Xianyi Zhang [Fri, 11 Nov 2011 14:26:49 +0000 (14:26 +0000)]
Merge branch 'loongson3a' of github.com:xianyi/OpenBLAS into loongson3b
traz [Thu, 10 Nov 2011 15:38:48 +0000 (15:38 +0000)]
Add conjugate condition to gemv.
Xianyi Zhang [Wed, 9 Nov 2011 19:28:22 +0000 (19:28 +0000)]
Support detecting ICT Loongson-3B CPU.
Xianyi Zhang [Wed, 9 Nov 2011 19:08:29 +0000 (19:08 +0000)]
Merge branch 'develop' of github.com:xianyi/OpenBLAS into loongson3b
traz [Fri, 4 Nov 2011 19:32:21 +0000 (19:32 +0000)]
Fix the compute error of gemv when incx and incy are negative numbers.
traz [Thu, 3 Nov 2011 13:53:48 +0000 (13:53 +0000)]
Add complete gemv function on Loongson3a platform.
traits [Tue, 18 Oct 2011 10:44:23 +0000 (18:44 +0800)]
Fixed #66 the bug in zgemv kernel with transpose matrix on 64-bit MingW (Windows).
traits [Tue, 18 Oct 2011 02:23:17 +0000 (10:23 +0800)]
Ref #65. Fixed 64-bit Windows calling convention bug in cdot and zdot.
According to 64-bit Windows calling convention, the return value is in %rax instead of %xmm0 in cdot kernel.
In zdot, the caller allocates a memory space for return value and sets this memory address to the first hidden parameter. Thus, the callee (zdot) should assign the result to this memory space and return the memory address in %rax.
Xianyi Zhang [Sun, 16 Oct 2011 14:56:19 +0000 (22:56 +0800)]
Ref #62. In OpenMP implementation, check the return value of omp_get_max_threads().
It makes sure the return value as same as blas_cpu_numbers which is an internal global variable to store the number of threads in OpenBLAS.
traits [Sun, 9 Oct 2011 09:25:44 +0000 (17:25 +0800)]
Ref #63. Fixed generating DLL bug on ming-w64.
Xianyi Zhang [Sun, 9 Oct 2011 07:14:48 +0000 (15:14 +0800)]
ref #62. Added the user friendly message with USE_OPENMP=1. The users should use OMP_NUM_THREADS.
When OpenBLAS is compiled with USE_OPENMP=1, it ignores OPENBLAS_NUM_THREADS and GOTO_NUM_THREADS flags.Therefore, you should use OMP_NUM_THREADS.
Without setting OMP_NUM_THREADS, a process will use maximal number of threads on a computing node. Thus, if there are 2 processes on the computing node, the thread will contend against other threads on CPU cores. As a result, the application will hang.
traz [Mon, 26 Sep 2011 15:21:45 +0000 (15:21 +0000)]
Adding conditional compilation(#if defined(LOONGSON3A)) to avoid affecting the performance of other platforms.
traz [Fri, 23 Sep 2011 20:59:48 +0000 (20:59 +0000)]
Modify aligned address of sa and sb to improve the performance of multi-threads.
Xianyi [Sun, 18 Sep 2011 09:00:29 +0000 (17:00 +0800)]
Merge branch 'hotfix-0.1alpha2.4' into develop
Xianyi [Sat, 17 Sep 2011 21:46:08 +0000 (05:46 +0800)]
Prepared the document for 0.1 alpha 2.4 version.
Xianyi [Sat, 17 Sep 2011 21:07:00 +0000 (05:07 +0800)]
Change the installation folder into /include and /lib.
Xianyi [Sat, 17 Sep 2011 17:35:12 +0000 (01:35 +0800)]
Refs #57. Continue to fix absolute path issue about shared library on Mac OSX.
Used $(CURDIR) instead of pwd in generating shared library on Mac OSX.
Add more tips about this issue. Thank Dr Kane O'Donnell.
Xianyi [Fri, 16 Sep 2011 23:55:59 +0000 (07:55 +0800)]
Updated the document for 0.1 alpha 2.4.
Xianyi [Fri, 16 Sep 2011 23:32:10 +0000 (07:32 +0800)]
Merge branch 'develop' into hotfix-0.1alpha2.4
Xianyi [Fri, 16 Sep 2011 23:21:11 +0000 (07:21 +0800)]
Output the installation tip after building complete.
Xianyi [Fri, 16 Sep 2011 19:05:26 +0000 (03:05 +0800)]
Bump the version number.
Xianyi [Fri, 16 Sep 2011 18:58:01 +0000 (02:58 +0800)]
Refs #57. The bug about absolute path of shared library on Mac OSX.
