2 ====================================================================
6 * Added CBLAS interface for ?omatcopy and ?imatcopy.
7 * Enable ?gemm3m functions.
8 * Added benchmark for ?gemm3m.
9 * Optimized multithreading lower limits.
10 * Disabled SYMM3M and HEMM3M functions
11 because of segment violations.
14 * Improved axpy and symv performance on AMD Bulldozer.
15 * Improved gemv performance on modern Intel and AMD CPUs.
17 ====================================================================
21 * Added some benchmark codes.
22 * Fix link error on Linux/musl.(Thanks Isaac Dunham)
25 * Improved s/c/zgemm performance for Intel Haswell.
26 * Improved s/d/c/zgemv performance.
27 * Support the big numa machine.(EXPERIMENT)
30 * Fix detection when cpuinfo uses "Processor". (Thanks Isaiah)
32 ====================================================================
36 * Added BLAS extensions as following.
37 s/d/c/zaxpby, s/d/c/zimatcopy, s/d/c/zomatcopy.
38 * Added OPENBLAS_CORETYPE environment for dynamic_arch. (a86d34)
39 * Added NO_AVX2 flag for old binutils. (#401)
40 * Support outputing the CPU corename on runtime.(#407)
41 * Patched LAPACK to fix bug 114, 117, 118.
42 (http://www.netlib.org/lapack/bug_list.html)
43 * Disabled ?gemm3m for a work-around fix. (#400)
45 * Fixed lots of bugs for optimized kernels on sandybridge,Haswell,
46 bulldozer, and piledriver.
47 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
50 * Improved LAPACK testing.
52 ====================================================================
56 * Improved the result for LAPACK testing. (#372)
57 * Installed DLL to prefix/bin instead of prefix/lib. (#366)
58 * Build import library on Windows.(#374)
60 * To improve LAPACK testing, we fallback some kernels. (#372)
61 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
63 ====================================================================
67 * Added OPENBLAS_VERBOSE environment variable.(#338)
68 * Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
69 (#294, Thank Olivier Grisel)
73 * Detect Intel Haswell for new Macbook.
75 ====================================================================
79 * Update LAPACK to 3.5.0 version
80 * Fixed compatiable issues with Clang and Pathscale compilers.
83 * Optimization on Intel Haswell.
84 * Enable optimization kernels on AMD Bulldozer and Piledriver.
87 * Support ARMv6 and ARMv7 ISA.
88 * Optimization on ARM Cortex-A9.
90 ====================================================================
94 * Support Open64 5.0. (#266)
95 * Add executable stack markings. (#262, Thank Sébastien Fabbro)
96 * Respect user's LDFLAGS (Thank Sébastien Fabbro)
99 * Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
100 We will fix the compuational error bug in bulldozer and piledriver kernels.
102 ====================================================================
106 * Support LSB (Linux Standard Base) 4.1.
108 * Include LAPACK 3.4.2 source codes to the repo.
109 Avoid downloading at compile time.
110 * Add NO_PARALLEL_MAKE flag to disable parallel make.
111 * Create openblas_get_parallel to retrieve information which
112 parallelization model is used by OpenBLAS. (Thank grisuthedragon)
113 * Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
114 * Change LIBSUFFIX from .lib to .a on windows.
115 * A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)
118 * Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
119 AMD Bulldozer. (Thank Werner Saar)
120 * Add Intel Haswell support (using Sandybridge optimizations).
122 * Add AMD Piledriver support (using Bulldozer optimizations).
123 * Fix the computational error in zgemm avx kernel on
125 * Fix the overflow bug in gemv.
126 * Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
127 is very large.(#214, #221, #246).
129 * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.
132 * Support Power7 by old Power6 kernels. (#220)
134 ====================================================================
138 * Improved OpenMP performance slightly. (d744c9)
139 * Improved cblas.h compatibility with Intel MKL.(#185)
140 * Fixed the overflowing bug in single thread cholesky factorization.
141 * Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
144 * Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
145 We will tune the performance in future.
146 * Auto-detect Intel Xeon E7540.
147 * Fixed the overflowing buffer bug of gemv. (#173)
148 * Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
152 ====================================================================
156 * Added NO_SHARED flag to disable generating the shared library.
157 * Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
158 * Export LAPACK 3.4.2 symbols in shared library. (#147)
159 * Only detect the number of physical CPU cores on Mac OSX. (#157)
160 * Fixed NetBSD build. (#155)
161 * Fixed compilation with TARGET=GENERIC. (#160)
163 * Restore the original CPU affinity when calling
164 openblas_set_num_threads(1) (#153)
165 * Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
168 ====================================================================
172 * Upgraded LAPACK to 3.4.2 version. (#145)
173 * Provided support for passing CFLAGS, FFLAGS, PFLAGS,
174 FPFLAGS to make. (#137)
175 * f77blas.h:compatibility for compilers without C99 complex
176 number support. (#141)
178 * Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
179 * Fixed zdot incompatibility ABI issue with GCC 4.7 on
180 Windows 32-bit. (#140)
182 * Fixed the generation of shared library bug.
183 * Fixed the detection bug on the Loongson 3A server.
