2 ====================================================================
6 * Support cmake on x86/x86-64. Natively compiling on MS Visual Studio.
7 (experimental. Thank Hank Anderson for the initial cmake porting work.)
9 On Linux and Mac OSX, OpenBLAS cmake supports assembly kernels.
14 On Windows MS Visual Studio, OpenBLAS cmake only support C kernels.
15 (OpenBLAS uses AT&T style assembly, which is not supported by MSVC.)
16 e.g. cmake -G "Visual Studio 12 Win64" .
17 Open OpenBLAS.sln and build.
19 * Enable MAX_STACK_ALLOC flags by default.
20 Improve ger and gemv for small matrices.
21 * Improve gemv parallel with small m and large n case.
22 * Improve ?imatcopy when lda==ldb (#633. Thanks, Martin Koehler)
23 * Add vecLib benchmarks (#565. Thanks, Andreas Noack.)
24 * Fix LAPACK lantr for row major matrices (#634. Thanks, Dan Kortschak)
25 * Fix LAPACKE lansy (#640. Thanks, Dan Kortschak)
26 * Import bug fixes for LAPACKE s/dormlq, c/zunmlq
27 * Raise the signal when pthread_create fails (#668. Thanks, James K. Lowden)
28 * Remove g77 from compiler list.
29 * Enable AppVeyor Windows CI.
32 * Support pure C generic kernels for x86/x86-64.
33 * Support Intel Boardwell and Skylake by Haswell kernels.
34 * Support AMD Excavator by Steamroller kernels.
35 * Optimize s/d/c/zdot for Intel SandyBridge and Haswell.
36 * Optimize s/d/c/zdot for AMD Piledriver and Steamroller.
37 * Optimize s/d/c/zapxy for Intel SandyBridge and Haswell.
38 * Optimize s/d/c/zapxy for AMD Piledriver and Steamroller.
39 * Optimize d/c/zscal for Intel Haswell, dscal for Intel SandyBridge.
40 * Optimize d/c/zscal for AMD Bulldozer, Piledriver and Steamroller.
41 * Optimize s/dger for Intel SandyBridge.
42 * Optimize s/dsymv for Intel SandyBridge.
43 * Optimize ssymv for Intel Haswell.
44 * Optimize dgemv for Intel Nehalem and Haswell.
45 * Optimize dtrmm for Intel Haswell.
48 * Support Android NDK armeabi-v7a-hard ABI (-mfloat-abi=hard)
49 e.g. make HOSTCC=gcc CC=arm-linux-androideabi-gcc NO_LAPACK=1 TARGET=ARMV7
50 * Fix lock, rpcc bugs (#616, #617. Thanks, Grazvydas Ignotas)
52 * Support ppc64le platform (ELF ABI v2. #612. Thanks, Matthew Brandyberry.)
53 * Support POWER7/8 by POWER6 kernels. (#612. Thanks, Fábio Perez.)
55 ====================================================================
59 * Improve OpenBLASConfig.cmake. (#474, #475. Thanks, xantares.)
60 * Improve ger and gemv for small matrices by stack allocation.
61 e.g. make -DMAX_STACK_ALLOC=2048 (#482. Thanks, Jerome Robert.)
62 * Introduce openblas_get_num_threads and openblas_get_num_procs.
63 (#497. Thanks, Erik Schnetter.)
64 * Add ATLAS-style ?geadd function. (#509. Thanks, Martin Köhler.)
65 * Fix c/zsyr bug with negative incx. (#492.)
66 * Fix race condition during shutdown causing a crash in
67 gotoblas_set_affinity(). (#508. Thanks, Ton van den Heuvel.)
70 * Support AMD Streamroller.
73 * Add Cortex-A9 and Cortex-A15 targets.
75 ====================================================================
79 * Add SYMBOLPREFIX and SYMBOLSUFFIX makefile options
80 for adding a prefix or suffix to all exported symbol names
81 in the shared library.(#459, Thanks Tony Kelman)
82 * Provide OpenBLASConfig.cmake at installation.
83 * Fix Fortran compiler detection on FreeBSD.
84 (#470, Thanks Mike Nolta)
88 * Add generic kernel files for x86-64. make TARGET=GENERIC
89 * Fix a bug of sgemm kernel on Intel Sandy Bridge.
90 * Fix c_check bug on some amd64 systems. (#471, Thanks Mike Nolta)
93 * Support APM's X-Gene 1 AArch64 processors.
94 Optimize trmm and sgemm. (#465, Thanks Dave Nuechterlein)
96 ====================================================================
100 * Added CBLAS interface for ?omatcopy and ?imatcopy.