OSX cann't use relative path in shared library. Thank Mr.Kane for this patch. The detail is in this link (https://github.com/xianyi/OpenBLAS/issues/57).
Xianyi [Fri, 16 Sep 2011 18:27:56 +0000 (02:27 +0800)]
Fixed #61 a building bug about setting TARGET and DYNAMIC_ARCH at the same time.
traz [Fri, 16 Sep 2011 17:50:40 +0000 (17:50 +0000)]
Complete all the complex single-precision functions of level3, but the performance needs further improve.
traz [Fri, 16 Sep 2011 16:08:39 +0000 (16:08 +0000)]
Add ctrmm part in cgemm_kernel_loongson3a_4x2_ps.S.
traz [Thu, 15 Sep 2011 16:08:23 +0000 (16:08 +0000)]
Complete cgemm function, but no optimization.
traz [Wed, 14 Sep 2011 20:00:35 +0000 (20:00 +0000)]
Fix some compute error.
traz [Wed, 14 Sep 2011 16:32:36 +0000 (16:32 +0000)]
Merge branch 'loongson3a' of github.com:xianyi/OpenBLAS into loongson3a
Zhang Xiianyi [Wed, 14 Sep 2011 15:52:51 +0000 (23:52 +0800)]
Fixed #58 zdot SEGFAULT bug with GCC-4.6. Thank Mr. John for this patch.
In i386 calling convention, the caller put the address of return value of zdot into the first hidden parameter.
Thus, the callee should delete this address before return.
Actually, I have fixed the same bug on x86/zdot_sse2.S (issue #32). However, that is not a good implementation which uses 3 instructions. Mr. John told me used "ret $0x4" to skip the first hidden address (4 bytes).
traz [Wed, 14 Sep 2011 15:32:25 +0000 (15:32 +0000)]
Use ps instructions in cgemm.
traits [Thu, 8 Sep 2011 16:57:47 +0000 (00:57 +0800)]
Merge branch 'hotfix-0.1alpha2.3' into develop
traits [Thu, 8 Sep 2011 16:55:04 +0000 (00:55 +0800)]
Merge branch 'hotfix-0.1alpha2.3'
Xianyi Zhang [Thu, 8 Sep 2011 16:39:34 +0000 (16:39 +0000)]
Refs #47. Fixed the seting parameter bug on Loongson 3A single thread version.
Xianyi Zhang [Tue, 6 Sep 2011 18:27:33 +0000 (18:27 +0000)]
Check the return value of pthread_create. Update the docs with known issue on Loongson 3A.
Xianyi Zhang [Tue, 6 Sep 2011 18:19:50 +0000 (18:19 +0000)]
Merge branch 'develop' into loongson3a
traits [Tue, 6 Sep 2011 06:14:07 +0000 (14:14 +0800)]
refs #55. Changed DTB_ENTRIES to DTB_DEFAULT_ENTRIES in x86 gemv_n kernel codes.
traz [Mon, 5 Sep 2011 16:31:40 +0000 (16:31 +0000)]
Merge branch 'loongson3a' of github.com:xianyi/OpenBLAS into loongson3a
traz [Mon, 5 Sep 2011 16:30:55 +0000 (16:30 +0000)]
Fixed #46. Initialize variables in cblat3.f and zblat3.f.
Xianyi Zhang [Mon, 5 Sep 2011 15:13:05 +0000 (15:13 +0000)]
Refs #47. On Loongson 3A, set DGEMM_R parameter depending on different number of threads. It would improve double precision BLAS3 on multi-threads.
traits [Mon, 5 Sep 2011 09:40:55 +0000 (17:40 +0800)]
Updated the doc about 0.1alpha2.3.
traits [Mon, 5 Sep 2011 09:37:07 +0000 (17:37 +0800)]
refs #55. Added DTB_ENTRIES into dynamic arch setting parameters. Now, it can read DTB_ENTRIES on runtime.
traz [Fri, 2 Sep 2011 19:41:06 +0000 (19:41 +0000)]
Fix an illegal instruction for strmm_RTLU.
traz [Fri, 2 Sep 2011 16:57:33 +0000 (16:57 +0000)]
Fix an error for strmm_LLTN.
traz [Fri, 2 Sep 2011 16:50:50 +0000 (16:50 +0000)]
Fix an error for strmm_LLTN.
traz [Fri, 2 Sep 2011 16:00:04 +0000 (16:00 +0000)]
Fix a compute error for strmm.
traz [Fri, 2 Sep 2011 15:28:01 +0000 (15:28 +0000)]
Fix stack-pointer bug for strmm.
traz [Fri, 2 Sep 2011 09:15:09 +0000 (09:15 +0000)]
Add strmm part.
traz [Thu, 1 Sep 2011 17:15:28 +0000 (17:15 +0000)]
Tuning mb, kb, nb size to get the best performance.
traz [Wed, 31 Aug 2011 21:24:03 +0000 (21:24 +0000)]
Using PS instructions to improve the performance of sgemm and it is 4.2Gflops now.
