  
	* Enable MAX_STACK_ALLOC flags by default.
	  Improve ger and gemv for small matrices.
	* Improve gemv parallel with small m and large n case.
	* Improve ?imatcopy when lda==ldb (#633. Thanks, Martin Koehler)
	* Add vecLib benchmarks (#565. Thanks, Andreas Noack.)
	* Fix LAPACK lantr for row major matrices (#634. Thanks, Dan Kortschak)
	* Fix LAPACKE lansy (#640. Thanks, Dan Kortschak)
	* Import bug fixes for LAPACKE s/dormlq, c/zunmlq 
	* Raise the signal when pthread_create fails (#668. Thanks, James K. Lowden)
	* Remove g77 from compiler list.
	* Enable AppVeyor Windows CI.

x86/x86-64:
	* Support pure C generic kernels for x86/x86-64.
	* Support Intel Boardwell and Skylake by Haswell kernels.
	* Support AMD Excavator by Steamroller kernels.
	* Optimize s/d/c/zdot for Intel SandyBridge and Haswell.
	* Optimize s/d/c/zdot for AMD Piledriver and Steamroller.
	* Optimize s/d/c/zapxy for Intel SandyBridge and Haswell.
	* Optimize s/d/c/zapxy for AMD Piledriver and Steamroller.
	* Optimize d/c/zscal for Intel Haswell, dscal for Intel SandyBridge.
	* Optimize d/c/zscal for AMD Bulldozer, Piledriver and Steamroller.
	* Optimize s/dger for Intel SandyBridge.
	* Optimize s/dsymv for Intel SandyBridge.
	* Optimize ssymv for Intel Haswell.
	* Optimize dgemv for Intel Nehalem and Haswell.
	* Optimize dtrmm for Intel Haswell.

ARM:
	* Support Android NDK armeabi-v7a-hard ABI (-mfloat-abi=hard)
	  e.g. make HOSTCC=gcc CC=arm-linux-androideabi-gcc NO_LAPACK=1 TARGET=ARMV7
	* Fix lock, rpcc bugs (#616, #617. Thanks, Grazvydas Ignotas)
POWER:
	* Support ppc64le platform (ELF ABI v2. #612. Thanks, Matthew Brandyberry.)
	* Support POWER7/8 by POWER6 kernels. (#612. Thanks, Fábio Perez.)

====================================================================
Version 0.2.14
24-Mar-2015
common:
	* Improve OpenBLASConfig.cmake. (#474, #475. Thanks, xantares.)
	* Improve ger and gemv for small matrices by stack allocation.
	  e.g. make -DMAX_STACK_ALLOC=2048 (#482. Thanks, Jerome Robert.)
	* Introduce openblas_get_num_threads and openblas_get_num_procs.
	  (#497. Thanks, Erik Schnetter.)
	* Add ATLAS-style ?geadd function. (#509. Thanks, Martin Köhler.)
	* Fix c/zsyr bug with negative incx. (#492.)
	* Fix race condition during shutdown causing a crash in
	  gotoblas_set_affinity(). (#508. Thanks, Ton van den Heuvel.)

x86/x86-64:
	* Support AMD Streamroller.

ARM:
	* Add Cortex-A9 and Cortex-A15 targets.

====================================================================
Version 0.2.13
3-Dec-2014
common:
	* Add SYMBOLPREFIX and SYMBOLSUFFIX makefile options
	  for adding a prefix or suffix to all exported symbol names
	  in the shared library.(#459, Thanks Tony Kelman)
	* Provide OpenBLASConfig.cmake at installation.
	* Fix Fortran compiler detection on FreeBSD.
	  (#470, Thanks Mike Nolta)


x86/x86-64:
	* Add generic kernel files for x86-64. make TARGET=GENERIC
	* Fix a bug of sgemm kernel on Intel Sandy Bridge.
	* Fix c_check bug on some amd64 systems. (#471, Thanks Mike Nolta)

ARM:
	* Support APM's X-Gene 1 AArch64 processors.
	  Optimize trmm and sgemm. (#465, Thanks Dave Nuechterlein)

====================================================================
Version 0.2.12
13-Oct-2014
common:
	* Added CBLAS interface for ?omatcopy and ?imatcopy.
	* Enable ?gemm3m functions.
	* Added benchmark for ?gemm3m.
	* Optimized multithreading lower limits.
	* Disabled SYMM3M and HEMM3M functions 
	  because of segment violations.

x86/x86-64:
	* Improved axpy and symv performance on AMD Bulldozer.
	* Improved gemv performance on modern Intel and AMD CPUs.