184 ====================================================================
188 * Fixed LAPACK unstable bug about ?laswp. (#130)
189 * Fixed the shared library bug about unloading the library on
191 * Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
192 Please use gcc and IBM xlf. (#134)
194 * Supported goto_set_num_threads and openblas_set_num_threads
195 APIs in Windows. They can set the number of threads on runtime.
197 ====================================================================
201 * Fixed exporting DLL functions bug on Windows/MingW
202 * Support GNU Hurd (Thank Sylvestre Ledru)
203 * Support kfreebsd kernel (Thank Sylvestre Ledru)
205 * Support Intel Sandy Bridge 22nm desktop/mobile CPU
207 * Improve the detection of SPARC (Thank Sylvestre Ledru)
209 ====================================================================
214 * Fixed the SEGFAULT bug about hyper-theading
215 * Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
217 ====================================================================
221 * Removed the limitation (64) of numbers of CPU cores.
222 Now, it supports 256 cores at max.
223 * Supported clang compiler.
224 * Fixed some build bugs on FreeBSD
226 * Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
227 Please use gcc >= 4.6 or clang >=3.1.
228 * Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
230 ====================================================================
234 * Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
235 * Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
236 * Fixed the build bug (MD5 and download) on Mac OSX.
237 * Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
238 * Fxied the compatibility issue for compilers without C99 complex number
241 * Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
242 * Test alpha=Nan in dscale.
243 * Fixed a SEGFAULT bug in samax on x86 windows.
245 ====================================================================
249 * Set soname of shared library on Linux.
250 * Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
251 this flag to control the library name, e.g. libopenblas.a,
252 libopenblas_ifort.a or libopenblas_omp.a.
253 * Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
254 The lib use single thread in GEMM function with small matrices.
256 * Used GEMV SSE/SSE2 kernels on x86 32-bit.
257 * Exported CBLAS functions in Windows DLL.
259 * Completed Level-3 BLAS optimization on Loongson 3A CPU.
260 * Improved GEMV performance on Loongson 3A CPU.
261 * Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)
263 ====================================================================
267 * Fixed missing "#include <sched.h>" bug on Mac OS X.
268 Thank Mike Nolta for the patch.
269 * Upgraded LAPACK to 3.4.0 version
270 * Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
271 SystemStubs does not exist on Lion. Thank Stefan Karpinski.
272 * Improved README with using OpenMP. Check the internal threads
273 count less than or equal to omp_get_max_threads()
275 * Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
276 * Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.
278 ====================================================================
282 * Fixed a bug about installation. The header file "fblas77.h"
284 * Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
285 * Try to handle absolute path of shared library in OSX. (#57)
286 Thank Dr Kane O'Donnell.
287 * Changed the installation folder layout to $(PREFIX)/include and
291 * Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
292 to i386 calling convention, The callee should remove the first
293 hidden parameter.Thank Mr. John for this patch.
295 ====================================================================
300 * Added DTB_ENTRIES into dynamic arch setting parameters. Now,
301 it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)
303 ====================================================================
308 * Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
309 (Refs issue #44 on github)
311 ====================================================================
316 * Stop the build and output the error message when detecting
317 fortran compiler failed. (Refs issue #42 on github)
319 ====================================================================
324 * Fixed blasint undefined bug in <cblas.h> file. Other software
325 could include this header successfully(Refs issue #13 on github)
326 * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
327 of CPUs or cores should be less than or equal to 64.(Refs issue #14
329 * Support "void goto_set_num_threads(int num_threads)" and "void
330 openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
331 * Added extern "C" to support C++. Thank Tasio for the patch(Refs
333 * Provided an error message when the arch is not supported.(Refs
335 * Fixed issue #23. Fixed a bug of f_check script about generating link flags.
336 * Added openblas_set_num_threads for Fortran.
337 * Fixed #25 a wrong result of rotmg.
338 * Fixed a bug about detecting underscore prefix in c_check.
339 * Print the wall time (cycles) with enabling FUNCTION_PROFILE
340 * Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
341 * Added install target. You can use "make install". (Refs #20)
345 * Fixed #28 a wrong result of dsdot on x86_64.
346 * Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
347 * Fixed #33 ztrmm bug on Nehalem.
348 * Work-around #27 the low performance axpy issue with small imput size & multithreads.
351 * Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
352 * Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
353 * Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)
355 ====================================================================
360 * Support "make NO_LAPACK=1" to build the library without
362 * Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
363 Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
364 * Added DEBUG=1 rule in Makefile.rule to build debug version.
365 * Disable compiling quad precision in reference BLAS library(netlib BLAS).
366 * Added unit testcases in utest/ subdir. Used CUnit framework.
367 * Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
368 * Imported GotoBLAS2 1.13 BSD version
371 * On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
372 zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
373 * Modified ?axpy functions to return same netlib BLAS results
374 when incx==0 or incy==0 (Refs issue #7 on github)
375 * Modified ?swap functions to return same netlib BLAS results
376 when incx==0 or incy==0 (Refs issue #6 on github)
377 * Modified ?rot functions to return same netlib BLAS results
378 when incx==0 or incy==0 (Refs issue #4 on github)
379 * Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
380 to use Nehalem codes.
381 * Fixed a typo bug about compiling dynamic ARCH library.
383 * Improve daxpy performance on ICT Loongson 3A.
384 * Supported ICT Loongson 3A CPU (Refs issue #1 on github)
385 ====================================================================