2 ====================================================================
6 * Added OPENBLAS_VERBOSE environment variable.(#338)
7 * Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
8 (#294, Thank Olivier Grisel)
12 * Detect Intel Haswell for new Macbook.
14 ====================================================================
18 * Update LAPACK to 3.5.0 version
19 * Fixed compatiable issues with Clang and Pathscale compilers.
22 * Optimization on Intel Haswell.
23 * Enable optimization kernels on AMD Bulldozer and Piledriver.
26 * Support ARMv6 and ARMv7 ISA.
27 * Optimization on ARM Cortex-A9.
29 ====================================================================
33 * Support Open64 5.0. (#266)
34 * Add executable stack markings. (#262, Thank Sébastien Fabbro)
35 * Respect user's LDFLAGS (Thank Sébastien Fabbro)
38 * Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
39 We will fix the compuational error bug in bulldozer and piledriver kernels.
41 ====================================================================
45 * Support LSB (Linux Standard Base) 4.1.
47 * Include LAPACK 3.4.2 source codes to the repo.
48 Avoid downloading at compile time.
49 * Add NO_PARALLEL_MAKE flag to disable parallel make.
50 * Create openblas_get_parallel to retrieve information which
51 parallelization model is used by OpenBLAS. (Thank grisuthedragon)
52 * Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
53 * Change LIBSUFFIX from .lib to .a on windows.
54 * A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)
57 * Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
58 AMD Bulldozer. (Thank Werner Saar)
59 * Add Intel Haswell support (using Sandybridge optimizations).
61 * Add AMD Piledriver support (using Bulldozer optimizations).
62 * Fix the computational error in zgemm avx kernel on
64 * Fix the overflow bug in gemv.
65 * Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
66 is very large.(#214, #221, #246).
68 * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.
71 * Support Power7 by old Power6 kernels. (#220)
73 ====================================================================
77 * Improved OpenMP performance slightly. (d744c9)
78 * Improved cblas.h compatibility with Intel MKL.(#185)
79 * Fixed the overflowing bug in single thread cholesky factorization.
80 * Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
83 * Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
84 We will tune the performance in future.
85 * Auto-detect Intel Xeon E7540.
86 * Fixed the overflowing buffer bug of gemv. (#173)
87 * Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
91 ====================================================================
95 * Added NO_SHARED flag to disable generating the shared library.
96 * Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
97 * Export LAPACK 3.4.2 symbols in shared library. (#147)
98 * Only detect the number of physical CPU cores on Mac OSX. (#157)
99 * Fixed NetBSD build. (#155)
100 * Fixed compilation with TARGET=GENERIC. (#160)
102 * Restore the original CPU affinity when calling
103 openblas_set_num_threads(1) (#153)
104 * Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
107 ====================================================================
111 * Upgraded LAPACK to 3.4.2 version. (#145)
112 * Provided support for passing CFLAGS, FFLAGS, PFLAGS,
113 FPFLAGS to make. (#137)
114 * f77blas.h:compatibility for compilers without C99 complex
115 number support. (#141)
117 * Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
118 * Fixed zdot incompatibility ABI issue with GCC 4.7 on
119 Windows 32-bit. (#140)
121 * Fixed the generation of shared library bug.
122 * Fixed the detection bug on the Loongson 3A server.
123 ====================================================================
127 * Fixed LAPACK unstable bug about ?laswp. (#130)
128 * Fixed the shared library bug about unloading the library on
130 * Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
131 Please use gcc and IBM xlf. (#134)
133 * Supported goto_set_num_threads and openblas_set_num_threads
134 APIs in Windows. They can set the number of threads on runtime.
136 ====================================================================
140 * Fixed exporting DLL functions bug on Windows/MingW
141 * Support GNU Hurd (Thank Sylvestre Ledru)
142 * Support kfreebsd kernel (Thank Sylvestre Ledru)
144 * Support Intel Sandy Bridge 22nm desktop/mobile CPU
146 * Improve the detection of SPARC (Thank Sylvestre Ledru)
148 ====================================================================
153 * Fixed the SEGFAULT bug about hyper-theading
154 * Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
156 ====================================================================
160 * Removed the limitation (64) of numbers of CPU cores.
161 Now, it supports 256 cores at max.
162 * Supported clang compiler.