Xianyi Zhang [Wed, 31 Aug 2011 10:21:37 +0000 (18:21 +0800)]
Fixed the bug about installation. f77blas.h works OK now.
traz [Tue, 30 Aug 2011 20:57:00 +0000 (20:57 +0000)]
Modify compile options.
traz [Tue, 30 Aug 2011 20:54:19 +0000 (20:54 +0000)]
Using ps instruction.
traz [Mon, 18 Jul 2011 17:06:53 +0000 (17:06 +0000)]
Merge branch 'loongson3a' of github.com:xianyi/OpenBLAS into loongson3a
traz [Mon, 18 Jul 2011 17:03:38 +0000 (17:03 +0000)]
Complete all the plura single precision functions of level3 on Loongson3a, the performance is 2.3GFlops.
traits [Wed, 13 Jul 2011 17:09:21 +0000 (01:09 +0800)]
Merge branch 'hotfix-0.1alpha2.2' into develop
traits [Wed, 13 Jul 2011 17:03:09 +0000 (01:03 +0800)]
Merge branch 'hotfix-0.1alpha2.2'
traits [Wed, 13 Jul 2011 17:02:19 +0000 (01:02 +0800)]
Update the documents for 0.1alpha2.2 version.
traits [Wed, 13 Jul 2011 16:54:23 +0000 (00:54 +0800)]
Fixed #44 a makefile bug when DYNAMIC_ARCH=1 and INTERFACE64=1.
Xianyi Zhang [Thu, 7 Jul 2011 06:25:51 +0000 (14:25 +0800)]
Merge branch 'develop' into loongson3a
traits [Tue, 28 Jun 2011 07:46:55 +0000 (15:46 +0800)]
Merge branch 'hotfix-0.1alpha2.1' into develop
traits [Tue, 28 Jun 2011 07:43:08 +0000 (15:43 +0800)]
Merge branch 'hotfix-0.1alpha2.1'
traits [Tue, 28 Jun 2011 07:42:09 +0000 (15:42 +0800)]
Refs #42. Output the error message when detecting fortran compiler failed.
traz [Fri, 24 Jun 2011 09:28:12 +0000 (09:28 +0000)]
Merge branch 'loongson3a' of github.com:xianyi/OpenBLAS into loongson3a
traz [Fri, 24 Jun 2011 09:27:41 +0000 (09:27 +0000)]
Fix compute error in ztrmm.
traz [Thu, 23 Jun 2011 21:11:00 +0000 (21:11 +0000)]
Add ztrmm and ztrsm part on loongson3a. The average performance is 2.2G.
traz [Thu, 23 Jun 2011 10:46:58 +0000 (10:46 +0000)]
Change prefetch length of A and B, the performance is 2.1G now.
Xianyi Zhang [Thu, 23 Jun 2011 08:08:23 +0000 (16:08 +0800)]
Merge branch 'release-v0.1alpha2' into loongson3a
Xianyi Zhang [Thu, 23 Jun 2011 08:07:34 +0000 (16:07 +0800)]
Merge branch 'release-v0.1alpha2' into develop
traits [Thu, 23 Jun 2011 07:18:40 +0000 (15:18 +0800)]
Merge branch 'release-v0.1alpha2'
traits [Thu, 23 Jun 2011 07:16:24 +0000 (15:16 +0800)]
Fixed #38. Released v0.1 alpha2.
traits [Thu, 23 Jun 2011 07:09:34 +0000 (15:09 +0800)]
Refs #37. Updated REAME about the compatible issue with EKOPath compiler.
Xianyi Zhang [Wed, 22 Jun 2011 05:19:39 +0000 (13:19 +0800)]
Refs #39. Moved the shared lib (dll) to top directory in MingW64 compiler environment.
traz [Tue, 21 Jun 2011 22:16:23 +0000 (22:16 +0000)]
Improve zgemm performance from 1G to 1.8G, change block size in param.h.
Xianyi Zhang [Tue, 21 Jun 2011 17:52:20 +0000 (01:52 +0800)]
Refs #39. It's unnecessary to include sys/mman.h file in blas_server_omp.c.
Xianyi Zhang [Tue, 21 Jun 2011 10:06:13 +0000 (18:06 +0800)]
Refs #38. Prepare the docs with v0.1alpha2.
