2 ====================================================================
6 * Improved the result for LAPACK testing. (#372)
7 * Installed DLL to prefix/bin instead of prefix/lib. (#366)
8 * Build import library on Windows.(#374)
10 * To improve LAPACK testing, we fallback some kernels. (#372)
11 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
13 ====================================================================
17 * Added OPENBLAS_VERBOSE environment variable.(#338)
18 * Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
19 (#294, Thank Olivier Grisel)
23 * Detect Intel Haswell for new Macbook.
25 ====================================================================
29 * Update LAPACK to 3.5.0 version
30 * Fixed compatiable issues with Clang and Pathscale compilers.
33 * Optimization on Intel Haswell.
34 * Enable optimization kernels on AMD Bulldozer and Piledriver.
37 * Support ARMv6 and ARMv7 ISA.
38 * Optimization on ARM Cortex-A9.
40 ====================================================================
44 * Support Open64 5.0. (#266)
45 * Add executable stack markings. (#262, Thank Sébastien Fabbro)
46 * Respect user's LDFLAGS (Thank Sébastien Fabbro)
49 * Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
50 We will fix the compuational error bug in bulldozer and piledriver kernels.
52 ====================================================================
56 * Support LSB (Linux Standard Base) 4.1.
58 * Include LAPACK 3.4.2 source codes to the repo.
59 Avoid downloading at compile time.
60 * Add NO_PARALLEL_MAKE flag to disable parallel make.
61 * Create openblas_get_parallel to retrieve information which
62 parallelization model is used by OpenBLAS. (Thank grisuthedragon)
63 * Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
64 * Change LIBSUFFIX from .lib to .a on windows.
65 * A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)
68 * Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
69 AMD Bulldozer. (Thank Werner Saar)
70 * Add Intel Haswell support (using Sandybridge optimizations).
72 * Add AMD Piledriver support (using Bulldozer optimizations).
73 * Fix the computational error in zgemm avx kernel on
75 * Fix the overflow bug in gemv.
76 * Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
77 is very large.(#214, #221, #246).
79 * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.
82 * Support Power7 by old Power6 kernels. (#220)
84 ====================================================================
88 * Improved OpenMP performance slightly. (d744c9)
89 * Improved cblas.h compatibility with Intel MKL.(#185)
90 * Fixed the overflowing bug in single thread cholesky factorization.
91 * Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
94 * Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
95 We will tune the performance in future.
96 * Auto-detect Intel Xeon E7540.
97 * Fixed the overflowing buffer bug of gemv. (#173)
98 * Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
102 ====================================================================
106 * Added NO_SHARED flag to disable generating the shared library.
107 * Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
108 * Export LAPACK 3.4.2 symbols in shared library. (#147)
109 * Only detect the number of physical CPU cores on Mac OSX. (#157)
110 * Fixed NetBSD build. (#155)
111 * Fixed compilation with TARGET=GENERIC. (#160)
113 * Restore the original CPU affinity when calling
114 openblas_set_num_threads(1) (#153)
115 * Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
118 ====================================================================
122 * Upgraded LAPACK to 3.4.2 version. (#145)
123 * Provided support for passing CFLAGS, FFLAGS, PFLAGS,
124 FPFLAGS to make. (#137)
125 * f77blas.h:compatibility for compilers without C99 complex
126 number support. (#141)
128 * Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
129 * Fixed zdot incompatibility ABI issue with GCC 4.7 on
130 Windows 32-bit. (#140)
132 * Fixed the generation of shared library bug.
133 * Fixed the detection bug on the Loongson 3A server.
134 ====================================================================
138 * Fixed LAPACK unstable bug about ?laswp. (#130)
139 * Fixed the shared library bug about unloading the library on
141 * Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
142 Please use gcc and IBM xlf. (#134)
144 * Supported goto_set_num_threads and openblas_set_num_threads
145 APIs in Windows. They can set the number of threads on runtime.
147 ====================================================================
151 * Fixed exporting DLL functions bug on Windows/MingW
152 * Support GNU Hurd (Thank Sylvestre Ledru)
153 * Support kfreebsd kernel (Thank Sylvestre Ledru)
155 * Support Intel Sandy Bridge 22nm desktop/mobile CPU
157 * Improve the detection of SPARC (Thank Sylvestre Ledru)
159 ====================================================================
164 * Fixed the SEGFAULT bug about hyper-theading
165 * Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
167 ====================================================================
171 * Removed the limitation (64) of numbers of CPU cores.