101 * Enable ?gemm3m functions.
102 * Added benchmark for ?gemm3m.
103 * Optimized multithreading lower limits.
104 * Disabled SYMM3M and HEMM3M functions
105 because of segment violations.
108 * Improved axpy and symv performance on AMD Bulldozer.
109 * Improved gemv performance on modern Intel and AMD CPUs.
111 ====================================================================
115 * Added some benchmark codes.
116 * Fix link error on Linux/musl.(Thanks Isaac Dunham)
119 * Improved s/c/zgemm performance for Intel Haswell.
120 * Improved s/d/c/zgemv performance.
121 * Support the big numa machine.(EXPERIMENT)
124 * Fix detection when cpuinfo uses "Processor". (Thanks Isaiah)
126 ====================================================================
130 * Added BLAS extensions as following.
131 s/d/c/zaxpby, s/d/c/zimatcopy, s/d/c/zomatcopy.
132 * Added OPENBLAS_CORETYPE environment for dynamic_arch. (a86d34)
133 * Added NO_AVX2 flag for old binutils. (#401)
134 * Support outputing the CPU corename on runtime.(#407)
135 * Patched LAPACK to fix bug 114, 117, 118.
136 (http://www.netlib.org/lapack/bug_list.html)
137 * Disabled ?gemm3m for a work-around fix. (#400)
139 * Fixed lots of bugs for optimized kernels on sandybridge,Haswell,
140 bulldozer, and piledriver.
141 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
144 * Improved LAPACK testing.
146 ====================================================================
150 * Improved the result for LAPACK testing. (#372)
151 * Installed DLL to prefix/bin instead of prefix/lib. (#366)
152 * Build import library on Windows.(#374)
154 * To improve LAPACK testing, we fallback some kernels. (#372)
155 https://github.com/xianyi/OpenBLAS/wiki/Fixed-optimized-kernels-To-do-List
157 ====================================================================
161 * Added OPENBLAS_VERBOSE environment variable.(#338)
162 * Make OpenBLAS thread-pool resilient to fork via pthread_atfork.
163 (#294, Thank Olivier Grisel)
167 * Detect Intel Haswell for new Macbook.
169 ====================================================================
173 * Update LAPACK to 3.5.0 version
174 * Fixed compatiable issues with Clang and Pathscale compilers.
177 * Optimization on Intel Haswell.
178 * Enable optimization kernels on AMD Bulldozer and Piledriver.
181 * Support ARMv6 and ARMv7 ISA.
182 * Optimization on ARM Cortex-A9.
184 ====================================================================
188 * Support Open64 5.0. (#266)
189 * Add executable stack markings. (#262, Thank Sébastien Fabbro)
190 * Respect user's LDFLAGS (Thank Sébastien Fabbro)
193 * Rollback bulldozer and piledriver kernels to barcelona kernels (#263)
194 We will fix the compuational error bug in bulldozer and piledriver kernels.
196 ====================================================================
200 * Support LSB (Linux Standard Base) 4.1.
202 * Include LAPACK 3.4.2 source codes to the repo.
203 Avoid downloading at compile time.
204 * Add NO_PARALLEL_MAKE flag to disable parallel make.
205 * Create openblas_get_parallel to retrieve information which
206 parallelization model is used by OpenBLAS. (Thank grisuthedragon)
207 * Detect LLVM/Clang compiler. The default compiler is Clang on Mac OS X.
208 * Change LIBSUFFIX from .lib to .a on windows.
209 * A work-around for dtrti_U single thread bug. Replace it with LAPACK codes. (#191)
212 * Optimize c/zgemm, trsm, dgemv_n, ddot, daxpy, dcopy on
213 AMD Bulldozer. (Thank Werner Saar)
214 * Add Intel Haswell support (using Sandybridge optimizations).
216 * Add AMD Piledriver support (using Bulldozer optimizations).
217 * Fix the computational error in zgemm avx kernel on
219 * Fix the overflow bug in gemv.
220 * Fix the overflow bug in multi-threaded BLAS3, getrf when NUM_THREADS
221 is very large.(#214, #221, #246).
223 * Support loongcc (Open64 based) compiler for ICT Loongson 3A/B.
226 * Support Power7 by old Power6 kernels. (#220)
228 ====================================================================
232 * Improved OpenMP performance slightly. (d744c9)
233 * Improved cblas.h compatibility with Intel MKL.(#185)
234 * Fixed the overflowing bug in single thread cholesky factorization.
