OpenBLAS ChangeLog
====================================================================
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)
====================================================================