2 ====================================================================
6 * Added some benchmark codes.
7 * Fix link error on Linux/musl.(Thanks Isaac Dunham)
10 * Improved s/c/zgemm performance for Intel Haswell.
11 * Improved s/d/c/zgemv performance.
12 * Support the big numa machine.(EXPERIMENT)
15 * Fix detection when cpuinfo uses "Processor". (Thanks Isaiah)
17 ====================================================================
21 * Added BLAS extensions as following.
22 s/d/c/zaxpby, s/d/c/zimatcopy, s/d/c/zomatcopy.
23 * Added OPENBLAS_CORETYPE environment for dynamic_arch. (a86d34)
24 * Added NO_AVX2 flag for old binutils. (#401)
25 * Support outputing the CPU corename on runtime.(#407)
26 * Patched LAPACK to fix bug 114, 117, 118.
27 (http://www.netlib.org/lapack/bug_list.html)
28 * Disabled ?gemm3m for a work-around fix. (#400)
30 * Fixed lots of bugs for optimized kernels on sandybridge,Haswell,
31 bulldozer, and piledriver.
32 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
35 * Improved LAPACK testing.
37 ====================================================================
41 * Improved the result for LAPACK testing. (#372)
42 * Installed DLL to prefix/bin instead of prefix/lib. (#366)
43 * Build import library on Windows.(#374)
45 * To improve LAPACK testing, we fallback some kernels. (#372)
46 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
48 ====================================================================
52 * Added OPENBLAS_VERBOSE environment variable.(#338)
53 * Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
54 (#294, Thank Olivier Grisel)
58 * Detect Intel Haswell for new Macbook.
60 ====================================================================
64 * Update LAPACK to 3.5.0 version
65 * Fixed compatiable issues with Clang and Pathscale compilers.
68 * Optimization on Intel Haswell.
69 * Enable optimization kernels on AMD Bulldozer and Piledriver.
72 * Support ARMv6 and ARMv7 ISA.
73 * Optimization on ARM Cortex-A9.
75 ====================================================================
79 * Support Open64 5.0. (#266)
80 * Add executable stack markings. (#262, Thank Sébastien Fabbro)
81 * Respect user's LDFLAGS (Thank Sébastien Fabbro)
84 * Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
85 We will fix the compuational error bug in bulldozer and piledriver kernels.
87 ====================================================================
91 * Support LSB (Linux Standard Base) 4.1.
93 * Include LAPACK 3.4.2 source codes to the repo.
94 Avoid downloading at compile time.
95 * Add NO_PARALLEL_MAKE flag to disable parallel make.
96 * Create openblas_get_parallel to retrieve information which
97 parallelization model is used by OpenBLAS. (Thank grisuthedragon)
98 * Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
99 * Change LIBSUFFIX from .lib to .a on windows.
100 * A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)
103 * Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
104 AMD Bulldozer. (Thank Werner Saar)
105 * Add Intel Haswell support (using Sandybridge optimizations).
107 * Add AMD Piledriver support (using Bulldozer optimizations).
108 * Fix the computational error in zgemm avx kernel on
110 * Fix the overflow bug in gemv.
111 * Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
112 is very large.(#214, #221, #246).
114 * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.
117 * Support Power7 by old Power6 kernels. (#220)
119 ====================================================================
123 * Improved OpenMP performance slightly. (d744c9)
124 * Improved cblas.h compatibility with Intel MKL.(#185)
125 * Fixed the overflowing bug in single thread cholesky factorization.
126 * Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
129 * Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
130 We will tune the performance in future.
131 * Auto-detect Intel Xeon E7540.
132 * Fixed the overflowing buffer bug of gemv. (#173)
133 * Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
137 ====================================================================
141 * Added NO_SHARED flag to disable generating the shared library.
142 * Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
143 * Export LAPACK 3.4.2 symbols in shared library. (#147)
144 * Only detect the number of physical CPU cores on Mac OSX. (#157)
145 * Fixed NetBSD build. (#155)
146 * Fixed compilation with TARGET=GENERIC. (#160)
148 * Restore the original CPU affinity when calling
149 openblas_set_num_threads(1) (#153)
150 * Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
153 ====================================================================
157 * Upgraded LAPACK to 3.4.2 version. (#145)
158 * Provided support for passing CFLAGS, FFLAGS, PFLAGS,
159 FPFLAGS to make. (#137)
160 * f77blas.h:compatibility for compilers without C99 complex
161 number support. (#141)
163 * Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
164 * Fixed zdot incompatibility ABI issue with GCC 4.7 on
165 Windows 32-bit. (#140)
167 * Fixed the generation of shared library bug.
168 * Fixed the detection bug on the Loongson 3A server.