235 * Fixed the overflowing buffer bug of multithreading hbmv and sbmv.(#174)
238 * Added AMD Bulldozer x86-64 S/DGEMM AVX kernels. (Thank Werner Saar)
239 We will tune the performance in future.
240 * Auto-detect Intel Xeon E7540.
241 * Fixed the overflowing buffer bug of gemv. (#173)
242 * Fixed the bug of s/cdot about invalid reading NAN on x86_64. (#189)
246 ====================================================================
250 * Added NO_SHARED flag to disable generating the shared library.
251 * Compile LAPACKE with ILP64 modle when INTERFACE64=1 (#158)
252 * Export LAPACK 3.4.2 symbols in shared library. (#147)
253 * Only detect the number of physical CPU cores on Mac OSX. (#157)
254 * Fixed NetBSD build. (#155)
255 * Fixed compilation with TARGET=GENERIC. (#160)
257 * Restore the original CPU affinity when calling
258 openblas_set_num_threads(1) (#153)
259 * Fixed a SEGFAULT bug in dgemv_t when m is very large.(#154)
262 ====================================================================
266 * Upgraded LAPACK to 3.4.2 version. (#145)
267 * Provided support for passing CFLAGS, FFLAGS, PFLAGS,
268 FPFLAGS to make. (#137)
269 * f77blas.h:compatibility for compilers without C99 complex
270 number support. (#141)
272 * Added NO_AVX flag. Check OS supporting AVX on runtime. (#139)
273 * Fixed zdot incompatibility ABI issue with GCC 4.7 on
274 Windows 32-bit. (#140)
276 * Fixed the generation of shared library bug.
277 * Fixed the detection bug on the Loongson 3A server.
278 ====================================================================
282 * Fixed LAPACK unstable bug about ?laswp. (#130)
283 * Fixed the shared library bug about unloading the library on
285 * Fixed the compilation failure on BlueGene/P (TARGET=PPC440FP2)
286 Please use gcc and IBM xlf. (#134)
288 * Supported goto_set_num_threads and openblas_set_num_threads
289 APIs in Windows. They can set the number of threads on runtime.
291 ====================================================================
295 * Fixed exporting DLL functions bug on Windows/MingW
296 * Support GNU Hurd (Thank Sylvestre Ledru)
297 * Support kfreebsd kernel (Thank Sylvestre Ledru)
299 * Support Intel Sandy Bridge 22nm desktop/mobile CPU
301 * Improve the detection of SPARC (Thank Sylvestre Ledru)
303 ====================================================================
308 * Fixed the SEGFAULT bug about hyper-theading
309 * Support AMD Bulldozer by using GotoBLAS2 AMD Barcelona codes
311 ====================================================================
315 * Removed the limitation (64) of numbers of CPU cores.
316 Now, it supports 256 cores at max.
317 * Supported clang compiler.
318 * Fixed some build bugs on FreeBSD
320 * Optimized Level-3 BLAS on Intel Sandy Bridge x86-64 by AVX instructions.
321 Please use gcc >= 4.6 or clang >=3.1.
322 * Support AMD Bobcat by using GotoBLAS2 AMD Barcelona codes.
324 ====================================================================
328 * Upgraded LAPACK to 3.4.1 version. (Thank Zaheer Chothia)
329 * Supported LAPACKE, a C interface to LAPACKE. (Thank Zaheer Chothia)
330 * Fixed the build bug (MD5 and download) on Mac OSX.
331 * Auto download CUnit 2.1.2-2 from SF.net with UTEST_CHECK=1.
332 * Fxied the compatibility issue for compilers without C99 complex number
335 * Auto-detect Intel Sandy Bridge Core i7-3xxx & Xeon E7 Westmere-EX.
336 * Test alpha=Nan in dscale.
337 * Fixed a SEGFAULT bug in samax on x86 windows.
339 ====================================================================
343 * Set soname of shared library on Linux.
344 * Added LIBNAMESUFFIX flag in Makefile.rule. The user can use
345 this flag to control the library name, e.g. libopenblas.a,
346 libopenblas_ifort.a or libopenblas_omp.a.
347 * Added GEMM_MULTITHREAD_THRESHOLD flag in Makefile.rule.
348 The lib use single thread in GEMM function with small matrices.
350 * Used GEMV SSE/SSE2 kernels on x86 32-bit.
351 * Exported CBLAS functions in Windows DLL.
353 * Completed Level-3 BLAS optimization on Loongson 3A CPU.
354 * Improved GEMV performance on Loongson 3A CPU.
355 * Improved Level-3 BLAS performance on Loongson 3B CPU. (EXPERIMENT)
357 ====================================================================
361 * Fixed missing "#include <sched.h>" bug on Mac OS X.