Xianyi Zhang [Tue, 21 Jun 2011 09:50:00 +0000 (17:50 +0800)]
Merge branch 'loongson3a' into release-v0.1alpha2
Xianyi Zhang [Tue, 21 Jun 2011 09:40:16 +0000 (17:40 +0800)]
Merge branch 'add_install_target' into develop
Xianyi Zhang [Tue, 21 Jun 2011 09:39:08 +0000 (17:39 +0800)]
Refs #20. Fixed the installation bug with DYNAMIC_ARCH=1.
Xianyi Zhang [Mon, 20 Jun 2011 10:40:05 +0000 (18:40 +0800)]
Merge branch 'add_install_target' into develop
Conflicts:
Changelog.txt
Xianyi Zhang [Mon, 20 Jun 2011 10:36:29 +0000 (18:36 +0800)]
Refs #20. Updated the docs.
Xianyi Zhang [Mon, 20 Jun 2011 10:35:35 +0000 (18:35 +0800)]
Fixed #20. Added install target in makefile. You can use "make install PREFIX=your_installation_directory".
Xianyi Zhang [Sun, 19 Jun 2011 04:07:31 +0000 (12:07 +0800)]
Updated gitignore file.
Xianyi Zhang [Sun, 19 Jun 2011 03:59:38 +0000 (11:59 +0800)]
Merge branch 'master' of github.com:xianyi/OpenBLAS into develop
Xianyi Zhang [Sun, 19 Jun 2011 03:55:29 +0000 (11:55 +0800)]
Fixed #27. Temporarily walk around axpy's low performance issue with small imput size & multithreads.
Xianyi Zhang [Sat, 11 Jun 2011 12:59:00 +0000 (05:59 -0700)]
Merge pull request #36 from pipping/master
Fixed the bug about USE_OPENMP=0 enabling OpenMP
Elias Pipping [Sat, 11 Jun 2011 12:36:16 +0000 (14:36 +0200)]
Make USE_OPENMP=0 disable openmp
Xianyi Zhang [Thu, 9 Jun 2011 14:59:49 +0000 (22:59 +0800)]
Fixed #35 a build bug with NO_LAPACK=1 DYNAMIC_ARCH=1 FC=gfortran. I forgot to test it with gfortran in last bug fixed commit.
Xianyi Zhang [Thu, 9 Jun 2011 03:38:59 +0000 (11:38 +0800)]
Fixed #35 a build bug with NO_LAPACK=1 & DYNAMIC_ARCH=1.
Xianyi Zhang [Thu, 9 Jun 2011 02:40:15 +0000 (10:40 +0800)]
Print the wall time (cycles) with enabling FUNCTION_PROFILE.
Wang Qian [Tue, 7 Jun 2011 04:53:25 +0000 (12:53 +0800)]
Fixed #33 ztrmm bug on Nehalem.
Xianyi [Fri, 3 Jun 2011 05:19:54 +0000 (13:19 +0800)]
Fixed #32 a SEGFAULT bug with gcc-4.6. According to i386 calling convention, The called funtion should remove the hidden return value address from the stack.
Xianyi Zhang [Mon, 30 May 2011 04:42:17 +0000 (12:42 +0800)]
Fixed #31 Shared library placement on Mac. Thank Mr.Viral B. Shah for this patch.
traz [Sat, 28 May 2011 09:48:34 +0000 (09:48 +0000)]
Fixed #30 strmm computational error on Loongson3A.
Xianyi Zhang [Fri, 27 May 2011 13:15:30 +0000 (21:15 +0800)]
Fixed the makefile bug about openblas_set_num_threads.
Xianyi Zhang [Fri, 27 May 2011 10:16:19 +0000 (18:16 +0800)]
Fixed a bug about detecting underscore prefix in c_check.
Xianyi Zhang [Fri, 27 May 2011 10:12:45 +0000 (18:12 +0800)]
Ingnore *.obj files in git.
traz [Fri, 27 May 2011 09:47:17 +0000 (09:47 +0000)]
Modify single precision compiler conditions, increasing single precision kernel code on Loongson3a.
traz [Wed, 18 May 2011 10:54:51 +0000 (10:54 +0000)]
Remove the useless code, modify code comments and format.
Xianyi Zhang [Tue, 17 May 2011 21:24:00 +0000 (21:24 +0000)]
Fixed #28. Convert the result to double precision in MIPS64 dsdot_k kernel.
traz [Sat, 14 May 2011 22:00:57 +0000 (22:00 +0000)]
Fixed #25 dtrmm and dtrsm computational error on Loongson3A.