====================================================================
Version 0.2.11
18-Aug-2014
common:
	* Added some benchmark codes.
	* Fix link error on Linux/musl.(Thanks Isaac Dunham)

x86/x86-64:
	* Improved s/c/zgemm performance for Intel Haswell.
	* Improved s/d/c/zgemv performance.
	* Support the big numa machine.(EXPERIMENT)

ARM:
	* Fix detection when cpuinfo uses "Processor". (Thanks Isaiah)

====================================================================
Version 0.2.10
16-Jul-2014
common:
	* Added BLAS extensions as following.
	  s/d/c/zaxpby, s/d/c/zimatcopy, s/d/c/zomatcopy.
	* Added OPENBLAS_CORETYPE environment for dynamic_arch. (a86d34)
	* Added NO_AVX2 flag for old binutils. (#401)
	* Support outputing the CPU corename on runtime.(#407)
	* Patched LAPACK to fix bug 114, 117, 118. 
	  (http://www.netlib.org/lapack/bug_list.html)
	* Disabled ?gemm3m for a work-around fix. (#400)
x86/x86-64:
	* Fixed lots of bugs for optimized kernels on sandybridge,Haswell, 
	  bulldozer, and piledriver.
	  https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List

ARM:
	* Improved LAPACK testing.

====================================================================
Version 0.2.9
10-Jun-2014
common:
	* Improved the result for LAPACK testing. (#372)
	* Installed DLL to prefix/bin instead of prefix/lib. (#366)
	* Build import library on Windows.(#374)
x86/x86-64:
	* To improve LAPACK testing, we fallback some kernels. (#372)
	https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List

====================================================================
Version 0.2.9.rc2
06-Mar-2014
common:
	* Added OPENBLAS_VERBOSE environment variable.(#338)
	* Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
	  (#294, Thank Olivier Grisel)
	* Rewrote rotmg
	* Fixed sdsdot bug.
x86/x86-64:
	* Detect Intel Haswell for new Macbook.

====================================================================
Version 0.2.9.rc1
13-Jan-2013
common:
	* Update LAPACK to 3.5.0 version
	* Fixed compatiable issues with Clang and Pathscale compilers.

x86/x86-64:
	* Optimization on Intel Haswell.
	* Enable optimization kernels on AMD Bulldozer and Piledriver.

ARM:
	* Support ARMv6 and ARMv7 ISA.
	* Optimization on ARM Cortex-A9.

====================================================================
Version 0.2.8
01-Aug-2013
common:
	* Support Open64 5.0. (#266)
	* Add executable stack markings. (#262, Thank Sébastien Fabbro)
	* Respect user's LDFLAGS (Thank Sébastien Fabbro)

x86/x86-64:
	* Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
	  We will fix the compuational error bug in bulldozer and piledriver kernels.

====================================================================
Version 0.2.7
20-Jul-2013
common:
	* Support LSB (Linux Standard Base) 4.1.
          e.g. make CC=lsbcc
	* Include LAPACK 3.4.2 source codes to the repo.
          Avoid downloading at compile time.
	* Add NO_PARALLEL_MAKE flag to disable parallel make.
        * Create openblas_get_parallel to retrieve information which
          parallelization model is used by OpenBLAS. (Thank grisuthedragon)
	* Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
        * Change LIBSUFFIX from .lib to .a on windows.
	* A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)

x86/x86-64:
	* Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
          AMD Bulldozer. (Thank Werner Saar)
        * Add Intel Haswell support (using Sandybridge optimizations).
          (Thank Dan Luu)
	* Add AMD Piledriver support (using Bulldozer optimizations).
        * Fix the computational error in zgemm avx kernel on
          Sandybridge. (#237)
        * Fix the overflow bug in gemv.
	* Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
          is very large.(#214, #221, #246).
MIPS64:
        * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.