169 ====================================================================
173 * Fixed LAPACK unstable bug about ?laswp. (#130)
174 * Fixed the shared library bug about unloading the library on
176 * Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
177 Please use gcc and IBM xlf. (#134)
179 * Supported goto_set_num_threads and openblas_set_num_threads
180 APIs in Windows. They can set the number of threads on runtime.
182 ====================================================================
186 * Fixed exporting DLL functions bug on Windows/MingW
187 * Support GNU Hurd (Thank Sylvestre Ledru)
188 * Support kfreebsd kernel (Thank Sylvestre Ledru)
190 * Support Intel Sandy Bridge 22nm desktop/mobile CPU
192 * Improve the detection of SPARC (Thank Sylvestre Ledru)
194 ====================================================================
199 * Fixed the SEGFAULT bug about hyper-theading
200 * Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
202 ====================================================================
206 * Removed the limitation (64) of numbers of CPU cores.
207 Now, it supports 256 cores at max.
208 * Supported clang compiler.
209 * Fixed some build bugs on FreeBSD
211 * Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
212 Please use gcc >= 4.6 or clang >=3.1.
213 * Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
215 ====================================================================
219 * Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
220 * Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
221 * Fixed the build bug (MD5 and download) on Mac OSX.
222 * Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
223 * Fxied the compatibility issue for compilers without C99 complex number
226 * Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
227 * Test alpha=Nan in dscale.
228 * Fixed a SEGFAULT bug in samax on x86 windows.
230 ====================================================================
234 * Set soname of shared library on Linux.
235 * Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
236 this flag to control the library name, e.g. libopenblas.a,
237 libopenblas_ifort.a or libopenblas_omp.a.
238 * Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
239 The lib use single thread in GEMM function with small matrices.
241 * Used GEMV SSE/SSE2 kernels on x86 32-bit.
242 * Exported CBLAS functions in Windows DLL.
244 * Completed Level-3 BLAS optimization on Loongson 3A CPU.
245 * Improved GEMV performance on Loongson 3A CPU.
246 * Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)
248 ====================================================================
252 * Fixed missing "#include <sched.h>" bug on Mac OS X.
253 Thank Mike Nolta for the patch.
254 * Upgraded LAPACK to 3.4.0 version
255 * Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
256 SystemStubs does not exist on Lion. Thank Stefan Karpinski.
257 * Improved README with using OpenMP. Check the internal threads
258 count less than or equal to omp_get_max_threads()
260 * Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
261 * Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.
263 ====================================================================
267 * Fixed a bug about installation. The header file "fblas77.h"
269 * Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
270 * Try to handle absolute path of shared library in OSX. (#57)
271 Thank Dr Kane O'Donnell.
272 * Changed the installation folder layout to $(PREFIX)/include and
276 * Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
277 to i386 calling convention, The callee should remove the first
278 hidden parameter.Thank Mr. John for this patch.
280 ====================================================================
285 * Added DTB_ENTRIES into dynamic arch setting parameters. Now,
286 it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)
288 ====================================================================
293 * Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
294 (Refs issue #44 on github)
296 ====================================================================
301 * Stop the build and output the error message when detecting
302 fortran compiler failed. (Refs issue #42 on github)
304 ====================================================================
309 * Fixed blasint undefined bug in <cblas.h> file. Other software
310 could include this header successfully(Refs issue #13 on github)
311 * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
312 of CPUs or cores should be less than or equal to 64.(Refs issue #14
314 * Support "void goto_set_num_threads(int num_threads)" and "void
315 openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
316 * Added extern "C" to support C++. Thank Tasio for the patch(Refs
318 * Provided an error message when the arch is not supported.(Refs
320 * Fixed issue #23. Fixed a bug of f_check script about generating link flags.
321 * Added openblas_set_num_threads for Fortran.
322 * Fixed #25 a wrong result of rotmg.
323 * Fixed a bug about detecting underscore prefix in c_check.
324 * Print the wall time (cycles) with enabling FUNCTION_PROFILE
325 * Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
326 * Added install target. You can use "make install". (Refs #20)
330 * Fixed #28 a wrong result of dsdot on x86_64.
331 * Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
332 * Fixed #33 ztrmm bug on Nehalem.
333 * Work-around #27 the low performance axpy issue with small imput size & multithreads.
336 * Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
337 * Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
338 * Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)
340 ====================================================================
345 * Support "make NO_LAPACK=1" to build the library without
347 * Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
348 Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
349 * Added DEBUG=1 rule in Makefile.rule to build debug version.
350 * Disable compiling quad precision in reference BLAS library(netlib BLAS).
351 * Added unit testcases in utest/ subdir. Used CUnit framework.
352 * Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
353 * Imported GotoBLAS2 1.13 BSD version
356 * On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
357 zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
358 * Modified ?axpy functions to return same netlib BLAS results
359 when incx==0 or incy==0 (Refs issue #7 on github)
360 * Modified ?swap functions to return same netlib BLAS results
361 when incx==0 or incy==0 (Refs issue #6 on github)
362 * Modified ?rot functions to return same netlib BLAS results
363 when incx==0 or incy==0 (Refs issue #4 on github)
364 * Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
365 to use Nehalem codes.
366 * Fixed a typo bug about compiling dynamic ARCH library.
368 * Improve daxpy performance on ICT Loongson 3A.
369 * Supported ICT Loongson 3A CPU (Refs issue #1 on github)
370 ====================================================================