172 Now, it supports 256 cores at max.
173 * Supported clang compiler.
174 * Fixed some build bugs on FreeBSD
176 * Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
177 Please use gcc >= 4.6 or clang >=3.1.
178 * Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
180 ====================================================================
184 * Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
185 * Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
186 * Fixed the build bug (MD5 and download) on Mac OSX.
187 * Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
188 * Fxied the compatibility issue for compilers without C99 complex number
191 * Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
192 * Test alpha=Nan in dscale.
193 * Fixed a SEGFAULT bug in samax on x86 windows.
195 ====================================================================
199 * Set soname of shared library on Linux.
200 * Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
201 this flag to control the library name, e.g. libopenblas.a,
202 libopenblas_ifort.a or libopenblas_omp.a.
203 * Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
204 The lib use single thread in GEMM function with small matrices.
206 * Used GEMV SSE/SSE2 kernels on x86 32-bit.
207 * Exported CBLAS functions in Windows DLL.
209 * Completed Level-3 BLAS optimization on Loongson 3A CPU.
210 * Improved GEMV performance on Loongson 3A CPU.
211 * Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)
213 ====================================================================
217 * Fixed missing "#include <sched.h>" bug on Mac OS X.
218 Thank Mike Nolta for the patch.
219 * Upgraded LAPACK to 3.4.0 version
220 * Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
221 SystemStubs does not exist on Lion. Thank Stefan Karpinski.
222 * Improved README with using OpenMP. Check the internal threads
223 count less than or equal to omp_get_max_threads()
225 * Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
226 * Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.
228 ====================================================================
232 * Fixed a bug about installation. The header file "fblas77.h"
234 * Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
235 * Try to handle absolute path of shared library in OSX. (#57)
236 Thank Dr Kane O'Donnell.
237 * Changed the installation folder layout to $(PREFIX)/include and
241 * Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
242 to i386 calling convention, The callee should remove the first
243 hidden parameter.Thank Mr. John for this patch.
245 ====================================================================
250 * Added DTB_ENTRIES into dynamic arch setting parameters. Now,
251 it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)
253 ====================================================================
258 * Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
259 (Refs issue #44 on github)
261 ====================================================================
266 * Stop the build and output the error message when detecting
267 fortran compiler failed. (Refs issue #42 on github)
269 ====================================================================
274 * Fixed blasint undefined bug in <cblas.h> file. Other software
275 could include this header successfully(Refs issue #13 on github)
276 * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
277 of CPUs or cores should be less than or equal to 64.(Refs issue #14
279 * Support "void goto_set_num_threads(int num_threads)" and "void
280 openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
281 * Added extern "C" to support C++. Thank Tasio for the patch(Refs
283 * Provided an error message when the arch is not supported.(Refs
285 * Fixed issue #23. Fixed a bug of f_check script about generating link flags.
286 * Added openblas_set_num_threads for Fortran.
287 * Fixed #25 a wrong result of rotmg.
288 * Fixed a bug about detecting underscore prefix in c_check.
289 * Print the wall time (cycles) with enabling FUNCTION_PROFILE
290 * Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
291 * Added install target. You can use "make install". (Refs #20)
295 * Fixed #28 a wrong result of dsdot on x86_64.
296 * Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
297 * Fixed #33 ztrmm bug on Nehalem.
298 * Work-around #27 the low performance axpy issue with small imput size & multithreads.
301 * Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
302 * Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
303 * Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)
305 ====================================================================
310 * Support "make NO_LAPACK=1" to build the library without
312 * Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
313 Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
314 * Added DEBUG=1 rule in Makefile.rule to build debug version.
315 * Disable compiling quad precision in reference BLAS library(netlib BLAS).
316 * Added unit testcases in utest/ subdir. Used CUnit framework.
317 * Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
318 * Imported GotoBLAS2 1.13 BSD version
321 * On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
322 zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
323 * Modified ?axpy functions to return same netlib BLAS results
324 when incx==0 or incy==0 (Refs issue #7 on github)
325 * Modified ?swap functions to return same netlib BLAS results
326 when incx==0 or incy==0 (Refs issue #6 on github)
327 * Modified ?rot functions to return same netlib BLAS results
328 when incx==0 or incy==0 (Refs issue #4 on github)
329 * Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
330 to use Nehalem codes.
331 * Fixed a typo bug about compiling dynamic ARCH library.
333 * Improve daxpy performance on ICT Loongson 3A.
334 * Supported ICT Loongson 3A CPU (Refs issue #1 on github)
335 ====================================================================