Power:
        * Support Power7 by old Power6 kernels. (#220)

====================================================================
Version 0.2.6
2-Mar-2013
common:
	* Improved OpenMP performance slightly. (d744c9)
	* Improved cblas.h compatibility with Intel MKL.(#185)
	* Fixed the overflowing bug in single thread cholesky factorization.
	* Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)

x86/x86-64:
	* Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
	  We will tune the performance in future.
	* Auto-detect Intel Xeon E7540.
	* Fixed the overflowing buffer bug of gemv. (#173)
	* Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)

MIPS64:

====================================================================
Version 0.2.5
26-Nov-2012
common:
	* Added NO_SHARED flag to disable generating the shared library.
	* Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
	* Export LAPACK 3.4.2 symbols in shared library. (#147)
	* Only detect the number of physical CPU cores on Mac OSX. (#157)
	* Fixed NetBSD build. (#155)
	* Fixed compilation with TARGET=GENERIC. (#160)
x86/x86-64:
	* Restore the original CPU affinity when calling
	  openblas_set_num_threads(1) (#153)
	* Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
MIPS64:

====================================================================
Version 0.2.4
8-Oct-2012
common:
	* Upgraded LAPACK to 3.4.2 version. (#145)
	* Provided support for passing CFLAGS, FFLAGS, PFLAGS,
	  FPFLAGS to make. (#137)
	* f77blas.h:compatibility for compilers without C99 complex
	  number support. (#141)
x86/x86-64:
	* Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
	* Fixed zdot incompatibility ABI issue with GCC 4.7 on
	  Windows 32-bit. (#140)
MIPS64:
	* Fixed the generation of shared library bug.
	* Fixed the detection bug on the Loongson 3A server.
====================================================================
Version 0.2.3
20-Aug-2012
common:
	* Fixed LAPACK unstable bug about ?laswp. (#130)
	* Fixed the shared library bug about unloading the library on
	  Linux (#132).
	* Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
	  Please use gcc and IBM xlf. (#134)
x86/x86-64:
	* Supported goto_set_num_threads and openblas_set_num_threads
	  APIs in Windows. They can set the number of threads on runtime.

====================================================================
Version 0.2.2
6-July-2012
common:
	* Fixed exporting DLL functions bug on Windows/MingW
	* Support GNU Hurd (Thank Sylvestre Ledru)
	* Support kfreebsd kernel (Thank Sylvestre Ledru)
x86/x86-64:
	* Support Intel Sandy Bridge 22nm desktop/mobile CPU
SPARC:
	* Improve the detection of SPARC (Thank Sylvestre Ledru)

====================================================================
Version 0.2.1
30-Jun-2012
common:
x86/x86-64:
	* Fixed the SEGFAULT bug about hyper-theading
	* Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes

====================================================================
Version 0.2.0
26-Jun-2012
common:
	* Removed the limitation (64) of numbers of CPU cores.
	  Now, it supports 256 cores at max.
	* Supported clang compiler.
	* Fixed some build bugs on FreeBSD
x86/x86-64:
	* Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
	  Please use gcc >= 4.6 or clang >=3.1.
	* Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.

====================================================================
Version 0.1.1
29-Apr-2012
common:
	* Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
	* Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
	* Fixed the build bug (MD5 and download) on Mac OSX.
	* Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
	* Fxied the compatibility issue for compilers without C99 complex number
	  (e.g. Visual Studio)
x86/x86_64:
	* Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
	* Test alpha=Nan in dscale.
	* Fixed a SEGFAULT bug in samax on x86 windows.