362 Thank Mike Nolta for the patch.
363 * Upgraded LAPACK to 3.4.0 version
364 * Fixed a bug on Mac OS X. Don't require SystemStubs on OS X.
365 SystemStubs does not exist on Lion. Thank Stefan Karpinski.
366 * Improved README with using OpenMP. Check the internal threads
367 count less than or equal to omp_get_max_threads()
369 * Auto-detect Intel Core i6/i7 (Sandy Bridge) CPU with Nehalem assembly kernels
370 * Fixed some bugs on MingW 64-bit including zgemv, cdot, zdot.
372 ====================================================================
376 * Fixed a bug about installation. The header file "fblas77.h"
378 * Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH.
379 * Try to handle absolute path of shared library in OSX. (#57)
380 Thank Dr Kane O'Donnell.
381 * Changed the installation folder layout to $(PREFIX)/include and
385 * Fixed #58 zdot/xdot SEGFAULT bug with GCC-4.6 on x86. According
386 to i386 calling convention, The callee should remove the first
387 hidden parameter.Thank Mr. John for this patch.
389 ====================================================================
394 * Added DTB_ENTRIES into dynamic arch setting parameters. Now,
395 it can read DTB_ENTRIES on runtime. (Refs issue #55 on github)
397 ====================================================================
402 * Fixed a building bug when DYNAMIC_ARCH=1 & INTERFACE64=1.
403 (Refs issue #44 on github)
405 ====================================================================
410 * Stop the build and output the error message when detecting
411 fortran compiler failed. (Refs issue #42 on github)
413 ====================================================================
418 * Fixed blasint undefined bug in <cblas.h> file. Other software
419 could include this header successfully(Refs issue #13 on github)
420 * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number
421 of CPUs or cores should be less than or equal to 64.(Refs issue #14
423 * Support "void goto_set_num_threads(int num_threads)" and "void
424 openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
425 * Added extern "C" to support C++. Thank Tasio for the patch(Refs
427 * Provided an error message when the arch is not supported.(Refs
429 * Fixed issue #23. Fixed a bug of f_check script about generating link flags.
430 * Added openblas_set_num_threads for Fortran.
431 * Fixed #25 a wrong result of rotmg.
432 * Fixed a bug about detecting underscore prefix in c_check.
433 * Print the wall time (cycles) with enabling FUNCTION_PROFILE
434 * Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1
435 * Added install target. You can use "make install". (Refs #20)
439 * Fixed #28 a wrong result of dsdot on x86_64.
440 * Fixed #32 a SEGFAULT bug of zdotc with gcc-4.6.
441 * Fixed #33 ztrmm bug on Nehalem.
442 * Work-around #27 the low performance axpy issue with small imput size & multithreads.
445 * Fixed #28 a wrong result of dsdot on Loongson3A/MIPS64.
446 * Optimized single/double precision BLAS Level3 on Loongson3A/MIPS64. (Refs #2)
447 * Optimized single/double precision axpy function on Loongson3A/MIPS64. (Refs #3)
449 ====================================================================
454 * Support "make NO_LAPACK=1" to build the library without
456 * Fixed randomly SEGFAULT when nodemask==NULL with above Linux 2.6.34.
457 Thank Mr.Ei-ji Nakama providing this patch. (Refs issue #12 on github)
458 * Added DEBUG=1 rule in Makefile.rule to build debug version.
459 * Disable compiling quad precision in reference BLAS library(netlib BLAS).
460 * Added unit testcases in utest/ subdir. Used CUnit framework.
461 * Supported OPENBLAS_* & GOTO_* environment variables (Pleas see README)
462 * Imported GotoBLAS2 1.13 BSD version
465 * On x86 32bits, fixed a bug in zdot_sse2.S line 191. This would casue
466 zdotu & zdotc failures. Instead, work-around it. (Refs issue #8 #9 on github)
467 * Modified ?axpy functions to return same netlib BLAS results
468 when incx==0 or incy==0 (Refs issue #7 on github)
469 * Modified ?swap functions to return same netlib BLAS results
470 when incx==0 or incy==0 (Refs issue #6 on github)
471 * Modified ?rot functions to return same netlib BLAS results
472 when incx==0 or incy==0 (Refs issue #4 on github)
473 * Detect Intel Westmere,Intel Clarkdale and Intel Arrandale
474 to use Nehalem codes.
475 * Fixed a typo bug about compiling dynamic ARCH library.
477 * Improve daxpy performance on ICT Loongson 3A.
478 * Supported ICT Loongson 3A CPU (Refs issue #1 on github)
479 ====================================================================