163 * Fixed some build bugs on FreeBSD
165 * Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
166 Please use gcc >= 4.6 or clang >=3.1.
167 * Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
169 ====================================================================
173 * Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
174 * Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
175 * Fixed the build bug (MD5 and download) on Mac OSX.
176 * Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
177 * Fxied the compatibility issue for compilers without C99 complex number
180 * Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
181 * Test alpha=Nan in dscale.
182 * Fixed a SEGFAULT bug in samax on x86 windows.
184 ====================================================================
188 * Set soname of shared library on Linux.
189 * Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
190 this flag to control the library name, e.g. libopenblas.a,
191 libopenblas_ifort.a or libopenblas_omp.a.
192 * Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
193 The lib use single thread in GEMM function with small matrices.
195 * Used GEMV SSE/SSE2 kernels on x86 32-bit.
196 * Exported CBLAS functions in Windows DLL.
198 * Completed Level-3 BLAS optimization on Loongson 3A CPU.
199 * Improved GEMV performance on Loongson 3A CPU.
200 * Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)
202 ====================================================================
206 * Fixed missing "#include <sched.h>" bug on Mac OS X.
207 Thank Mike Nolta for the patch.
208 * Upgraded LAPACK to 3.4.0 version
209 * Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
210 SystemStubs does not exist on Lion. Thank Stefan Karpinski.
211 * Improved README with using OpenMP. Check the internal threads
212 count less than or equal to omp_get_max_threads()
214 * Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
215 * Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.
217 ====================================================================
221 * Fixed a bug about installation. The header file "fblas77.h"
223 * Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
224 * Try to handle absolute path of shared library in OSX. (#57)
225 Thank Dr Kane O'Donnell.
226 * Changed the installation folder layout to $(PREFIX)/include and
230 * Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
231 to i386 calling convention, The callee should remove the first
232 hidden parameter.Thank Mr. John for this patch.
234 ====================================================================
239 * Added DTB_ENTRIES into dynamic arch setting parameters. Now,
240 it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)
242 ====================================================================
247 * Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
248 (Refs issue #44 on github)
250 ====================================================================
255 * Stop the build and output the error message when detecting
256 fortran compiler failed. (Refs issue #42 on github)
258 ====================================================================
263 * Fixed blasint undefined bug in <cblas.h> file. Other software
264 could include this header successfully(Refs issue #13 on github)
265 * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
266 of CPUs or cores should be less than or equal to 64.(Refs issue #14
268 * Support "void goto_set_num_threads(int num_threads)" and "void
269 openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
270 * Added extern "C" to support C++. Thank Tasio for the patch(Refs
272 * Provided an error message when the arch is not supported.(Refs
274 * Fixed issue #23. Fixed a bug of f_check script about generating link flags.
275 * Added openblas_set_num_threads for Fortran.
276 * Fixed #25 a wrong result of rotmg.
277 * Fixed a bug about detecting underscore prefix in c_check.
278 * Print the wall time (cycles) with enabling FUNCTION_PROFILE
279 * Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
280 * Added install target. You can use "make install". (Refs #20)
284 * Fixed #28 a wrong result of dsdot on x86_64.
285 * Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
286 * Fixed #33 ztrmm bug on Nehalem.
287 * Work-around #27 the low performance axpy issue with small imput size & multithreads.
290 * Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
291 * Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
292 * Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)
294 ====================================================================
299 * Support "make NO_LAPACK=1" to build the library without
301 * Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
302 Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
303 * Added DEBUG=1 rule in Makefile.rule to build debug version.
304 * Disable compiling quad precision in reference BLAS library(netlib BLAS).
305 * Added unit testcases in utest/ subdir. Used CUnit framework.
306 * Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
307 * Imported GotoBLAS2 1.13 BSD version
310 * On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
311 zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
312 * Modified ?axpy functions to return same netlib BLAS results
313 when incx==0 or incy==0 (Refs issue #7 on github)
314 * Modified ?swap functions to return same netlib BLAS results
315 when incx==0 or incy==0 (Refs issue #6 on github)
316 * Modified ?rot functions to return same netlib BLAS results
317 when incx==0 or incy==0 (Refs issue #4 on github)
318 * Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
319 to use Nehalem codes.
320 * Fixed a typo bug about compiling dynamic ARCH library.
322 * Improve daxpy performance on ICT Loongson 3A.
323 * Supported ICT Loongson 3A CPU (Refs issue #1 on github)
324 ====================================================================