====================================================================
Version 0.1.0
23-Mar-2012
common:
	* Set soname of shared library on Linux.
	* Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
	  this flag to control the library name, e.g. libopenblas.a,
	  libopenblas_ifort.a or libopenblas_omp.a.
	* Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
	  The lib use single thread in GEMM function with small matrices.
x86/x86_64:
	* Used GEMV SSE/SSE2 kernels on x86 32-bit.
	* Exported CBLAS functions in Windows DLL.
MIPS64:
	* Completed Level-3 BLAS optimization on Loongson 3A CPU.
	* Improved GEMV performance on Loongson 3A CPU.
	* Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)

====================================================================
Version 0.1 alpha2.5
19-Feb-2012
common:
	* Fixed missing "#include <sched.h>" bug on Mac OS X.
	  Thank Mike Nolta for the patch.
	* Upgraded LAPACK to 3.4.0 version
	* Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
	  SystemStubs does not exist on Lion. Thank Stefan Karpinski.
	* Improved README with using OpenMP. Check the internal threads
	  count less than or equal to omp_get_max_threads()
x86/x86_64:
	* Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
	* Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.

====================================================================
Version 0.1 alpha2.4
18-Sep-2011
common:
	* Fixed a bug about installation. The header file "fblas77.h"
	  works fine now.
	* Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
	* Try to handle absolute path of shared library in OSX. (#57)
	  Thank Dr Kane O'Donnell.
	* Changed the installation folder layout to $(PREFIX)/include and
	  $(PREFIX)/lib

x86/x86_64:
	* Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
	  to i386 calling convention, The callee should remove the first
	  hidden parameter.Thank Mr. John for this patch.

====================================================================
Version 0.1 alpha2.3
5-Sep-2011

x86/x86_64:
	*  Added DTB_ENTRIES into dynamic arch setting parameters. Now,
	   it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)

====================================================================
Version 0.1 alpha2.2
14-Jul-2011

common:
	* Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
	  (Refs issue #44 on github)

====================================================================
Version 0.1 alpha2.1
28-Jun-2011

common:
	* Stop the build and output the error message when detecting
	  fortran compiler failed. (Refs issue #42 on github)

====================================================================
Version 0.1 alpha2
23-Jun-2011

common:
	* Fixed blasint undefined bug in <cblas.h> file. Other software
	  could include this header successfully(Refs issue #13 on github)
	* Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
	  of CPUs or cores should be less than or equal to 64.(Refs issue #14
	  on github)
	* Support "void goto_set_num_threads(int num_threads)" and "void
	  openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
	* Added extern "C" to support C++. Thank Tasio for the patch(Refs
	  issue #21 on github)
	* Provided an error message when the arch is not supported.(Refs
	  issue #19 on github)
	* Fixed issue #23. Fixed a bug of f_check script about generating link flags.
	* Added openblas_set_num_threads for Fortran.
	* Fixed #25 a wrong result of rotmg.
	* Fixed a bug about detecting underscore prefix in c_check.
	* Print the wall time (cycles) with enabling FUNCTION_PROFILE
	* Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
	* Added install target. You can use "make install". (Refs #20)


x86/x86_64:
	* Fixed #28 a wrong result of dsdot on x86_64.
	* Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
	* Fixed #33 ztrmm bug on Nehalem.
	* Work-around #27 the low performance axpy issue with small imput size & multithreads.

MIPS64:
	* Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
	* Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
	* Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)

====================================================================
Version 0.1 alpha1
20-Mar-2011

common:
	* Support "make  NO_LAPACK=1" to build the library without
	  LAPACK functions.
	* Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
	  Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
	* Added DEBUG=1 rule in Makefile.rule to build debug version.
	* Disable compiling quad precision in reference BLAS library(netlib BLAS).
	* Added unit testcases in utest/ subdir. Used  CUnit framework.
	* Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
	* Imported GotoBLAS2 1.13 BSD version

x86/x86_64:
	* On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
	  zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
	* Modified ?axpy functions to return same netlib BLAS results
	  when incx==0 or incy==0 (Refs issue #7 on github)
	* Modified ?swap functions to return same netlib BLAS results
	  when incx==0 or incy==0 (Refs issue #6 on github)
	* Modified ?rot functions to return same netlib BLAS results
	  when incx==0 or incy==0 (Refs issue #4 on github)
	* Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
	  to use Nehalem codes.
	* Fixed a typo bug about compiling dynamic ARCH library.
MIPS64:
	* Improve daxpy performance on ICT Loongson 3A.
	* Supported ICT Loongson 3A CPU (Refs issue #1 on github)
====================================================================